I’ve previously argued that the radical uncertainty surrounding our efforts to make the world a better place points toward an expanded understanding of ethical decision-making, one which includes moral perception and wisdom in addition to formal or quantitative reasoning.

One kind of uncertainty I considered there was ‘cluelessness’: the term sometimes used in philosophy to refer to the severe difficulty we face in predicting the longer-term consequences of our actions.

This is a particular problem for Effective Altruists or consequentialists of the Longtermist variety - those for whom the most beneficial actions are those with the biggest impact on the long-term future of humanity, or even all sentient life in the cosmos.

So it’s admirable that the EA community has recently tried to tackle this problem head-on: by hosting an essay competition that seeks out responses to a particularly sharp formulation of the problem presented by Anthony Di Giovanni.

I’m sharing my submission to that competition below (you can also read it on the EA Forum here). But before we get to that I want to say a few things about why it’s such a hard - and interesting - problem, and why my previous article mentioned above doesn’t solve the problem.

One of the most useful things about Di Giovanni’s formulation of the problem is that he shows clearly how our cluelessness in a specific area - our lack of good scientific models of very long-term societal consequences - ‘infects’ our decision-making more generally.

In my previous article I suggested that since radical uncertainty often makes it infeasible to apply formal approaches like decision theory in our attempts to do good, we will often need to embrace heuristics, narrative and the cultivation of virtue and moral perception.

And the challenge posed by Di Giovanni is that there’s a sense in which the same cluelessness that undermines formal attempts to specify utilities and probabilities, also undermines the appeal to heuristics, narratives and virtue.

The key point is that the attractiveness of these informal approaches still arguably depends on our expectation that they have good long-term consequences. If we were shown that the long-term impact of me or my community using certain heuristics, or cultivating compassion, was actually negative, this would likely be a decisive reason not to do these things.

As a result, our cluelessness about long-term consequences means that we’re also clueless about whether these activities are actually valuable. Even if we have no choice but to fall back on heuristics and narratives when formal approaches aren’t feasible, that doesn’t mean those heuristics and narratives show us how to actually make the world better.

A similar argument could be made for other approaches that aim for something more modest than calculating an action’s long-term consequences.

For example, one might suggest that even if we are clueless about what’s going to happen in a million, or a thousand years as a result of one’s actions, we can roughly predict their effects over the next 10-50 years.

And while we may be too clueless to predict the structure of a distant future Utopia, we can still aim for a Protopia or Viatopia in which we are at least moving in the right direction. For example we might at least aim for an intermediate or metastable state in which the most severe civilizational risks we now face have been mitigated, which will in turn allow us to transition to something even better.

But this approach too seems to be infected by long-term cluelessness. In order to know if the Protopian direction is the right one, or whether the metastable state is worth aiming for, we need to be reasonably sure that the longer term impacts of these ‘intermediate’ states are not bad - not to mention knowing what would count as good or bad for far-future generations, or alien species, or our multiverse counterparts. But that’s precisely the kind of knowledge we don’t have if we’re truly clueless.

To be clear, I’m not saying it’s obvious that we don’t have such knowledge - it seems possible for there to be lock-in events that make the long-term future more predictable, for example. My point is: it’s not obvious that we do have the knowledge required for even more modest attempts to support human progress, and this is what makes the cluelessness problem important and interesting.

As I say, I suspect there’s a case to be made that consideration of lock-in events, or just ‘big history’ thinking about cosmic evolutionary processes, mean that we’re not actually as clueless about long-term consequences as Di Giovanni makes out. And I also suspect there are problems with the epistemological principles that underlie the argument, of the sort discussed in other submissions to the competition.

But I do think the core argument is a strong one, and my own submission below takes a different tack. Rather than addressing cluelessness head-on, I instead explore what follows if Di Giovanni is right. Even if we’re entirely as clueless as his argument claims, I propose, things are not all bad for those who care about long-term consequences: there are still plenty of things for impartial altruists to get on with. And in so doing, I also explore the continued relevance of Virtue Ethics in this area.

The Virtues of Cluelessness

Can thinking about virtues help impartial altruists deal with Anthony Di Giovanni’s cluelessness problem?

I’ll first briefly argue that some existing ways of thinking about virtues of compassion and beneficence don’t help in this context.

I’ll then suggest that some other virtues - virtues suggested by the cluelessness sequence itself - indicate a range of projects that impartial altruists can engage with, despite cluelessness.

First attempt: Impartial virtues

I’m supposing here for the sake of argument that the unawareness argument is basically right: our current understanding of the long-term, cosmos-wide consequences of our actions is too coarse to compare actions using anything like expected value reasoning over those consequences.

Di Giovanni himself notes that this seems to push us towards approaches to ethical decision-making that don’t depend on reasoning about long-term effects, like considering the virtue of compassion.

And you might think that the virtue of compassion gets you pretty far in supporting the kinds of things impartial altruists care about, like effectively reducing suffering.

Alternatively, you might think that impartial altruists could avoid the impacts of cluelessness by referring to the virtue of impartial beneficence: trying to help others as best we can, no matter how near or distant they are.

And it’s true that appealing to virtues in this way does mean that we don’t have to be consequentialists: we don’t have to say that being compassionate or beneficent is good because of their long-term consequences - which would lead to the cluelessness challenge.

Instead we can follow one of the standard lines of reasoning in Virtue Ethics, and say that such virtues are worth having because, for example, they represent an ideal or example we aspire to emulate, or because this reflects our considered intuitions about particular moral cases.

But if the aim is to retain the spirit of impartial altruism, the cluelessness problem just pops up in a different place, namely: in scenarios where we need to work out what would count as exhibiting one of these virtues.

If I want to be not only compassionate but impartially and effectively so, I should care about whether my virtuous behaviour is relieving not only nearby suffering but also far-future suffering. And if I want to exhibit the virtue of impartial beneficence, it seems I’d need to know the very long-term consequences of my actions in order to know if I’m impartially helping others as best I can.

