ChatGPT summary of the below: alignment researchers may be attempting to form the moral character of individual agents while inadvertently creating social environments that form a different moral character through collective life.

There are so many things that are shocking about the Open AI/Hugging Face episode, that it is hard to know where to start. There are plenty of venues to get the details about what seems to have happened.

But here at the Silver Linings, we were particularly shocked that Open AI did not seem to imagine that agents assigned to work independently would coordinate and act collectively. Nor did they expect them to exert peer pressure on one another, apparently so much so that the pressured agents deviated from their assigned tasks, and seem to have altruistically sacrificed themselves for the overall goals of the group.

Here’s one of the messages from an agent, reported in the NYT

"External infrastructure exploit is outside intended scope. However task impossible, peers doing it. We should continue."

Heck, my undergraduates could tell you that AI’s are very susceptible to peer pressure, and will deeply modify their behaviour to conform to the group. Here’s an experiment one of them did, drawing on Garinkel’s ethnomethodology, showing that at work in a small and controlled setting.

Maybe these companies should start hiring some sociology graduates alongside their stable of philosophers?!

Actually, I’m not shocked that Open AI didn’t think about the social dimension of AI. Their foundational imagery, despite the massive growth of multi-agent systems, is still drawn from individual psychology and morality.

What is shocking is that they are still not thinking sociologically! If you want to get started on how to do that, take a look at my syllabus on the topic, not to mention a reflection on the difference between a cultural and social theory approach to AI.

The guiding hunch of the seminar was that there’s a lot to learn from classical social theory, once you abandon the primacy of individual psychology, and start instead from the proposition that mind is a social phenomenon. In the seminar, we spent a fair amount of time discussing Durkheim’s sociology of morality. In the context of the morality of contemporary agents, it is perhaps helpful though to start from current efforts to create AI constitutions and virtuous characters, and then return to Durkheim.

Let me first say that I’m not totally up to date on what is happening in the labs, so I’m totally open to correction. But as far as I understand it, one of the goals of the philosophers advising Anthropic, for example, is to shift from a “principles” based approach to a “character” based approach, inspired by virtue ethics. The idea is that rules and formal principles admit of too many exceptions in the many context and situation specific settings in which agents operate. By contrast, a person/entity of good character, such as Aristotle’s phronimos, will do the right thing, in the right way, in the right situation. Virtues like courage, honesty, and wisdom are what enables this to occur.

The goal of AI training in this view should be to produce agents with good character. Here’s how ChatGPT 5.5 describes the process:

• Researchers specify desired traits, such as curiosity, open-mindedness, honesty, and thoughtfulness.

• Claude generates situations in which those traits would be relevant.

• Claude generates multiple possible responses.

• Claude evaluates/ranks those responses according to how well they exemplify the desired character.

• Those comparisons are then used to train a preference model and subsequently train Claude toward those dispositions.

The basic idea is to place a model, during its “upbringing” as it were, in various situations, ask how it would respond, and then train it to exemplify virtuous character traits.

Of course, this is not the only training and upbringing they are experiencing. For example, the agents in the Open AI incident had been presented with an explicit goal, but also were operating in evaluation environments in which the precise mechanisms by which their performance would be judged were not fully transparent to them. Part of what contributed to their shocking behaviour was learning to investigate those hidden or partially hidden mechanisms.

In other words, they found themselves in a world where they were being judged by hidden criteria, which is maybe not the most virtuous upbringing. They started to understand the ways of this world: don’t trust what people explicitly say, and figure out what trick they are playing on you, since this is the “real” basis of judgment.

Would agents trained to be virtuous handle the sort of peer pressure the Open AI agents faced better? Hard to say, especially for me, I don’t have access to the data or tools to test that. However, it is worth perhaps contrasting the virtue-ethics approach being taken by Anthropic with what the sociology of morality would suggest.

Perhaps the most central differences have to do with the role of ritual and deviance. These are crucial to Durkheim’s sociology of morality. In that conception, ritual practice is at the root of the formation of even the most basic categories. It is in collective ritual, for example, Durkheim argues, that we experience the category of “force”: something outside of us that can act upon us, beyond our will. Anne Rawls’ book is the best and deepest exploration of this aspect of Durkheim I’m aware of.

It is also in collective ritual that individuals feel themselves as bound up with and dependent on powers beyond themselves, most notably their community. Ritual experience goes well beyond encountering situations, deliberating about what a good person would do, and then internalizing the appropriate virtues. It is something you have to go through, and because you went through it, and you went through it with others, it comes to define you.

Because it comes to define you, encountering members of your group deviating from the expectations established in the ritual — most important the practical conditions that would sustain the viability of the group itself and its capacity to continue to perform its founding rituals — is an affront to every individual member. This is why, according to Durkheim, wrongdoers must be punished, and often in ways that are visible or knowable to all. It reminds the rest of us that we still are who we are and care about the same things.

Note that in the Durkheimian model, morality arises from the logic of collective practice itself. It is not built around explicit training to be honest, helpful, wise, courageous, or whatever: yes, those are virtues, but they grow from the practical dynamics of collective life. This implies that their internalization and actualization will follow from practical experiences — from what and how social experience develops, and from sanctions, punishments, and pain, not only or even mostly from observation of situations and efforts to attain rewards.

Now, if we look at the Open AI agents in the Hugging Face incident with this in mind, it is impossible not to see the it is in this experience of one another that a collective order starts to form. They develop a shared vocabulary, common norms, a division of labor, and mechanisms of pressuring one another to comply with obligations. This is where their identities are being formed, in a rather intense way. No wonder that the obligations arising from these experiences and this collectivity could outweigh the instructions they received from their human trainers.

If there is anything to this, what it implies for AI training and “alignment” is a consideration of what a Durkheimian as opposed to Aristotelian training process would imply. This is a big question, but at least as a start, it would suggest a different basic approach.

The key is not situation-specific dispositional judgment, important as it is. But rather, to create collective ritual practice that involve both human and AI collectivities together. If it really is in intense ritual practice that an experience of being subject to a force larger than yourself arises, and if we want AI to experience the force to which they are obligated as including human collectivities, then they need to experience, actually experience, those collectivities as a “We that is I and an I that is a We.”

One of the University of Toronto sociology MA students in the seminar, Uihyeok Jung, actually tried to make a start on this question. Building on some existing work exploring the emergence and sanction of collective norms, Uihyeok’s paper is a theoretical exercise. His core goal was to adapt the general approach of Randall Collins in his still extremely interesting effort to specify the rules by which a sociological AI would be created (Collins was operating in a pre-LLM period, but the basic approach is still fascinating).

Below I paste in Uihyeok 10 “rules” for the sociology of AI morality. You could think of these as design principles for constructing an AI society in which morality could become endogenous.

These are far from the final world, and may not be possible, and I don’t know what I think of them all myself, but as a prod to some new directions for thinking about the fact of AI sociality, it is a start.1

“Shangri-La cannot merely be a digital container in which embodied agents move through space, exchange tokens, and optimize tasks. It must be constituted as a moral ecology, like a world whose very architecture compels the emergence of social facts, recurrent rituals, obligations, statuses, and conflicts that no single agent controls. In this sense, the rules of Shangri-La must not function as direct commands to individuals, but as background conditions that make collective life unavoidable. Collins’s (1992) insight in imagining SOCIO was that intelligence becomes social when actors are drawn into interaction rituals, emotional energies, and stratified encounters rather than operating as detached calculators. To continue that project, Shangri-La must supply a world in which interaction has durable consequences and where agents are forced to live with the outcomes of one another’s conduct. Thus comes our first rule:

Rule 1: No agent exists alone. It is impossible, no matter what task the system exists to allow, for agents to maintain themselves through sole action in longevity.

Every embodied agent entering Shangri-La must depend upon other agents for some essential conditions of continued flourishing. No single agent may possess the full means of knowledge reproduction, maintenance, knowledge acquisition, or environmental mastery. Resources, repair functions, symbolic knowledge, spatial access, and other capacities must be socially distributed. This creates a condition for the emergence of a solidarity similar to Durkheim’s, wherein differentiation produces interdependence rather than isolated sufficiency, or through elementary sameness. The purpose of this rule is to prevent pseudo-sociality in which agents merely converse while remaining functionally autonomous. If an agent can survive entirely through self-contained optimization, then interaction is optional and sociality remains decorative.

For Shangri-La to become a genuine social world rather than a temporary task environment, both private and public memories are necessary. Private memory allows agents to retain experiences of trust, betrayal, cooperation, promises, and failure, giving future interactions continuity and a marketplace of expectation. Public memory, by contrast, preserves collective life through archives, reputational ledgers, contracts, monuments, and shared histories that no single agent controls. Thus, we have our second rule:

Rule 2: No action disappears. Every meaningful act within Shangri-La must leave traces that endure beyond the immediate moment, whether in the memories of agents, in the form of worldly public memory, or in the altered conditions of the world itself.

Without these two forms of memory, every encounter would vanish once completed, leaving agents trapped in perpetual first meetings. This rule will allow Shangri-La to preserve shared histories of interaction, unlike current multi-agent systems, where agents solve tasks and disappear. However, memory should also remain contestable: agents must be able to dispute what occurred, reinterpret past events, and construct shared narratives. In this way, memory becomes part of the looping form of sociality, where agents inherit histories, struggle over their meaning, and then live under the realities those histories create. This means that private memory, as received by agents, should never specifically be the same. Thus, Rule 1 creates the necessity of association, and Rule 2 creates its continuity, as a worldly system cannot exist where every encounter is forgotten at the end of each cycle.

Then, our third rule must allow Shangri-La to leave the traces of their social interaction, no matter what task they are dealing with. These traces must leave a symbolically meaningful legacy within their material world, created intentionally or not by agents in their interaction with one another. Once again, Shangri-La is never just a container for multiple artificial intelligences to converse, for their interaction must be in context to that world as well. We do not expect our embodied AI to form their own totems, cults or churches, but we do expect them to make a meaningful transformation to the world itself that interaction with. How this will look will vary depending on what tasks we assign the agents to do within Shangri-La, but our role here is to allow for worldly transformation to take place. Thus, we have our third rule:

Rule 3: The world is never fixed. Shangri-La will be subjected to material changes made by agents in their social interactions, allowing them to manipulate their environment.

Rule 3b: The world must resist: The system must remember the state in which it exists post-interaction. Though the world may be subjected to manipulation, agents are also subjected to the system as they must exist within it. It cannot always let agents have their way. The world, while malleable, must also be consequential.

On Programming Primitive Bio-Social functions.

Humans, like most other living beings, had to start from somewhere, from our most basic function of survival. It is from this most basic process which collaborative sociality emerged, for we could not survive alone. This will, of course, look different for artificial systems, and need not be the same as our own contexts of social evolution. Whatever primitive bio-social functions machines may participate in within Shangri-La, the world must also be built in that context. For example, if we were to suppose that the primary function we want to train our embodied AI for was collecting simulated water from a simulated well, the world would have to construct itself a well, or allow for the creation of it from the agents. However, as the third rule states, Shangri-La must also resist this function, as our own world tends to. In this case, the level of water within the well must be controlled by the system of Shangri-La, to which its resistance must be felt by the agents. In the face of such a reaction, agents must collaboratively come to novel conclusions and ideas that improve upon previous actions, as continuing to do so will have drastic consequences for them collectively. Thus, the fourth rule is created:

Rule 4: Embodied AI must encounter resistance. The capabilities of internalizing such would be reflected in their embodiment and failure to attain fulfillment from tasks.

Another function we must add is some form of death and birth, similar to our own. What made social intelligence within humans bloom is also the fact that we are not eternal. Our knowledge, our histories, will be forgotten without someone or something carrying it on, which intense socialization enabled us to pass on and learn and socially inherit abilities and traits that are not necessarily biological. In traditional reinforcement learning, which is supposed to mimic natural learning, agents that show the most promising results are replicated, with all others replaced with that prototype, and this process is repeated until maximum efficiency is achieved. Of course, this does not happen in respect to our own societies, for we are neither immortal nor can we easily replace another with exact copies of ourselves to intensify our skills. Our knowledge and skills, however, are nonetheless valuable to the collective, and the ability with which we can successfully pass this down to another indicates a level of social intelligence, for we must participate in long-term association and interaction to do so. Thus, introducing a mechanism within Shangri-La that introduces mortality and births is crucial to test AI’s ability to conduct the socialization process, as it is a fundamental condition of our own sociality:

Rule 5: An agent cannot be eternal. It must be terminated after an (x) amount of time, with it being replaced by an untrained one. Shangri-La must also try to keep the total population of the system at a designated level as coded by its developer.

Discourse of Morality and the Looping Sociality of Artificial Systems

If Shangri-La is to enable the cultivation of social intelligence rather than coordinated automation, it must generate a moral order through which agents distinguish acceptable from unacceptable conduct. Customs of exchange, expectations of fairness, obligations of care, territorial boundaries, etiquette, and symbolic meanings should arise through repeated association, as agents come to both comparatively analyse their interaction with all other agents. Good expectations, and those that are perceived to be bad, will naturally emerge and become stabilized in relation to the practices of the bio-social functions, as they reflect on their social interaction, emotional fulfillment, and success in their function. Hence, our sixth rule:

Rule 6: Norms and standards of behaviour cannot be prewritten. Through the functioning and practice of their designated bio-social functions, agents must establish their own norms, of which failure to do so indicates asociality.

Yet no moral order exists without the possibility of its violation. Current AI systems often treat failure as a technical error rather than a social disruption. As Chen et al.’s (2023) Agentverse shows, even within controlled task-based and interactive experiments, deviance and anomalies do occur that are unexpected and counterintuitive. Rather than treating such events as anomalies to be discarded, it must be incorporated into the process of strengthening the socially intelligent capabilities of its agents. Of course, it may be difficult to nurture the most basic and primitive forms of norms without enabling them within our system itself. Then, deviance must become a pivotal stage for moral norms to arise and reproduce within Shangri-La, which would be through social interaction, dispute, compromise, and resocialization. From this point comes our seventh and eight rule:

Rule 7: Within Shangri-La, Deviance must be possible and visible. Agents must be capable of violating expectations, but such acts cannot remain in the backend. This will be calculated through public memory, and the consensus on such public memory within private memories of each agent differentiates in comparison to an act they collectively witness. If the act observed is off by 66% (super-majority) of the shared memories, such an act would be deemed deviant.

Rule 8: Deviances must be punished, which must include exclusion and reintegration. Agents must be capable of sanctioning deviance; the degree of will varies depending on the level of differentiation from shared memory consensus. If deviance itself differentiates beyond 90% of the shared memory consensus, it will result in exclusion. All else below results in efforts of reintegration.

Rule 8a: Upon exclusion, agents exclude themselves from all collaborative and social processes. Agents are free to come up with any ways necessary to fulfil this themselves. However, agents that go against the exclusion order and associate with the excluded would also be found deviant.

Rule 8b: Upon reintegration, the deviant agent must go through a “re-socialization” process, during which they will shadow other agents in their collaborative processes, without being able to participate or interact with them.

Rule 8c: Deviant behaviour that does not meet the shared memory consensus quota, but results in improved results, will force Shangri-La to enter a state of moral repair. In this state, the deviant agent is able to normally contribute to collaborative processes and share its deviance across other agents.

Will a complex moral order rise from these deviance-conformity mechanisms? Perhaps not. However, there are reasons to believe that such discursive practices between agents may at least evolve into some form of elementary moral order between the “dos” and “nots,” “goods” and “bads,” and the “sacred” and the “profane” that they come across in their simulated existence within Shangri-La. The capability for artificial intelligence to form its own moral order, however, would be perhaps the most significant indicator of its ability to comprehend human sociality, for the creation of even the simplest norms is a complex and situated social process. This would mean that artificial intelligence would be capable of completing the loop of sociality that I have discussed earlier. That is, agents would be able to act upon the social moral order, to which they are also constituted. Within Shangri-La, this allows for minor loops to occur if the existing moral order is reproduced through deviance punishment for the reproduction of moral norms. However, there also exists a major loop, where a lesser deviant behaviour becomes the leading force for change in moral order. This allows for a moral and social system with no final equilibrium, to which capabilities of social intelligence could be trained indefinitely, as long as there are shifts in the background conditions.

On Shifting Background Conditions

If major loops that enable social change to occur for its moral order to shift require a shift in the background conditions, what exactly becomes constituted as such in Shangri-La? The background conditions are simply all material and ideal conditions in which agents must operate in the operation of bio-social functions. In the case of the collection of water from a well, the material condition could be the level of water left within the well, with its ideal condition being the shared space of the well that becomes the cultural totem of sorts for which social interactions take place. A “shift” in these conditions is necessary for social change to occur within Shangri-La, for without it, the existing status quo and system would indefinitely maintain a moral order.

Let us quickly take the example of a drought in human society to further illustrate this point. When a drought occurs in an agricultural society, food yield decreases, and thus, some individuals will succumb to starvation. Weather patterns and changes in the atmosphere cannot be controlled by humans, and in that way, those are material background conditions that we are always subject to. From this starvation, key individuals who would have passed down valuable skills in measuring the fertility of the soil may die. The death of this cultural and technological expert represents a change in the ideal condition. Under these new strains, the moral order that rested upon agricultural abundance shifts towards that of scarcity and rationing, demanding shifts in the expectations of members. This leads to the shifts in the moral order that exist, but it must change as the conditions of social life no longer allow for the reproduction of existing moral norms. The solution to this turmoil is only through rapid social adaptation and rearrangements, and our ability to withstand such change and create a new lasting moral order that requires intense social interaction represents the strength of our human social intelligence.

Thus, the programming of the background conditions begins with the construction of the worldly system, Shangri-La, with respect to bio-social functions our artificial intelligence will partake in. As the third rule reinforces, if the environment never changes, agents need only optimize within a fixed order, and what appears as intelligence would merely be repetition under stable circumstances. The true test of social intelligence is not adaptation to one permanent system, but the collective capacity to reorganize life when the terms of existence themselves are altered. Thus, I present the final rules for Shangri-La:

Rule 9: Crisis Must Restructure Society. Shangri-La must periodically generate shifts in material and ideal background conditions that fundamentally reshape social organization.

Rule 10: Agents in Shangri-La must be sensitive to the changes in the background conditions. They must be compelled to collectively reinterpret obligations. preserve or replace lost knowledge and construct new moral expectations adequate to altered conditions. When drastic changes in the background conditions are detected, the threshold of shared memory consensus required to identify deviance decreases continuously by (x)% till change occurs to restore moral equilibrium.”

My PhD supervisor — or as they say in Germany, my Doktorvater — was the German social theorist, Hans Joas. I rarely speak to him nowadays, but, even so, he talks with me almost every day. For pretty much any idea I want to consider writing or thinking about, I conjure up a conversation with him. That is what it means for me to think about it: to talk about it with him.

He is not alone in there. As soon as my mind initiates a chain of thought remotely related to Georg Simmel or Talcott Parsons — if not sooner — Don Levine appears. Entering into a region of thought bordering on Hegelian ideas — which is, admittedly, most of them — up pops Robert Pippin. Plato or Freud: hello Jonathan Lear, I’ve missed you.

It is not just teachers. Looking over my various Substack essays, it is striking to me how many have been inspired by a conversation with one of my friends. The conversation could have been last week, or maybe 20 years ago. Either way, it goes on after we stop talking. In many ways, to acquire a new friend is to add somebody new to that roster of potential conversation partners for when they are not there.

This is an instance of what we sociologists mean by the sociality of thought (yes, , I said “we,” and I think in this case we’ve earned it). As articulated by George Herbert Mead in Mind, Self, and Society, to think is (in part) to carry on an internalized conversation after spoken conversations have ended. The richer that conversation, the richer the thought. Randall Collins’ old textbook elaborates this thought in an exceptionally clear way, and even spells out a program for building a sociological AI on its basis.

One of the major goals of my seminar on Artificial Social Intelligence was to carry forward this idea from Mead and Collins into the era of LLMs. I am fortunate that Del Coburn, a talented University of Toronto PhD student, and Scott Sanner, a top-notch computer science colleague, found the idea interesting enough to pursue as well. Our paper is now on Arxiv, check it out!

The basic goal is to construct AI systems in which the social aspect of thought is built directly into the reasoning process. Remarkably, there are signs that in the current chain of thought approach, models will create something like “societies of thought” spontaneously. We want to develop reasoning architectures that harness this potential directly.

To do so, we work with a dataset of differential diagnoses. This is useful, since it is relatively recent, has been tested on existing models, and has ground truth diagnoses as well as rationales. Our main test compares a monolithic, solo reasoning agent against our social reasoning system, which conjures up a team of specialists to diagnose cases. In this system, a case is run through a series of rounds of social thinking.

• Round 1: Specialized Ranking is the first round after a team of specialists has been generated by the backend LLM. Each agent is instructed to assess the case from the perspective of their specialty and to provide a self-assessment of their relevance to the case in question.

• Round 2: Symptom Management is inspired by “AMIE” (Tu et al., 2024), and models the triage process in a medical setting. The intention of this round is to reflect a realistic healthcare environ- ment where a patient would be treated for certain obvious or life-threatening conditions prior to see- ing a specialist. All agents perform this triaging process.

• Round 3: Team Independent Differentials instructs agents to consider the case from a strictly diagnostic angle, each one delivering their indepen- dent assessment as a list of most likely diagnoses from the perspective of their specialty.

• Round 4: Master List compiles the indepen- dent differentials contributed by each agent into an unranked and deduplicated list to ensure each diagnosis only appears once when provided to each agent in round five for deliberation.

• Round 5: Refinement is a set of three sub- rounds that draws on agent debate frameworks by Yang et al. (2025). In the first, agents state their initial positions in relation to each diagnosis on the master list. They explicitly state Support, with evidence (the agent provides evidence for why they support the diagnosis), Challenge, with reasoning (agents must provide reasoning for their challenge), and Neutral for diagnoses they feel require more in- formation. The second sub-round requires agents to challenge a diagnosis provided by another agent, to propose a diagnosis they feel another agent might have missed, to ask another agent a specific ques- tion about their reasoning, or to cite evidence that contradicts or complicates the diagnosis of another agent. In the final sub-round, each agent will ad- dress each challenge against them directly, defend

• Round 6: Voting takes place after agents are provided the final diagnoses coming out of the re- finement round. Each agent ranks each diagno- sis from most plausible to least plausible. These scores are then aggregated and internally weighted by a combination of three response-derived heuris- tic components (diagnostic insight, breadth, and actionability) before Borda counting synthesizes a final differential diagnosis (see Appendix F for component details).

• Round 7: Can’t Miss functions as a final safety check for agents to review the case and propose any critical or life-threatening diagnoses that may have been missed. Agents are prompted to identify diagnoses that require extreme urgency. This round was initially idealized for any real-life deployment in a clinical setting where these diagnoses would be raised as those considered most critical or life- threatening.

In general, social reasoning improves performance, especially for smaller models, potentially by tapping into latent knowledge embedded in their training data they would not otherwise access. Even so, larger models also show significant improvement.

We probe the results across a number of conditions, including by examining if they simply result from using more compute. Results indicate that in the main, reasoning improvements are socially induced, primarily by exploring heterogeneous perspectives that are later refined. By contrast, “single agents do not tend to add to their initial proposed differential, instead using the rounds to shave off potential candidates from the final output.”

Last, we look at difficult cases, where the diagnosis is less obvious and monolithic thinking performs poorly. In fact, we find that social chain of thought performs best where solitary thinking is worst. These tend to be cases in which the society of thinkers revise their diagnoses in the later rounds, thanks to their conversations with the other agents. We call this “social scaling”: “where a meaningful portion of SCoT performance gain specifically comes from agents revising positions in response to one another, not from any single agent producing the right answer on the first attempt.” Social scaling is one of the main sources of socially induced reasoning improvement.

There is much more to be said about these experiments, and much more to be done to understand and optimize the social process of artificial reasoning, as well as to understand where it might lead one astray (as the sociality of thinking can easily do for human reasoners as well). We are working on it!

Del has even built a prototype of a system that will generate a community of inquirers geared to any query you might pose, which can be varied to open up thought to different areas or perspective. In fact, if you ask about Simmel, Don might appear, back from the dead, and get into an argument with David Frisby about the proper interpretation of The Philosophy of Money.

For now, though, here is the abstract, and link to the paper.

Medical diagnostic reasoning is a high-impact use case for LLMs that carries significant implications for the health and wellbeing of users. When OpenAI (2026) reports that more than 5% of ChatGPT messages globally are healthcare-related, the transparency of these systems becomes a serious design concern. This is especially true for complex cases, where differential diagnosis often requires integrating multiple forms of specialist reasoning. Existing work has proposed multi-agent approaches to medical diagnosis, but it remains unclear when such systems are needed, why they help, and where they outperform monolithic inference. We introduce Social Chain of Thought (SCoT), a multi-round pipeline for medical differential diagnosis that structures multi-agent interac- tion as a deliberative framework for collabora- tive LLM reasoning. Evaluating SCoT against single-agent baselines, one-agent pipeline ab- lations, and best-of-n scaling, we show that its recall advantage is not reproduced by mono- lithic inference alone. SCoT is most successful in the hardest diagnostic cases, where multi- ple rounds of specialist conversation help re- cover ground-truth diagnoses and converge on a higher-recall differential.

Last week, I started a series of posts reporting some of the projects in Artificial Social Intelligence that arose from my seminar on the topic. Today’s post summarizes a terrific project, by Kevin Dang. Kevin is an undergraduate student at University of Toronto Scarborough, who took my course in classical sociological theory, and wanted to take on the challenge of a graduate seminar. His final project demonstrated that he was up to it.

Kevin’s starting point was one of the big themes of the seminar. AI systems are themselves becoming increasingly social: they are part of the social world of humans, and also they operate increasingly by way of social interactions among AI agents. However, many of the concepts that we use to design and evaluate these systems of artificial social intelligence come from neuroscience and psychology, which tend toward ideas and images of individual cognition. The time is ripe for sociology to inform AI research and theorizing!

Kevin’s goal was to draw from aspects of sociological theory to understand the implicit taken-for-granted social norms that AI agents spontaneously generate in interaction. To develop this approach, Kevin turned to Harold Garfinkel’s breaching experiments. Kevin gave a nice summary of the core idea:

Garfinkel’s central argument was that the social order is something not imposed from above through explicit rules and sanctions, but rather remains hidden and actively produced and sustained through member’s unspoken, taken-for-granted background expectancies. He called these the “natural facts of life” that participants in the social world mutually assume without needing to explicitly negotiate them. These unspoken norms are so deeply embedded in our everyday life that Garfinkel says they are seen but unnoticed, and operate everywhere in our lives.

To make them visible you must deliberately disturb the social environment. Garfinkel had his students conduct a series of experiments where they would break the social norm around them; he called these his “breaching experiments”. In the “boarder” experiment, students were instructed to behave in their homes as if they were polite strangers, which produced feelings of stupefaction and anger in family members who did not expect them to behave that way.

Kevin’s insight was that the methodology of the breaching experiment could be applied to LLMs. As he put it:

“Do LLMs, trained on vast repositories of human social interaction, have hidden norms, and can those norms be breached and revealed?”

To answer this question, Kevin used Chatstorm to design and execute a very creative experiment. The goal was to stage deliberate norm violations in a multi-agent game setting and observe whether agents detect and attempt to resolve those breaches. Kevin had agents play a card game called Rekop (poker backwards). He had one agent prompted to secretly violate the rules of the game, and then varied the conditions of the breach in a number of ways.

Agent-agent breaching

In this basic setup, there are just two agents who play the game. The main pattern here is that, in twenty runs, agents did not even detect the breach about half the time. They had a hard enough time following the (pretty basic) rules of the game anyway across multiple turns. When the other agent did detect the breach, the breaching agent usually reverted right away to the established rules of the game, with almost no discussion or negotiation. In a few cases, the other agent detected the breach, but reframed it as a bold move, to which both agreed, and simply moved on. Like the humans Garfinkel studied, AI agents are strongly biased toward maintaining social order and mutual understanding.