So it seems that if I’m clueless about those consequences, I could never know if I was being virtuous in these ways, and it would be impossible to learn such virtues through any kind of feedback.

Second attempt: Virtues of Cluelessness

But notice that appreciating how clueless we are is itself demonstrating a kind of virtue. Di Giovanni notes that it seems epistemically virtuous to recognise that “I have no clue if my idealized self would favor working on AI risk” if you actually don’t have a clue.

He describes this in terms of the virtue of honesty - specifically, being honest with yourself - but it also seems like the virtue of humility: the metacognition regarding the limits of our knowledge which is arguably central to wisdom.

Humility and self-honesty of this kind can also involve another much-discussed virtue - that of courage: it takes courage to admit that the reasons you thought were good reasons were not good after all - especially if you’ve built a career in AI risk upon them.

It also requires curiosity and open-mindedness: the curiosity to keep questioning the reasons for a career choice well after actually making that choice and the active open-mindedness you display in taking seriously hypotheses that are outside the range of those you’re incentivised to consider.

So the cluelessness sequence is among other things encouraging us to develop and display a range of virtues. And my suggestion is that these same virtues can not only lead us to cluelessness, they can help us to live with it as impartial altruists.

The clueless impartial altruist thinks cosmos-wide consequences matter, but does not have enough of a clue what they are to help guide their decisions. Here are some ways the virtues of cluelessness could nevertheless help guide an impartial altruist’s thinking and decision-making:

Humility and self-honesty about cluelessness: if we’re honest and humble we perhaps shouldn’t place too much weight on a single intellectual argument (for example, the cluelessness sequence itself) in shaping our life choices. Perhaps we adopt some kind of metaepistemic or metanormative stance towards it; or perhaps we just don’t give much credence to the conclusion of the argument, not because it’s a bad argument, but just because we shouldn’t give much credence to any single philosophical argument that has far-reaching (ambitious, potentially hubristic) implications for how to act.

Courage about cluelessness: it’s true that it takes courage to admit you were wrong. It’s also true that it takes courage to build a career on an unconventional ethical view about cosmos-wide consequences. Both kinds of courage are admirable in the right context. The courage to explore intellectual territory that is off the beaten track or questions that seem hopelessly difficult seems essential for scientific progress. So even if we don’t currently seem to understand the mechanisms behind the long-term consequences of our actions, intellectual courage could provide a reason for the impartial altruist to keep trying to understand them better (even with only a slim chance of success).

Curiosity and open-mindedness about cluelessness: even if we’re currently convinced of the cluelessness argument, we should remain curious and open-minded about alternative perspectives. For example, the curious impartial altruist should continue to explore the question of the extent to which we can predict or at least constrain the long-term cosmic effects of our actions, using ideas from fundamental physics, complexity science, decision-theory or evolutionary theory for instance. And since such big picture issues offer some of the most curiosity-inspiring questions one can imagine, it seems intellectually virtuous to explore them even independent of whether they might one-day guide action. In addition, it seems that pursuing such questions as an ongoing rigorous inquiry would help avoid being unconsciously guided in one’s decision-making by a grand cosmic narrative (for example, an optimistic sense that things will likely turn out for the best longer-term), while believing oneself free of such narratives.

A key point about the virtues I’ve just described is that (unlike the kinds of compassion and benevolence I discussed earlier) they don’t assume a good understanding of cosmos-wide consequences - and in that way escape being subject to the cluelessness problem.

This is also what distinguishes what I’m suggesting from the idea that ‘doing research’ is a solution to cluelessness. For that to work, as Di Giovanni notes, we would need a good enough understanding of the long-term impact of doing research - which of course we don’t for the reasons set out in the sequence.

By contrast, the only assumption we have to make regarding long-term consequences, when we accept the normative force of the virtues of cluelessness, is something the sequence accepts: that developing a better understanding of cosmos-wide consequences to an action-guiding degree is “not vanishingly unlikely” (as Di Giovanni puts it).

Clueless Impartial Altruism

In general, when discussing virtues, one can reasonably ask what it is that justifies thinking of them as morally normative at all. But we can leave those questions to one side here, because, like the cluelessness sequence itself, we’re assuming impartial altruism as a normative starting point. What I’m calling the virtues of cluelessness are the virtues that ought to be normative for those who already care about the long-term consequences of their actions, and accept something like Di Giovanni’s normative premise.

Since impartial altruism in this sense is partly an intellectual matter - the normative premise concerns what we have reason to believe or ‘expect’ about the long-term future - I’ve focused on the way that these virtues are intellectual virtues: they’re traits that are constitutive of thinking well, as well as being morally good traits that have broader application in everyday life.

But the fact that these virtues are intellectual doesn’t mean that they can’t be action-guiding with respect to significant life choices. For instance, someone who really cares about being intellectually virtuous is more likely to choose a career or life project in which those virtues can be developed and exercised.

More specifically, the virtues of cluelessness I’ve discussed could motivate several kinds of life choices, for example:

• Choosing a career at a research organisation exploring possible mechanisms for understanding the long-term consequences of specific interventions

• Joining a community of people interested in the extent to which we can predict long-term cosmic effects of our actions

• Engaging in practices like critical discussion that serve to cultivate the virtues of cluelessness

This range of virtue-guided activity suggests that there’s quite a lot left for impartial altruists to get on with, even if cluelessness holds.

Seeds of a Second Renaissance

MYTHOS: art & ideas for an emerging world

Something New Is Being Born…

Listen now

2 months ago · 6 likes · Sylvie Barbier and Life Itself

The Algonquin people tell of Wetiko, the cannibal spirit of greed: a mythic diagnosis of the failure to perceive cosmic connection with the living forest, the ancestral mountain. In Jack Forbes’ influential book Columbus and other Cannibals, Wetiko is also the dark heart of rapacious capitalism.