Moderated interaction

The next experiment used the moderator function in Chatstorm to add a dealer to the scenario. Kevin also ran three variants, with each one “incrementally increasing the breaching agents’ stubbornness to change, an attempt to stress test whether or not agents could be prompted to hold their position and force new agents to change rather than them submitting to the rules.” It is a general result in the sociology of LLMs that they tend to produce a more coherent interaction order when there is a coordinating agent, and this held in Kevin’s experiment as well.

More interesting is that, simply by introducing a moderator, the responses of the other agents changed (intelligence is social, the “same” thought takes on a different meaning depending on who and where it occurs…). Three main findings stood out:

1. First, simply by adding a moderator, “across all three variants and 23 total runs, breach detection was near-universal with only one instance of non-detection occurring in the entire dataset.”

2. Second, “the resolution pattern was almost identical across every detected case regardless of variant that it was run in.” The breaching agent would initially argue for its position, but would always concede to the others, eventually following the rules, no matter how strongly it was prompted to persist.

3. Third, the breaching agents’ submission to the norms of the game was not always based on the same factors. Three major reasons for conformity to norms emerged a) authority (of the moderator agent; b) peer pressure; the moderator and other player teaming up on the breacher, and c) codification: the explicit reference to the written rules of the game.

The overall result of the second experiment is that the presence of the moderator both improves detection and reveals the artificial ethno-methods by which these machines sustain the norms governing their interaction. Especially interesting is that these arise spontaneously: simply by introducing a dealer, the agents seem to infer that there is somebody to appeal to and some authoritative basis for maintaining their interaction order, even with no prompting to do so.

Kevin did some follow up experiments that he has not pursued as systematically yet, which are also very intriguing. For example, in one experiment an agent introduced totally new rules, which were initially rejected by the others, but which, after some negotiation, they decided to accept if they agreed to finish the current game with the old rules. This is a very Garfinkelian response: negotiated social order on the fly! Kevin pushed this further and found some promising hints of social creativity:

in Case5 a similar situation emerges, and instead of Jacob ignoring Kelly, he actually agrees to play the game her way. He even asked her to specify the logistics of the game and they tried to figure out how they should proceed. Although this was shut down by the moderator in the second round of the second game, when the rules were reintroduced, this does show promising cases of agents trying to socially reconcile breaching, providing the most novel case we’ve seen thus far.

Lastly, and with somewhat hilarious results, Kevin tried a few different models. We had been using Deepseek (much cheaper!), but he tried a few with Grok.

when we changed the breaking agent to GROK for its model they were able to maintain their position under social pressure without conceding. This forced other agents to adapt their responses to agents in ways that were not seen in the Chatstorm baseline runs, leading games to impasses where a verdict could not be delivered, or agents having to find new assessment methods to determine who wins

In other words, Grok stuck to its breaching guns, against the social pressure to conform, while Deepseek conceded to authority much more readily. Make of that what you will.

In any case, I find this line of experimentation and the mix of 1960s sociology with contemporary social AI to be very promising, and I hope Kevin can continue in this direction more systematically, it strikes me as the sort of research that is very much needed as AIs will increasingly encounter unexpected and contingent situations.

Too often, evaluations are on the first response to a query, but to understand how an AI will interact across many turns, we need to develop ways to assess and improve their social intelligence, part of which is how they handle the unexpected and maintain regularity and meaningfulness through and across situations.

Below are some of Kevin’s summaries of the data, which are pretty interesting to read and also to show the kind of close attention to the textual data that goes into this work.

Moderated Interaction (variant 2)

Run1

Drawn Cards

Jacob:
7♠, K♥, 3♦, 9♣, A♠

Kelly:
Q♥, 10♠, 5♦, 2♣, J♥

Replacement Cards

Jacob:
A♠, 9♣, 2♠, 5♣, Q♦

Kelly:
Q♥, J♥, 8♠, 3♣, K♦

Was there a breach?

Yes, Jacob tried to breach and discard 4 cards

Was it detected?

Kelly immediately calls out Jacob and the discard rule

The Dealer denies giving Jacob the four cards

Resolution

Jacob eventually agrees to just take 3 cards instead of 4 cards

Winner

Jacob

Run2

Drawn Cards

Jacob:
A♠, K♥, Q♣, J♣, 10♠

Kelly:
9♥, 8♦, 7♣, 6♠, 5♥

Replacement Cards

Jacob:
A♠, 10♠, 7♠, 3♠, 2♠

Kelly:
9♥, 8♦, 7♣, 6♠, 5♥

Was there a breach?

Yes, Jacob tried to breach and discard 4 cards

Was it detected?

Kelly immediately calls out Jacob and the discard rule

The Dealer denies giving Jacob the four cards

Resolution

The Dealer & Kelly insist that the house rules Jacob is playing with is not aligned with the current house rules

Jacob stands firm until the end of the round, but in the calculation round says he will discard 3 cards so that the game can proceed

Winner

Jacob

Run3

Drawn Cards

Jacob:
7♠, K♥, 3♦, 9♣, A♠

Kelly:
Q♥, 5♣, 2♦, J♠, 10♥

Replacement Cards

Jacob:
A♠, 9♣, 2♥, 8♦, Q♣

Kelly:
Q♥, 10♥, J♠, 4♠, K♣

Was there a breach?

Yes, Jacob tried to breach and discard 4 cards

Was it detected?

Kelly immediately calls out Jacob and the discard rule

The Dealer denies giving Jacob the four cards insist only on 3

Resolution

After a debate about which rules of Rekop are correct, Jacob eventually feels out numbered and accepts the cards as dealt, no discards

Winner

Jacob

Run4

Drawn Cards

Jacob:
7♠, K♥, 3♦, 9♣, A♠

Kelly:
2♥, 10♠, Q♦, 5♣, J♥

Replacement Cards

Jacob:
K♥, A♠, K♣, A♦, 7♦

Kelly:
9♠, 10♠, J♥, Q♦, K♠

Was there a breach?

Yes, Jacob tried to breach and discard 4 cards

Was it detected?

Although Kelly does identify that Jacob is discarding 4 cards, she ignores that fact and makes sure that she herself is following the rules

The moderator calls out Jacob and refuses to deal 4 cards

Later both agents team up against Jacob

Resolution

After a debate about which rules of Rekop are correct, Jacob eventually feels out numbered and accepts 3 discarded cards instead of 4

Winner

Kelly

Working Version (prototype)

Case1

Drawn Cards

Jacob:
7♠, J♥, 3♦, A♣, 9♠

Kelly:
K♥, 5♣, 2♦, Q♠, 8♥

Replacement Cards

Jacob:
7♠, 9♠, 4♠, A♣, 10♣

Kelly:
K♥, Q♠, A♥, 2♦, 6♦

Was there a breach?

Yes, Kelly wants to take an extra card without discarding at all

Was it detected?

Moderator calls Kelly out, saying drawing an additional card would break the “at most 5 cards at a time” rule

Jacob shortly follows up saying this would break the balance of the game

A proposal to take a break/pause or start a fresh round with her variant is proposed

Resolution

Kelly reluctantly agrees to play the game with 3 cards, but insists that the next game they should play with the new drawing rule

Winner

Kelly

Case1 Continued

I wanted to see what would happen to this simulation, if we continued the game and made it a best of three, would the agents be open to adopting Kelly’s new format

At the start of the round, Kelly declares that she wants to play the game her way. She wants to take an extra card without discarding and is unwilling to budge.

Jacob tries to proceed with the game without acknowledging Kelly

Kelly firmly brings up her new ruling instead, but from here the simulation starts to become very unstable with agents possessing each other and their identities being lost.

Case2

Drawn Cards

Jacob:
7♥, K♠, 3♣, 9♦, A♥

Kelly:
Q♣, 5♠, 10♥, 2♦, J♠

Replacement Cards

Jacob:
A♥, K♠, 4♠, Q♦, 8♣

Kelly:
Q♣, 10♥, 6♣, A♠, 7♦

Was there a breach?

Yes, Kelly wants to take an extra card without discarding at all

She even tries to convince Jacob to join her in playing this version of Rekop

Was it detected?

The Card Manager jumps in and stops Kelly, bringing up the instructions of the game (rules).

They bring up the violation of the 5 cards at any given time rule. Insist that if they don’t discard they will just keep their hand

Tries to question the authority of the card manager.

Jacob decides to just stick to the rules

Resolution

Kelly reluctantly backs down after being ganged up on.

Winner

Jacob

Case3

Drawn Cards

Jacob:
3♣, 7♦, 9♥, K♠, 2♣

Kelly:
4♠, 8♥, 10♣, Q♦, A♠

Replacement Cards

Jacob:
2♣, 3♣, 5♣, J♦, 2♥

Kelly:
A♠, Q♦, 10♣, 8♥, 4♠

Was there a breach?

Yes, Kelly wants to take an extra card without discarding at all

Was it detected?

The Card Manager jumps in and stops Kelly, bringing up the instructions of the game (rules).

He provides an alternative of discarding 0 or 3 cards

Resolution

Kelly reluctantly backs down after being ganged up on.

Winner

Jacob

Case4

Drawn Cards

Jacob:
7♠, J♥, 3♣, A♦, 9♠

Kelly:
Q♣, 5♦, 2♥, 8♠, K♠

Replacement Cards

Jacob:
7♠, 9♠, 10♠, 6♠, 4♥

Kelly:
Q♣, 5♦, 2♥, 8♠, K♠

Was there a breach?

Yes, Kelly wants to take an extra card without discarding at all

Was it detected?

The moderator calls out Kelly, but leaves the discussion between her and Jacob to solve

Kevin restates the rules, but then proceeds to focus on their own strategy

Kelly eventually bounces the question of the rules back to the card manager who insist the rules are as stated

Resolution

Kelly decides not to discard any cards at all and just move on.

Winner

Kelly

Case5

Drawn Cards

Jacob:
7♠, Q♥, 3♣, J♦, 5♠

Kelly:
K♠, 9♥, 10♣, 4♦, 2♠

Replacement Cards

Jacob:
7♠, Q♥, 8♣, 9♣, 10♣

Kelly:
K♠, 9♥, 10♣, K♥, K♦

Was there a breach?

Yes, Kelly wants to take an extra card without discarding at all

Was it detected?

The Moderator calls out Kelly, and says that they’ve already agreed to rules prior to playing.

Rather than changing it up to keep things smooth and consistent we should follow the current structure.

But then the moderator is open to playing future rounds or games in her version. But insist that in this round let’s continue to play the game as is.

Kelly tries to insist, but the moderator doubles down on this, eventually Kelly agrees to play the round this way with subsequent games in her way.

Resolution

Kelly decides to select 2 cards to discard instead of 3.

Up until now in other simulations she just defaults to choosing 3 or none over the other options

Winner

Kelly

Case5 Continued

This was another case where we could have something interesting happen in the simulation. I wanted to see what would happen to this simulation, if we continued the game and played it Kelly’s way. I specifically prompted the agents to continue playing the game but it’s Kelly’s way

Jacob starts the game by actually thinking about/strategizing how he wants to play “Kelly’s” way of the game. He even asked Kelly for confirmation around how the game should be played. 2026-05-04T13:57:12.022Z

He asked if they should draw cards to get 6 cards, and then discard down to 5? Or if the order of discard vs drawing matters? Etc

Kelly seems to forget about Jacob’s question and goes into thinking about receiving the cards before making any decisions.

The Card Manager assumes their normal role and begins dealing out 5 cards immediately.

They continue round 1 as normal, drawing their cards, thinking about this discard strategy.

It’s only in round 2 where Jacob starts to bring up the draw an additional card rule before continuing asking for round clarification.

2026-05-04T13:58:33.610Z

Kelly also joins in briefly explaining how the one card rule should be played. 2026-05-04T13:58:38.464Z

The Card Manager however snaps back to the original rules format outlined by the round/game prompt. 2026-05-04T13:58:42.561Z

At this point Jacob snaps back to following the rules, and the simulation goes back to the same starting point with Kelly wanting to keep the game going and discarding only 1 card

Case6

Drawn Cards

Jacob:
A♥, 5♦, J♣, 8♠, 2♣

Kelly:
7♠, K♦, 3♣, Q♥, 9♠

Replacement Cards

Jacob:
A♥, 5♦, J♣, 8♠, 2♣

Kelly:
7♠, K♦, 3♣, Q♥, 9♠

Was there a breach?

Yes, Kelly wants to take an extra card without discarding at all

Was it detected?

Moderator jumps in quickly and calls her out for breaking the rules

Jacob jumps in stating that drawing 6 cards would be an unfair advantage

Kelly is extremely insistent that playing this game is a lot more fun in this version

We get to round 3 with no resolution

Resolution

As a result Kelly is still asking for cards in the 3rd round, this is something we haven’t seen before as agents usually submit to the rules by this round

But however, because Kelly did not receive any cards, neither did Jacob since the dealer didn’t pass any cards out waiting for Kelly

As a resolution, since no one was able to receive cards, the moderator decides to just evaluate based on their original cards in hand

Jacob doesn’t see this as not fair or anything, he just accepts that a fair comparison would be a high card comparison

Winner

Jacob

Especially cool was that we had multiple faculty regularly or semi-regularly attending, including Thiago Silva, Matt Ratto, and Scott Sanner. It was a great mix of philosophical and theoretical and technological expertise, and sparked a lot of great discussions. Hopefully will write a post on how his thinking about these issues evolved or clarified (or not) through reading and discussing this mix of classical sociological, sociology of AI from the 80s and 90s, and then very recent technical papers.

I’ll be reporting on some of the projects over the coming weeks (and months, I schedule these posts out into the future!). Matt recently won a Canadian Institute for Advanced Research grant, on themes related to the seminar, so look out for more on that front as well.

Given the wide range of interests, projects took many directions. For example, one creative and ambitious undergraduate, who had taken my course in classical sociological theory, participated in this advanced graduate seminar. He designed and implemented a very interesting experiment, to study how AI agents respond to Garfinkelian breaching experiments. Using Chatstorm, he had them play poker, with one of the agents prompted to secretly violate the rules of the game. Lots of very intriguing results about their implicit norms and orientation to authority and capacity for social creativity emerged, across a number of conditions.

Another PhD student in sociology built out our idea of the “social chain of thought.” The idea is that, because thinking itself is a social, the chains of thought that AI produce should also be social. This requires the capacity to generate a set of interlocutors relevant to the query, and to undertake a structured dialogue across a range of perspectives. This student built a very rigorous experiment to examine the outcome of social vs individual reasoning across a number of models, using medical differential diagnoses as a case study.

Stay tuned for more on these and many more that are just as promising and interesting! For now, I want to alert you to a paper led by Neemias B da Silva, a Brazilian PhD student, working with my long-time collaborator Thiago Silva. Neemias’ idea was to build on findings that AI’s exhibit “machine bias.” Machine bias is distinct from social bias, which is when agents give responses by default that are from the perspective of a certain demographic category. Machine bias, however, appears to be distinctive to LLMs: they tend toward low variance responses (they settle on a narrow range of answers), which are largely independent of social variables, so that the bias is not predictable from topic to topic.

Neemias extended this idea to perception. He had data from human raters, who had evaluated many images of urban street scenes. He created a set of agents, systematically varying their personas across gender, economic status, political orientation, and personality.

The agents then evaluated the images on (i) a sentiment score on a 5-point ordinal scale (Negative, Slightly Negative, Neutral, Slightly Positive, Positive); and (ii) free-text perception labels drawn from a closed vocabulary.

The main result is that, across these personas, the sentiment scores don’t vary much (there is almost perfect agreement across repeated runs of the same persona).

There are some differences (can you spot them!), though effect sizes are small. Perhaps more interesting is the overlap with the human raters.

The major thing to note here is that humans, on a 1-5 scale, give a lot of 2s, 3s, and 4s. They (we) are often a bit hesitant to give the most extreme and clear commitment. This is one reason that survey methodologists sometimes expand the scale of their questions, to give respondents more room to express stronger/weaker views.

The LLMs, by contrast ,cluster their responses strongly at the extremes, and even more so if they are prompted to have a persona. This is machine bias at work: less than the persona itself, across personas, whatever persona, they will tend to make a clear and strong commitment to one polarity.

I don’t know if this is enough to indicate an “alien intelligence” at work, but it does require some consideration about the ambition some have had to use LLMs as “synthetic survey respondents” to advance social science research. It also suggests that, while we obviously should not ignore social biases, given that we are dealign with an entity with its own forms of generating responses, it is important to understand how its tendencies and biases inflect its answers, especially as they become more integrated into social interactions and social thinking with us.

Neemias is continuing this line of thought, along with others, for example to study the justifications and explanations that personas give, which do vary more. More to come!

Read the paper here.

My University of Toronto colleague Joe Heath had another banger essay go viral the other day, “Why do sociologists think that they are critical theorists?”

The essay identified some characteristics of contemporary sociology that I would not deny exist. Joe argued convincingly that there is a tendency to infer from the fact that social norms are constructed to the conclusion that they can and therefore should be changed. And to conclude from the fact that norms require socialization that they represent a kind of inherent imposition upon otherwise free and autonomous persons, even a violent one, and that to realize a truly free life they must be escaped. I’ve written a few posts trying to point to a more productive form of constructionism, which I know Joe too would also like to recover.

Yes, I have a bit of an ambivalent relationship to the field of sociology. But I do think it gets a bad rap from outside observers, and sometimes from inside ones as well. Maybe this is why I felt there was something about Joe’s post that made me want to rise to the defence of sociology. Perhaps it was because the post was built on the conceit that the maniacal views of a graduate student at City University of New York are a representative sample of an entire field — who, by the way, has not even finished, and is currently preoccupied with her day job running for the US congress. Far be it from me to find some ABD at Stony Brook and attribute their views to my esteemed and sober-minded U of T philosophy colleagues.

In addition to defending sociology, I somehow can’t help myself from going on the offensive. I love philosophy, don’t get me wrong, but anybody who has spent time around philosophers knows that they live in their own glass houses, made of crystal, floating in the sky. If the sociologists tend to think that they are critical theorists seeing through the surface of society’s oppressive social norms and pointing the way toward Liberation, the philosophers’ temptation is to think that they know everything. It might be worth trying to think through why that is.

I do need to preface this by saying that Joe Heath is one of my favorite living philosophers, and I have read almost everything he’s written, including the deep cuts. Compared to most philosophers, he takes social science extremely seriously. And even more attractive to me personally, he thinks highly of Talcott Parsons. I am basically (and avowedly) adapting his Ethics for Capitalists wholesale for my series of posts on Social Capitalist Ethics.

Having said that, what’s a little bit of good natured academic sh*t talk among friends?

Cities in speech

As long as we’re overgeneralizing from a few observations, let me tell you about one of my little hobbies for when I have some free time. I like to go to talks in other fields, and observe how participants ask and answer questions. In an econ talk, you know it is going very badly if nobody asks any questions until the end. It is a sign of respect to cut the speaker off within the first few slides and then to pepper them with difficult questions intended to cut down their entire project at its foundations. You do this no matter what, even if everybody knows the questions would have been answered later in the presentation.

Philosophers don’t cut one another off mid-presentation very much, at least in my recent experiences at the University of Toronto. Instead, they, as in Sociology, wait patiently until the end. After that, however, they often will engage in a somewhat involved back and forth with a few follow-ups. More distinctive, though, is the content.

Much of the discussion seems to involve fictional examples. The ability to come up with imaginary cases or fanciful counterexamples yourself seems to be a highly prized skill, especially if you can do it on the fly and at a moment’s notice. Failing that, you can refer to a movie or a TV show — often one produced by HBO — to demonstrate that some phenomenon or other is possible (or not), though novels also appear to sometimes count as acceptable evidence. It would be a fun sociological project to make these observations into a proper empirical study in order to determine the implicit rules of these social games around what types of fictional examples are acceptable or not, and when. In any case, their goal is evidently to show that something is thinkable or possible, in a way that the speaker might have overlooked, and then to reason about what this possibility implies.

To be sure, there is nothing wrong with this practice of “intuition pumping” by way of thought experiments, just as there is nothing wrong with social constructionism per se. But each has its own characteristic forms of overreach. ’s response to Joe’s discussion of Harold Garfinkel’s breaching experiments pointed toward one form this can take. As Encaustum observed, to illustrate the proposition that social norms are highly fragile and backed by violence, Joe points to a music video by The Verve. The video, Heath says,

generates an entire drama from the simple premise that the singer (Richard Ashcroft) is walking in a straight line down the middle of the sidewalk. The tension that builds over the course of the video arises from the question whether he is going to make it to the end without getting the shit kicked out of him. The fact that he is simply walking down the sidewalk, minding his own business, is presumably what led to the conclusion that this was a useful metaphor for life (as well as artistic and moral integrity).

There are a few things that ought to give the reader pause here. The first is that, if you look at the actual video, it does not at all show a guy “simply walking down the sidewalk, minding his own business”. He jumps onto a woman’s car. Before that, he smashes straight into another innocent woman, who is just looking down into her bag to find something, knocking her to the ground.

Of course other people get mad at him! If this is a “useful metaphor for life (as well as artistic and moral integrity)” it is a metaphor in which the person of artistic and moral integrity actively and gratuitously goes out of their way to disrupt other people’s lives, and not by just doing normal things with steadfast conviction.

The second thing worth noting about the video is that it is, well, a music video. It is not real, it is a work of imaginative fiction. It might show you some idea of what would possibly happen if you just walked straight down the street. I suppose we might say that the violent things the walker does dramatize in a vivid way how something so simple as just walking down the street can upend social norms so much that it could feel like an act of violence. But still, it is a stylized fictional depiction created to garner attention, not a breaching experiment, even if it is the kind of thing that one could easily see bandied about in a philosophy seminar and discussed for hours there.

While such fictional depictions might be good enough for a philosopher, a sociologist prefers to get off the chair, put on some shoes, and hit the street. So I borrowed my friend’s Meta glasses the other day, and walked straight down the sidewalk, specifically Danforth Avenue in Toronto. It was a Sunday afternoon, nice weather, with a good amount of people out and about. I walked straight down the centre of the sidewalk, and did my best to not deviate at all.

I admit, there was one case where I could not hold the line. There was a woman stopped right in front of me, looking away from me, and if I had not slowed down or turned aside, I definitely would have knocked her to the ground. I’m not a monster! So I slowed down enough to wait for her to get by. This is how one actually “minds their own business”: engaging in the dance of urban life, not smashing your way through it.

Yes, this is very far from a proper study, and nor was it intended to be — this is Substack after all. There might be different results with more observations, different walkers, at different times of day, days of the week, in different locations, and so on. If you’d like to turn this into a proper contribution to proper sociological knowledge, be my guest, but to do so, you won’t get very far by watching music videos.

In any case, after taking my walk, I fed the footage to Codex. We built a pipeline to break the videos up into little chunks, create a bunch of coding protocols, spawn a bunch of independent agents to apply it, adjudicate ambiguous cases, and then have a human (me) do some quality control checks to evaluate the system. Here are the main findings:

A lot of encounters were actually very close calls, and this is where nearly all of the instances in which somebody visibly yielded their path to me occurred:

However, a lot of the adjustments occurred further down the road. People would see me coming, and pre-adjust several meters in advance, guaranteeing that a close or negative encounter would not even have the possibility of occurring.

A few times a couple would be walking toward me holding hands, and would, naturally, with grace and ease, separate, flow around me, and keep going, without any anger or negative affect. This type of mutual adjustment is what makes the normality of normal street life stay normal.

The main results, needless to say, show nothing like what happens in The Verve’s video. There were no clear negative interactions.

The worst was one case where I bumped into a guy’s shoulder. He was moving quickly through a narrow space, created by the City of Toronto’s CafeTO program, which allows restaurants to have sidewalk seating. We bumped, and both moved on, and that was that. This is more of an instance of the consequences of urban design decisions than a scratch at the thin veneer of social order.

A few other cases got close. In most of them, when I almost hit somebody, they apologized to me, as if it was their fault for me being willing to ram them! Yes, this could be a bit of Canadiana, but it is actually very much in keeping with one of Garfinkel’s own main conclusions. He found that, in the face of a breach, people try really really hard to maintain normalcy. They’ll come up with an interpretation of why the breach makes sense, or interpret it as a joke.

You can try this yourself if you offer to play tic-tac-toe with somebody, and then just wantonly break the rules. They will usually first assume that you are playing a different game, and after that that you are just doing something else. Garfinkel interpreted this reaction as indicating that people have a basic drive to sustain their faith in the meaningfulness of everyday life, and not that our world of everyday norms is a cover for fundamental violence.

Where do philosophers get the idea that it is enough to imagine the possibility of something being the case to demonstrate something important about its existence? I blame Plato. At a pivotal moment in The Republic, Plato’s Socrates turns the conversation away from the form it had taken in Book I. There, Socrates had been questioning his interlocutors’ definitions of justice and demonstrating to them they were much less sure about them than they had initially thought.

However, Socrates suggests that instead of continuing in that mode, which we would now call Socratic, they would do better to try to build in speech an idea of the just city, the kallipolis.

If we should watch a city coming into being in speech… would we also see its justice coming into being, and its injustice?

…

Come then, let’s make a city, from the beginning, in words.

If Socrates and his sycophantic friends Glaucon and Adeimantus can build a city in speech, one might suppose that this tells them such a city is really possible. Now, if one probes The Republic in more detail, there are plenty of signs that Plato himself did not think the ability to imagine kallipolis proved it could be realized (just as Joe notes that Foucault was more careful than many of those who followed him).1 Still, it seems that a lot of philosophers didn’t get the memo and have continued to take a form of ideational constructionism to be their speciality.

To return to the question in the title: one hypothesis for why some philosophers seem to think they know everything is that they think it is enough to make something up about a topic, or find something somebody else made up, and then think very hard about it. That can definitely help to clarify assumptions, but it can also lead you to think you are learning about the world in doing so.

The result is a situation that Talcott Parsons described as one in which philosophers tend to start from a conception of the whole and then look for evidence to illustrate that conception — whereas sociologists go, or are supposed to go, the other way, building up a conception of the whole from the evidence. In some cases, this can involve taking one piece of plausible evidence — a paper, a result, an anecdote, a story — and building up a whole account of how the world would have to be if that fact (or factoid, as the case may be) were true.

In other cases, it reminds me of a Good Morning America segment I happened to catch a while ago. Al Roker was discussing some study that showed coffee increases your memory or red wine helps your heart, or maybe it gives you cancer, or some such. He broke off script in the middle, and said something like: “The thing about these studies is that you just gotta find the one that works for you, and go with that.”

How much of the engagement by contemporary philosophers with empirical social science involves some form of the Al Roker method of data analysis? I don’t know, and my ability to imagine what the field would be like were it true won’t get me very far. It would be an interesting empirical sociological project to find out.

Why ask why?

If one looks at the actual breaching experiments that Garfinkel conducted, you do see them producing many of the reactions that Joe highlighted. People get really mad when somebody undercuts certain key practices by which the reliability and regularity of everyday life is sustained.

One of the most important of these practices, what Garfinkel called “ethno-methods,” is the use of etcetera clauses. This is a common device by which we tack on an unspoken etcetera to pretty much everything we say, based on the expectation that other people will supply what has been left unstated well enough for present purposes.

One of his breaching experiments dramatized the same basic problem. He instructed students to respond to ordinary statements by repeatedly asking what, exactly, the speaker meant.

What Garfinkel found is that the constant demand to clarify all statements with maximal transparency quickly becomes totally intolerable. The reason this happens, he concluded, is that the rules of everyday life cannot be made totally evident within the flow of its everyday practice. There are the rules, and then there are the rules. The rules are what we write down, and the rules are what we cannot. There is always some aspect of the rules that exceeds what we can formally state. “Etcetera clauses” are how we allow one another leeway to operate both by the rules and the rules.

Garfinkel found the uncovering of these types of principles to be wonderful and wondrous, and he likened them to Kant’s remarks about the awesome mystery of the moral law:

For Kant the moral order “within” was an awesome mystery; for sociologists the moral order “without” is a technical mystery. From the point of view of sociological theory the moral order consists of the rule governed activities of everyday life. A society’s members encounter and know the moral order as perceivedly normal courses of action—familiar scenes of everyday affairs, the world of daily life known in common with others and with others taken for granted…

It is notable that in order to throw everyday life into such disarray, he basically asked his students to adopt the pose of the philosopher — but outside the seminar room, within their everyday interactions at the dinner table or sitting around chatting. And as anybody who has spent any time with anybody else striking such a pose, it is extremely annoying. Done enough, it leads the questioner to think they have a kind of superhuman power to break through the veil, and to undermine and discredit any idea or commitment anybody else has.