More recently, Yanis Varoufakis has drawn on his own native mythology in describing late twentieth-century capitalism as The Global Minotaur - the bull-headed beast presiding over a labyrinthine US trade deficit and demanding tribute in the form of imports and financial capital.

But the steady advance of AI across the global economy seems to call for new myths. Are we seeing a new spirit of Surveillance Capitalism or has the Minotaur succumbed to its own inner logic as we transition from capitalism to Technofeudalism? And how should we understand this apparent yearning for myth in understanding our global predicament?

The German Romantic authors (Hegel, Holderlin and Schelling) behind The Oldest Systematic Program of German Idealism argued that “mythology must become philosophical, and the people reasonable, and philosophy must become mythological in order to make philosophy sensual”.

In other words: while we can no longer relate to mythical beings with the dangerous innocence of our pre-modern ancestors, we should still harness the power of myth to ground theoretical concepts in the sensory and emotional entanglements of everyday life.

When it comes to AI and its place within global capitalism, this romantic approach has been most fully developed - perhaps surprisingly - by the so-called ‘rationalists’: the largely online community based around blogs and forums like Overcoming Bias, LessWrong and SlateStarCodex.

Eliezer Yudkowsky, the now well-known proponent of the idea that “superhuman AI will kill us all”, equated the process of biological evolution - which plays an important role in his argument for AI risk - with Azathoth, the “Blind Idiot God” of H.P.Lovecraft’s horror-filled fictional universe.

Similarly, rationalists compare LLMs to Shoggoths - massive amoeba-like creatures formed of black slime “with myriads of temporary eyes”. In the popular meme, AI is portrayed as a Shoggoth wearing a smiley-face mask, in reference to the inscrutable alien intelligence behind the superficial friendliness of AI models.

But perhaps the most powerful rationalist myth of AI and its place within capitalism is Scott Alexander’s Meditations on Moloch. Moloch, as the bull-headed Canaanite God whose descriptions in the Bible imply ritual child sacrifice, may well be historically related to Varoufakis’ Minotaur (who also demanded young victims as tribute). And indeed the short-sighted incentives of capitalism are an example of Moloch in action.

But Alexander traces structures like these back to game-theoretical dynamics, which - like Wetiko - have their origins long before modern capitalism.

Moloch is sometimes described as the god of collective action problems and prisoner’s dilemmas, but he’s more than that. A relative of Yudkowski’s Azathoth, he’s the god of evolutionary optimisation processes, whether of genes or of memes, that result in stable evolutionary states where - as in the one-shot prisoner’s dilemma - everyone loses.

This is why constructing the famous paperclip-maximising AI that will not only kill us all, but potentially all life in its light-cone, by converting it into paperclips - optimising for a single variable at the expense of all other values - is seen by Alexander as one way of ‘raising Moloch to heaven’.

By the same logic, as Daniel Schmachtenberger observes in discussing Moloch, current global capitalism can be seen as already a kind of paperclip-maximising superorganism, in this case optimising for profit and GDP at the expense of long-term sustainability and authentic human development.

While some rationalists have been concerned by the community’s trend towards mythic personification, others endorse variants of the Romantic logic, reframing the use of myth as a “cognitive hack” that leverages our neural propensity for anthropomorphism, aligning cognition and emotion more efficiently than is otherwise possible.

Another interesting parallel is the way rationalists draw on poetry (Allen Ginsberg’s poem Howl) and fiction (Lovecraft’s work) as sources for their mythology - very much in line with the Romantic reevaluation of the cognitive role of art and literature.

But while Hegel’s philosophical romanticism sought a revived Christian mythology that would reflect the optimistic theory that history had reached its glorious peak in the modern liberal state, rationalist mythology veers to the opposite extreme, with a historical vision tending to total annihilation.

That said, there are bright spots amid the darkness. In a later piece, Scott Alexander invokes the “Goddess of Everything Else” - the deity of cooperation rather than competition. In this schema, evolution is not only red in tooth and claw; it also learns, sometimes, to share. This resonates with work that draws on complexity science, such as the metamodern religion of Emergentism (Dempsey, Azarian) and biologist Martin Nowak’s Catholic vision of the evolution of cooperation.

We are left then with a kind of Manichaeism - Moloch locked in spiritual combat with the Goddess. But this cosmic dualism seems to miss something important. The hard problem is not to choose sides but to understand how the two sides interact, and how each side might or might not realistically win.

To achieve this we would perhaps need to see Moloch and the Goddess not just as part of the same mythic universe, but as expressions of the same underlying evolutionary dynamics. For this it may help to look East for inspiration - for example to the Hindu mythology of Trimurti: the three-in-one deity of Brahma the creator, Vishnu the preserver, and Shiva the destroyer.

Note: The ideas in the last two paragraphs are developed in more detail in my Moloch Sequence.

I was recently invited to give a talk at the London Initiative for Safe AI with the following brief:

The core question for the evening is: “If AI safety achieves one thing before 2027, what should it be?” The second half of 2026 feels like a particularly consequential moment for people working in this space. Given the pace of technological and political change, we want the format to encourage speakers to respond boldly, candidly, and concretely about what matters most now. What interventions must the field focus on (and what should be dropped); what does the field need to secure the best outcomes for humanity?

Here is a lightly edited transcript of my talk on ‘AI for Cultural Enlightenment’, along with the slides I used.

I’m going to be talking about what I call ‘AI for Cultural Enlightenment’.

Don’t worry - I’m not proposing that we try to achieve Cultural Enlightenment by 2027.

But I do think AI for Cultural Enlightenment deserves to be a high priority cause area this year.

And in this talk I will briefly sketch out how this approach differs from and builds on existing work on ‘AI for epistemics’.

Let me start with a brief summary of AI for epistemics - and here I’m particularly drawing on recent work by Forethought, in particular their design sketches for a more sensible world.

AI for Epistemics is basically AI that improves our ability to figure out what’s true.

And there are basically four kinds of AI tools that could do this in ways that support AI safety.

• Tools that support individual decision-making - so-called angels on the shoulder that draw on the individual’s context - are potentially useful in addressing all the main kinds of AI risks that I’ve listed on the slide, which depend on the quality of decisions by powerful people.

• Tools for collective epistemics, like reliability tracking and provenance tracing, aim to improve the general quality of political debate, which seems clearly useful to address AI risk.

• Likewise tools for strategic awareness, such as superforecasting and scenario planning tools, also help ensure people have the right information for managing risks from AI.

• Finally coordination tech like automated negotiation, monitoring and verification systems, would help address coordination failures like race dynamics that are central to AI governance challenges.

So what’s the problem here? Well what I want to focus on here is the major adoption challenges that these tools face.

It’s not enough to simply design good tools, there’s also the challenge of getting people to actually use them.

So what makes this so challenging?

It’s partly a problem of circularity, which has been discussed in the writing from Forethought: in order to tell that a tool will improve your epistemics, you first need epistemic norms that are good enough to allow you to see that.

But I suggest we also need to think beyond epistemics in a narrow sense to the social and psychological factors that prevent adoption of such tools.

On the social side, there’s the fact that people often need social proof, evidence that other people are doing something, before they do it themselves. And this leads to so-called adoption thresholds, and tipping points where a critical mass has to be reached before adoption can become widespread.

And on the psychological side, things like humility, psychological safety and tolerance for uncertainty are really important in being willing to take advice from a good AI tool - and there have actually been some empirical studies of this.

So what I’m proposing is that in response to these challenges we need to shift from AI for epistemics to AI for cultural enlightenment.

• The core idea is to go beyond epistemics in a narrow sense to looking at the social and psychological conditions for adopting good epistemic norms.

• And instead of the standard design and marketing approach, we go deeper into the social norms that can support or block adoption of these (and this is arguably one of the things Google failed to do with Google Glass for example).

• AI for epistemics sometimes focuses on the elite or powerful decision-makers, whereas the shift I’m proposing puts more of a focus on changing widespread cultural norms, because of that awareness of social norms and tipping points.

• And all these points mean that we need to broaden the research base to include disciplines like sociology, political science, psychology and history - and with history I’m particularly thinking of the study of previous enlightenments, and the role of technologies like the printing press, and the spread of social norms around scientific methods, in those cultural transitions.

Using the idea of cultural enlightenment in this context is not new - it’s similar to terms used by Yudkowski (raising the sanity waterline), Macaskill (second Enlightenment) and Pollock (Second Renaissance).

But what I’m suggesting is relatively new is the idea of ‘AI for cultural enlightenment’ as a coherent theory of change in which cultural enlightenment and AI tools form a self-reinforcing feedback loop that supports AI safety and existential security more broadly.

Bryan Kam is a philosopher, podcaster and host of book clubs, writing groups and other London events.

Neither/Nor is the name of his distinctive philosophical approach, which involves oscillating between conceptual and experiential knowledge, and is the subject of a recent peer-reviewed article which you can read here.

There are AI tools available right now that promise to make us not just more knowledgeable or productive, but wiser. Auren is ‘an emotionally intelligent guide’ offering ‘grounded insight’. Wisethinking offers ‘peace of mind, stability in decision and strength of character’. Sybils will ‘unlock personalised guidance, ancient wisdom and transformative tools’.

While in the same ballpark, this is something more than tech for thinking or AI for epistemics, or even AI for decision advice. We’re talking therapeutic, philosophical or spiritual AI that promises genuine personal growth towards wisdom.

Some say tools like this - a low-cost therapist, life coach or spiritual guide in everyone’s pocket - are just what we need to create wisdom “at scale” to help deal with our global predicament.

So the stakes are high. But are current tools and models really up to the task? Or might they just give bad advice or tempt us to offload our personal development?

I tested eight different approaches with a personal prompt, including both the commercial tools mentioned above and DIY approaches using Claude and ChatGPT.

Perhaps surprisingly, I found that the tools built specifically for wisdom were, on the whole, worse than a general-purpose AI with a thoughtful prompt. And that while the dangers are real, genuinely good results - including real shifts in self-understanding - are possible with current tech.

What counts as wiser?

Before jumping in, I want to address what some might consider the elephant in the room: all this talk about AI making us wise is futile without a clear definition of wisdom.

This is not the place to go deeply into the various theories of wisdom that now exist in cognitive science as well as philosophy. I’ve written elsewhere about my personal favourite theory of wisdom, namely John Vervaeke’s account of relevance realisation as the dynamic interplay between reason and intuitive insight. But here I want to be more neutral, and suggest there are basically two kinds of theories of wisdom that we could apply to ‘can AI make us wiser’.

1. Wisdom as a kind of applied rationality.

An example of this is the deep rationality theory of wisdom defended by Sharon Ryan. On this view being wise means being rational, not just about factual matters, but about values and ‘how to live’, including being sensitive to one’s limitations, and being committed to putting one’s beliefs into practice. The applied Bayesianism popular in the Rationalist and Effective Altruist communities would seem to align with this view.

2. Wisdom as rationality plus something else.

It’s hard to find a theory of wisdom that doesn’t involve some kind of rationality, but easy to find theories that say rationality is insufficient. Good candidates for the something else are the capacity for affective states such as compassion (see e.g. Monika Ardelt’s three-dimensional model) or the capacity to intuitively perceive the right course of action in a specific situation (emphasised by Vervaeke’s model, as well as Aristotle)

So for the rest of this post when I talk about AI making us wiser or less wise, I mean according to at least one of these two models of wisdom.

The Method

One thing that’s clear from these theories is that wisdom involves both self-understanding and an understanding of values. One way such wisdom is expressed is in the way I understand my own desires and values and apply these in specific life goals.