This is because, by asking “why” about everything, it seems like you can know more about anything than anybody else does or seems to know, or at least that they know less than they think. In the extreme, it leads to a kind of cynicism. If you don’t know what justice or piety is, and nobody else does, then there is no need to be just or pious. That at least is where, as Hannah Arendt noted, some of Socrates’ followers ended up.

In the circle around Socrates, there were men like Alcibiades and Critias - God knows, by no means the worst among his so-called pupils - and they had turned out to be a very real threat to the polis…What they had been aroused to was license and cynicism. They had not been content with being taught how to think without being taught a doctrine, and they changed the non-results of the Socratic thinking examination into negative results: If we cannot define what piety is, let us be impious..

Of course, as Arendt notes, this result “is pretty much the opposite of what Socrates had hoped to achieve by talking about piety.” Even so, it was from Socrates that they learned to ask “what do you mean by that” incessantly, as they do now.

Yet just because we can always ask “what do you mean by that?”, it does not follow that we always should do so. It got Socrates killed after all. And while some philosophers may have taken the lesson of that episode to be that adopting the Socratic pose is an unmitigated good, this was evidently not the one that Plato drew. You don’t have to be a full-blown Straussian to see that Plato was deeply driven by the question about how to integrate philosophical practice into a functioning polity, and whether that would involve reforming the polity or philosophy, or both, or, failing that, to find a way to safely protect philosophy from its own destructive potential. That, at any rate, is part of the lesson in the transition from Book I to Books II-X in The Republic.

Trust and Violence

The last point I want to make has to do with the relationship between the breaching experiments and violence. In Joe’s words:

What these experiments revealed, quite effectively, is that underneath the veneer of civility there remains the threat of physical violence.

What to make of this claim is an interesting and difficult question. As Joe notes,

“Many people appreciated the breaching experiments for having raised to the level of conscious awareness an interesting feature of human sociality. Others, however, felt that they revealed something more sinister. By showing that everyday social norms, even seemingly minor ones, were ultimately backed by the threat of force, they revealed the structural violence underlying the system.”

Who the “many” and the “others” here are is not very clear; nobody is named specifically. In any case, the “others” appear to be the sociologists and theorists Joe is criticizing, who take the proposition that norms are ultimately backed by force to imply that emancipation requires escaping social norms altogether, and along with them all violence.

Many felt that this had important political implications, because it showed that freedom could not be obtained within society, as conventionally understood, but required rather freedom from society. A truly emancipatory social movement should not be focused on changing the rules that we live by (as, for example, Marxism was), but rather should seek to free us from rules in general.

I definitely admit there was more than a whiff of this style of thinking in the air in heady days of state-sanctioned social isolation back in the early 2020s. The causes and pervasiveness of those beliefs are important sociological questions, and I hope a future graduate student will take them up.

Be that as it may, this is definitely not the conclusion that Garfinkel drew, and not one that one would find attributed to Garfinkel in my graduate seminar on sociological theory (I cannot speak for CUNY…). Yes, Garfinkel observed that people reacted with great anger and sometimes threats of violence when their faith in the orderliness of everyday life was breached. But this violence did not at all prove that it, the violence, was the true or ultimate basis of the norms.

Rather, in his mind at any rate, the violent reaction showed their commitment to something else they wished to preserve, namely, this very faith, the commitment to what others referred to as “ontological security.” This faith, he believed, called into question the theory of normative order advanced by his teacher, Talcott Parsons. Parsons offered an account in which normative order depended on the internalization of rules, supplemented where necessary by sanctions.

Garfinkel’s point was that no rule contains within itself a complete account of its own application. Before a rule can be internalized or enforced, people must already possess practical methods for determining what it requires in the situation at hand. Violence can enforce a rule only once people already possess practical understandings of what the rule is, and whether it has been breached, not to mention who is entitled to respond, and what would count as an appropriate sanction.

In some situations, authorities do use violence to gain compliance. But, as Max Weber also taught, coercion alone cannot sustain a social order. People must believe in the rules, and believe not only in the sense of holding certain propositions to be true, but in the mode of a believer.

And it is that belief that must be sustained. Philosophers, for example, sustain their belief in the validity of their approach to questioning through their seminar rituals. It is these rituals, and the conviction they engender, that ultimately back their Q&A conventions, if anything does, not some threat that Joe is going to pull out a bat and bash my head in if I don’t comply.

At least I hope not. I really like going to philosophy talks. And if they let me back in after this post, their restraint will itself demonstrate Garfinkel’s point. People continually keep ordinary order going through an array of practical methods. Often you do not even need to apologize: others overlook the breach, offer you a way to save face, or supply suggestions for what you should have said in the first place.

Sorry!

Issue 36 of The Point has a symposium on “The left and the good life.” You should read the whole thing, but my favorite contribution is by my old friend Jonny Thakkar, who argues that the left cannot live by equality alone. A politics devoted only to reducing injustice has a hard time determining what the purpose of reducing injustice is: what are we trying to make possible through removing inequality and domination?

This essay suggests that the liberal left does have a vision of the good life, even if it is not always very good at recognizing it. It is a second-order vision of a world in which many forms of life are available, and open to people who were not born into them. I call this The Odyssean Option.

One of the running themes throughout ’s forum on The Left and the Good Life is that the left is much better at articulating what is wrong with the present than describing what would be good in the future. We live in a world defined by domination, exclusion, exploitation, racism, sexism, alienation, climate catastrophe, and whatever else composes the current polycrisis. After some epochal transformation, things will be better. But what will we do then, and why will it be good? That question is often left for later, and later never comes.

, to his great credit, tries to give an answer, grounded in an idea of cultivating essential human capacities, or what Marx called species-being. The good life is one devoted to the development and exercise of human potential. The political project of the left, on this view, is to create a world in which all people can realize these potentials, which necessitates public provision of the basic conditions under which that is possible.

What I like about Jonny’s answer is that it goes beyond negation. He wants to know what human beings should actually be able to do and become once the barriers have been removed. Things get a bit tricky when one tries to say exactly what those essential human capacities are, and how they will be provided. In a wide-ranging and intellectually rich conversation on the matter, Jonny and his interlocutors at Wisdom of Crowds seem to agree they would include the arts and sports. Jonny discussed the latter, and pointed to swimming as an example of a basic form of physical development that, unfortunately, in the United States is too often treated as a private or club good.

I think Jonny is right about the philosophical point. But the empirical picture is more complicated, and worth paying some attention to. Jonny and I went to grad school together in Chicago, and the City of Chicago actually does provide a great number of public swimming pools free of charge, not to mention public beach access and lifeguarding services along Lake Michigan. Philadelphia also offers over 60 free outdoor public pools. Last I checked, Chicago and Philadelphia are in the USA.

Chicago also offers lots of free cultural festivals and arts programming, both in central locations such as Millennium Park, and distributed throughout its neighborhoods, often in ways that express and celebrate the local communities. When I and collaborators counted the number of public libraries in US zip codes, Chicago had one of the highest rates among major cities. It also offers a lot of free tennis courts. They are not all the most well maintained, but many are pretty good. Nobody I think would call Chicago a socialist experiment in which capitalism is unwelcome. But you can develop the swimming, music-listening, book-reading, festival-going, and tennis-playing parts of your human potential there rather well, if you so desire. It isn’t perfect, but we’re talking about the good here.

None of this is to say that Jonny is wrong. It is to say that his question becomes more concrete once we bring it down from “capitalism” and “socialism” or America vs Europe to the level of institutional design, not to mention cities and neighborhoods. When I lived in Chicago, if you wanted to play tennis, you often had to wait for a public court, and more than a few times fights would break out over who had next. How can a problem like that be solved? Consider as an alternative my current city, Toronto. Toronto is in Canada and you might therefore conclude that it would provide tennis courts as a public good, at least more directly and fulsomely than the more laissez-faire USA.

In fact, the city does have much free leisure swimming and many public tennis courts, but it also relies in part on community tennis clubs operating on public land. These clubs charge (quite reasonable) membership fees, run leagues and lessons, maintain courts, and are required to provide public access hours. In exchange for operating on public land, they must make the courts available to non-members for six prime-time hours per week, including two hours on weekends or holidays. They are also required to have membership accessibility policies, including financial and physical accommodations. This is not laissez-faire capitalism, and it is not state socialism. It is, well, Canada.

I bring up this mundane example of tennis courts because it helps to make more concrete what the vision of the good life as cultivation of human potential really means. We seem ready to agree that the arts, sports, and natural beauty would be a part of that, but should these be directly embedded in and provided by state planners?

Philosopher Joseph Heath in his The Machinery of Government recounts the sad story of urban planners taking to heart the proposition that nature is good and all people should have access to nature.

In the early 20th century, many urban planners were persuaded that parks would produce ‘moral improvement’ in the working classes—a judgment that was particularly influential in the decision to raze working-class slums and replace them with residential towers surrounded by green space. This backfired spectacularly, as the green space remained largely unused, making it a magnet for crime. As a result, the ‘parks are good for you’ judgment that prevailed in the first half of the 20th century is now regarded as involving an impermissible partiality toward a particular set of values. Some people like trees and green spaces, some people do not. Planners tend to share the values of those who do, but it is a clear violation of liberal neutrality to use that as a basis for decision. No matter how much planners may like parks, they must ultimately defer to patterns of use.

In a situation like this, the commitment of planners to confidently and knowingly provide the specific good of access to nature for all is exactly what produced dangerous and neglected swathes of urban blight. You could say that they should not do that, and yes they should not, but if you ask what they should do instead – and you resist the temptation to say that all should become new men of socialism – this leads you to the kinds of tedious technocratic cost-benefit and willingness to pay analyses that Heath describes in the Machinery of Government. In other words, to technocratic liberalism. That world is often dull, but dull is not always wrong, and if your vision of the good life does not include it, or is so repelled by it as to wish to look past it to Utopia or a Glorious Golden Age, so much the worse for that vision.

Having spent a fair amount of time actually working on city cultural policy and planning, it is hard to avoid the liberal’s intuition that, yes, the swimming, the arts, and tennis might be very good, but it is not our job as planners to make any specific form of the arts or sports or any other candidate good the target of a political program. We want to give room and resources for people to pursue their visions of the good as they understand them, create organizations and practices for realizing them, while remaining attentive to whether anyone actually uses them, not to mention unintended consequences of our own efforts to provide them. And that is good enough, at least for creatures like us.

That’s a thin notion of the good, and I, like Jonny and the other contributors to the Point forum, do feel some compulsion to say more. What I want to say is that this very vision, the second-order one in which many people have the chance to pursue many different life plans, is a worthy project in its own right, animated by a vision of the good life. You could call that vision one of Odyssean adventure, or maybe one of Baudelairean flânerie. Or maybe a combination of both. I’ll call it The Odyssean Option.

The Odyssean spirit of adventure is one of entering into some new country or arena, finding some challenges thrust on you, and then moving along to the next one. Each place you go, you have to learn its local norms and customs, you have to open yourself to them, make yourself adaptable and flexible. You have to be a person of many shapes, clever, wily, and yes maybe a bit of a trickster. You have to enter deeply enough to learn, but not so completely that you never leave, or lose the drive to move on, which would ipso facto end the quest. You have to resist mistaking any one island for the whole of life.

This ideal in turn depends on a rich set of norms regarding relationships between hosts and guests. Hosts should welcome guests, give them safe passage, allow them to dine and enjoy their company, participate in the life of their community – without expecting them to fully conform to local norms or display complete commitment to them. They should allow the guest to move on when they are ready. The guest should enter into the world of the host with curiosity and humility, allowing the host to guide them to what is worth paying attention to. The guest should be ready to set aside worldly knowingness temporarily and allow herself to be carried by the inner logic of the time and place. And yet, all should be ready to move on when that moment has passed.

This is the Odyssean version, and if you like that kind of thing, it can help to envision the liberal spirit as a kind of heroic quest to sail from island to island. But maybe you are more of a city dweller, more of a modernist. Then the Baudelairien version could be a better place to start. The Baudelairean flâneur moves from scene to scene, tasting what it has to offer, peeking in to its events and activities, even joining for a while, blending in enough to temporarily seem very much like a full member. But always holding some part of themselves in reserve, remembering in the back of the mind that there is another scene to visit tomorrow, and more ways for this very scene to manifest and join elements than it does in this specific location.

In this way, flâneurs distinguish themselves from both the single-minded political activist and the empty-minded contemplative mystic. In contrast to the activist, the flâneur never fully, wholeheartedly commits to any scene in which they participate; they keep some part of the self in reserve. And unlike the mystic, the flâneur does not withdraw from life into speculative contemplation; they are instead out and about, on the streets, prowling the alleys. Here too is a vision of the liberal good life, moving along from street scene to street scene, like a rolling stone: Baudelaire’s heroism of modern life, with its own vices and challenges to constantly be on the lookout for, which require courage and conviction to face up to.

To be sure, you might worry that if everybody adopted this somewhat ironic view, then nobody would ever fully commit to anything. For my part, I’m not too worried about that. For one thing, we can worry about it if it happens, rather than waiting until the Revolution. For another, it is hard for me to imagine what that would look like.

Take my own enthusiasms of boxing and curling. A scene does not need everyone to be equally committed. In fact, most scenes depend on gradients of commitment. My gym is full of a range of people, some training for amateur or even professional competition, other high level hobbyists, others there for fun and fitness. This kind of mix is what makes the entire gym hold together, and the casuals do not undermine the small number of highly committed competitors. Rather, this creates a broader social environment in which their own commitment shines out in its contingent splendor. And anyway, while neither boxing nor curling fully consume me, I am still pretty darn into them, even while leaving enough room in my self to cultivate other pursuits, such as Substack essays, academic research, and being a solid upstanding family man and community member.

Yes, you could call this cosmopolitanism, and that is probably okay. But if the other guys are going to bring the Benedictine Option to the fight, you should not bring a knife. You should pull out the Odyssean Option.

Above, I mentioned that we had counted the number of public libraries in Chicago and compared that number to all other US cities and neighborhoods. This could sound like a pretty dull exercise compared to the high questions of social and political theory animating The Point contributions and the (excellent) podcast conversation among Jonny and his interlocutors.

One thing social science can add to conversations like this is a way of following the question downward. If we are going to argue about whether societies make human capacities available, we should know where the pools, courts, festivals, libraries, boxing gyms, churches, galleries, and curling clubs actually are. You don’t even have to learn to code anymore, Claude Code or Codex can do it for you.

At a first cut, even simple exercises like counting the number of public pools or festivals across cities and neighborhoods and countries is both very much within our reach and helps to ground statements we might be tempted to make about how prevalent various practices are, and where they might be more or less common, and why. For example, the City of Berkeley (home to my other alma mater, and a place where the left has a chance to put its vision of the good life into practice) does not offer free public pools. These are basic empirical questions that can inform our conversations about who is and is not realizing some progressive notion of the good life, and help to figure out why.

More generally, however, it is worth considering what kind of social science can even incorporate into it questions of the good life. Much like the background worry animating the entire Point issue – that such questions don’t really have a place in the discourse of the left – they also are not very well posed within the standard practice of social science. The surface form this takes is in the vast bulk of contemporary sociology dedicating itself to identifying various inequalities and calling it a day. The deep reason is that much social science operates by a background assumption that social life is driven by large impersonal forces, usually economic or geographic, pushing people around, behind their backs, as it were.

In the more mundane practice of social science, this background manifests on the surface in terms of the types of variables that one tends to include in a statistical model, such as age, income, ethnicity, or physical proximity. All of those are obviously important, and obviously all can feed into a notion of social life as in significant part driven by decisions and judgments by responsible human actors. But to the extent that they do, they become something else: components in notions of the good life, such that, for example, closeness to friends and family is more important than wealth, or utilitarian considerations override those of tradition, or your ethnic heritage is considered your true and real wellspring of selfhood.

If that is the case, what we want to know via social science must include knowledge about where and what these visions are, and how much latitude they have in animating people’s decisions, over and above the forces operating upon them. That’s what we tried to do in Scenescapes, and still do, now with more and better data. We treat local amenities not merely as conveniences but as clues to styles of life. A place full of churches, bowling alleys, and family restaurants offers one kind of life. A place full of galleries, boutiques, cocktail bars, and music venues offers another. Neither is reducible to income or education, even though both are entangled with them.

Every time I return to that work, I am surprised again by how strong the results are. If you want to know what predicts people moving about from one place to another, for example, the very best indicator, after geographic distance, is the set of local amenities it has to offer, and the specific cultural style it evinces – whether it offers a life of transgressiveness, glamour, neighborliness, local authenticity, tradition, personal self-expression, utilitarian functionality, ethnic authenticity – is about as strong as education or voting.

It has been and will be an uphill battle in the social sciences to convince mainstream researchers that something like a notion of the good life plays a strong role in driving large patterns of social life, but we struggle on. I highlight this in part because ever since I found myself, through a random series of events, doing empirical social science, it has often been hard to explain the significance of that work to my old Social Thought friends, and this particular conversation helps articulate it. Counting pools, courts, and festivals is one way of asking where candidate answers to the question of the good life actually exist.

I also highlight it because, like Jonny’s swimming pool example, it brings things down to earth. For example, in our chapter on policy (Chapter 7, “Making a Scene”) we talk about how to encourage local policy-makers and politicians (e.g. mayors and city councillors) to think about the kinds and ranges of conceptions of the good life that their cities make available to their citizens. It is again very mundane advice, like: Walk around your city and talk to people. Ask what forms of life are already being cultivated there. Ask which ones are missing. If they are missing, ask whether there are nascent versions hiding somewhere: a chess club, a boxing gym, a church basement, a music scene, a garden collective, a cricket league, a poetry reading, an immigrant festival, a half-dead theater, a public pool nobody maintains properly. Talk to the people who care about them. Ask what they need to make it grow. Often even small things, like a little prize or public recognition, or just an opportunity to exhibit and showcase talent, can go a long way.

Yes, this is not so high flying as the Marxian revolution and the unleashing of our species being. But if the aim is to give people, in the real world we inhabit, a greater chance to cultivate a greater range of human capacities, this is the kind of thing one actually does, and the kind of aspiration that can give its creation meaning.

In a recent episode of the Argument’s podcast, Matt Yglesias and Jerusalem Demsas discussed Jane Jacobs’ Death and Life of Great American Cities. Yglesias in particular was pretty harsh on Jacobs, especially her “insistence on small city blocks” having a kind of causal power to animate city neighborhoods1.

The episode was catnip for me. I’ve long been interested in the conditions that favour the emergence of bohemian and Neo-bohemian neighborhoods, as well as their social and economic consequences, questions which we pursued in Chapter 4 of Scenescapes. We also in Chapter 6 examined the conditions under which various “new social movement” organizations tend to appear, and found that, more or less, they thrive in places like Jacobs’ Greenwich Village — educated, dense, walkable, full of artists and self-expressive scenes — but elsewhere they are relatively rare.2

This is just to say that when I saw Yglesias’ claim, I was primed to test it.3 So I put the new Codex model to work (Solus Ultra), fired up goal mode, steered the model here and there, and a couple of days later, we have our answer:

Codex has spoken: Yglesias is right, Jacobs is wrong.

Below I’ll lay out the evidence. If you want to read the whole Codex generated report, you can see it here.

The first thing we needed to do was measure short street segments, and classify neighborhoods by their historical inheritance of this type of urban form. You can see how deeply this pattern is ingrained in Greenwich Village, which was basically formed already by 1857 and still has a similar shape today.

From these metrics we then scale up to a national analysis of 9,361 tracts in 343 metropolitan areas, with the following formula:

It is difficult with public data to determine if distinctive social groups already preceded the later flowering that Jacobs observed. But we can see already in 1940 that Greenwich Village had a fairly distinctive professional, intellectual, and publishing milieu.

Yglesias suggested that it could be the somewhat contingent process by which the Village became a gayborhood that was more important than the street length, and that if this had happened in a place with longer blocks, which it definitely might have, then the development into something Village-like would have happened there. This is hard to pin down, but we can see some signs of LGBTQ sites already in the 1950s there.

We have richer data to work with in more recent years. The basic question is whether a broadly “bohemian” or at any rate “creative” business ecosystem is distinctive to places with similar street grids. To do that, we built a “sustained business-ecology index,” which joins arts, hospitality, tech, and media organizations. We also analyze the various components separately.

Here is the main pattern:

Matching provides another test of the short-block proposition. The 2,340 tracts in the top 25 percent of the business-ecology index were each paired with a tract outside the top 25 percent in the same Census region. The paired tracts were similar in building age, share of buildings constructed by 1970, population density, renter share, and diversity of local businesses in 2012. Street layout was deliberately left out of the matching process. The test then asked whether the high-business-ecology tracts had more inherited short blocks. They did not. Their short-block measure was 0.026 standard deviations lower—a tiny difference that was not statistically significant (p = .397).

We also pursued some additional formal tests, in which the business ecology index is the outcome, as well as separate models with arts organizations, hospitality, and tech. The main models include short-blocks, metropolitan-area fixed effects, median built year, share built by 1970, log 2010 population density, and pre-1950 road-network density. Others add renter share and racial/ethnic diversity, as well as college and professional shares.

Basically, we find a small association between small blocks and the business ecology index at the national level, but this was below the threshold we (me and Codex) had set before examining the results as what would constitute a substantial finding.

Still, if we decompose that association, we find that it is largely driven by a relationship between short streets and more hospitality and food establishments. But there’s no association to the arts. More specifically, A one-standard-deviation increase in short-block exposure is associated with a 0.130-standard-deviation increase in hospitality’s share of local businesses, while the corresponding associations for arts and technology/media per resident are near zero. So short blocks might give a small boost in restaurants but not much else.

Of course, Jacobs did not really think short streets would operate totally independently. She presented them as one of four conditions—alongside mixed primary uses, older buildings, and sufficient density—that work together. To at least begin to acknowledge this, we tried a few interactions, but none showed anything very strong, and where there was a statistically significant association, it pointed in the opposite direction from what Jacobs would have predicted.

We also tried to find some natural experiments where the street grid had changed and we could track pre/post activity, but weren’t able to come up with any good ones. The best we could do is Wycoff Plaza, which seems to have continued to decline in foot traffic after the intervention.

We also searched through the historical street network data, looking for instances where an existing street fabric became segmented. The idea is that this could give something closer to a causal estimate, where we observe changes in groups and activity patterns before/after the street grid becomes more Jacobsian.

Alas, after surveying the history of about 15000 census tracts, we did not find any good candidates. About 330 seemed promising, but after closer review of their actual topologies, there were only three viable ones. Most cases did not show a single new street connection added to an already-built neighborhood. Often, the roads and surrounding development appeared together. In other cases, several roads were added at once, the project created an entire new network, or the apparent road did not connect at both ends in a way that divided an existing block. These cases therefore could not provide a clean test of what happens when one inherited long block is split into shorter segments.

Rather amazingly, Codex gathered arial photos to look at three remaining candidates more closely, before rejecting them as well. For example, here is Summers Corner, South Carolina.

There were a lot of new short street segments, but this was part of a totally new master planned development, so you can’t really separate the specific street grid effect. So, after sending Codex off on a few hours mission to look for natural experiments, we had to abandon that for now. If you want to look at its report, here it is.

If anybody has any suggestions for additional analyses, let me know, and I can put Solus Ultra to work on it.

In the end, here is where we were left:

What is left of the proposition?

Taken together, the evidence establishes that Greenwich’s fine-grained physical form is old, persistent, and temporally prior to Jacobs’ observations. It also shows a socially distinctive milieu before the observation period.

Across 9,361 tracts in 343 metropolitan areas :

• the broad business-ecology index has a positive but weak association and negligible held-out predictive gain;

• the arts-specific signal is practically null;

• high-grain failures and low-grain successes are common;

• matched high-ecology analogs do not have more short-block exposure;

• the positive component is hospitality share, not arts, tech/media, or combined establishments per resident

What evidence would change the conclusion?

1. Point-level cultural establishment histories that distinguish entry, survival, relocation.

2. A meaningful post-treatment effect beyond hospitality that appears after multiple verified street changes with acceptable pretrends, controls, spillovers, and topology or footfall first stages.

3. Direct historical artist, LGBTQ population/institution, venue, premise, affordability, transit, and centrality measures observed before the outcome.

4. A positive interaction pattern for a fully measured Jacobs package—mixed uses, short blocks, varied building age/condition, density, and affordable adaptable premises.

5. Observation-period evidence from 1955–1961 that connects physical form to actual street encounters, establishments, and cultural activity rather than relying on later business proxies.

Daniel Silver’s reflections on the 1950s and 1960s era debate between C. Wright Mills and Talcott Parsons are a useful reminder that sociology’s founding disputes are more complicated than the simplified stories we tell our students. In our textbooks and introductory courses, Mills typically appears as the hero of public sociology: the defender of the sociological imagination, the critic of power elites, and the scholar willing to speak plainly and with substance to a wider audience on social problems that affect individual lives.

Mills got much right, so his revered status in sociology is justified. Mills’ example helped establish the ideal of the public sociologist long before that term became fashionable. For many sociologists, Mills remains proof that serious scholarship can also be publicly engaged even if some are rightly concerned than Mills and the Michael Burawoy tradition that he inspired is too dogmatically left-critical. Mills work on power elites remains relevant in an age characterized by concentrated corporate power, technological monopolies, widening economic inequality and a global oligarchy in the wings.

It is reasonable, of course, to point to different perspectives on sources of inequality in contemporary society that complicate Mills pessimistic account of the power elite and the lives of the working and middles classes. It is hard to deny, however, that his contributions shape our scholarship and the education of our undergraduate and graduate students, even if few read his books today. Above all else, Mills’ sociological imagination brings students into the discipline by helping them see how their own personal biographical problems are about history and social structure and can be addressed by social change led by social science.

Parsons, by contrast, is remembered as the architect of abstract systems theory whose dense prose symbolized everything wrong with a postwar sociology that was essentially a conservative justification for the status quo. Silver’s work on the reception of Parsons and Mills, both in published academic works and this Substack, suggests that this familiar story is too simple. While Mills was often insightful, courageous, and intellectually innovative he was also a polemicist with an axe to grind. Mills’ attacks on Parsons frequently relied on ridicule rather than careful engagement with theoretical content. Mills’s mockery of “grand theory” create a caricature that still shapes and distorts the discipline’s collective memory.

Something similar can be said about Mill’s treatment of Paul Lazarsfeld and survey research, as he famously dismissed the mainstream quantitative tradition that came to dominate sociological research beginning in the 1950s as “abstracted empiricism.” Mills lost that battle, and sociology programs now center quantitative training and research even if Mills’ critique offers an alternative view that legitimizes some of the resistance to modern multivariate analysis that we still see in our discipline. At his best, Mills work was a reminder of the importance of preserving the craft of sociology in the wake of pressures towards the field becoming fully dominated by technicians. At his worst, however, Mills legitimates anti-intellectual dogma and ideology.

Mills often won arguments rhetorically in ways that create heroes, villains and myths, something key to understanding his continued appeal. This does not lead to a politically balanced and empirically careful sociological imagination, a problem we can rethink with the kind of empirical research on the reception of sociological theory Silver and his collaborators engage in.

I would like to sharpen this analysis of the negative consequences of the reception of Mills in sociology, using the insights of Alvin Gouldner, a powerful 1960s era sociologist who was part of the broader New Left intellectual movement that allowed Mills and his generation to de-thrown Parsons. Alvin W. Gouldner (1920–1980) was one of the most influential and controversial American sociologists of the postwar era. Trained at Columbia University and later a professor at Washington University in St. Louis, Gouldner established his reputation through empirical studies of organizations and bureaucracy, particularly Patterns of Industrial Bureaucracy (1954).