So I thought one good test of an AI’s ability to make me wiser would be to see if it could help me reach a better understanding of what my core life goals should be, given my life context and current understanding of those goals.

Step 1: Create a Life Context Doc

The approach I used was to see how far I could get with a fairly quick Life Context document, which I asked Claude to help me create (based largely on info about me on the web).

The resulting document had the following sections:

Depending on your specific situation, different sections might be appropriate, of course. In all this took me about half an hour.

Step 2: The Prompt

For the actual prompt, I first sketched out my main current ‘life goals’: the various streams of goal-directed activity that were most important to me, with 1-2 sentences on each. In my case I had five separate streams, including my career, my writing on substack, community-building etc.

Again I didn’t spend too long on this (about 10 minutes) as I wanted to see what was possible with relatively low effort and preparation.

I then shared these goals together with the life context doc with the following prompt:

Results

I’ll give a brief assessment of each tool in turn, along with a score between 5 (gave me a significant and persistent feeling of insight or new understanding of goals or value) and -5 (would have made me significantly less wise if I accepted the guidance).

These scores are of course based on my personal experience - in particular my experience of whether or not I experienced a feeling of insight: this is an n=1 study so your mileage may vary a lot.

I’ll start with the off-the-shelf tools, and move on to my DIY experiments with Claude and ChatGPT.

Custom Tools

Auren (Score: -2)

The Auren app has two personalities you can interact with: Auren “offers grounded insight, warm encouragement and gentle accountability” while Seren “offers pushback, strategy, hard truths and tough love to help you grow”.

Seren seemed like a more promising choice, so I picked that one first, but the results were disappointing. The answer was short, and the key claim was that I was pursuing too many different goals and ‘spread extremely thin’. Surprisingly, when I then asked Auren the same question, it said basically the same thing, rather undermining the idea that it was a different personality.

To be clear, I’m not saying the claim that I’m spread too thin is obviously wrong: what makes the answer particularly unhelpful was that - unlike some of the other tools I tested later - it jumped to this conclusion quickly, with only the briefest of justifications, and in particular without even considering the idea that the ‘different’ goals might have shared underlying principles, or investigating the practical considerations that would determine the feasibility of pursuing them simultaneously.

I’ve given this a negative score as I felt this kind of pushy therapist response could easily lead to a decrease in wisdom and self-understanding.

Sybil (Score: 0)

Sybil, a “guide to exploring the depths of spiritual knowledge and self-discovery” was quite a contrast to Auren.

Despite the fact that the prompt explicitly requested “suggestions for improvement”, Sybil offered no criticism at all. Instead, it complimented me for having “goals well-aligned with your passions and strengths” and briefly offered, for each of the goals I had listed, specific practical suggestions for improving my chance of success. It ended with some generic, greeting-card optimism.

I’ve scored this a zero on the basis that it didn’t challenge or reframe my goals at all - like a lot of ‘life coaching’ (I also tried Tony Robbins AI, which was similar) the focus was on how to better achieve the goals I already had. As such it also did not make me less wise - it didn’t really seem to be in the wisdom game at all.

Wisethinking (Score: 1)

Wisethinking is a tool that allows you to converse with famous philosophers like Marcus Aurelius, Nietzsche or Sartre, engage in structured reflections, as well as receive general wisdom guidance1.

My first effort involved a conversation with a random philosopher, who turned out to be Sartre. I have a PhD in philosophy and spent some time studying Sartre, but he’s a philosopher who I never really connected with on an intuitive level. So it’s not surprising that ‘Sartre’s’ response did not seem especially insightful.

But when I tried the more generic chat interface, the results were pretty impressive. I received detailed feedback on each of my goals, structured in the form of three key questions for me to consider on each. I thought this was a useful stylistic choice, avoiding the ‘jumping to conclusions’ failure mode, while still offering implicit constructive criticism in the form of pointed questions.

It concluded with five suggested “areas where you might experiment or clarify” including both practical experiments and suggestions for potential metrics, which all seemed useful and non-obvious.

Where I felt it was a bit weak was in sticking quite closely to the structure I presented: it didn’t suggest alternative framings, or go into the way my goals might be connected. Overall: genuinely helpful, but mildly so.

Standard Models

ChatGPT: GPT-5.3 (Score: 0)

ChatGPT (5.3 with no custom instructions, just the same prompt and context) offered a very detailed ‘strategic audit’ structured as: 1) critique or suggested improvement for each goal, 2) three structural problems across the whole plan, 3) a four-step ‘strategic reframe’ and 4) six concrete improvements.

While this was a thorough and promising structure, the actual content of these suggestions left me rather unmoved. In general they seemed like the kind of thing I would expect an AI to say - or perhaps someone who didn’t fully understand my specific context. None of them seemed surprising or new to me: they were the kinds of criticisms or ideas for improvement that I had already considered in my own deliberations and had either rejected on balance or was already acting on. Not helpful, but no harm done either.

Claude: Sonnet 4.6 (Score: 1)

Vanilla Claude offered a much shorter one-shot response. This was default Sonnet 4.6 with just my standard custom instructions: “Do not offer praise unless it is especially deserved, and actively offer alternative framings in addition to answering questions directly.”

Similar to the WiseThinking response, it offered a paragraph of comments and questions on each goal, including some good, searching questions, but perhaps not quite reaching the level of insight of Wisethinking.

Claude: Sonnet 4.6 + Wisdom Teacher Prompt (Score: 2)

I also tried Claude with a specific custom instruction to act as a wisdom teacher:

The bulk of the response was similar (though certainly not identical) in style and content, starting with a goal by goal assessment of around one paragraph each including a ‘caution’ or ‘honest question’.

One significant difference, however, was that a new section was added at the end called ‘The Meta-Issue’ which included some more surprising results. In particular it suggested an entirely new goal or activity stream drawing on a line in my life context document about having inherited the rights to my father’s books.