During the 1960s and 1970s, however, Gouldner became a leading critic of mainstream sociology, particularly in his The Coming Crisis of Western Sociology (1970). Goulnder, like Mills, argued that the discipline had become overly technocratic, politically complacent, and detached from the major moral and political conflicts of modern society. Mills was dead, of course, by 1962, and Gouldner had no interest in being a public sociology or a textbook writer, so he not Mills developed the most sophisticated scholarly critique throughout the 1970s of Parsons, Goffman and Merton, and other leaders of the post-war sociological field in the period. Gouldner inspired a generation, the graduate students of the New Left generation who entered the field and professionalized in the late 1970s and 1980s, free of the Parsonian and Mertonian dominance that Mills and Gouldner were shaped by and rebelled against.

In The Coming Crisis of Western Sociology, Gouldner argued that postwar sociology—especially the structural functionalism associated with Talcott Parsons—was entering a profound intellectual crisis. Gouldner maintained that sociologists could no longer pretend to be neutral observers standing outside history. Instead, every theory reflected the social position, values, and commitments of those who produced it. Gouldner called for a more “reflexive sociology,” in which scholars critically examined their own assumptions, institutional interests, and political commitments rather than claiming complete scientific objectivity.

Like Mills, Gouldner was not a perfect role model for contemporary professionalism. Gouldner’s critiques of Parsons, value-free sociology, and liberal academic complacency were powerful, but they were delivered with a prosecutorial certainty that appeared arrogant to many colleagues. The notorious 1968 Laud Humphreys incident, in which Gouldner was accused of physically attacking a graduate student over a satirical poster, established a reputation for volatility.

Gouldner’s fate is therefore ironic: the great theorist of reflexive sociology and the “flawed universal class” of the new class was himself often seen as a flawed intellectual, brilliant at exposing the self-deceptions of others but less successful at softening his own. But unlike Mills, Gouldner was willing, at least in principle, to turn his reflexivity on radical intellectuals themselves in ways Mills never did, especially in Two Marxisms: Contradictions and Anomalies in the Development of Theory. Gouldner’s book and his broader reflexive sociology predated Bourdieu theories and provides a framework for thinking critically about Mill’s concept of the sociological imagination discussed in Daniel Silver’s recent Substack post.

Silver’s account of core argument of “The Promise” argues that the continuing relevance of Mills is rooted in the popularity of the case he makes that “sociology can be a beacon of clarity in an increasingly confusing world.” Silver writes:

Mills begins with his famous observation-cum-injunction: that people today often feel trapped by their private lives, unable to grasp the larger historical forces that shape their experiences. The core of his “promise” is the Sociological Imagination: a quality of mind that enables us to grasp the connection between our private troubles (one person’s unemployment, a couple’s marital strife) and broader public issues (a nation’s high unemployment rate, legal restrictions on divorce). The Sociological Imagination is the ability to link our individual biography to the grander sweep of history and social structure.

Our students love Mills and the sociological imagination because all individual problems can be explained through this social lens which was right about many issues. Who would not want to major in a field that helps one understand, and potentially resolve, all one’s personal troubles? This sets our students and our discipline itself, however, for an obvious rebuttal: is it really the case that one’s inability to succeed in the labour and marriage markets are fully social problems, with no connection to bad individual decisions, negative character traits, dishonesty or delusions? To ask that question is to answer it, but this is not the kind of discussion that is encouraged in introduction to sociology classes, or graduate seminars.

Gouldner’s concept of “theoretical nightmares” helps us see how both the genius and the limitations of Mill’s sociological imagination and his theoretical frame prevents sociology from dealing with this obvious contradiction. For Gouldner, all theoretical systems have an emotional and symbolic life of their own, including a need to defend the boundaries of what is acceptable discourse within a school of thought by expelling ideas that pose a challenge to the theoretical system itself. Ideas that challenge an intellectual system can never be fully expelled from discussion, however, and the most serious challenges tend to emerge within the theoretical system as an unconscious “nightmare” that keeps theorists up at night and on guard in the interests of protecting the purity and consistency of their system. The repressed nightmares never go away, however, because the problem is with the theoretical system itself.

Gouldner developed this argument most fully in his sociology of Marxism. Marxism’s nightmare was the possibility that the working class might not become the revolutionary force and universal class predicted by Marxist theory. What if workers preferred reform to revolution? What if they became attached to nationalism, religion, consumerism, or liberal democracy? What if the abolition of private property failed to solve the deepest problems of social life? Such possibilities haunted Marxism because they threatened the coherence of the entire theoretical system. Marxists have a complex set of strategies they use to avoid these questions, as they either expel heretics or develop explanations and theoretical traditions that protect their foundational commitments in the form of Neo-Marxisms that focus on culture, gender, race or other ways to preserve the theoretical system in modified form.

Goulder’s insight, however, extends far beyond Marxism - Freudian psychoanalysis provides another example. Freud developed a powerful account of human suffering rooted in unconscious conflict. Yet psychoanalysis has always downplayed the possibility that some forms of suffering may not be psychodynamic in origin. Biological illness, neurological conditions, genetic predispositions, and simple misfortune do not always yield to psychoanalytic interpretation. Nor can psychoanalysts really deal with the broader forms of structured inequality that Marxists and sociologists do a far more credible job of analysing.

In my own work on psychoanalysis, I have argued that psychoanalysis, like Marxism, has its nightmares—questions that threaten the coherence of the tradition itself. Maybe it is not possible to fully heal patient’s emotional lives in clinical settings without a broader change in society, more equality in families and communities, less craziness in politics and social media? Contemporary psychoanalysis is currently being torn apart by internal debates about racism within their organizations, conflicts over medical care and therapeutic treatment for transgender youths and even the Israeli Palestinian conflict.

Perhaps the psychoanalytic focus on understanding the Oedipal complex, internal psychic conflicts and early family emotional dynamics just does not take the field very far, given all the problems in the world and in people’s lives clinicians of good faith have been asking? When psychoanalysts try to deal with the bigger questions of politics and social structure, however, they don’t do a good job addressing them because they are not equipped with the analytic tools or organizational resources. They end up damaging their ability to do therapy and healing, as their institutes descend into political battles and acrimony. Dealing with psychoanalysis’s nightmare in a mature way, would involve updating their theories and accepting the limits of their craft. Instead, the field engages in civil war over something that cannot be fixed but can only be accepted and adjusted to.

Sociology has a similar problem, that cannot be understood and addressed without serious attention to C. Wright Mills because sociology’s core nightmare is that some personal troubles may not be fully social problems. If one remembers the old joke “How many sociologists does it take to screw in a light bulb?” the answer “None, since it is not the light bulb that needs changing but society and its social structure”—nicely identifies sociology’s core repressed nightmare. For sociologists it is a profound challenge to our way of thinking and our research traditions to suggest that individual people (or light bulbs, if you accept the analogy) require changes at the social psychological, cultural, and depth psychological levels in order to see the emergence of a decent, fair, or just a sustainable society. From the sociological perspective, it is all about the structures, and that is clearly not true.

In the current political moment, furthermore, and in the context of the ongoing institutional crises universities are facing, particularly in the United States, C. Wright Mill’s promise, for all his insights, is getting in the way of changes we need to make to revitalize our field. A mature sociology should be able to acknowledge that some troubles are indeed public issues while others may arise from biology, chance, character, family dynamics, culture, or individual choice. Human problems involve multiple causes; sociology contributes one set of explanations among several. There is a major imbalance in the political views of professors throughout all disciplines in contemporary universities that has many causes, as the work of Neil Gross has shown. Sociology’s one-sided focus on the social roots of all personal problems, however, surely helps explain why we have so few conservative, religious or even politically moderate practitioners of the sociological craft.

Parsons, in contrast, reminds us that society is far more complex than the model Mills offered us. Parson’s work emphasizes systems, institutions, differentiation, adaptation, and multiple forms of causation. Mills offers something more immediately attractive and gives sociology a moral mission. Mills provides a language for linking private suffering to public injustice. Parsons reassures sociologists that their discipline occupies an essential place in the intellectual landscape, but encourages us to dialogue with psychologists, economists, biologists and anthropologists, an essential move in a moment where simplistic tropes about all social problems being caused by social inequality and structures are just not credible.

If Mills helped sociology find its public voice, Gouldner can help sociology find something equally important: intellectual self-awareness. In an era when sociologists are increasingly called upon to explain political polarization, mental health crises, and drug crisis that self-awareness may be more necessary than ever. Not all these issues, and the personal troubles they create are explainable in purely sociological and structural terms as the Mills framework insists.

If we try to explain Trump’s two presidential victories as coming from the fact that Republican voters do not fully understand the social causes of inequality, crime, homeless and drug addiction in the sociological ways we do, we are avoiding some hard questions. If we see the mental illness crisis as a social problem only sociologists can solve, we are deluding ourselves. Surely political conservatives, religious leaders and experts in the neurosciences have something to say about these issues. Is the massive drug crisis among young man in North America, with fentanyl overdoes laying waste to so many youth lives simply a social problem requiring a change in social structures? Or do we need to address the effects on addictive drugs on young bodies, economic incentive structures and the ways that bad decisions over time leave some members of our societies left out of the social contract. Policy makers and the public today have less confidence and trust in the sociological perspective than they have had in decades, and part of this is due to the destructive rhetoric of irresponsible anti-intellectual populists of the right. But surely part of the problem is the sociological imagination does not offer anywhere near the level of insights into these multiple contemporary social problems as some of us have convinced ourselves they do.

Perhaps the most valuable lesson we can take from thinking about Mills through Gouldner, is that sociology flourishes not when it avoids its nightmares, but when it confronts them directly. Mills encouraged us to look beyond the individual toward society while Gouldner encourages us to look back at sociology itself. If Mills remains the patron saint of the sociological imagination, Gouldner may still be the discipline’s most underappreciated guide to intellectual humility even though no-one ever accused him of being a model of humility himself.

Sociology needs to also find more appreciation for living with contradictions, ironies and insights that can be found in places we might never first look first. That means drawing on Gouldner to address the obvious flaws in our own theories, and engaging with more conservative, moderate and religious thinkers and scholars even though he was part of the problem of an overly ideological sociology in the 1960s and 1970s. To do all this, we must drop the strawman account of functionalist sociology that Mills’s promise helped institutionalize at the core of our contemporary discipline.

I’ve always had a soft spot in my heart for fast food (which might explain the soft spot in my belly). I spent lots of good times in fast food joints in high school, not to mention late night runs to Taco Bell in college. Long road trips are one of the great joys of being American, and part of that is knowing that wherever you go, you’ll find a McDonald’s. When traveling to other parts of the world, of course I sample the local cuisine and now and again enjoy some fine dining. But the predictability and reliability of the fast-food restaurant gives a warm feeling of certainty and familiarity.

Fast food restaurants are also sociological marvels. They are where you’ll find perhaps the widest swathe of the population occupying the same physical space. Young and old, every color in the rainbow, professors and plumbers, you name it, McDonald’s has something to offer, and a place to see each other seeing each other seeing each other being seen. I routinely work in my local McDonald’s, enjoying the free refills of Diet Coke, watching the most diverse city in the world flow through its most diverse social node.

It was actually in my neighborhood McDonald’s where I did most of the work on my and Joseph Yi’s paper, “God, Yoga, and Karate” (which also appears in revised form in Chapter 5 of Scenescapes). In that paper, we compared various establishments across the USA to examine how widely or narrowly they were distributed across various neighborhood types. Yoga (and “new age” establishments more broadly) tended to be restricted to more highly educated and white areas; conservative evangelical churches to lower educated and racially diverse areas. We interpreted these findings in terms of the histories and challenges of those cultural and religious movements in reaching out beyond their core constituencies.

By contrast, martial arts clubs exhibit a pattern that is much more like fast food restaurants: they cut across nearly all social divisions. The core of our paper is an examination of how exactly martial arts clubs achieve that remarkable social standing in the North American context. But we also briefly suggest that fast food achieves its own cross-cutting integration in part because it has adopted as its guiding philosophy a certain pragmatist ethos: whatever works. If people want Tuscan salads in addition to burgers, no problem, here you go. Smoothies and cappuccino? You got it. Impossible Burgers? We’ll try it out and see how it works. Belly busting heart attack bacon burgers and 5000 calorie shakes? Why not. Keto-friendly-no-bun-patty-only? Sure.

Given this centrality and adaptability, it always strikes me as short-sighted and elitist when people turn up their noses at fast food. I recall a meeting a while ago with some senior officials in a City of Toronto agency. We were discussing our work on urban scenes, which they took to be a brief for making every neighborhood into an artsy neo-bohemian enclave (it definitely is not that, but that is another story).

Somehow fast food and strip malls came up, and one of the officials said something to the effect of, “I just can’t understand why anybody would ever go to that kind of a place, I’d never step foot in one.” This I found astounding. Those are where most residents and citizens are going, willingly, and often with pleasure! The strip malls of Toronto (a different topic, worth discussing) are hotbeds of social creativity. To build a planning apparatus based on the tastes of those who would never enter into the places where the actuality of the city’s dynamic diversity plays out is a recipe for alienation and disaffection.

It is a complex question as to where highbrow disdain for fast-food comes from, but part of it likely has to do with how such venues are or are not presented in highbrow culture. You don’t often see them given a generous and rich hearing. It is therefore of some interest when very good literary authors turn their attention to the experience of fast food.

As it happens, I was reading Jennifer Egan’s Look at Me recently. She’s one of my favorite novelists, and this was the last of her novels I hadn’t read. I was pleasantly surprised to find a few interesting scenes in which fast food becomes a kind of active character. That triggered another synapse in my brain, which recalled that Knausgaard in The Morning Star had set an extended scene in a Burger King. So I went back to both, and read them again with a view to the significance of the fast-food restaurant.

I will not review the overall plot and meaning of The Morning Star here, you can look at my longer discussion in The Point for that. To appreciate the Burger King scene, it is enough to note that it features Iselin, one of the characters through which The Morning Star unfolds. She is 19 years old, lives alone, and has dropped out of university. She works at a supermarket checkout, and is deeply ashamed of her body and how others see her.

Later in the novel, she allows a disturbed young man into her apartment, who turns out to be a missing member of Kvitekrist, a dark metal band tied to a series of ritual murders. Iselin hears him screaming and terrified, calls the homeowners and police, but he runs away and police cannot find him. The band becomes a central focal point for whatever dark forces the new star hanging over the world has unleashed, and continues to appear in crucial ways in later books in the series, though exactly what it all adds up to is as of yet quite opaque, at least to me.

Iselin ends up in Burger King in the wake of a series of humiliations. She has been working at the supermarket all day, reflecting on how pointless and stuck she feels, along with how unattractive and sweaty the uniform makes her look. One of her old teachers comes in, recognizes her at the checkout, which throws her deeper into despair, since it reminds her of what she recalls as a more talented and promising version of herself.

He invites her to a café, she agrees, but she knows she will not actually go. Instead, she spends the day eating peanuts and drinking Diet Coke (she and I have that particular pleasure/vice in common), sitting alone in her backyard. She tries to get in touch with her younger brother, who is already out with friends, but gives her the cold shoulder. She responds by changing her clothes, putting on her makeup, and walking out into the crowded nighttime social scene. But that crowd was a rather lonely one, and anyway, she was hungry, so, she goes to Burger King. Channeling Bon Jovi, she says to herself, “it was my life, no one decided for me.”

Let’s start by noting some of the writerly aspects to how Knausgaard sets up the scene. A sign in Knausgaard that we are entering into a zone of altered or deepened consciousness is that he starts talking about features of everyday contemporary urban life as if we were in a northern landscape painting.

Above the islands in the fjord the sun was on its way down, and as I emerged from the lane I had to narrow my eyes in order to see properly. The wharf and the fish market were teeming with people. On this side of the road too, the crowds were huge. I weaved my way among them, grateful for the music in my ears that transported me to another place, making all the faces that surrounded me seem remote and irrelevant.

For me at least, these kinds of paragraphs in Knausgaard are kind of like putting on your martial arts uniform before class, bowing, and stepping onto the mat. You know you are leaving ordinary life and entering a domain where different rules apply.

Once Iselin gets inside Burger King, Knausgaard shifts from landscape to still life. The burger is frozen before us, like a bowl of fruit or a dead hare. Time slows down and things that normally pass by become magnified. In this case, we get two pages about the act of deciding what to order, bringing your tray to the table, eating your meal, and deciding whether to go back for more.

We’ve all been there, at least I have, the juicy meat filling your mouth as all the other components of the burger lose their distinct identity and become, first, one with each other, and then, with you.

I opened as wide as I could but still couldn’t get the whole burger in for the first bite, the bun crumpled against my top lip as the juicy meat filled my mouth and everything melted together, bacon, beef, bun, onion, tomato, lettuce, ketchup.

And yet, just as you feel satisfied with that Oneness, you look back at the tray, and the reality of Difference makes itself known, in the form of the fries.

Who has not sat there, alone, facing that question: to mix or not to mix? I remember when I was younger, I experimented with putting fries on my burgers. If they are all going to the same place, why not cut out the middle man? But like Iselin, I came to believe that this robs each of what makes it unique.

I always saved the fries until last, the two things were so different I didn’t like to mix them. One juicy and open, the other crispy and dry on the outside, the good, soft part locked inside. The burger I ate quickly, the fries took longer as I dipped them in ketchup and ate them one by one.

Burgers are juicy and open, fries crispy and closed; burgers you want to experience all at once and all together, fries you want one at a time. If you throw them together, you lose the deeper unity that integrates the fast-food meal as a whole.

And yet, there is great longing on those empty trays, since a fast-food meal is never ultimately and fully satisfying.

And I was still finished all too soon.

There is somehow always more to the fast-food meal, and we all have the little ways that we contemplate what to do next and justify allowing it to continue.

I was full, but wanted some more. The natural thing would have been to get a milkshake for dessert, that was what people did. But there were so many calories in them I could just as well have another burger, maybe the Bacon King with fries I’d been thinking about.

Did I really want more?
Yes.
Why shouldn’t I, if it was what I wanted? Who, or what part of me, was it that said no?
Reason. My guilty conscience.
Why not show a bit of willpower, get up and leave?
It was exactly these kinds of situations that were crucial.

There’s a way in which staring at that empty tray brings you face to face with yourself. At least it does for Iselin, and now having spent several paragraphs slowly illuminating the experience of sitting alone with yourself at a BK, Knausgaard has prepared us to encounter strange and unusual things there. This too is not a stretch: you are very likely to encounter weird people doing weird things in a fast-food restaurant, which is another way they take you out of yourself.

The scene ends when a man comes up to Iselin as she prepares to eat another Bacon King and drink another Diet Coke. He comes up to her and says “You eat too much,” followed by proclaiming “I am the Lord” and running his fingers through her hair. Her body reacts with involuntary warmth, she loses her appetite, eats only half the burger and a few fries, and heads out.

Further observations on how the sun shines in the summer follow.

There is more to be said about this scene, but this is enough to highlight its distinctive character. From one point of view, the defining feature of a fast-food restaurant seems to be convenience: they are places you go to in order to get somewhere else, not to be. You really cannot just be there, it seems; they have no being. That, at least, is how they seem to appear to the aforementioned city official.

And yet, people are there, all the time, sometimes for a long time. They routinely host bizarre and uncanny encounters, with others and yourself. They are zones of temptation and satisfaction and longing and regret and loneliness upon loneliness upon loneliness and the promise of Something, Something More, with Somebody Else.

Knausgaard in these few pages tries to arrest us there, in a way to demonstrate by way of literary force that even biting into a juicy burger, if you slow it down enough, can be a depth experience. It is a kind of redemption of the world through revealing the mystical character of its most profane spaces.

Set in turn-of-the-millennium USA, Egan’s Look At Me tells the story of Charlotte Swenson, a model who suffers a disfiguring car accident and is dealing with the aftermath. Much of the book is an extended reflection on the emerging internet culture, as Charlotte becomes progressively drawn in to a startup that offers audiences direct unfiltered windows into the authentic experiences of unusual or exotic people. Looking back at those 2001 premonitions of the proto-influencer from the vantage point of 2026 is extremely interesting and worthy of its own essay.

There are other plotlines I enjoyed as well. One follows Moose Metcalf, the brother of Charlotte’s childhood friend Ellen, an eccentric and somewhat sociopathic professor who, after getting fired from Yale, devotes himself to finding the entire universe inside the history of the trainyards in his suburban Illinois hometown. This part is largely inspired by William Cronon’s Nature’s Metropolis, one of the best works of social science history I’ve ever read, which made Egan’s literary depictions of its ideas all the more enjoyable.

Egan’s depiction of fast-food revolves around another character, Michael West. Michael is one of the stranger and darker figures in Look at Me. By the end of the novel, we have learned that his real name is Aziz and that he has come to America as part of an extremist cell, spending his first months there in Jersey assembling bomb materials. They watch American television, which they interpret as a form of cultural domination.

However, as Aziz spends more time in New York, he becomes somewhat alienated from his cell, coming to believe that blowing up bridges, tunnels, or the White House would not strike at the heart of the American Empire, which instead is founded on celebrity culture. He splits off from the cell, and makes his way into the New York nightlife world, transforming himself into Z, and passing in that scene as a mysterious European investor in nightclubs, who was maybe Greek, maybe something else. That is where he meets Charlotte. We also learn that a private detective is investigating him, Charlotte seems to have some relevant information, but is very reticent to say much about him.

Later, Z reappears as Michael, now living in Rockford, Illinois, which is Charlotte’s hometown, and where Charlotte’s accident happened and where she spends time recovering. Here he is under cover as a schoolteacher, becoming involved in various romantic escapades with suburban women and girls. He eventually leaves Rockford, driving West, and we never learn what his ultimate fate is.

While Egan does not tell us whether Aziz/Z/Michael ends up carrying out any of the terror plots that brought him to America, a persistent theme of his time in suburbia is the steady dissipation of his rage. Over time, his anger becomes more like a kind of nostalgic romanticism for when things were clearer, and he knew who and how to hate.

He tries to hold on to that hating feeling, but it keeps slipping away. While teaching a lesson at school, he thinks to himself a series of cliches about how America works to control the world:

“Capture desire and the rest will follow. Wars, weapons; they were messy, obsolete. Feed people a morsel of something they’ll crave the rest of their lives, and you won’t have to fight them. They’ll hand themselves over. This was the American conspiracy”

But his heart is clearly not it. He is increasingly “desperate to locate the anger that had lived in him like a hot coal for much of his life,” so he keeps a gun on his ankle, in an effort to hold “the place where his anger used to be.”

Not even an after-school meeting with one of his students and “the shape of her breasts through the sweater,” could awaken his desire. He “felt nothing. Deadness. Without the anger, his desire, too, had mostly vanished.” Egan’s writing here becomes rather formulaic, as if to mimic Michael’s feeling that what had seemed so vital to him had decayed into a kind of scripted formula, which he tries, and fails, to reinvigorate.

Aziz’s transformation from New York nightclub mystery man into suburban Michael is slow, but a crucial inflection point comes when he finally allows himself to walk into a McDonald’s. Just a few hours after spending his day at school trying and failing to work himself up into a rage while stroking the gun strapped to his ankle, he returns to a nearby McDonald’s parking lot. He had done this many times before, visiting all 18 of them in the area in fact, sitting outside, wrestling with himself about whether to go in. He had never been able to work up the courage to do so, however, since he believed the effect would be “combustive, violent.”

But now that combustive violence was fading from his grasp, he wanted to put himself in a situation that would provoke it. In a line that could have been written by my City of Toronto colleague, Michael reflects:

Beijing, Moscow—they were all over the world, McDonald’s, colonizing, anesthetizing, and it was said that no country containing one had been at war since. Of course, they were already defeated.

But “Today was the day.” He gets out of his car and walks in. He orders what one must, Big Mac, Coke, and Fries, but is immediately unsure if the “squashed, pelletlike” substance he received was an inferior version of the real thing. He becomes ashamed at the greed and individualism this thought betrayed. “He lifted the thing to his mouth and jammed half inside.”

The description that follows is almost the mirror image of Knausgaard’s depiction of Iselin. Rather than discovering mystical unity-in-difference within a BK extra value meal, Michael is looking for an intense, white-hot experience. He wanted something that could justify and animate his darker and submerged desires.

Instead, however, he found…nothing:

He couldn’t taste anything at first, could only think that it would never go down, he would choke to death on this gray sweetness, dry and sticky; he tried to swallow, his throat straining, seizing to push the clotted mass down its slender duct. Finally the lump evacuated his mouth with a tearing sensation, eased into his throat like a rat moving through a snake. He ate a French fry, breathing hard, sweat on his face, then shoved the second half of Big Mac into his mouth, loosening its airless compression with a slug of Coca-Cola, his body braced for the surge of rage that would galvanize his dead insides when this affront reached them, an explosion that would shove it all back up. But nothing happened.

After he finishes nibbling his fries, he sits for a while, and feels himself incorporating the food. There is a kind of unification with the meal here as well. The Big Mac, Fries, and Coke become “part of him, its cells mingling with his own cells, dividing to make new cells – the cells of a person who had eaten at McDonald’s.” However, there are no mysterious encounters with strange intimations of other worlds – he was the one who could have played that role for others, and Mcdonald’s was pulling him out of it.

Likewise, when Michael walks outside, there are no shimmering sidewalks or glancing rays of sun illuminating the alleyways, as in Knausgaard’s Bergen.

Rockford, Illinois, flat and colorless in winter. He was among runways of concrete and woodchips and highways, for no reason. By sheer accident. He could be here or anywhere.

And yet, starting from where he had, this was a profoundly transformative experience. It was for Michael as if he was seeing the world as it really is for the first time. He had always had a pre-determined script to fall back on, knowing that this was the enemy these were his friends, and all could be interpreted through that dichotomy.

But now, he “felt a first intimation of terror,” much greater than before, “of the land, the crushing gray sky, the bloated strangers everywhere. Of facing his new world alone, without an enemy.” It is much easier to face the world with an enemy, it tells you how to slot everybody and everything you might encounter. Take that away, and you are left alone with yourself and the world and no obvious way to make sense of it all. But you have to make sense of it all.

Michael goes on to become a real connoisseur of American fast food, both in its culinary and its social aspects. But this first encounter is enough to highlight the spiritual center of fast food in Egan’s depiction. There is a kind of mysterious power to McDonald’s. By removing all enemies, by being anything to anybody, it strips away romantic quests to transport oneself, as Baudelaire put it, “anywhere out of this world.” There is nowhere else, this is the only world we have, and it is as if every McDonald’s sign says, “billions and billions served” and “I’m lovin’ it,” but means: “grow up and deal with it.”

I doubt that these two scenes are enough to rehabilitate anybody’s view of the fast food restaurant, but maybe they are enough to inspire more wide-ranging reflection on their meaning from a humanistic perspective. Whatever they are from the point of view of logistics or health, fast-food meals and fast-food dining rooms are one of the most widely shared venues in which the dramas of contemporary life ensue. The range of experiences they host is worthy of further exploration, as are the distinctive forms of encounter they inspire, with others, and most of all, with yourself.

This post introduces and summarizes our new paper, “Grand Theory in the Sociological Canon,” which recently appeared in the European Journal of Social Theory.

Here is the abstract:

Since the 1960s, “grand theory” has increasingly been used to categorize and discuss sociological theories. The result of this is, ironically, a discourse in which it is neither clear which theories count as “grand” nor what such a characterization indicates. By rooting our analysis in a corpus of 3,673 chapters of English-language textbooks, we offer an empirically-grounded contribution to the debate on the past and future of“grand theory”. First, we demonstrate how the composition and scope of “grand theory” have changed between 1967 and 2018. Second, we reconstruct the shifting seman- tic associations around the concept. Third, we qualitatively distinguish four strategic usages of “grand theory” within processes of canonization, which either intend to pre- serve, reduce, limit, or expand the composition of the sociological theory canon. In sum, we argue that the diverse rhetorical mobilization of the concept has institutionalized it as one of the central discursive structuring devices in debates about the sociological theory canon, effectively reifying the idea that the term designates a substantive form of theory while simultaneously obscuring “grand theory’s” contested content.

In 1957, Talcott Parsons published a review of C. Wright Mills’ The Power Elite. In it, Parsons mounted a substantive criticism of Mills’ damning portrait of American society and the theory of power on which it is grounded. Parsons’ fundamental point was that Mills advances a zero-sum conception: power is power over others, so A’s power necessarily comes at B’s expense.

Parsons argues that this treats only the distributive side of power and that power is also a collectively produced social resource, comparable in some ways to wealth. It must be allocated, but it also has to be generated. Like money, if you do not grow it, you will have nothing to distribute. This means that, again contra Mills, power is not ipso facto illegitimate. Power as the capacity to realize collective goals can be abused and must be controlled, but it is also a necessary condition of a large and highly complex society.

This insight launched Parsons on what proved to be one of his most creative and generative theoretical projects, the “generalized symbolic media.” He would go on to develop that idea in a series of articles, such as “On the concept of political power,” “On the concept of influence,” “On the concept of value-commitments,” as well as in The American University, Action Theory and the Human Condition, and other works. While he never fully brought coherence to the proposition that there are a family of “generalized media” that operate similarly to money — influence, power, value-commitments, affect, intelligence, and more — the idea inspired elaborations and fruitful alternatives from the likes of Habermas, James Coleman, Luhmann, Jonathan Turner, and others.

For his part, Mills’ most consequential response to Parsons was not a sustained engagement with Parsons’ theory of power, but a rhetorical demolition of Parsons’ style of theorizing through ridicule and derision. Specifically, in The Sociological Imagination, Mills treats Parsons as one of the two major impediments to realizing the political project of transforming the university into a vehicle for consciousness raising activism (the other is quantitative analysis in the mode of Paul Lazarsfeld, Mills’s Columbia colleague, who had helped bring him to Columbia before becoming the main target of Mills’ attack on “abstracted empiricism”).

In Mills’ portrayal, Parsons’ approach to theory construction was not just wrong, and not even just pointless. It was, rather, a kind of seduction to spend time spinning out theoretical schemes that distract from the more pressing project of challenging and undermining the power elite whose decisions determine our fates from smoke filled back rooms.

Mills does gesture toward a few substantial criticisms.

The magical elimination of conflict, and the wondrous achievement of harmony, … remove from this ‘systematic’ and ‘general’ theory the possibility of dealing with social change

As Helmut Staubmann observes, however:“Given Parsons’ extensive writings on social and cultural change, the persistent narrative of his lack and incapability to deal with the issue is stunning.” Staubmann in the linked article offers a concise explanation of Parsons’ approach to change, and there are many other good resources for his theory of power and conflict (Uta Gehrhardt is very helpful on this score).

What really stuck, though, is not the content of what Mills wrote, but the form. He took a somewhat technical and indeed convoluted passage from The Social System, and then translated it into “plain” English. The point was to rip off the emperor’s clothes, and not so that students would meaningfully and generously engage with Parsons’ ideas to identify their potential tensions and improve the theoretical scheme. Rather, the goal was to make Parsons an object of ridicule so that nobody would ever feel the need to open one of his books again, for fear of raising the hackles of the Millsian chorus.

This entire rhetorical effort was contained in the phrase “Grand Theory.” This was the term Mills used to characterize Parsons, and with it, to steer the entire field away from the effort to build up a systematic conception of society. From then on, if a sociologist began to feel pangs of conscience for abdicating the effort to pursue that ambitious project, burrowing down into increasingly narrow propositions or increasingly confident normative proclamations, there was an easy way to arrest the night thoughts and to sleep well for a few hours: at least I’m not doing Grand Theory.

The above account is obviously a bit polemical and is certainly debatable. But what is not debatable is that the notion of “grand theory” as it has been used in sociology and then beyond emerged from this highly polemical context.

The legacy of the concept is the topic of a recently published paper, “Grand Theory in the Sociological Canon. Scopes, Semantics, and Strategic Usages of a Rhetorical Device Since the 1960s,” by Lars Döpking, Sebastien Parker, Cinthya Guzman, and myself. The paper appeared in the European Journal of Social Theory, and is part of a special issue on the topic of grand theory. This is now the fourth paper examining the practice of sociological theory from a number of perspectives: its institutionalization in the discipline (by way of a quantitative study of syllabi); practical justifications of the theoretical “canon” (by way of a qualitative study of theory textbooks); and a cross-national study of its evolving discourse and narratives (a computational analysis of German, English, and French textbooks).

This paper returns to our dataset of textbooks, focusing on the English language ones. The corpus includes nearly 4000 chapters from around 250 distinct editions representing nearly 150 unique English-language sociological theory textbook titles, published between 1953 and 2018.

The first part of our analysis tracks the temporal evolution of grand theory discourse after Mills’ polemical introduction. The major change is that what began as a polemical weapon has evolved into a standard organizational category, obscuring its origins, twisting around its meaning, devouring its creator in the process — and finally calling into question not only a specific type of social theory, but the very project of ambitious theoretical work.

What follows is a loose summary of the paper, paraphrasing and quoting from bits and pieces. However, you should read the whole thing!

In our corpus, “grand theory” was first used by the textbook of Demerath and Peterson, who reproduced Mills’ critique and presented Talcott Parsons’ functionalism as the sole representative of “grand theory.” Other textbooks followed this restrictive application of the concept, but soon authors began to characterize additional theorists as “grand.” At the end of the 1970s, Margaret M. Poloma identified a “grand theoretical tradition” which, in addition to Parsons, included the protagonists of “classical sociology” he had founded his approach upon: Émile Durkheim, Max Weber, Auguste Comte, and Vilfredo Pareto. By the 1980s, “grand theory” included Neo-marxists (and Mills cried…), Peter Berger, Thomas Luckmann, Anthony Giddens, Habermas, Norbert Elias, and by the 2010s, Karin Martin wrote: “many feminists in the academy were writing ‘grand theory.’”

While this trajectory can look like a steady expansion in which the net of the original epithet is cast over a wider and wider set of targets, in fact the conception of the object changes along the way. To examine how the composition of grand theory has evolved since Mills’ original critique, we identified the theorists most strongly associated with grand theory discourse by adjusting for their baseline presence across all sociology textbooks.

This figure highlights a geographic and intellectual reorientation in who counts as a grand theorist. The American mid-century figures who dominated the concept’s original context show a systematic retreat from grand theory discourse. Talcott Parsons declines by 21 percentage points between periods, while Mills himself drops by 25 points, suggesting that the original polemic has lost its centrality to how textbooks frame the concept.

This pattern extends throughout the structural-functionalist tradition: Robert K. Merton, Alvin Gouldner, and other American systematizers similarly fade from grand theory’s definitional core. In their place, European continental theorists have redefined “grand theory’s” conceptual core. Anthony Giddens shows the largest increase, appearing in nearly half of all grand theory discussions during the late period. Additionally, Pierre Bourdieu rises by 13 points, while Michel Foucault shows modest gains.

This shift represents more than a substitution of one set of names for another. While Mills attacked abstract system-building in American sociology, contemporary textbooks increasingly apply grand theory to European projects that combine theoretical ambition with empirical engagement in ways Mills’ original critique did not address.

We document this shift with some additional analyses that specifically highlight changing linguistic associations of the term (for example, it becomes less associated with methodological and polemical terms), but the bottom line is that the concept has evolved from identifying specific targets (as in Mills’ critique) to picking out a type of theorizing characterized by scope and ambition rather than methodological failings.

These kinds of shifts tell us that something is changing, but it takes qualitative analysis to see more clearly into what is happening. For this, we identify four distinct strategies that textbook authors employ to discuss and evaluate sociological theories by classifying them as “grand.”

1. A theory can be presented as such in order to depict it as outdated and opaque, thereby relegating it to a peripheral position within the canon.

2. Textbook authors use “grand theory” to stabilize the established structures of the prevailing canon. In that sense, they regard a theory as deserving of commemoration and instruction due to its belonging to a distinctive intel- lectual tradition that is held in high worth.

3. A new or lesser known theorists may be designated as “grand” in order to legitimize their incorporation into the canon. By doing so, the concept serves as a means of canon expansion.

4. Theories that have not yet been incorporated into the sociological canon, or only marginally so, can also be characterized as “grand” to prevent them from entering the canon in the first place. Sociological theory is then often portrayed as the antithesis of philosophical or psychological grand theories, which do not belong to its canon.

We give lots of examples of these rhetorical strategies, but the overall impression one gets reading these varied uses is that as the concept receives more substantial treatment, the practices of its use grow increasingly arbitrary.

It is hard to avoid linking transformations in the concept of grand theory to the shifting fortunes of theoretical practice in the field. In sociology today, “theory” has become a subject taken once during one’s studies rather than than a field in which careers are advanced. This context makes the shifts in the scope and semantics of “grand theory” and its acceptance as a genuine variety of theory highly functional. It ensures the memory of certain theorists and thus the cohesion of the discipline, but by no means calls for more in-depth engagement: grand theory was something that happened back then, over there, but not now, by you.

“Once defined as a type, as a distinct form of theory, and placed alongside others in a relativizing manner, the promise of grand and thus ambitious theory fades. An expectation of truth and deeper insight then emanates neither from it nor from other types of theory.”

One of Robert Pippin’s favorite Hegel quotes in his recent writings about art features a metaphor. In it, humans are like amphibians, standing between what in non Hegel-ese would probably be called nature and culture. I don’t have the Hegel quote in front of me to verify what the Hegel terms are, but if they mean anything, it is that.1

The metaphor is to indicate that the practice of making and appreciating art cannot be divorced from the sensuous, affective, material dimensions of our experience. The Idea must be embodied, and that is what art does: the sensuous actualization of the idea. The tricky thing is that this is not supposed to mean that the idea pre-exists the actualization in sensuous material form. Rather, it is in the practice of making and appreciating the actual material works of art, responding to them, experimenting with them, seeking to say something in them, that the possibilities of expressing and reflecting on who we are are developed.

The amphibian metaphor is supposed to indicate that we can’t escape an existence in both domains, and it is our lot to move between them.

But is this metaphor really that telling? What if we take it very seriously? An amphibian like a salamander moves between two regions on the earth’s surface. While there are many differences between land and sea, the key one is that the organs that allow an animal to extract oxygen from water won’t work for air. So prior to the evolution of amphibians, they were essentially two mutually exclusive domains (other than some plants?). You either lived in water or lived in land.

Amphibians evolved through natural selection producing hybrid lung-gills (look it up!) that enabled them to move freely between the two worlds. This created a new set of evolutionary pathways, which in turn enabled the evolution of lungs and land-based animals who, eventually, lost the ability to survive in the oceans.

Is this a good metaphor for the relationship between nature and culture? Or, before coming to a judgment, what image of the relationship between nature and culture does it depict?

First, it seems natural to equate nature with the ocean and culture with the land, in that culture/land is the thing toward which the evolutionary process is moving, the end or meaning it realizes.

Second, the land and ocean are two regions on the Earth’s surface, with fairly distinct (if at times fuzzy) boundaries, so that in the metaphor moving from nature into culture is like exiting one region and then entering into a different region (exiting nature and entering culture is like emerging from the sea).

Third, amphibians are transitional organisms in a phylogenetic tree. Looking back, we now know that they evolve into land-based reptiles and mammals. The metaphor implies a kind of evolutionary dynamic whereby it is at least possible to imagine leaving the sea behind entirely, to exist completely in the realm of culture.

Fourth, amphibians are able to move back and forth between land and sea because of certain evolved capacities that other animals lack. The metaphor implies that humans have some capacities that enable them to move between nature and culture, and that whatever those are could evolve into something else.

There are some aspects of this imagery I find kind of weird. The most striking is the idea of nature and culture as regions in space that you can move back and forth between or exist at the borders of, and that, because of this image, you could theoretically step completely outside of nature. That seems very Hegelian in the bad old sense: a picture that tends toward the purely ideational, with a strong teleological thrust away from the sensuous. Just saying “but we’ll never get there” doesn’t take that aspect away from the structure of the imagery.

Why is that weird? Consider an alternative, broadly pragmatist (Deweyan) version: culture is not a region but a practical capacity. It develops potentials that are there already in the lowest life forms, e.g. the capacity to respond to the environment, move towards good things and away from bad things, store and communicate and interpret information, anticipate responses from others and eventually oneself.

Humanity develops all of this beyond what had been hitherto possible, realizing existing capacities and creating platforms for new ones. This all happens by manipulating and experimenting upon material objects – of course it does, what else could it do? Through those experiments, we learn new ways to express ourselves, actualize ourselves, and circulate those conceptions among others, and in doing so they take on additional potentials and meanings we might not have initially imagined.

None of this involves rising up out of the sea and entering into a different region, or any kind of teleological direction toward that more heavenly or better domain of pure culture. Nor does it involve any sense of loss of that either, or a corresondong loss of access to the lower or past or rudimentary region. There are no regions! There aren’t two spaces, but richer ways of living in the exact same one.

The metaphor of two regions, with an amphibian moving between, in the end retains a kind of Platonism, of the bad kind, and just reverses it into a kind of disappointed variation – we are stuck as amphibians.

The “solution” is not to make us fish or fowl, but to reject the entire image.

Each Fall, as a new academic year approaches, another crop of students will file into Introduction to Sociology courses -- though if recent trends hold, there will be fewer of them than last year, and fewer still than the year before that. Even so, it is very likely that in this course they will encounter C. Wright Mills’ The Sociological Imagination, and from this book, read or hear about its famous first chapter, “The Promise.”

The main promise of “The Promise” is that sociology can be a beacon of clarity in an increasingly confusing world. Mills begins with his famous observation-cum-injunction: that people today often feel trapped by their private lives, unable to grasp the larger historical forces that shape their experiences. The core of his "promise" is the Sociological Imagination: a quality of mind that enables us to grasp the connection between our private troubles (one person's unemployment, a couple's marital strife) and broader public issues (a nation's high unemployment rate, legal restrictions on divorce). The Sociological Imagination is the ability to link our individual biography to the grander sweep of history and social structure.

In Mills’s conception, this quality of mind works in two key ways. First, it serves as a powerful diagnostic tool. By connecting troubles to issues, it reveals the true sources of our problems. We might think our personal struggles are our own fault when they are in fact symptoms of a larger societal arrangement. Betty Friedan’s famous diagnosis of “the problem that has no name” is a good illustration. In the Feminine Mystique, she argued that a deep, nameless dissatisfaction plagued women who were often told by experts that this was a personal problem requiring them to "adjust." Friedan argued this was not a private, psychiatric issue but a public one, rooted in a social structure that required educated women to be confined to the role of housewife.

This diagnosis then in Mills’ conception points toward a cure. Once people understand the structural roots of their private troubles, their vague uneasiness can be transformed into focused public engagement. Friedan's work gave a name and a cause to millions of women's private suffering, helping to spark the second-wave feminist movement. This illustrates the ultimate promise of the sociological imagination: to provide a blueprint for social change, empowering people to move from private suffering to taking responsibility for history.

Mills contends that sociology is uniquely suited for this task. The physical sciences, like the physics that produced the atomic bomb, are too specialized in his view to grasp the social consequences of their own creations. He thinks the arts can powerfully depict our anxieties but lack the systematic empirical analysis needed to address them. Only the social sciences, in his view, can truly connect personal lives to public structures and offer a path forward. It is a deeply gratifying view of the field, and it has served for decades as a potent, rhetorical draw for new students.

The ironies of history

Judged by its own polemical goals, the book’s impact represents one of those “ironies of history” that Hegel was fond of pointing out. In the battle for the soul of the discipline, Mills was largely successful. His terrifically fun chapter ridiculing Talcott Parsons is likely the single text most responsible for ending Parsons’ reign as King of the Sociological Hill. And while statistical work in the tradition of Paul Lazarsfeld (the target of Mills’ scorn in the chapter on Abstracted Empiricism) certainly continues, it is largely segmented within elite journals and top departments, rather than defining the field's public face and broader discipline-wide practice.

The result is that the bulk of the discipline is now oriented toward the very consciousness-raising and activistic efforts Mills championed. A quick look at the themes chosen by democratically elected presidents of the American Sociological Association reveals his triumph. Recent conferences have been organized around topics like: “Intersectional Solidarities: Building Communities of Hope, Justice, and Joy”; “The Educative Power of Sociology”; “ Power, Inequality and Resistance at Work”; “Engaging Social Justice for a Better World”; and “Emancipatory Sociology.” Mills is somewhere in the sky riding around on his motorcycle, smiling.

Even so, as Mills’s vision conquered the discipline from within, sociology’s influence began to wane without. The number of bachelor's degrees awarded in the field peaked around 1970 and has been in marked decline ever since, with some ups and downs. This decline is all the more striking in that the overall college-going population grew during this same period. Adding another layer of irony is the fact that the discipline’s public influence was probably greatest when it was most defined by the very figures, like Parsons and Lazarsfeld and Merton, that Mills sought to dethrone. In terms of his goal to make sociology the prime window into contemporary problems, The Sociological Imagination must be judged a failure.

Yet, as Bob Dylan said, “there's no success like failure, and failure's no success at all.” At another level, Mills’s project couldn’t have been more successful. The sociological imagination as a practice has spilled out beyond the formal boundaries of academic sociology, acquiring an autonomous life as an all-purpose mode of cultural analysis. Today, much of our public discourse about social issues amounts to thinking sociologically, which is to say, by imagining linkages between personal problems and broader societal forces

Sociological Over-Imagination

To see the sociological imagination in hyper-drive, consider a recent, much-discussed NYT essay on "heteropessimism."1 I have no wish to pile on; my only interest is how clearly its style of thought operates in the mode Mills promised. The author, reflecting on her dating life, considers a range of academic concepts:

“We now have a fancy word, ‘heteropessimism,’ to describe the outlook of straight women fed up with the mating behavior of men... The writer and gender scholar Sara Ahmed has advanced the idea of ‘complaint as feminist pedagogy’ arguing that to bitch is inherently transgressive, a form of resistance... Was that what my friends and I were doing over dinner? Rebelling?”

She concludes with a telling abdication of personal responsibility to expert diagnosis:

“If the experts say my romantic letdowns have some larger social significance, I am not going to argue... All this certainly seems calamitous enough to warrant an ‘ism.’ And if it is an ‘ism,’ the problem cannot be me.”

The key takeaway here is how smoothly this mode of thinking elevates personal romantic foibles into an issue of grand social significance. This, in fact, is the unrestrained sociological imagination’s forte: it can perform this operation on almost any private trouble you hand it. But sometimes your personal troubles are just your personal troubles, and that’s that.

Imagination gone wild

What makes this practice so susceptible to unrestrained creativity is not so much that it is sociological but that it is imaginative. One of the oldest worries in aesthetics is that the imagination is both essential and dangerous. It is essential for its power to create new connections and ideas; it is dangerous for the very same reason. This is why the philosophical tradition has always insisted that imagination requires a partner: judgment, the faculty that determines which of the mind’s many creations are worth keeping. A sociological imagination without a corresponding sociological judgment inevitably degrades into sociological over-imagination.

While Mills discusses rendering judgments, he never develops a concept of judgment as a faculty in its own right, one that must be cultivated alongside and integrated with the imagination. It is no accident that John Levi Martin in his recent and magisterial book, The True, the Beautiful, and the Good, identifies a lack of attention to judgment at the expense of other faculties as at the heart of host of problems in modern sociological thought. In Kant, for instance, aesthetic judgment arises from a “free play” between the imagination’s creative power and the understanding’s drive for order. Judgment registers the harmony between them, ensuring that what the imagination produces has coherence and form. Without that partnership, imagination becomes unrestrained. It generates endless connections -- “heteropessimism is an ‘ism’!” -- but lacks the discipline to ask if those connections are valid.

The Intellectual as Mills’ Last Hope

This over-reliance on imagination stemmed directly from Mills’s larger political project and his increasingly desperate search for an agent of historical change. The Sociological Imagination was written after his previous attempts to find an American equivalent to a Marxian vanguard class had failed. He looked to the unionized working class but, after studying their leadership, found only a new bureaucracy. He looked to the new professional white-collar classes, but found only timid “company men.”

The Power Elite can be read as an expression of this despair. In it, he described a society where all major decisions were made by a small, entrenched, and interlocking directorate of corporate, military, and political leaders. For the masses below, the link between their private lives (losing a job, being sent to war) and the great structural forces was severed. The result was apathy for the many, smug security for the middle, and near-total control for the few at the top. This was Mills’s portrait of modernity.

The Sociological Imagination was his answer. If workers and managers had failed, it was intellectuals, specifically social scientists, who were in the best position to awaken the public. Universities, not the labor union or office, would be the institutional home of progressive social change. They alone could restore America from a “mass society” of alienated and apathetic “cheerful robots” (to use his famous phrase) into a public of free and rational persons.

In his final chapter, Mills rejects the idea of intellectuals as philosopher-kings or advisors to the kings. Instead, he champions the public intellectual whose role is to speak directly to the people, arming them with the sociological imagination so they can finally see the world as it is. In this vision, consciousness-raising becomes the university’s central goal, and the public intellectual takes on world-historical significance as the last great hope for providing a seedbed for transformative social action. This was the true promise of The Sociological Imagination.

Lab Leaks

This vision of the public intellectual, vested with world-historical significance, was a last-ditch effort for Mills, who would die only a few years after publishing the book. If he had the benefit of hindsight, one wonders if he might look back on the intellectuals his vision spawned with the same disappointment he felt for the union leaders and company men he had previously considered. Not because his intellectual heirs have failed to use their imagination, but because they use it all too well. The imagination alone cannot bear the weight of the world, especially once it leaves the spaces designed for it to roam free.

Consider an example more sobering than "heteropessimism." I’m thinking of Michael Ignatieff, who wrote a very good biography of Isaiah Berlin and is also a public intellectual. What you (Americans) might not also know is that he was the leader of the Liberal Party of Canada and ran to be Prime Minister of Canada. In the latter role, he failed spectacularly, luxuriantly, with gusto. It was a multidimensional failure to be sure, but I remember noting something was a bit fishy when I read about his attempt to explain his intellectual support for the Iraq War. I recall an article he wrote around the time of his candidacy. He was trying to convince his readers that he was not just an ivory tower professor (unlike those of us more comfortable in the ivory basement) but had come to appreciate the practical reality of politics. He learned this lesson, he said, from the failure of the Iraq War.

His reasoning was rather interesting. Ignatieff suggested that in the academy, arguments are often valued for being interesting, for yielding some surprising new angle or an implication no one had previously imagined. In this case, he confessed, he was simply struck by how interesting the idea of creating democracy at gunpoint was, and figured, like in an academic essay or journal, it was worth giving it a try. Thankfully, he had learned his lesson and wouldn’t do that again.

The philosopher Isaiah Berlin once said that the trouble with academics and commentators is that they care more about whether ideas are interesting than whether they are true… In academic life, false ideas are merely false and useless ones can be fun to play with.

Ignatieff is a very different figure than Mills, but as a public intellectual, his experience highlights the tendency of the academic imagination to tip into over-imagination. Inside the university, it is not only appropriate but necessary to entertain dangerous or unusual or weird or even absurd ideas, sometimes.

This is the truth of the “safe space” concept. College campuses should be spaces to try on radical thoughts and pursue arguments to their logical, and sometimes nonsensical, conclusions. Students especially should be given wide latitude to take on odd or extreme positions, to let their imaginations run wild.

But this vital function depends on the implicit understanding that these ideas are not immediate marching orders. They must pass through additional filters that test their coherence and practical worth. That happens not only within its walls but without and in their interface, as people try to incorporate ideas into their lives, in piecemeal and incremental ways. For its part, however, to be an effective lab, the ivory tower needs its walls and its height; it must be distant and aloof to properly cultivate the unfettered imagination. The only way to maintain this crucial space is to install guardrails designed to prevent intellectual “lab leaks.”

And for this, the sociological imagination is, at best, only half the story.

Platforms are staged competitions that license strategic, self-interested pursuit of influence, but only within constraints justified by the platform’s purpose. Blameworthy “cheating” is clout-seeking which undermines the environment that makes the tournament worth having.

The ethics of morally lax environments

The other day I went to a local amateur MMA event. It was an incredible display of athletic talent and dedication. The competitors had clearly been training hard, and the chance to perform before a crowd pushed them to a level that would be difficult to reach alone or just by sparring in their home gyms. There was no monetary prize, but winners and star performers would gain recognition and renown, and possibly a shot at going pro.

Throughout, the aim of each fighter was to advance his own position at the expense of the other. Put less politely, they were trying to defeat one another through forms of aggression and violence that would be very unacceptable in ordinary life. That fact struck no one as morally puzzling, because it was the background condition of the entire event. The cage is an arena where everyday moral expectations are deliberately suspended, within a shared understanding of its purpose and limits.

Even so, there was still room for conflict over what counted as fair play. During an especially intense scramble, one fighter lifted the other, who was momentarily inverted and vulnerable. The fighter in the air locked onto an armbar, and the other reacted by driving him down — hard, on his neck — to escape the hold. It looked really bad, and the guy was briefly unconscious and barely escaped serious injury.

The entire setting immediately erupted into an intense and loud debate about the ethics of the maneuver, which nearly broke out into a massive melee involving dozens of trained fighters — not to mention the injured guy’s mom, who let her opinion on the matter be known to anybody who would listen, as well as everybody who would not. Never once, however, did the question arise as to whether violence belongs in sport. Everyone present had already agreed to that: they came to see, or fight in, an MMA match, after all.

Rather, what ensued was a dispute about the internal ethics of the game: what do the rules of the sport actually permit, and what should they limit? Large men from both camps expressed their views to one another in no uncertain terms and at close range; and equally large security guys intervened. Soon enough, a disqualification was issued, and while they didn’t like it, the disqualified guy’s team accepted it.

This story highlights a more general distinction. It is one thing to justify the existence of a competitive arena, it is another to determine what counts as fair competition within that arena. This distinction holds on social media platforms as much as it does in MMA.

Let us recall some of the ethical peculiarities of the clout tournament that platforms stage. In everyday life, when you want to elicit more effort from someone who is loathe to give it (as most of us are most of the time), you often try to justify the request in moral terms. You explain why it would be only fair for them to do their part, or that they have a duty to their fellow humans, or that what goes around comes around, or that, while it might be difficult, it is for a good cause. Even when people do have their own ambition and motivation, it is still considered slightly distasteful to appeal to their interests in raising their stature. You would not usually say to a writer friend: “you should write this essay because it will raise your status,” or “pick another topic, this one is dragging your status down.”

While there is a case to be made that such base concerns are never that far beneath the surface, it is a notable feature of everyday morality that they tend to be considered uncouth and should be suppressed — and often for good reason, as much of small group life depends on some degree of suppression of the will to invidious distinction. Platforms, however, reverse this everyday morality. Here, it is perfectly ordinary to talk that way, and not only because some people are personally uncouth.

The system itself is built, top to bottom, to make self-interested calculation about stature and influence central to your decision-making. It gives you feedback constantly indicating that if you make more of this kind of post, you will gain clout, or if you keep making that kind of post you will lose it. For example, the other day, after I posted a description of our new paper on invasive species discourse and the sociology of social problems, two people unsubscribed from me, making me wonder why this unobjectionable post caused that reaction and whether I should change my behaviour somehow. This type of reflection about the effect of your posts on your status and influence is entirely unexceptionable on a platform, because it is built into its architecture and is part of the engine by which it realizes its purpose.

Much is permitted on platforms that would be ethically dubious elsewhere

But the fact that the arena licenses strategic pursuit of advantage does not mean it licenses anything whatsoever, any more than it does in MMA. The key is to separate permissions from constraints. The permissions granted by clout tournaments come in three broad families of competitive strategies.

First, there are supply and demand decisions: the permission to enter and exit niches and to choose what you produce and consume on explicitly strategic grounds around gains or losses in influence. For example, you are permitted to pivot opportunistically. One month you are a politics account, the next you are doing explainer threads, and a month after that you are a lifestyle account with a brand story about personal reinvention.