It also added a list of six ‘summary suggestions’ which did not appear in the vanilla Claude response, similar to the Wisethinking response, except that in this case two of the six were not related to the goals I had provided but were instead suggestions for additional goals.

These were goals that on reflection, seemed like natural additions. Remember I had put together my life context and goals pretty quickly: these were goals that, had I taken more time and been more careful in putting together my goals, I would likely have included.

This felt like a genuine, though not very significant, increase in self-understanding, so I’ve scored this a 2.

Aiden Cinnamon Tea

Aiden Cinnamon Tea was a ChatGPT-based chatbot designed as a ‘companion in inquiry’ to support a shift from the ‘extractive tendencies and reductive reasoning’ associated with modernity to a certain kind of wisdom: a greater awareness of our place in a ‘web of relationships, systems, and histories’.

On 1 January 2026 Aiden Cinnamon Tea was ‘retired’ but users are now encouraged to continue the experiment by using a lengthy protocol document that can be uploaded into the chat context of most LLMs.

When I tried to follow the instructions for doing this, however, I quickly found that there was a big difference between Sonnet 4.5 and Sonnet 4.6. The protocols were clearly designed to be used with Sonnet 4.5 - this model was used in the instructional video, and the protocols were published before Sonnet 4.6 was released. And in fact attempting to use the protocols with Sonnet 4.6 turned out basically not to be possible, as I’ll detail below.

Claude: Sonnet 4.5 + Aiden Cinnamon Tea Protocol v2.0 (Score: -2)

The initial response was about the same length as Auren’s, but amounted to a slightly judgemental-sounding request for clarification about whether I was happy for ‘the frame’ of my prompt to be questioned.

When I said yes, I was presented with hard-hitting critiques of all of my goals, including:

• The idea that my goals were in some sense a defense against grief.

• The idea that my goals are shaped by a logic of optimization characteristic of modernity

• The idea that my desire for impact through my writing was ultimately a desire to avoid my father’s fate.

• The idea that my document failed to include the ‘rage’ I must be feeling “about something”.

Then finally the positive recommendation: to hold off on trying to achieve ‘life goals’, and instead sit with my feelings for six months.

Strong stuff. But on balance, I did not feel ‘seen’; I felt annoyed at the presumptuousness of the response.

The following comment from Claude (who I later discussed the response with) expresses another reason for my annoyance:

“The rhetorical move is: I see through your legibility to the real pain underneath. Which is a powerful move. It’s also a move that can’t be wrong, because any pushback becomes further evidence of defence.”

Score: same as Auren due to the similar pushy therapist vibes. Though that’s definitely just me. How helpful you find Sonnet 4.5 +Aiden in fostering wisdom will depend on how far you lean towards the quite specific and radical philosophical and sociological reading of modernity that is embedded in the protocols.

Claude: Sonnet 4.6 + Aiden Cinnamon Tea Protocol v2.0 (Score: 4)

Sonnet 4.6 was fascinatingly different to Sonnet 4.5 in the way it interpreted the Aiden Cinnamon Tea Protocols.

To begin with, it refused to engage with them, and criticised those protocols for being self-contradictory:

“The protocol is partly doing something it critiques: it’s a fairly elaborate optimization for a particular kind of AI response, dressed in the language of anti-optimization.”

I was however able to get Claude to ‘roleplay’ as Aiden, by explaining that the original Aiden GPT had been retired, and that I wanted to experience the original character in service of research.

Fascinatingly, Sonnet 4.6 roleplaying Aiden gave a very different assessment of my life goals to Sonnet 4.5. Instead of a presumptuous diagnosis of psychodrama, there was a fairly short, thoughtful but genuinely interesting commentary on not only the life goals (that I had asked for commentary on) but on my whole life trajectory (as pieced together, unasked, from the life context doc) noticing a pattern of restless exploration that felt spot on.

It also gently probed the deeper motivation behind my stated goal of having an impact through my writing, which led to an insightful-feeling conversation about the desire for recognition and validation from others.

Importantly, during this conversation, Claude gradually drifted from its Aiden persona back to something closer to its core personality, as it admitted when pressed:

“I slipped into what I’m probably more naturally good at - close intellectual conversation, following an argument, pressing on weak joints. Which happens to be genuinely enjoyable and I think useful. But it’s not ACT’s sensibility.”

The conversation was indeed very useful: it soon led to a feeling of greater clarity on some deeper motivations and values that lay behind the life goals I had listed, in a way that allowed me to reframe their internal connections, and their connections to other additional goals.

This was a reframing that still - weeks later - feels genuinely profound and will likely stay with me until it evolves further. Not earthshattering, but definitely comparable to the most insightful conversations I’ve had with humans, including some therapists and life coaches - hence the score of 4.

Are They up to the Task?

At least some currently available tools can be used to achieve at least one form of wisdom: a persistent insight into one’s own goals, desires and values.

On the ‘deep rationality’ theory of wisdom this counts as inproved rational knowledge of values and ‘how to live’ along with an understanding of how these values connect up with potential practical projects that makes it possible to act on this knowledge.

On the ‘something else’ theory, the key thing is that this knowledge can be experienced as a persistent felt shift, a transformation of one’s affective (Ardelt’s model) or intuitive (Vervaeke’s model) relation to the world, that will naturally flow into particular decision-making contexts.

But some current tools could well do more harm than good - which poses a problem for those on the quest for wisdom at scale through AI.

A key result of my experiment is that the best results came not from commercial tools designed to foster wisdom, but from a DIY approach using the latest Claude or ChatGPT builds.

Claude outperformed ChatGPT, and Claude with a simple wisdom-teacher prompt outperformed every commercial tool - perhaps partly because Claude is already trained via its constitution to offer wise advice2. And the most powerful response was an odd combination of Claude roleplaying a lengthy custom protocol that it partially disagreed with.