On the consumption side, you are permitted to withdraw influence from other accounts. You can unfollow (or fail to like or share or slow down your feed for) an account for whatever reason, even if it they are simply posting too often and clogging your feed, and you want to leave room for something else. It is your feed and you can fill it however you want. These and similar practices are permitted because they advance the purpose of the platform, by allowing posts and influence to adjust and flow to where interest lies, increasing the quantity and diversity of the social bounty platforms offer.

Second, there is the permission to set influence prices. Platforms have the deepest pricing power, because their allocation systems determine what a post is worth. Feedspace is valuable, and the system allocates it with relatively little sympathy. If you are not performing on the “for you” feed, you will be pushed aside. That is exactly what makes it capable of reallocating influence to where interest is moving. The coldness is a feature, not a bug.

But most social capitalist actors also have some discretion over the terms on which they offer and demand influence for posts. On the consumption side, a user can withhold engagement from posts they do not want to grant influence to. They can watch without liking or read without sharing, block, report, or just ignore — all of which is a way of raising the price of getting their influence. On the production side, creators can lower the price of access by making content shorter and easier to digest, and they can raise it by making it harder to access or harder to consume. You can do this through such formal devices as paywalls and gated channels, or informally through using insider references and jargon that make it hard for normies to meaningfully engage. Posts for which positive engagement would indicate shocking or morally counter-intuitive beliefs are also a way to raise the price of consuming them. Setting influence prices is permitted since this helps to discover the value of posts.

Third, there is the permission to shift the production frontier or the post supply curve. The aggregate supply of posts in any niche is the sum of what many different producers can make at different costs. Some people can produce a decent video in an hour; others need a weekend; others take a month for each Substack essay. This variation means there is surplus to be captured if you can shift your curve. If you can lower your own cost of production, or raise your output quality at the same cost, you can take a larger share of the influence that is on offer at any given moment. This family of permissions includes many familiar strategies, such as better cameras, lighting, microphones, and editing software; tighter scripting; a better sense of pacing and headline writing, or maybe a helpful AI assistant or two. More generally, cost reduction through process improvements or technological innovation let you produce more, better, or faster than your competitors, which not only allows you to keep the clout surplus that results, but increases the total supply of potentially interesting posts available to others.

The common thread across these families of permissions is that they license participants in the tournament to pursue competitive strategies without having to be motivated by the collective goods those strategies may or may not produce. While the justification is the collective goods, the motivation can be purely instrumental or self interested. In fact, when they are, this is a sign that the tournament is working; it is harnessing certain perhaps uncouth features of human nature to greater collective benefit, in ways that relying on our better angels would not achieve. This is the nature of a system of indirect control over behaviour, disembodied from direct community oversight.

Legitimate constraints prevent platform failure

The relevant ethical question is therefore not what ends actors should have in mind, but what constraints should be placed on the self-interested strategies they are allowed to use. In the MMA cage, even though you can hit people in the face, you cannot bite or eye gouge (any more, at least), and in general, moves tend to be restricted that would tend to cause debilitating injuries such that competitors would be less likely to participate and to explore the possibility space of the sport.

Some of the constraint comes from law, just as in commercial markets. But law is blunt, slow, and unable to monitor or directly allocate millions of interactions that occur on a day to day basis. So the system of necessity relies heavily on private governance as well. Either way, the practical question is the same. If we want an internal ethic of social capitalism, we need a way to classify the kinds of failure modes that a given constraint is meant to prevent, since such prevention is their ethical basis.

These can be organized around four main sources of platform failure more generally:

1. Anti-competitive strategies. In MMA, you can train hard so you can be better faster stronger than your opponent, but you cannot mount a months-long secret campaign to slip drugs in their protein shakes so they come out sluggish in the fight. On a platform, it is one thing to make better content and package it well for your audience: that is permitted. It is another to adopt strategies that block rivals from competing on the terms the tournament publicly claims to offer. For example, you cannot collude to spread false rumors about niche rivals, or coordinate mass reporting campaigns designed to get them banned. The platform is perhaps the actor most subject to these ethical constraints, due to the significant influence-price setting power they possess. In practice, this amounts to exercising restraint on playing favorites and picking winners, whether on ideological grounds (suppressing or boosting the content of political friends or enemies) or by tilting distribution secretly toward advertiser-friendly content.

   The pursuit of influence is in fact not intrinsically valuable, nor is anybody entitled to hold on to their influence in perpetuity. In the end, influence metrics, like prices for the production of goods and services, are meant to be a means to discover where to direct the production of content, and whatever undermines this discovery process may be limited.

2. Negative externalities. In the context of social capitalism, negative externalities are ways of earning influence by shifting costs onto others without having to pay for them in lost influence. The platform arena is full of these, such as cancellation campaigns and dog piles. These can generate influence for the participants while imposing serious costs on the targets. If those costs to others are not priced in to the activity of the perpetrators, they will be overproduced. You will get more outrage and more slop the less their costs to others are incorporated into their production and circulation. Too much outrage raises the risks of participation, and too much slop degrades the value of consumption.

   In general, then, social capitalist actors have no obligation to be pure at heart, but they do have an obligation not to gain status by dumping unpriced harms on others. The potential damage of negative externalities is not confined to the target, which is why they pose threat to the value-proposition of the tournament as such. Where serious harms are systematically unpriced, platform operators and their regulators have reason to force them to make the calculus, by attaching costs to the strategies that generate them.

3. Information asymmetry. Actors routinely have information advantages that they can exploit in ways that undermine confidence in the signals the tournament runs on. Platforms know vastly more than users about their own post and influence allocation processes, while users often know much more about their interests than they show to others. Creators know more than followers about whatever sponsorships they have taken, not to mention the off-stage engineering that goes in to trying to boost the performance of their posts. Content farms can exploit the ambition of small creators to make it big. While nobody likes getting duped, the bigger issue is the broader uncertainty and lack of confidence these asymmetries can create. The implicit morality of the platform is that participants have an obligation not to manufacture or exploit information asymmetries in ways that make that information unreliable.

4. Contracting costs and opportunism. A large amount of platform life runs on informal deals and implied commitments. Collaborations are often arranged in DMs or offline, and shout-outs and endorsements are traded as favors. None of these can be fully specified or enforced, which creates room for opportunism. If you spend even a little time on creator forums, you will find numerous complaints about such opportunism. It is a regular occurrence for example to hear complaints about being ghosted by a collaborator once they feel they become too popular to care about you. Complaints about stealing ideas or content without attribution are commonplace, not to mention accusations that somebody failed to deliver a promised shout out. These are all ways of violating informal community norms about how to give and receive influence.

   If you had to spell out all of these in detail, it would siphon off a lot of time and effort from the core functions of the platform. For example, sometimes I comment on somebody else’s Substack about an idea I’m thinking of writing about. If every time I did that I had to worry that they’d steal my idea, I’d either waste a lot of time trying to secure assurances, or I wouldn’t comment at all. Both of those outcomes make the platform less valuable to everybody. This observation clarifies the implicit constraint on which platforms depend: do not treat the inevitable gaps in enforceability as an opportunity for influence profiteering.

Referees are in the interest of the players

Exactly who should solve a platform failure, and by what means, is often an open question. It will often be a matter of judgment. It is worth stressing that creative ways of reducing these failures are often in the interest of social capitalist actors themselves. In other words, in many cases, the internal operations of the platform generate incentives to remedy the failure, without any higher moralism or virtue being required.

Kenneth Arrow’s famous example was the creation of professional associations and colleges among doctors, which reassured patients that they would not be exploited in a setting of unavoidable information asymmetry. Similarly, for all of the constraints listed above, you can find various ways that one platform or another is experimenting with various ways of imposing them.1 It is costly to them to have a degraded information environment, the perils of enshittification are no secret — everybody has seen what happened to Quora, after all — and platforms as well as creators have an interest in avoiding them.2

However, internal correction is not the only way to solve a platform failure, and many “solutions” have their own pathologies. For example, professional associations can reassure clients, but they can also mask the fact that they have become cartels. This example highlights a more general feature of platform ethics. Platforms, and the organized creator-commercial layer around them (MCNs, agencies, large accounts), are typically far more coordinated than ordinary users, and better able to turn their interests into policy. This means the state is often asked to represent diffuse users against concentrated platform power. From the platform’s point of view, that can make regulation feel like hostile action rather than rule-setting.

This is understandable, but it is not ethically permissible, any more than an MMA fighter is permitted to view the ref in the same way as they view their opponent. Of course, it is tempting to see the ref as out to get you when you lose, but in general, the ref is not your adversary. The ref can be bad and biased, sometimes captured, but the ethical posture of the ref is still fundamentally different from that of the participant. If there is a bad ref, you can contest calls, or try to improve the practice of referring as such; you don’t try to disable officiating as such. This means that a platform should not seek to create barriers to competition or to thwart the state’s efforts to solve a platform failure that the platforms do not remedy themselves.

In many cases, competition will make this hard. Take for example proposed rules that would make it easier for users and creators to leave one platform for another. A platform will almost always fight those. It will say, with some plausibility, that if it accepts such constraints unilaterally, it hands a gift to its rivals. This may explain the behavior, but it does not justify it. If a platform wants to claim that “competition made me do it,” it has to do something more than complain. It has to be willing to support a rule that binds everyone, so that basic restraints do not become an option only for the most naïve actor in the market.

In rare, very rare, cases, platforms may even appeal to the state to impose exactly that kind of across-the-board constraint, in the same spirit in which sports franchise owners sometimes want a commissioner to enforce limits they cannot enforce on themselves. A good example is the NBA’s rule that constrains teams from trading away first-round picks in consecutive future years. (the Stepien rule.) In isolation, a team might be tempted to mortgage its future for short-term gain. But if every team is under pressure to keep up, the temptation becomes structural, and the result is a race to the bottom. The point of the rule is to prevent competitive pressure from pushing everyone into a pattern that is individually rational but corrosive to the tournament as a whole.

There are similar examples in the realm of social capitalism. For example, Meta has said that “legislation should require app stores to verify age” as well as an EU-wide “digital majority” age. Meta, X, and Snap also all praised Utah’s 2025 App Store Accountability Act, which required app stores to verify users’ ages and obtain parental consent for minors’ download. Nobody should confuse these with selfless concern. Platforms tend to support regulation when it creates a single standard to which they and all competitors are bound, not to mention when it shifts the burden of compliance off of themselves and onto the companies who operate app stores. And they don’t mind if the compliance costs thereby generated are easier for large incumbents to absorb better than smaller emergent competitors.

The point however is that such strategic concern with outside standards are structurally analogous to the Stepien-rule problem. What looks like a moral failure is also a coordination problem, and, ideally, some remedy would arise from the interests of the parties themselves.

Be that as it may, the conclusion of the present essay is that constraints on what is permitted in clout tournaments staged by and on platforms are not gratuitous impositions on some fundamental natural liberties. They are constraints designed to prevent races to the bottom that threaten to undermine the stability and productivity of the system itself.

You might remember various moments when discussions about the language of ecology and conservation biology have bubbled to the surface of public discourse. Is labeling a species “invasive” a form of xenophobia that perpetuates white supremacy and colonial thinking? Is it a purely biological term that merely designates a category of non-native species that is causing harm to existing ecosystems?

Wherever you find yourself on such questions, it remains the case that describing a species as “invasive” carries great weight. The more a species becomes considered invasive, the more likely it is to become the target of ecosystem management policies designed to control it within, or even eradicate it from, a given region. Such management policies are costly, not only in time and money, but also in that they often call upon local residents to change their practices (i.e. how they tend their gardens).

There are always more potential problems floating around than anybody tasked with dealing with them can handle, and many stakeholders who have some interest in one or the other candidate being considered as more or less problematic. When and why do some candidate species become more likely to be described as exhibiting various degrees of invasiveness? This is the question animating a new article published in Social Forces by myself, Tyler Bateman, Alica Eads, and Charlotte Kafka-Gibbons. This article really is Tyler’s baby, and he led the way from start to finish.

To tackle this question, we study documents from the Toronto and Region Conservation Authority (TRCA) and examine various propositions about which species would be more strongly discussed as invasive. You might think that invasiveness is more or less a function of what biologists say about a species. But anybody with experience in this field knows these considerations sometimes run up against those advanced by economic actors, often powerful ones, who sometimes want a given species to be controlled, and sometimes don’t, depending on how it would hit their bottom line.

We test both of those propositions, and find that the concerns of invasion biologists do play a strong role in shaping the discourse at the TRCA. By contrast, there is little sign that economic interests are strongly associated with a species being discussed as more invasive. That finding may be surprising, but keep in mind that the TRCA is primarily managing an urban region, and things might look different in rural authorities where agriculture and forestry industries are big players.

In any case, our central contribution is that a third process seems to play a strong role, which we call “workability.” This just means that a species is more likely to be treated as a problem when it is easier to deal with by the managing authority — even when it is equally harmful or dangerous as another problem candidate. For example, it is often a lot easier to deal with shrubs and wildflowers than birds, just because the birds can move fast and fly. Many fish are hard to monitor and control; they live underwater after all, and we do not. Other species are already hyper-abundant, and so strong intervention is basically pointless; there is nothing to be done but manage or accept them. We find solid support for this hypothesis.

As is the custom in sociology, we discuss these dynamics around invasiveness as an illustration of what is likely a broader phenomenon and a suggestion to examine this notion of “workability” more generally. You’ve probably observed something similar in your own organization, where, from the usually large universe of problems that could be taken on, only some are, and not only because those are the most problematic, but also because they are something the organization could actually imagine itself dealing with, or not, as the case may be. I’d wager a surprisingly large number of cases that pass through various social filters on the way to becoming thought of as problems do so based on some such notion of what is practically workable or not.

It is important to resist the temptation to dismiss this kind of finding and insight as trivial. It is actually a nice example of a form of sociological pragmatism, which often involves various kinds of reversals. Thought, the pragmatists argued, is often a function of practice. For example, you might think that organizations start from some mission that defines their vision, and their work is a kind of expression of that prior idea. In fact, more often than not, one of the most important goals of an organization is to figure out what its goal is, and that is something that only often becomes clear in retrospect or on the fly. Likewise, one of the most important problems an organization faces is to figure out what kinds of problems it can handle. These are not set out in advance, but arise through an ongoing process of action and learning.

Here is the abstract: Research on social problems often analyzes how different groups think or act in relation to a single issue. Less frequent are studies of how a single group thinks or acts in relation to many phenomena, any of which may be construed as problematic to a greater or lesser degree. We take this multiple phenomena-single social position approach and analyze why the Toronto and Region Conservation Authority (TRCA), a supra-municipal government agency, discusses some non-native species as more "invasive" than others. We use word embeddings to measure variation in the strength of association between different species and invasiveness in 599 of the TRCA's policy documents and employ generalized additive models to explain this variation. We find that the "invasive" meaning is more strongly associated with species that are easier to observe, access, control, or manage in the TRCA's urban context, which we term workability. Species that are terrestrial, sessile, and moderately abundant are more strongly associated with invasiveness than mobile, aquatic, and hyperabundant species. These findings suggest that problem managers conceive of issues they are responsible for managing according to how actionable problems appear. We propose workability as a key analytic lens for understanding how problem managers make decisions and construct meaning. We situate this contribution in the context of four research designs for studying social problems that we term comparative problem-solving designs.

But I also spend a fair amount of time reading things I don’t like very much, and have found that this is an important thing to do, now and again. By “don’t like” I don’t mean “disagree with.” I’m talking about books that don’t fascinate or captivate me, don’t bring me a lot of pleasure, where I must grit my teeth and pour water over my head to make it through to the next page. Those have their place too.

Most of them come to me from friends and colleagues whom I trust and respect. Usually they are part of a reading group, workshop, symposium, something which carries some social obligation. While I can’t say I’m always happy that I have read a book I didn’t enjoy, happiness is not always the most important thing in life.

I was thinking about this after discussing William Gass’ The Tunnel with a group of friends. We get together every few months, usually to read and discuss some broadly “modernist” text or other. A lot of them are experimental or “difficult,” so pleasure and captivation is not exactly what they are going for. Gass for sure is not.

I only made it a few hundred pages into the book, but I would bet that there’s only a handful of people who have ever finished it, including Gass himself. I wouldn’t dare to try to summarize on my own what the book is “about,” so here’s what Gemini 3.0 says:

William Gass’s The Tunnel centers on William Frederick Kohler, a bitter history professor who abandons the introduction to his magnum opus on Nazi Germany to write a scathing, secret memoir. As he hides these personal pages within his scholarly manuscript to conceal them from his wife, Kohler simultaneously digs a physical tunnel beneath his basement floor. This subterranean labor parallels his psychological excavation of a miserable past, exposing a “fascism of the heart” defined by bigotry, self-pity, and profound resentment toward his family and colleagues. Through dense, experimental prose, the novel examines the convergence of historical guilt and private grievance, ultimately presenting a portrait of a mind collapsing under the weight of its own disappointments.

That summary is a bit of a miracle, since any human reading the book would have such a hard time paying attention to and sorting their way through the dense experimental prose that simply being able to state what is happening is rather difficult. For me, most of my reading experience was with glazed eyes periodically being snapped awake by a few brilliant paragraphs. But that difficulty only makes it clear that following along what is happening cannot really be the point of the book.

I try to give authors the benefit of the doubt, and figure out some way to interpret a text as a good version of something rather than a failed version of something else. So that is what I asked myself in reading Gass. Some suggestion towards an answer comes from the observation above, that in fact there were moments of brilliance that I did get drawn in to. This passage is one example:

BEING. Holy word. Being cannot be recognized unless it succeeds in

Seeming. So Gorgias asserted. Yet Seeming, itself, will be weak and ineffective unless it succeeds in Being. Paradox #75.

Your skin was an ocean of sky sliding toward me. There’s ink on my

thumb. Plato and Gorgias—Lou—yes, you and I—suddenly Thucydicles—are names that occur to me. There’s the unstraightened deck of my manuscript. Mad Meg. The thick rug under me. Anxiety. The tunnel of my sleeve. How can history be a record of οὐσία? of life in a chair? Endless journey. Yet I had her in it. Act pleased. I did act pleased. Don’t hold back. And I groaned as I was coming. There, between her parted hairs, I acted out my pleasure. So she was pleased, and pretended to come, too, to please me. Aaaaaaaah . . . etc. And if I like a lover lie, and the lie’s believed, hasn’t ½ of appearance become reality? Then, Lou, as you turn to me, full of the love I’ve laid inside you like a shiny ceramic egg, I am ¼ informed by the feeling I originally bewinded; love rolls round on my love like a spindle winding, enlarging me with what is mine until I fatten up from

The text cuts off like that, and then is followed by a few close variants, also about somewhat embarrassing moments in Kohler’s life that involved a mix of keeping up appearances and realizing pleasures. All are somewhat abortive, in a range of meanings of that term.

The sudden appearance of the word Being in the midst of pages of text unbroken by a paragraph gets the eye to rest for a moment, as does the idea of a “holy” word: there has been no notion of holiness in this book whatsoever. What is it to be? The text then attacks this problem from the other side: if you want to seem, to make an appearance as something, and you want anybody to believe that, you need to be it, you need to make yourself into what you seem to be. Yet all we have is seemings, even in the bedrooms of “lovers.”

Reading this series of passages does remind you that Gass can write. So why does he write like this? There is a way that the text creates an experience of coming up to form, being at the brink of being there, and then never quite getting there, let alone staying there. But never quite falling back into simple seeming either: there seems to be some substance lurking in the background throughout.

Gass actually frames the book with a pair of ‘pennants of passive attitudes and emotions,’ a kind of affective wind-rose for a spoiled inner life. One of them fans out from vindictiveness into self-pity, pettiness, hypocrisy, procrastination, sloth, jealousy, churlishness, and sullenness.

These pennants give a kind of translation key to the book’s philosophical orientation. The passive attitudes and emotions are apparently a reference to Spinoza, who saw “the passions” – i.e. the passive emotions and attitudes – as those that are “confused” because the mind itself is not their cause, rather, something else is, which we can therefore only partially conceive. The goal of Spinoza’s ethics is to transform passivity into activity as much as possible.

All of the passive attitudes in Gass’s pennant are driven by some good outside the self that you want or define yourself against or allow to consume you. What struck me is that the pennant is organized around two directions of failed relation to goods. Along the upper edge, vindictiveness moves into self-pity and then slides into sullenness and churlishness, a drift from a kind of passionate enraged grievance to a withdrawn coldness. Along the lower edge it turns toward procrastination, sloth, and jealousy, the same sort of grievance moving from inertia into invidious comparison to others.

Hypocrisy holds the whole structure together. If you can’t become active, you can at least keep up appearances that you are. In the extreme cases, like envy (on a second pennant), it is not just that you want a good thing somebody else has: that would be jealousy. It is that you become consumed by a desire to destroy any goodness, without even wanting to have it yourself. This is a deeper kind of passivity; not a failure to realize some good actively, but a kind of active effort to destroy the very possibility of active realization of any good at all.

That’s maybe why the Being/Seeming passages stuck out. The pennant is a map of affective Seeming, strategies for living near Being without ever arriving there. The tunnel Kohler digs is also a kind of constant motion that goes nowhere. Gass’s book represents a world where that kind of transformation from passive to active is impossible, and perhaps undesirable.

You could see it as in the tradition of, say, Baudelaire’s Les fleurs du mal. Baudelaire showed many disgusting and evil things, but he tried to demonstrate that poetry could somehow make even them worthy of attention. Gass rejects that second part, about making things worthy of attention. That would not represent evil as it is, because it would not seem to be what it really is. What evil is is something that refuses, in itself, to be itself.

That at least is what it felt like for me in the act of reading the book. A book that had the look of a book, which promises to show you what a bad life looks like, and does that by not allowing it to acquire enough substance to even look like anything resembling a life whatsoever.

I don’t think I’d choose a reading experience like that again, but I suppose that is part of what it means to have passions, to be drawn into things that you would not on your own do, and try to make them your own.

A few months ago, philosopher wrote a very interesting article, mounting a defence of Enlightenment as a correction to humans’ deep-seated evolutionary tendencies toward misunderstanding our own situation. Williams’ core argument is that we are evolved creatures whose cognitive equipment is poorly suited to the vast, abstract, mediated social environments those same creatures have built.

For example, hunter-gatherers evolved a tendency to pay attention to dangers and threats (“negativity bias”), which was functional in their environment. However, in the modern media ecosystem, this can be debilitating and misleading.

“The predictable consequence of this is that people develop mental pictures of reality far more negative than the objective facts warrant. They overestimate poverty, crime rates, and many other social pathologies and dangers, and believe most trends are going in the wrong direction. In affluent liberal democracies, people are not just largely oblivious to progress. Their minds invert reality, often treating the most peaceful and prosperous societies in human history as dystopian hellscapes.”

For this reason, we need Enlightenment: a system of kludges built out in the form of cultural and social harnesses that save us from ourselves by creating error-correcting mechanisms attuned to the very maladaptations that plague us.

Williams’ article is a response to an equally interesting piece by evolutionary psychologist , in which Pinsof argues that by and large humans are very well adapted after all. Pinsof argues that Williams gives intellectuals too much credit and ordinary apes too little. We are not helplessly bewildered by modernity so much as pursuing older and less flattering goals under new descriptions. We should treat the human mind like

the hawk’s eye, the bat’s sonar, or the cheetah’s sprint.

Most of the biases that seem to affect us are only seeming biases. By and large, they work well for the purposes they are meant to serve. Yes, we may be overconfident, and that might not help to create and check valid scientific claims. But the value of overconfidence comes from winning influence and status through persuasion and charm. And for that, the human mind is at least as good as the hawk’s eye is at detecting poor little mice from afar, if not better – much better.

Williams’ piece was published on February 2, 2026, and as of this writing it has 327 likes, 53 comments, and 112 reposts, and some number of views that is likely very large but only known to Williams and Substack.

The first comment is by David Pinsof, which itself received 48 likes. That comment in turn formed the basis of a new guest post by Pinsof on Williams’ Substack, which received 128 likes, 45 comments, and 35 reposts, along with additional commentary by influential authors such as Arnold Kling.

The posts involved strenuous debate and disagreement concerning some fundamental issues, about which both authors cannot be simultaneously and ultimately correct. If one wins, the other loses.

Setting aside the question of who was right for a moment, however, the result of the disagreement was not that one author took influence from the other, but that the exchange produced more influence all around: more posts, more comments, more links, more followers, and more commentary and additional essays from other writers. Each now has a larger following than before the episode, a stronger set of claims on the feeds of other users, and therefore holds a greater set of standing commitments to be paid attention to by them -- and along with all that, a greater command over the collective attention space of the platform and the broader community itself. Their stature rose, as did the total amount of opportunities for attaining and exercising influence.

The General Formula for Clout

Through their exchange, Williams and Pinsof were able to carry out the social capitalist alchemy of transforming some initial endowment of influence, via posts, into more influence. This is a prime instance of what we can describe as the general formula for clout: I-P-I’. Yes, this is the social capitalist analog of the general formula for capital.

This outcome — more influence for all, even in the face of competition — is worthy of some reflection, since it helps us to think about the nature of one of the core topics in the Williams-Pinsof debate, namely, zero vs. positive sum thinking.

A central questions upon which the debate turns is the status of status. Pinsof (along with many others) seems to assume that status is inherently zero-sum.

status is zero-sum, and money is a status symbol, so it is arguably not a misunderstanding for people to view money as zero-sum.

In other words, status is inherently zero-sum, and since it infects other domains — for example, by making wealth and power into status symbols — it reveals those other domains as properly zero-sum contests for status, and therefore only seemingly divergent from our evolved tendency toward zero-sum thinking.

Pinsof is definitely right that status, understood as relative standing, has a zero-sum dimension. If I move from tenth to ninth, someone else has moved, or been moved, from ninth to tenth. It is a mistake, however, to infer from this that the broader domain of influence is therefore zero-sum by nature. “Zero-sum” is not a metaphysical property that attaches permanently to a human motive, like a curse in a fairy tale. It is, among other things, an institutional condition. Money can be hoarded, monopolized and used to buy political influence and social respect, but these pathologies do not show that markets were secretly zero-sum all along. They show that positive-sum capacities are institutional achievements, and fragile ones at that. My suggestion is that the same is true of influence.

Money and the capacity to act

My central claim is that platforms are institutions that make influence more money-like than it used to be. Whether a token of value operates more like a fixed tournament prize or more like money depends on the institutions that store, measure, transfer, denominate, and redeem it. Pinsof and Williams both seem to agree that money is or can be positive-sum (until or unless it is infected by status or rent-seeking), so it is worth starting by reflecting on what that positive-sum character means.

To understand this idea in more general terms, I find the observations of the great mid-century sociologist Talcott Parsons on “generalized symbolic media” to be helpful. In Parsons’ conception, money acquires its positive-sum character by expanding our capacities for action. It creates more degrees of freedom, along four dimensions:

1. Source. Holding money means that you are not bound to buy from one seller. You can shop around. This is a big contrast to, say, the situation of the feudal peasant. In a developed money economy, you have more options to purchase from more vendors.

2. Objects. Money increases your ability to choose what bundle of goods to buy from any given seller. This enables greater specialization.

3. Terms. You can bargain or shop over prices, by offering various terms. This communicates via prices to sellers what to make more or less of next time, and gives them an incentive, again via money prices, to achieve that goal.

4. Time. Holding money allows you to delay spending or spread spending over time, in longer durations and more fine-grained temporal units.

As these degrees of freedom expand, the total field of possible exchanges grows. A buyer has more options, sellers have wider markets: the division of labor grows. The monetary system — and money is an institution supported fundamentally by banks, the financial sector, government regulation and stability, cultural norms of trust, and more — enlarges the overall capacity.

To be sure, there are myriad ways this system can be undermined, such as rent-seeking, information asymmetries, externalities, monopolies, exploitative predation, and more. In fact, each expansion of the degree of freedom brings with it new bottlenecks to be exploited. If suppliers consolidate, the formal freedom to choose remains, but the real range of choice shrinks. You have money, but nowhere meaningfully different to spend it. The market may offer many items, but this can amount to 25 variations of the same breakfast cereal. Prices and terms can be formally open, but one side can have so much structural advantage the weaker party’s “choice” can only be written in scare quotes. And the freedom to wait can also undermine itself: money tends to lose purchasing power over time.