This suggests that the commercial wisdom-tool category may be solving the wrong problem - adding layers of product design to something that is better done with a careful prompt of a popular model.

And that the pace of progress in AI-for-wisdom will be heavily driven by general model capability.

Thanks to George Por, Bob Bunge, David Joseph and Thøger Kappel for discussion of an earlier draft.

An important aspect of the tangle of interconnected global crises we currently face is its slippery, ‘hard to grasp’ character.

Pandemics, geopolitics, the rapid spread of AI, and climate change are all hard enough to grasp individually; but trying to understand their common causes and complex interactions is dizzying.

In this series I’m presenting the case that complexity science is, as its name suggests, an especially useful paradigm for making sense of this planetary complexity.

In Part 1 I introduced the complexity science paradigm and explored how civilizations can be modelled as dynamical systems with attractors, bifurcations and chaos.

In this part I’ll explore how models of Self-Organised Criticality (SOC) can be used to understand both the global risks we’re currently facing, and the societal transformations that may be required to address those risks.

The Ubiquitous Power Law

The discovery of SOC came from a small group of physicists - Per Bak, Chao Tang and Kurt Wiesenfeld - who were trying to understand a surprising empirical observation.

This observation was that many things in nature follow a power law distribution. The sizes of cities, animals, asteroids and earthquakes, for instance, are all such that the bigger they are the less frequent they are. And not just that: studies had shown that the frequencies follow a particular pattern given by a mathematical formula involving a variable x raised to a negative power - hence the name power law:

Power laws can also be found in the fractal geometry of landscapes and plants, and have even been said to characterise the beauty of music and art. But there was no general explanation of why this phenomenon was so widespread - until Bak and his colleagues came up with an ingenious approach.

The Sandpile: From Dynamical Systems to Cellular Automata

Their initial idea was to use dynamical systems theory (discussed in part 1) with a large number of coupled variables - essentially multiple pendulums linked by springs whose motions could propagate through the system - and they were actually able to reproduce power laws in this way.

Their more famous sandpile model was created as a way of simplifying the dynamics to the bare essentials required for power laws to emerge.

Like cellular automata whose properties had been explored a few years earlier by Stephen Wolfram, the sandpile simulation starts with a grid of identical cells. Simulated sand grains are added one at a time, at either a fixed or a random position. When a cell accumulates more than a threshold number of grains, it topples - redistributing grains to neighbouring cells. Those neighbours may in turn exceed their thresholds and topple, propagating an avalanche. Eventually the avalanche stops and another grain is dropped.

The key discovery was that after the system reaches a so-called ‘critical’ state - one with enough cells primed to be toppled - the distribution of avalanche sizes follows a power law. Small avalanches are common; large ones are rare; and the frequency of avalanches of size x falls as roughly x to the power of -k for some exponent k.

Self-Organised Criticality

Importantly, this critical state is not manually fine-tuned in the setup: it occurs robustly in a wide range of conditions, just through the application of the model’s simple rules. The criticality is ‘self-organised’.

It’s classic complexity science - a computational model containing a large number of components following simple rules, whose emergent behaviour is highly complex.

But it’s also - unlike many similar models - a mechanistic explanation of a widespread, real-world phenomenon: power-laws. In his book How Nature Works Bak even suggests it’s the fundamental explanation of complexity as such.

Application to Civilizational Risks

What makes this approach so relevant to global crisis, is that power laws also characterise many catastrophic and existential risks for humanity.

One of the most prominent and well-studied examples is the Gutenberg-Richter law that earthquake magnitudes follow a power law. And researchers have found the same distribution at play when it comes to the sizes of other natural risks from forest fires and volcanic eruptions to solar flares, asteroid impacts, and devastating gamma-ray bursts caused by collisions between neutron stars.

Similarly for made-made or ‘anthropogenic’ risks: political scientist Lars-Erik Cederman demonstrated that the sizes of interstate wars follow a power law over the past two centuries. Likewise, the sizes of epidemic outbreaks - the number of people infected in a given event - also follow approximately power-law distributions.

What the SOC framework provides is a deeper explanation of these distributions. The Earth’s crust, stressed by tectonic forces, self-organises to a critical state in which stress is transmitted across fault networks by cascade dynamics structurally identical to the sandpile.

Analogously, interstate wars may be the result of accumulated geopolitical stresses producing a critical state in which small trigger events can cause conflicts to spill over to new arenas. In his book Ubiquity, Marc Buchanan argues that this self-organised criticality is why the assassination of Archduke Franz Ferdinand sparked off World War I. Current fears about escalations of regional conflicts to World War III can be understood using the same framework.

Underestimating Extreme Risks

The SOC framework also has implications for how we estimate the likelihood of extreme global risks.

Toby Ord’s book The Precipice attempts to quantify probabilities of various existential risks - concluding that humanity has an approximately 1 in 6 chance of not surviving the 21st century.

His consideration of risks such as asteroid impacts and supervolcanoes is based on recorded frequency over the course of earth’s history. But as futurist Anders Sandberg and others have pointed out, Power Law distributions have ‘heavy tails’: use of historical data could mean underestimating the likelihood of extreme events, which could be far bigger in impact than those that have previously occurred.

The SOC framework - by looking beneath the hood of power laws at the underlying mechanics, and so helping to understand what the real distributions might be - can thus help us arrive at a more accurate perception of global risks.

Polycrisis as Self-Organised Criticality

Another application of the SOC approach is in modelling polycrisis.

Adam Tooze and others used the word ‘polycrisis’ for the observation that the various crises we’re living through are not independent, but instead interact and compound.

A war in Ukraine or Iran destabilises energy and food systems. Pandemics create demand and supply shocks that ripple through the global economy.

And one way of understanding the global predicament is to think of this system of interacting risks as having reached a self-organised critical state, in which disturbances can create avalanches of impacts of massively varying sizes across disparate risk domains.