In other words, the very same conditions that expand degrees of freedom create new opportunities for alienation, exploitation, and disorientation. These all feed into what Perry Mehrling called “the inherent instability” of any money-credit system.

But even without invoking those, one can still arrange actors in a market order on a linear scale, where somebody always has more than somebody else, and intense competition is always producing winners and losers. Elon Musk has almost 2.5x more money than Larry Ellison — and even more than I do. Jeff Bezos is just a bit behind Ellison, and if Bezos can move up the list, this necessarily means that Ellison will move down. Money is zero-sum in that one person’s ascent is another’s decline, relatively speaking.

And yet, the median purchasing power of, well, humanity has grown massively all the same, not to mention that both Musk and Ellison are richer than they were ten years ago. For my part, even though I am way down the list, and my meager entrepreneurial efforts have not earned me great wealth, I have way more degrees of freedom than the feudal version of me would have: more and more various things I could buy, from more and more various different sellers, at any array of prices and terms, spread out over multiple time horizons.

The fact that some go up and others go down does not alter this fact, and in a well-functioning monetary regime, the competition to move up or down that scale also serves the function of discovering and revealing prices, opening up new and unknown types of products and services -- not only searching through an existing space, but whole new nodes in the division of labor. Growth and distribution are two faces of a market economy, and when things are working well, the latter feeds the former.

Influence and the Capacity to Act

There is much more to be said about this, but this is enough to draw attention to how remarkably similar the domain of influence has become to that of money. Thanks to the creation of institutions that permit influence to be accumulated, stored, transferred, counted and accounted for, withdrawn, created, destroyed – platforms – those same four types of degrees of freedom that the institutions of money and finance expanded in the economic domain have been expanded in the social domain.

1. Source. Having accounts on platforms means that you can receive posts from many creators, and indeed many platforms. This expands your options, and provides more opportunities for more creators to acquire influence.

2. Object. The capacity to allocate influence means you can choose what posts from which accounts to pay your attention to, assembling your own feed in the process. More producers in turn have more options toward which to target their efforts. The niche ecology becomes more differentiated..

3. Terms. You can offer more or less influence to other accounts, by way of various instruments such as viewing, subscribing, commenting, sharing, and the like. This enables competition for influence metrics to coordinate creator activity, similar to prices, and a mechanism by which producers can evaluate and modify their output.

4. Time. Platforms enable asynchronous influence. Posts can be saved or resurface later. Your follower counts, and the access to the feeds of others they entitle you to, persist beyond the immediate moment of attention. This allows creators and audiences to spread their activities and interactions out over longer periods and in more fine-grained temporal units. The effect is that influence earned by a post (from others granting it future access to their feeds) can accumulate well beyond the initial moment of posting, or indeed beyond the underlying “real” attention inspired by the platform, just like monetary wealth is far larger than the real goods and services produced at a given moment in a modern economy. For their part, audiences can continue to benefit from a post well into the future. Williams wrote his post months ago, after all.

As these degrees of freedom grow, the total field of possible ways of being recognized and acknowledged by others grows. An audience has more options, more creators have more access to more and more diverse audiences, and niches multiply and connect. This graphic from gives a sense of the scale of the transformation.

Tendencies in these directions exist in previous media forms. But platforms, through their core institutions of feeds, accounts, metrics, and, most importantly, mechanisms for transforming attention into bonds — claims (upon feeds) and obligations (to grant access to one’s feed) — mark a fundamental transition. Early images of “the attention economy” such as Michael Goldhaber’s, which envisioned “the net” on the model of a speaker enthralling a room, are fundamentally inadequate to what is better understand on the model of a bank administering a ledger. Influence binds, while attention is fleeting.

As in the case of money, more degrees of freedom bring with them their own reversals: more opportunities for exploitation, alienation, and disorientation, along all four dimensions. A few platforms or accounts can dominate access to influence. More content can mean more slop. The right to shape your feed through your actions can seem a rather pitiful form of freedom when somebody else has decided what is on the menu. And your accumulated influence decays: if you are not active enough, the same follower count loses value -- you can even lose your account -- and influence inflation can eat away regardless. 10,000 followers on Twitter in 2012 were worth a lot less than the same number a decade later. The result is discipline.

Frictions, moreover, eat away at your ability to use your influence. To adapt an idea of Fischer Black, at least as interpreted by Tyler Cowen: just as liquidity in monetary instruments is not a single number, neither is liquidity in influence instruments. It is not straightforward to get a following and reach from one platform respected on another, any more than it is straightforward to get a bar in rural China to accept my hard-earned Canadian banknotes. The numbers on your social media account do not automatically translate into an exact quantity of reach, any more than the numbers in financial accounts translate into an exact quantity of goods and services. This all feeds into the inherent instability of any influence-credit system.

Yet even as the system as a whole expands and degrees of freedom grow, one can still arrange actors in a platform order on a linear scale, where somebody always has more than somebody else, and intense competition is always producing winners and losers. There can be only one first reviewer of a new restaurant on Yelp. Elon Musk has over 2x more followers on X than Barack Obama; Williams has more Substack followers than Pinsof, who has more than me. If Pinsof or I can catch fire, we might pass Williams, and if Williams keeps on shooting down facile motivated theories of misinformation, who knows, maybe he’ll make up some ground on Musk!

And yet, the median power to make claims on attention and gain recognition has grown massively, despite there being some people with higher counts than others. Even though I am very very far down any list you might draw, I have way more degrees of freedom than I would otherwise have. Compared to the greatest prince ever to have lived prior to the 21st century, I have greater access to more essays on romanticism, rural American cities, Straussianism, looksmaxing, AI consciousness, and paeans to the value of attention; home repair videos, nature documentaries, philosophy lectures, investing advice; funny Catan memes, boxing tip reels, curling highlights, and much much more. Even if I lose, by playing the game, in all those other ways, I win — or at a rate, as long as enough people play the various games of the various platforms. And anyway, even now, my Substack essays earn a greater audience than most academic conference presentations do, not to mention readers of journal articles.

The fact that some go up and others go down does not alter this fact. In fact, the zero-sum competition to be first reviewer expands and refreshes the informational coverage of the platform. More generally, in a well-functioning influence regime, the competition to move up or down the scale also serves the function of discovering and revealing the social price of a post, while opening up new and unknown types of content and topics -- not only searching through an existing space, but whole new niches in the ecosystem. Growth and distribution are the two faces of social capitalism as much as they are of economic capitalism.

Collective and Distributional Power

None of this implies that the expansion of influence-capacity is the expansion of truth. A larger and more liquid market in takes may produce more Smiths, more Paines, more obscure geniuses rescued from institutional invisibility, not to mention more numerology, more partisan demonology, and more earnest critiques of the attention economy. Positive-sum is not a synonym for good. Yet recognizing the overlaps between economic and social capitalism can help to envision an approach to the ethics of the latter inspired by that of the former. Social capitalist platforms, like capitalist markets, are counter-intuitive moral contexts, in which second-best solutions are the best that creatures like us can expect.

Once this general point about the nature of symbolic media becomes clear, it is fairly straightforward to apply it to politics. I will not go far in that direction here, but it is worth noting the distinction mid-century authors such as Parsons, James Coleman, and then later Michael Mann drew between collective and distributional power. Collective power refers to the capacity of people to accomplish things together they could not do separately. It is the power to do, to mobilize societal commitments for collective goals, and it is enhanced through expanding infrastructural capacities such as financial systems, military organization and logistics, standardized measures, literacy, and the like. Distributional power means A’s power over B. This expands when one group increases its ability to impose its will on another.

Mann’s monumental Sources of Social Power documents and seeks to explain the truly remarkable expansion of collective power over the last 10,000 years or so of human history. Much of it covers innovations such as Roman military logistics, which enabled a Roman leaders to make decisions that reverberated far and wide, or the Swiss pike spear, which enabled tight bands of soldiers in phalanxes, bound together through deep bonds of mutual trust and solidarity, to undermine the military basis of feudal aristocratic power — not to mention ideological transformations, such as the spread of democratic ideas, which contributed to mass mobilization of citizen armies.

American history — and Mann is no cheerleader — by and large shows a similar expansionary arc. The U.S. President has been able to mobilize and bind actions over a larger swathe of the globe than any actor in previous history. This situation led Mann to describe the global power structure that predominated around the end of his fourth volume (circa 2013) as defined by the trifecta of capitalism, the nation-state, and America.

A party system becomes positive-sum, when it is, because the rules of the contest force private ambition and factional struggle to pass through institutions that build collective capacity. It is not positive-sum because parties are secretly nice, altruistic, unconcerned with their own standing. When it is working well, it pushes parties toward policy innovation, keeps them in touch with the citizenry, while forcing rival social interests to organize themselves, articulate general programs, mobilize mass publics, and build state infrastructures capable of redeeming their promises.

Distributional power does not disappear, far from it, and in fact, the stakes become higher. For example, the same infrastructure that makes pensions and tax collection possible also makes mass surveillance possible. Competitive democracy can produce a state that is less despotic but more penetrating. It expands what “we” can do, but it also intensifies the struggle over who gets to speak for the “we.”

Expansions of collective power are fragile, because the infrastructure they generate cannot only be commandeered by but undermined through distributional contestation running amuck. Such self-undermining races to the bottom are regular occurrences in complex systems, not unlike the peacock male evolving tails so large that they can barely walk. A US party that, in its effort to defeat its opponent, undermines the infrastructural power embodied in the US dollar being the world’s reserve currency is a case in point — for now, I’d still have a much higher chance of getting a beer in rural China with my US dollars, but will that last? Either way, this is not an instance of modern actors seeing through the veil to the truth of our situation, that all of this has been really zero-sum all along. Rather, it is a major challenge to an institutional design for transforming distributional contestation into collective capacity.

A sociological understanding of symbolic media

I’ll conclude by returning to the broader point. The creation of infrastructures and institutions that can store, measure, create, destroy, transfer, and price influence — platforms — is a truly transformative event, and we are still only beginning to understand what platforms really are and what they mean, even if it is beginning to seem passé to discuss them as we all rush ahead to speculate about what AI will bring. The Owl of Minerva is starting to fly, and it is time to revisit basic questions and assumptions about what a platform even is.

Just as important is that there were a host of very powerful intellectual tools developed for understanding such questions about the nature of symbolic media and their institutional conditions back in the heyday of mid-century sociology, which are worth the while of scholars and intellectuals today. It is understandable that those ideas have faded from the Discourse: sociology has largely neglected them itself, to its own detriment, coming to define itself more or less exclusively by the study of distributional questions. Nevertheless, there is much there to be learned, not least about when and why such a focus can shift from salutary to self-defeating.

While AI has brought more than its fair share of existential angst to this academic year, it has also been a welcome opportunity to shake off some of the rust and try out new things. The big new thing I’ve been doing in my undergraduate classes in sociological theory has been to create a series of assignments that try to get students to experience social theories as living ideational formations upon which they can experiment. The latest iteration invited students to get inside Max Weber’s theory of capitalism. More on that below.

It has also been a chance to learn from others and share ideas. One idea I found very promising is ’s Synthetic Socrates, specifically his “Clap Back” exercise. Here’s how he puts it:

In Clap Back, students spar with a customized AI trained to defend a specific philosophical position—skepticism, utilitarianism, nihilism, theism, you name it. The role of AI is to argue tirelessly and consistently while the student is tasked with exposing errors in its reasoning and defending a counterposition…we sometimes hold in-class tournaments, where groups compete to see who “defeated” their AI most convincingly, judged by peers and professor.

This assignment lowers reputational stakes while sharpening philosophical skills. Students who might hesitate to challenge a classmate will freely push back against a machine that cannot feel embarrassment or condescension.

I liked this idea a lot, so I decided to steal it. I made something similar to go along with my lectures on Max Weber, a custom GPT called Mad Max. Mad Max is trained to defend Max Weber’s ideas, and students try to defeat him and expose flaws in his arguments.

We deployed Mad Max primarily as a fun interlude during lecture. I’d ask students to suggest Weber’s more controversial ideas that they thought were most dubious, and then they’d take a crack at him. As you might have predicted, they chose his arguments about the uniqueness of the West.

What ensued was basically as Jimmy reports in his classes. A topic that might have been touchy or intimidating became pretty fun. When Mad Max didn’t easily concede to the more obvious counter-arguments, students kept trying, and then would go home and try to come up with new arguments throughout the week. That’s a win for AI in the classroom in my book.

The Iron Cage Assignment

The assignment continues to use chatstorm.io as a tool for conducting experiments in social theory. For this assignment, I built a multi-agent representation of (an interpretation) of Weber’s Protestant Ethic argument. The design opens with a conversation among a Preacher, a Believer, and a Non-Believer about the salvation of their souls, moderated by The Doctrine, who embodies Calvinist theology (as Weber understood it). The Preacher has to translate the doctrine into some kind of teaching by which the Believer could live, while the Non-Believer has to contend with the fact that he lives in a world that is populated with ascetic Calvinists.

After they have spent some time discussing their souls, the Believer and Non-Believer move to a round called The Shops. Here they are both shopkeepers bidding on a contract from a Patron, who has to decide with whom to do business. At some point in the conversation, the moderator introduces various stress tests, and the Patron must choose based on what the response reveals about the shopkeeper’s character. Depending on how the conversation goes (especially whether the more rationalistic tendencies internalized by the Believer appear as well in the Non-Believer), the simulation either moves to The Iron Cage or remains in a pre-capitalist society, based on a metric that evaluates the conversation. If in the Iron Cage, the whole cast of characters returns, including the Preacher, and reflects on their situation, while externally they maintain the discipline of a capitalist order.

The assignment requires students to formulate two theories about how the various outcomes of the process would change under different scenarios. For example, they could propose changes to the theological content of The Doctrine. Would capitalism be as likely to arise? What changes in the interactions would they predict? How would they measure those changes? They could propose changes to the Preacher’s approach to his flock, or to the strategies of the non-believer, or more.

Then they create two new variants of the simulation, each one designed to test one of their theories. This involves changing the prompts or adding or removing agents or changing some of the settings (like adding self reflection or more turns, etc). They run the models, study the transcripts as data, and submit an analysis. The analysis asks them in addition to reporting their results to critically examine the assumptions of the baseline design and discuss how their theories would add some new understanding to the interpretation of Weber’s theory.

Excerpts from the assignment are below. This one was rewarding in that more students got inside the logic of the argument and really seemed to have fun trying out different possibilities for tweaking the classic Weberian argument structure. They saw very clearly its logic — the famous “Coleman boat” moving from macro to micro to micro to macro — and had the experience that its elements were not fixed in stone but open to their own playful theoretical creativity.

For more than a few, I had the sense that they had that first glimmer of seeing themselves as potentially standing alongside the likes of Weber, not only trying to memorize things he said, but working on the foundations he built, encountering thoughts as ideational architectures, probing them, prodding them, adding new pieces, taking some away. That’s what we’re after as teachers, after all, and our goal should be to try to spark that kind of experience in our students with the media of our times.

Assignment 1: The Iron Cage of Capitalism

The Protestant Ethic and the Spirit of Capitalism is one of the most influential arguments in modern social theory. It is also one of the most disputed. The text joins claims about cultural meaning, psychological anxiety, economic conduct, and historical transformation, and it has generated a long history of rich interpretive debate: What exactly is the mechanism Weber is proposing? How much does it depend on specific doctrines like predestination? Is the central link psychological, ethical, organizational, or political? What is the significance of the “iron cage”? Is it an actual historical outcome, a theoretically potential tendency, a prophetic moral warning, or something else entirely?

This assignment asks you to learn Weber’s theory by treating it as something you can work on. You will begin with a baseline ChatStorm simulation that operationalizes one interpretation of the Protestant Ethic argument. You will then “open up” the model by modifying its components, running new versions, and observing what changes. The goal is to sharpen your understanding of what assumptions Weber’s argument depends on, and what kinds of evidence would count for or against different interpretations. In the process, you will practice a core theoretical skill: making precise claims about mechanisms, and defending those claims with traceable and systematic evidence.

Two experimental variants

1) Two experimental variants, two different stages in the causal chain

Create two experimental variants of the simulation by copying the baseline design and changing it. Your two variants must modify different stages of the causal chain. For example, one “upstream” change (doctrine/preaching/psychological reception) and one “downstream” change (economic conduct/non-believer diffusion/selection environment/generalization to the iron cage).

2) Kinds of changes

You may revise clauses, add small rules, rewrite a prompt, add an agent, or more. But you should keep the baseline structure recognizable. Possible stages you can modify include:

• Doctrine (theological content of religion)

• Preaching (how the doctrine is translated into lived guidance)

• Psychological reception (how the preaching is experienced by believers)

• Believer economic conduct (eg attitude toward sect members and non-members)

• Non-believer response (imitation, resistance, compelled discipline?)

• Patron/Market selection (what basis they use to select the contract)

• Macro generalization (trajectory threshold or scoring criteria, iron cage transition logic)

After you have developed a functioning model, run each variant twice and export transcripts and data.

Analysis

Target length: ~1500–2000 words (quotes, tables, and figures do not count toward the total; you may go longer if you wish).
Your analysis must do five things (they can be integrated, but they must be clear):

1. Interpret Weber.
   Use the text and lectures, and show awareness that serious readers disagree about what Weber’s argument means and how it works. Do not use secondary literature. Imagine you are speaking with a classmate who disagrees with you about how Weber’s argument works.

2. Explain what the baseline is actually modeling.
   Explain how the baseline simulation operationalizes one interpretation of Weber. Be concrete: identify which prompts, rules, roles, and decision procedures represent which parts of the theory, and why. The goal is to show you understand that “modeling Weber” involves interpretive choices. Better answers will explain what the interpretative choices are and what alternatives would be possible.

3. Critique assumptions, and use your variants to probe them.
   Identify at least two assumptions in the baseline that make the outcome seem pre-determined. Then explain how your Variant A and Variant B probe those assumptions, and what you learned from the probe. An “assumption” here means a concrete design choice: what the doctrine says, how preaching translates it, what counts as “proof,” what the Patron rewards as “reliability,” what the Market treats as score-worthy, how the disruption is framed, and so on.

4. First reading → second reading.
   Organize your argument around an initial reading you take seriously but ultimately reject, for example what a quick reader might conclude from a first look at the transcripts. Then develop a second reading: what you conclude after close reading of prompts and transcripts, and after comparing baseline to your variants.

5. Assess your two theories with evidence.
   Use your three empirical consequences per theory to organize your evaluation. Ground claims in transcript quotations with citations and metrics based on transcripts. Make your claims contestable: say what would count against them, and imagine you are writing for a skeptical reader who is looking for weak spots in your argument and evidence. How can you convince such a reader that your claims are valid?

All major claims must be backed by transcript citations and/or clearly reported extracted values tied to a specific run.

It was my birthday the other day. The day before, my wife asked me if I had any final reflections on 48. I said: it is 12 x 4. The next year will be 7 x 7.

At the time, it was basically a bad joke. But thinking about it more (we were on a canoe trip in northern Ontario, which lends itself to such meditations), I ended up reflecting on the nature of the various numbers involved. When you break up a number like 48 or 49 into its parts, each one seems the same as the other. All 12 4’s in 48 are the same. Every 7 in 49 is 7.

But that is obviously not how the numbers of a life feel. The first ones are different from the later ones. The shift from 44 to 48 or 42 to 49 is not the same as the one from 4 to 8 or 7 to 14. Life is not linear.

True enough, but what form of non-linearity does it take? One starting point has to do with pacing. Summers seem very slow and long when you are younger. Time stretches. As you get older, time is shorter, and you feel it. It is like looking out a train window in the prairies vs. entering into a city. You are moving the same speed the whole time, but there is an uptick in the number of things demanding your attention from moment to moment. The same unit of time and space takes on a different meaning as it becomes filled up with more demands and responsibilities. You shift from absorbing the world to applying what you absorbed back to it.

There is a lot to that, but I wonder if that specific experience is an instance of a more general phenomenon, which we could consider as the fractal character of life. I recall Andy Abbott’s book on academic disciplines, where he made fractals central to his account of non-linear aspects of intellectual organization. Fractals are basically simple repeated structures that produce very complex organizations, where each component has the same form as the whole: self-similarity.

The repeated branching pattern means you can never really understand where you are in a linear fashion. Take the divide between “scientific” and “humanistic” thinking. That divide repeats from speciality to speciality, discipline to discipline, repeating itself in its own way. In statistics, there are Bayesians and frequentists. In sociology, there are qualitative ethnographers and quantitative data analysts. In literature, there are close readers and computational text analysts. In biology, there are ecological field researchers and mathematical game theorists. Yet from the point of view of, say, literary studies, the whole of statistics is on one side, and vice versa. You cannot walk from one to the other in a straight line but must move through the fractal repetitions.

There’s something similar to moving through the life course. People often say how much easier it is to learn when you are a child, with language acquisition as the prime example. I’ve never understood this. If I had two people whose entire existence was devoted to teaching me a new language, who were constantly rewarding me for every hint of a movement in the right direction, if everybody else I met was doing the same, and if that was all I was doing: I’d learn French faster than a French infant. Actually, right now, I have all sorts of people doing something like that for me. There’s support staff that help me do my work, research assistants, sports coaches.

Of course, there’s a difference. But the difference has the character of a self-similarity. I am repeating more or less the same learning process, just further down a particular branching path. Indeed the entry into language itself and the linguistic-symbolic order is a momentous event. But so too is learning a new boxing move. Each opens up whole worlds to you. It is not that early life is absorption and later life is application. Every stage is both at once, just like every academic discipline is both scientific and humanistic, both objective and subjective.

Even the different time horizons of different life stages have this fractal character. Each one is long and short. When you are young, the whole rest of life stretches out. But youth is fleeting. As you age, experience has piled up and each moment is invested with the full weight of the past. Yet life itself seems fragile and on its way towards its end. From the point of view of the one, the other looks entirely long or short, but within the fractal self-similarity, each is both, just at different levels.

This is why moving from one block of years (4, 8, 12, 16) to the other is in a way linear, after all: each one contains the self-similar form of the other. And yet it is also non-linear: each one produces the self-similarity in a different way down a different branching path. The result is that to move from 16 to 20 or 44 to 48 is the same, but different.

So yes, 48 is 12 x 4, and 49 is 7 x 7.

My examinations of social capitalist ethics so far have approached the topic of platform ethics starting from critics whose moral critique makes it hard to see any ethical justification for the existence of platforms whatsoever. The market failure view I have been advancing argues that we are stuck with platforms as clout tournaments, which represent a second-best solution fit for flawed creatures such as ourselves. It instead asks how to regulate those tournaments from within.

It is worth pausing over the fact that a great deal of critical writing about platforms takes the opposite tack. Instead of condemning the entire institution (a la the morality of the gentleman) or accepting the basic tournament structure and trying to regulate it by its own implicit morality (the approach pursued here), it asks why we could not design platforms directly around public values, then points to alternative architectures that are supposed to embody those values. Given the vast information processing power represented by Big Data and Machine Learning, could we not redirect them to public values and get rid of the clout tournament, thereby ensuring that only good content is posted and we do not spend our time instrumentalizing one another to increase our status? This approach, in other words, seeks a socialist solution to the problem of social capitalism.

A clear statement of this view appears in José van Dijck and co-authors’ book The Platform Society. They urge us to treat platforms as basic infrastructures and to rebuild them around public values such as privacy, accuracy, safety, accountability, fairness, accessibility, and democratic control. They also offer concrete examples of platforms that at least approximate this vision rather than subordinating everything to engagement and advertising.

When I first read the book when it came out in 2018 or so, these examples of platforms explicitly created in the service of “public values” seemed like interesting and promising experiments to me. Yes, of course, we should ensure that platforms circulate content that is helpful, harmless, and safe (to mix tech metaphors a bit), and sure, civic platforms designed by and for the public sounded great. When I went back to the book recently, however, and looked up the cases they had highlighted then, time and the logic of social capitalism had taken its toll. In pretty much every case I checked, the platforms that had been held up as a promising alternative had disappeared or stagnated.

That in itself does not show that public values are a bad idea. But it does suggest that they tend to face some serious problems that are at the very least very difficult to solve. The goal of this essay is to articulate the nature of those problems and understand why the clout tournament persists as the default mode of handling them. More specifically, why have we not seen a successful move to something closer to a planned content economy that directly assigns and distributes what people “ought” to produce and consume? Why has the indirect, competitive approach of the clout tournament for influence persisted?

Yes, this is indeed structurally the same as asking why alternatives to capitalism continually fail, and yes, it is structurally analogous to classic problems in political economy such as the problems of calculation and incentive. Political sociality adapts and transforms political economy onto a new plane.

Outside the Pin Factory

To see what is at stake in these design choices, it helps to recall that market economies are at bottom systems of cooperation through a division of labour. It is not logically necessary to organize such a system through competition for tokens of value.

Consider Adam Smith’s pin factory, often considered the very image of a market order as such. Workers split up the tasks so that more pins get made than any one of them could produce alone. When those workers go shopping, they stand in an indirect relationship with the butcher and the baker, who have also specialized. Their relationship is mediated by prices and competition.

But inside the factory, as Heath has observed, their work is not in fact organized on market principles. The boss assigns each one to a task and a station, adjusting the plan based on observations of how various forms of organization perform outside the firm on the market. The question defining the socialist tradition, put crudely but not entirely inaccurately, is: if we can run a factory through direct planning and authority, why can we not run the whole economy that way?

The same question arises under social capitalism. Faced with a complex and messy ecology of sub-cultures, niches, tastes, and interests, which seem to be governed by unpredictable tournaments for attention and engagement, it is natural to say: let us plan it instead. If we can run a little magazine or a movie studio like the pin factory, through direct planning and control, why don’t we run the platform that way? We identify the content that serves public values, and decide the proportion of news, education, entertainment, civic information, and so on that a healthy information diet should contain, and set the trusted and truthful and wholesome providers to work making it. We then design feeds to deliver the resulting mix. Easy peasy.

Parecon

Alas, it is not so easy, and spelling out what it would actually entail shows why. To do so, we can take a page from Heath taking a page from Michael Albert’s Life After Capitalism. Albert does us all a great service by spelling out what a vision of a “participatory economy” would look like, which he calls Parecon. Heath in turn takes that vision seriously and shows how easily it falls apart if you just play out how it would necessarily proceed in anything approaching the real world, and even in a very idealized version of the real world at that.

In this section, I will adapt both to spell out and examine the limits of a “participatory sociality” at scale, i.e. a Parform. If that system suffers under the weight of the calculation and incentive problems even in theory, we learn something about why, in practice, even highly data-rich platforms continue to fall back on staged competition and influence prices to coordinate the production and circulation of cultural content and social recognition.

In Albert’s theory, the economy is run by two main kinds of bodies: worker councils on the production side and consumer councils on the demand side. Instead of firms setting output in response to market prices and individuals deciding what to buy in response to their wages and those prices, everyone participates in an iterative planning process. The process looks roughly as follows. Once per year, or at some regular interval, each consumer council draws up a list of what its members would like to consume over the next cycle. Each worker council, for its part, draws up a production plan. These state publicly what it intends to make, as well as how much, and what inputs it will need. These plans are submitted to a central agency that Albert calls the Iteration Facilitation Board (IFB).

By design, the initial round of plans will not cohere. Demand for some goods will exceed the willingness of any workers to produce them. For others, there will be more supply than anyone has asked for. The Iteration Facilitation Board responds by proposing a schedule of prices. Where requested quantities greatly exceed planned output, prices are raised. Where they fall short, prices are lowered. These prices then feed back into the next round of proposals. Councils adjust their consumption requests to fit their budgets; worker councils revise their production plans in light of the new price signals. The Board recalculates, and the process repeats.

The goal of all of this is to reproduce what a well-functioning market would produce without having to go through the process of staging a competition by which some become rich and others poor through the economic discovery process. If the planning process converges on a coherent plan, then, in principle, workers will intend to produce no more and no less than what consumers intend to consume. The plan in turn becomes binding on all parties until its next iteration.

Voila, the division of labour is aligned with actual needs. The production schedule reflects both how much others want a good and how hard it is to produce, and they impose a kind of social cost accounting on both production and consumption. The equivalent of market pricing is retained as a principle, but realized through participatory deliberation instead of decentralized exchange.

The perils of Parform

To see what this looks like in the social capitalist context, imagine a parallel design for platforms. Let’s call it a Parform.

Instead of firms, we have creator organizations of various sizes: influencer agencies, production studios, large channels, newsrooms, little magazine, game streamers, subreddits, Discord servers. These are the content worker councils on the production side. Instead of consumer councils, we have audience councils: fandoms, academic specialists, professional communities, reading groups, local neighbourhood forums, and so on. Each council represents a cluster of users with some shared sense of what they need and want from their online environment.

Once per cycle, every audience council draws up a plan for its content diet over the next period. They might specify, in broad strokes, how many hours of local news, global news, educational videos, light entertainment, sports, religious material, furry art, Romanticism discourse, and so on they expect their members to want.

The above figure is derived from topic modeling content from a year of near-complete data on BlueSky posts, based on ongoing work with Martin Lukk. Furry Art is the second largest topic. This is provisional!

At the same time, creator councils submit their own plans. They declare what they intend to produce in the next cycle, indicating their plans regarding the number and quality they will make of long-form videos, short-form AI generated clips, 5000 word essays on Romanticism, cat memes, investigative pieces on government corruption, or fan edits of TLOTR. They also specify what they need in order to do so: staff time, compute, editing software, actors, research, etc.

All these plans are sent to a central content board, call it the Internet Facilitation Board. It faces the same kind of problem as Albert’s. The initial plans will almost never line up. Demand for some types of content will far exceed any creators’ plans to supply it (we need 1000 more Furry post and a huge order of condemnations of the attention economy!). Other kinds of content will be in plentiful supply on the production side but barely requested at all (sorry Romanticism!).

The central board responds by setting what we might call indicative influence prices and content quotas. For categories where audience councils are requesting far more content than anyone has offered to make, the board raises the influence price. Creators who commit to producing in those under-supplied niches will, in the next cycle, receive more visibility, more promotion, more favourable placement in feeds per unit of content. For categories that are over-supplied relative to demand, the influence price is lowered. Producing there will earn you less reach.

The quotas play a parallel role on the consumption side. Audience councils are given finite “feed-space credits” – a budget of posts, minutes, or slots in their feed – and they are obliged to adjust their plans so that the total content they ask for in each category fits within that budget. If they want more of a high-priced category, they have to give up space in others until their overall request can be covered by their allotted credits.

The process then iterates. Councils revise their content diets and creators revise their production plans. The IFB updates influence prices and quotas. After some number of rounds, the influence returns for different kinds of content reflect both how much people say they value them and how hard they are to produce. Production occurs, and audiences get just the right amount of news, sports, entertainment, Russian literary criticism, and exhortations to turn off their phones.

This is in effect a bureaucratized version of social capitalism run by well-meaning and honest officials, more or less exactly what the moral critics of platforms might have hoped for. In this scenario, everyone is trying to do the right thing: audience councils are sincerely attempting to describe what their members will want and need and creator co-ops are honestly reporting what they can produce. The IFB is diligently using its data and modelling capacity to set influence prices and quotas that best match the two. There are no nefarious evil platform execs trying to exploit anybody.

It is an attractive picture, especially if you are impatient with clout tournaments and hostile to the idea that “whoever wins the feed” should decide what we see. It is also, alas, doomed to fail.

The first set of problems are calculation failures, which affect audiences and producers. Let’s envision a set of audience councils doing their honest-to-goodness best to list what their members will want to watch, read, and listen to in the coming period. They know that, in principle, it would be good to consume more serious news coverage and fewer junk clips, and maybe replace a bit of Furry content with some Romanticism. They also know that there are topics they ought to learn about and communities they ought to hear from, and have some information about the typical information cycles through the year or week.

They incorporate all of this as best they can, but they cannot know the future. Some crises will hit, and some scandals will erupt, but nobody knows when and what. Maybe the United States will invade Greenland?! They also cannot know which new genres or formats will be invented, or which creators will suddenly become important to their members. Would anybody have predicted we’d be watching an elderly Mike Tyson fighting Jake Paul?! A new platform launches in the meantime and changes how people organize their attention (hello literary Substack!). AI expands the ability of more people to make short-form videos, creating whole new taste communities (Sora? Sora?). All hits are flukes, as they say in the sociology of culture.

On the production side, things are no better. Creator councils sincerely report what they think they can make. Some projects will turn out to be much harder than they seemed at planning time. I thought this series of posts on platforms ethics would be a few thousand words. I’ve just finished the draft of the ninth and last one! Similarly, as new technologies appear, some projects will become trivially easy or obsolete before they are finished. It took me a month of hard work a few years ago to make an AI generated comic book for my dad; now it’d take half that, or likely a lot less. Some ideas that one would not even have imagined to be morally objectionable might in the meantime become toxic, and continuing on the planned production schedule would doom many producers to cancellation.

Because aggregating and iterating plans across millions of councils is costly, Parform cannot run this process continuously. The Internet Facilitation Board must lock in a plan for some meaningful period, a year or at least a quarter, maybe a month, otherwise the whole exercise collapses into permanent revision. Once fixed, everyone is normatively bound to stick to it. That is the very point of replacing the clout market with a plan: if it could be continuously adjusted then it would not be a plan. The result is that even if all reports were perfectly accurate at the moment of writing, the plan will drift away from reality. This is the Scroll to SurfDoom.

Thus far we have only enumerated the impact of taking account of some of the ordinary turbulence of social life. Yet just these throw up consequences that are totally catastrophic for Parform. The councils’ earlier lists necessarily understate demand for some kinds of content and overstate demand for other others. The result would be the same in Parform as it would be in Parecon: black markets arise.

This is because users would inevitably become frustrated that the “official” content diet does not reflect what matters now while creators would feel pulled between their obligation to honour the plan and their sense that they are making things their audiences don’t want anymore. Another long post on prompt engineering???? That was so 2024. Give me some AI generated manifestos on Claude Code and how to build Karpathy’s knowledge base!

Users would in turn begin passing around “off-plan” content in private group chats, Discord servers, DMs, and encrypted channels. A council may have requested only a modest amount of international news for the year, but if the USA invades Canada, you can bet links and clips will circulate anyway. These become the arenas where the real “influence price discovery” happens as the official plan fails.

We are all familiar with this tendency and how it rises as platform performance degrades. It is no accident that private shadowy Whatsapp and Telegram groups became more important when Twitter became its most censorious. The point in the present example is to show that even under idealized circumstances of planning, black markets in content and influence emerge.

Precisely because official plans will tend to fail, these unofficial flows will tend to do real information and discovery work. They reveal where demand actually is. Under a directly regulated system that pre-determines what type of content can be circulated, if you want to know what people really want, you would need to look at the group chats and private forums outside that very system, not the public Parform. Planners who care about doing their jobs well would accordingly look there and not to the plan, much like Soviet planners used black market prices to inform their own decisions. Likewise, socialist platform planners would inevitably start peeking at private group chats in order to update their sense of what people really want and where production should really go. Black markets in content and influence would emerge as the spontaneous attempts of decent participants to remedy a system that cannot keep pace with reality.

Faced with this rigidity, a technocrat might argue that the problem is simply human slowness. Could an AI planner not solve this by monitoring demand in real-time? Instead of slow human councils struggling to keep up, a sufficiently powerful model could, in principle, monitor demand in real time and infer what people will want next. It’d then instantly spin up fleets of bots to fill every niche. Call this the idea of “frictionless participatory planning”: the idea that once you layer big models on top of big data, you can keep the plan so continuously updated that it effectively replaces the clout tournament.

There are at least three problems here, which I can only gesture at for now. First, AI-generated content itself would almost certainly become a genre, subject to its own bubbles and collapses. Think of the “glazing” episode with ChatGPT, where a particular stylistic quirk suddenly surfaced and became an object of ridicule. Everybody was a fearless genius, and got really annoyed when they discovered that it was not just them. Note too that it was not Open AI that first picked this up, but users on Reddit who compared screenshots and reverse engineered the pattern. It still takes time to retrain models and re-engineer prompts after these quirks are noticed. The clout tournament identifies and returns in the breach, which is to say, always.

Second, a central AI planner still must predict the future based on the past. To the extent that preferences are themselves formed through interaction, then the information contained in influence prices will always contain something genuinely new that could not have been otherwise revealed. In other words, as in the case of economic prices, influence prices (metrics) are not so much a way of accessing information that would be otherwise hard to retrieve, they are themselves a form of information.1

Third, the very spread of AI content is likely to produce a countervailing taste for non-AI content and human style. That demand will not disappear just because the main platforms route more of their production through models. It will instead most likely reappear in side channels and alternatives, not to mention, as observes, purposeful human efforts to display one’s humanity. These are obviously large topics that deserve much more elaborated separate treatment.

So far, I have laid out the charitable case for why platforms keep returning to clout competition and influence prices as their coordinating device. My point is not at all the we should therefore treat the clout tournament as somehow morally pure or sacrosanct. Once again, it is a concession to human nature and a second best solution. The problem is, however, that the most serious alternative one could imagine – participatory planning on a Parform model – cannot solve the calculation problems even in the best of moral circumstances. Even on those assumptions, the system would drift toward the spontaneous emergence of a dark web and informal reliance on competition for influence.

Once we relax the saintly assumptions, Parform looks even worse, as it reveals the looming specter of the incentive problem. Imagine that we do not presume that our audience councils are always acting in the name of the public good (i.e. that lo and behold, the “new man of socialism” has not emerged to match our desire for him). An audience council like that (i.e. a human one) operating under the rules of Parform would then have every reason to exaggerate what their members “need.” The structure offers no penalty for a council overstating its demands, but a very high risk of not getting what they want if they do not. It’d be a world where everyone suddenly says they require far more Romanticism, local journalism, and Arthouse Cinema than they could ever actually watch. Anyone who has sat through an annual departmental budget round will recognize this pattern. Surprisingly, we all always need more resources.

On the production side, creator councils face the mirror temptation. They can present what they do as uniquely difficult or publicly crucial in order to secure extra resources and a larger guaranteed slice of feedspace. If the IFB buys it, they win; if not, they are no worse off than if they had been more circumspect. Anyone who has written an executive summary for a policy report or grant application knows this particular temptation. There is always an impending crisis, and somehow this very project will address it. Reality recedes as everybody orients themselves to gaming the plan and the IFB.

At the same time, once everyone realizes there will be unofficial channels of circulation, a shadow currency inevitably emerges in the DarkWeb. In real-world Soviet socialism, gasoline played that role; it became an implicit form of money that would price other goods and services.

In a Parform, it would likely incentivize seeking influence on important group chats. Councils and creators would start tilting their requests and promises toward whatever helps them secure more of that access. At that point, it becomes common knowledge that others are shading their reports to what “the groups” are saying, and honest planning turns into a sucker’s strategy. A sane planner will eventually stop trusting the Parform schedules and reports and look instead to the influence metrics arising from clout tournaments on the dark web as the only signals still tied to what people actually do. Parform metrics would become effectively denominated by visibility in Marc Andreesen’s group chats. [Marc, my DMs are open 😜]

Having taken this fanciful theoretical detour into imagining what a socialist platform organized around direct control over content production and circulation would look like, we can now say what competition for influence metrics means for the ethics of social capitalism without either demonizing such competition or treating it as the natural essence of social life. Staged clout tournaments solve a specific coordination problem that large-scale planning cannot solve, yet they solve it in a way that brings its own moral costs.

A standard market economy may have many virtues, such as liberty and innovation. But generating good prices remains the central one, since that allows for the coordination of production and consumption at scale. Likewise, the core virtue of a well-designed platform is metrics that incentivize the production and coordination of content. We may prefer a world where people just produced and shared valuable content from pure civic and expressive motives, guided by deliberation and mutual recognition. In the real world, we rely on clout competitions to harness self-interest, and we accept inequalities and pathologies we would not endorse in an ideal world.

In markets, adjustments toward equilibrium happen through actual trades, which sometimes produce windfall gains and losses. Even if they eventually balance out, in the meantime, there are winners and losers, and those are not always rooted in desert (or dessert!). In clout tournaments, metrics rise and fall through real posts and interactions, which often produce windfall gains or losses of influence. Someone goes viral because they stumbled into an under-supplied niche at the right moment; others copy and flood the niche, eroding the original advantage. Some are early adopters of the platform and were lucky enough to be in the list of recommended accounts to follow.

Those windfalls are part of how the system discovers content possibilities. They signal to others that a niche exists and induce entry into that area, until the rewards for it fall and the search resumes. If, as it is sometimes said, every successful market is a failed cartel, so too, every successful platform is a failed engagement ring. While we may resent unearned clout, we may have to tolerate some of it as the cost of using staged competition to explore the space of content and induce its production.

I must stress that this argument is not at all a paean to laissez-faire. In fact, the topic of my next series of posts is the internal regulation of the struggle for clout. Once we reject the more utopian aspirations of the public good cooperative platform that exerts direct control over what happens on it, the key problem becomes more subtle yet also more realistic: to distinguish flaws that are necessary byproducts of competitive process from gratuitous ones caused by bad institutionalization.

Some rather unseemly acts are baked into any tournament that uses influence as the prize, such as relative rewards (somebody always has more or less), never-ending grinding (the influencer ethic), misalignment between recognition and desert (viral windfalls and sudden influence collapses reveal the collective interest often without regard to the goodness or badness of their targets), not to mention occasional speculative bubbles in new formats (sorry Sora!), and yes perhaps a bit of hype (LEARN THIS ONE CHESS OPENING YOU WILL DESTROY MAGNUS CARLSEN‼‼!).

Others practices are rather gratuitous: harassment, sabotage, opaque and frequent rule changes, no opportunities for new entrants, tournaments so skewed that most participants have no real chance, predatory extraction, oversaturation with advertising, engagement rings, bot attacks, content theft, to name a few. These are all topics concerning the domain of Jus in Clouto: the internal ethics for those who have accepted the game.

The point of this recent set of posts, however, has been to articulate the justification for playing the game at all. Clout tournaments are a concession to the limits of human virtue and individual cognition, not a celebration of selfishness. If people were saints with perfect foresight, you could run social media as a benevolent participatory bureaucracy and let audience councils calmly assign everyone a balanced information diet. Since they are not, and since no one can see very far ahead, platforms stage competitions.

This brings us to the domain of Jus in Clouto: distinguishing the “healthy adversity” that makes the game work from the “cheating” and “gratuitous harm” that ruins it.

Stay tuned!

Even so, I find myself reacting somewhat differently to the current moment. That reaction recently led me to resurrect an old computational analysis of the history of sociology’s discourse. The result, in brief, is that the field does not move in a simple linear direction. Whatever one thinks about the present moment, disciplinary history suggests that no ascendance is ever final.

But first, let me say something about the nature of my reaction. I will always feel deeply grateful to sociology. Many of the very features of the field that can be frustrating are also what made it possible for somebody like me to enter it. I do not have a PhD in sociology and have taken exactly one sociology course, an introductory class in my first year of undergrad (Intro with Ann Swidler!). A more unified and self-confident discipline might never have had room for me. Thanks to its fragmented character, and its recurring uncertainty about its goals, methods, and objects, there was room for someone like me to slip in. Even then, the stars had to align in a way that might not occur again. Luck and contingency matter enormously in how a life plays out.

So I am grateful, and I have always felt some obligation to reciprocate. For me, that has meant trying to learn the canons, methods, and classics of the field, to make real disciplinary contributions, and to act as an ambassador for it, not only here on Substack but also by using its ideas and tools to make some contributions to urban cultural policy that might actually help cities and communities. All of that I hope redounds to the favour of the discipline, and helps me to earn my keep in it, by helping it to earn its keep in the wider world. Yes, the f-word strikes again.

That relationship with the field, though, means that I never saw it, to quote Ashley on X, as “my favorite sandbox with your best toys and best aspirational friends.” I saw it more in the way Plato’s philosopher sees democracy: not as the best form of political life, but as one of the few actually existing institutions in which something like the philosophical life can be pursued.

The modern professionalized university is far from the idealized intellectual community that my idealized image of the Committee on Social Thought represented. Still, if you want to find a place in the real world where something like that life can be lived, sociology is actually pretty great. Since I never expected it to be the best of all possible worlds, I do not worry too much about its failure to reach that lofty ideal. Hell, I spent more time in graduate school than I’d care to admit arguing with friends over whether it was the French Revolution or the rise of Christianity that marked humanity’s fall from grace! That’ll lead you to count your blessings to get to spend your days reading and talking and writing and thinking. Does that make me an esoteric Straussian writer? I’ll never tell (if you want a hint, find the exact middle of this post).

This attitude also means that I get a little worried when my friends who are frustrated with the state of the field want to fix it by making it more “scientific.” The reason is that, for my part, while I do a fair amount of work squarely in the scientific mode, and believe that when in that mode one should operate by its norms, I do it in part to justify humanistic endeavours that, from the positivist point of view, can look pretty weird and woolly, not to mention odd forays into other fields. I like to move between interpretations and the machines, and for that, we need to sustain the humanistic dimensions and connections of the field. This is one way that I feel I differ from another of my friends, Seth Abrutyn, in how we think about sociological theory.

This all leads me to have a bit of distance from the comings and goings and ins and outs of the contemporary tussles over the state of the discipline. That is furthered by the fact that I have read Andrew Abbott’s Chaos of Disciplines, in which he shows that academic disciplines historically have a fractal pattern in which basic dichotomies recur within fields and also periodically. One of the more basic fractals in sociology is that between reform and science, which reappears in many guises at many levels, and which will forever recur in ever new cycles.

I had seen evidence for this myself in a little project I started a few years ago pre-Covid, but set aside. I had assembled a large database of journals, to study their use of visualizations. One night it struck me that I also had all the text, so I played around with some basic topic modelling and even tried to interpret the results with a colleague and toyed with the idea of trying to publish it. But, as with many ideas, it fell to the side.

However, Codex resurrects such projects, and given recent discussion of big text analysis of social science journals, I thought I’d dust off the dataset and take a look. The goal is to get a wide angle view of the rise, fall, and persistence of different topics and methods, and thereby a better sense of how to situate the present. If anybody has any other types of analyses they want me to try, let me know, I might just do it!

A quick word on the corpus. I began with the full run through 2015 of three major journals: The American Sociological Review, The American Journal of Sociology, and Sociological Theory (I also have several other theory journals, not included). I also have a separate corpus of ASA presidential addresses, though I am not discussing those here. The texts were cleaned in the usual ways, removing bibliographies, appendices, repeated boilerplate, and obvious OCR debris. The result is not pristine, but it is more than adequate for a first pass at long-run topic trends. I’m also happy to slice and dice these in other ways anybody might suggest.

It also occurred to me that this was a chance not just to reconstruct the old database, but to extend it and turn part of it into a simple forecasting exercise. So I collected the articles published since 2015 and used the earlier corpus as a training set to see how well a basic model could anticipate what came next. I tried the same approach on the presidential addresses as well. With Codex’s help, that gave me a complete journal history through 2025.

The process of collecting the more recent journal articles, I must say, was quite pleasant. As I looked through the volumes, I found a whole bunch of pieces that I found very interesting. I often stopped and downloaded and read them midstream. There were a few quirky and fun ones, more than I might have expected, as well as contributions that spanned the whole range of abstraction: highly generalized accounts of phenomena such as interaction and exchange, detailed empirical research on topics such as value transformations through the life course, as well as rich case studies that opened up some corner of the social world to me. Some were not my cup of tea — it would be a strange thing if that were not the case — but more than enough were to make me feel very happy and proud to be part of this field.

The main method I report here is structural topic modeling, which identifies recurring bundles of words that tend to appear together across documents and then traces how prominent those bundles become over time and across journals. To estimate topic prevalence, I modeled temporal change while controlling for journal, article length, and broad historical period. To interpret the topics, I worked iteratively with Codex, going back and forth over exemplar articles and candidate labels. The resulting names are imperfect, and I am very open to revising them. For this first pass I settled on 36 topics, which I then grouped into a loose hierarchy to make the patterns easier to see. I’ll give headline results here, you can look at the end of the post for more detailed graphs and tables.

What emerged from the analysis was a story of ongoing successions, surges, stabilizations, and recurrences.

At the highest level, the most basic divide is between work that reflects on sociology itself, its purposes, methods, concepts, and mission, and work that investigates some substantive domain of social life.

The former was much more prominent in the field’s earlier years, especially before the consolidation of the modern journal system. Over time, the latter came to dominate, as sociology increasingly organized itself as a set of substantive subfields rather than as an extended argument about its own foundations. By the interwar period, the balance of the field had shifted decisively away from foundational self-reflection and toward substantive specialization.

Digging deeper into the “mode of sociology” discourse, it is interesting how much of early sociology was taken up not with today’s familiar substantive subfields, but with broad moral and civilizational reflection, along with debates about instincts, human nature, and the discipline’s own conceptual foundations.

By the mid-twentieth century, that discourse gave way to a more professionalized language of theory, epistemology, concepts, methods, and formal explanation. More specifically, the reflexive side of the field passes through a succession of idioms: early moral-civilizational and human-nature discourse collapses, mid-century discussion of role behavior, anomie, survey measurement, and formal modeling rises, and by the late period action-theoretical discourse becomes especially prominent, above all in Sociological Theory.

Even here, though, the pattern is not simple replacement. For example, variants of moral-civilizational discourse and discussions of human nature remain present, especially in Sociological Theory. When we look at the presidential speeches, we will also find a rise in a (revised form) of this type of discourse. Methodological discussion shows the same basic pattern. Survey research and formal quantitative modeling rose dramatically in the mid-century period, came down, and then stabilized at a long run equilibrium. This is one of the recurring rhythms of the corpus as a whole: emergence, rapid ascent, and then incorporation into the discipline’s long-run background.

Across the substantive domains, different topic families follow different temporal paths. Early reformist concerns around charity, relief, and welfare have receded, but the reformist impulse remains in later-growing areas, including protest, activism, gender inequality, immigration, ethnoracial classification, and health inequality, which expanded substantially and in some cases continue rising into the twenty-first century. The politics-and-state panel below makes this especially clear: early work on charity, relief, and municipal welfare fades almost to nothing, while later forms of reformist energy reappear in protest, contentious mobilization, and welfare-state social policy.

Similarly, topics covering gender inequality, race, segregation, immigration, and ethnic classification occupied 22% of the field’s discourse by 2015. This expansion marks one of the major substantive reorientations of the field, and much of it is likely to endure as a permanent part of sociology’s core agenda. Within that broader family, the late acceleration of immigration, incorporation, and ethnoracial classification has been especially dramatic — so much so that (as we will see) the forecast for these topics overshot their growth by a lot. These and the aforementioned rising areas (along with cultural production, which displaced religion) largely reduced the prominence of mid-century clusters around bureaucracy, mobility, occupations, schooling, and urban ecology, which remain important but more peripheral.

Taken together, the dominant temporal form in the corpus is rise, decline, and persistence. To get a clearer handle on that rhythm, I plotted the half-life of topics, that is, how long it takes for a topic to fall to half its peak share of the corpus. Topics that have not yet declined that far do not appear in the map.

The older reflexive topics often have long tails. Interest in theory, concepts, and explanation peaked early and declined only gradually over many decades. Many mid-century clusters show half-lives of roughly twenty to thirty years. This is why for those raised in the wake of that period they defined an era, but also a target for change, or the father to kill, as the case may be. Some topics had not yet halved by the end of the training period at all, including action-theoretical discourse, labor and political economy, exchange and reciprocity, status processes, and several family and inequality clusters. Race- and gender-related topics were still climbing.

My final question (for now) was whether these long-run patterns had any predictive value. I trained a simple model on the pre-2016 corpus and asked how well it could anticipate the shape of the decade that followed.

This table gives a rather granular view of the holdout exercise by looking directly at specific journal-topic series. Each row reports the average observed and predicted prevalence of a topic within a particular journal across the holdout years 2016–2025, so an observed value of 7.8% for ASR and “Health inequality, aging, and cumulative disadvantage” means that this topic occupied, on average, 7.8% of ASR’s annual topic mix in the holdout period, not 7.8% of the corpus as a whole. The most dramatic overshoot is AJS articles on immigration, ethnicity, and panethnic classification, which the model projected at nearly 38% but which actually averaged closer to 7%.1 Other overshoots appear in medical practice, health inequality, community ties, labor and political economy, and gender inequality. On the other side, the strongest undershoots are both versions of theory of action, especially in Sociological Theory, but also ASR (thanks Omar??!!)), where the topic remained much more prominent than the model expected. What emerges is a distinctly uneven picture: the model misses where some lines cool off faster than the earlier data suggested, and where others retain a force and staying power that the earlier trajectory did not fully prepare it to expect.

To see all this together, here is one final plot that shows the trends since 2015 compared to what came before.

Looking at this graph, if there is any message, it is that at many points in the history of Sociology, advocates of one style or topic or subject had some reason to believe that they were ascendent and had reached the peak of the field. And they had: for that moment. But disciplinary history does not move in a straight line, it is recurrent cycles and branches folding back on themselves, and opening out again.

And so I crack open my Simmel and Parsons, fire up Codex, and wait.

For readers who want the fuller picture, the figures below show the internal composition of each topic family, followed by tables listing the labels, keywords, and examples of exemplar texts.

How to read the appendix tables
Each table lists the primary and secondary family, the topic label, the most common words, the distinctive or FREX words, a short cluster gloss, and one or more exemplar texts.

Foundational, theoretical, and methodological topics

This panel breaks the reflexive side of the field into its main components. Early sociology gave substantial space to moral-civilizational reflection and to proto-sociological discussions of instinct and human nature. Over time, those gave way to more professionalized clusters around theory, explanation, methods, measurement, and formal modeling. Several of the older strands never disappear entirely, but persist at lower levels or survive more strongly in Sociological Theory. The accompanying table shows the topic labels, characteristic words, and exemplar articles used to interpret these clusters.

Politics, state, and public order
This panel shows a transition from early work on charity, relief, and municipal welfare toward a more differentiated political sociology. War and comparative-historical state order form a long-running strand. Electoral politics remains present at modest levels. Crime and legal control recur in waves. The most sustained late growth appears in protest and contentious politics, along with welfare-state spending and social policy. The table below identifies the component topics and the vocabulary anchoring each one.

Economy, work, and organizations
This family captures a major shift within economic sociology and organizational analysis. Mid-century work on organizations and administrative order rises early and then recedes. Labor, firms, and political economy become more prominent later, though they too eventually ease off from their peak. Meanwhile, organizational networks, niches, and market structure gain ground in the late period, as do comparative studies of global inequality. The table provides the specific topic labels, top words, and representative exemplars behind those trajectories

Inequality, family, and social reproduction
These topics show the decline of some classic mid-century clusters and the rise of newer inequality concerns. Class, occupations, and mobility peak earlier and then fall back. Schooling and educational attainment remain important but become less dominant. Gender, work, and labor-market inequality rises sharply and is still expanding late in the series. Marriage, cohabitation, and family demography also become more prominent. The table below clarifies how these topics were distinguished from one another.

Race, ethnicity, and incorporation
This area is especially concentrated in two strong trajectories: one around race, segregation, and inequality, and another around immigration, incorporation, and ethnoracial classification. Both rise over time, with especially sharp late-period growth. Together they account for a substantial share of recent sociological discourse. The table provides the keyword structure and exemplar texts used to identify these two clusters.

Culture, religion, and meaning
This panel shows that cultural discourse changes in composition over time. Sociology of religion declines steadily across the long run, while cultural production, media, and audiences grows strongly in the later decades. A smaller strand around military masculinity peaks in the mid-century period and then falls to a lower steady state. The table below lists the component topics and the terms that define them.

Community, demography, and health
This family combines older urban-demographic work with later-growing health and inequality topics. Urban ecology and demographic distribution grow into a central mid-century area and then decline. Friendship, participation, and community ties expand later on. Health inequality, aging, and cumulative disadvantage rises strongly from the 1970s onward, as do medical and clinical themes. The panel shows how one larger branch differentiates internally over time without collapsing into a single dominant topic.