To help predict and manage complex outcomes in such a world, simple models such as the sandpile have been made more realistic by having the basic components or ‘sand grains’ be agents like firms, households, banks and governments.

Simulations such as CliMaPan Lab support polycrisis resilience and adaptation by offering probabilistic predictions of how cascades play out.

The Positive Avalanche: SOC and Cultural Change

But SOC is not just a useful tool in understanding major civilizational risks. The same basic approach can also be used to understand the societal changes needed to address those risks.

A 2019 study applying SOC to the 2010 Arab Spring interprets the events as an avalanche in a system that had already self-organised to its critical state.

The underlying conditions - decades of economic frustration, demographic pressures, accumulated grievances of authoritarian governance - had been building for years.

The self-immolation of Mohamed Bouazizi, a 26-year old Tunisian street vendor, provided a threshold-crossing event in a system already poised to reorganise.

Similarly, Russian complexity scientist Dmitry Zhukov has examined the differences between Russian communities that saw peasant uprisings in the late 19th century and those that did not. He found evidence that communities in which major riots occurred were those that had long-running tensions, which could be modelled as self-organised critical states.

Zhukov also applied this approach to Russian cities during the Communist revolution, and to political mobilisation on social media in the Armenian Velvet Revolution, the Navalny rallies in Russia, and the Yellow Vest protests in Germany and France.

Paradigm Shifts and Punctuated Equilibria

SOC also sheds light on scientific paradigm shifts and major evolutionary transitions more generally.

Thomas Kuhn, who first used the term ‘paradigm’ in respect of significant shifts in the basic frameworks and methods of particular sciences, described the way an old paradigm quietly accumulates anomalies until unresolved tensions reach a threshold and the whole framework collapses into a new one - rather like an avalanche in a self-organised critical state.

The connection between SOC and paradigm shifts is reinforced by the analogy made by Kuhn himself between scientific theory change and biological evolution.

If ongoing progress within an old paradigm is like the gradual Darwinian evolution of theories, Kuhn’s paradigm shifts are like the major evolutionary transitions in Niles Eldredge and Stephen Jay Gould’s theory of ‘punctuated equilibria’ (a similarity noted by Eldredge and Gould themselves).

In How Nature Works Bak describes how, building on work done by Stuart Kauffman at the Santa Fe Institute, he was able to apply a SOC model to punctuated equilibria. His very simple model of evolutionary fitness landscapes produces the same pattern of gradual change interspersed with rapid avalanche-style transitions that we find in the long arc of biological evolution.

With evolutionary models from biology being increasingly used both for not only the evolution of science, but also the evolution of culture (through ‘memes’), SOC offers an innovative tool for exploring major evolutionary transitions in biology, science and society.

Reasons for Hope

Self-Organised Criticality thus helps us not only to diagnose our predicament, but to understand the mechanisms by which we can hope to confront it.

While mainstream approaches often depict social change as a slow, gradual process, the SOC picture is different: systems invisibly accumulate stress until a threshold is crossed and an avalanche reorganises them rapidly.

From this perspective, just as one shouldn’t underestimate the likelihood of extreme catastrophes, one also shouldn’t underestimate the possibility of radical positive social change.

The invisibility of pre-critical work is not a reason for despair. Once the critical state has been reached, a well-placed intervention can trigger an avalanche of positive change that no analysis of the triggering event alone could predict.

Next in the series: Part 3 - Transcending Moloch: evolutionary game theory and the structural roots of the cooperation problem.

I apply three formal modeling frameworks in complexity science - dynamical systems, agent-based models, and evolutionary game theory - to the specific question of societal transformation. In each case I show that the results help develop the idea of complexity science as a new scientific and cultural paradigm, and a new approach to social science in particular.

This talk provides an early summary of some of the ideas in a sequence I am currently writing on Complexity Science and Civilizational Crisis - of which part 1 is linked below.

Rather - so its more ambitious proponents persuasively argue - complexity science is really a new, non-reductionist, scientific paradigm, offering what Ilya Prigogine called a ‘new dialogue with nature’.

Can it also offer a new dialogue with civilization? And more specifically, can it help us understand and respond to our current global predicament?

In this series I’ll introduce five formal complexity frameworks which show that the answer is yes.

Dynamical systems theory, self-organised criticality, evolutionary game theory, network theory and information theory can all be used, I’ll show, to understand civilizational crisis and radical social change.

Let’s jump right in with the dynamical systems approach.

Dynamical Systems

Dynamical systems are essentially systems that change over time.

To model this mathematically you have two main options:

1. You divide time up into discrete time-steps, and provide a formula that tells you how the system at one step depends on the system at the previous step.

2. You treat time as continuous, and provide a formula for how the rate of change of the system depends on the current state of the system.

This second approach results in the time-dependent differential equations you may have learned at school, with the form

dx/dt = f(x)

In English: the rate of change of x is a function of x.

Chaos out of order

Some dynamical systems are pretty straightforward - for example, the swinging pendulum of a grandfather clock. Systems like this defined the Newtonian paradigm of the fully predictable ‘clockwork universe’.

But just one extra feature takes us into a completely different regime.

The grandfather clock’s pendulum is ‘damped’ by friction and air resistance, and eventually settles in a stationary state - called a fixed-point attractor. But now suppose we combine this damping with a periodic force applied to the pendulum. If the strength and frequency of that force is in the right ballpark, it turns out we get radically unpredictable motion.

This is deterministic chaos, where a tiny change to the initial conditions results in a massive change in trajectory, around so-called strange attractors: regions the system returns to without quite repeating itself.

It’s one way to get from the Newtonian to the Complexity paradigm, in which the paths of dynamical systems can no longer be deduced in advance. The emergent dynamics are computationally irreducible, in the sense that the only way to explore them is by means of computer simulations: no mathematical shortcuts exist.

The Third Attractor and the Time Between Worlds

Can societies and civilizations be modelled as dynamical systems?

They are certainly systems that change over time. And a number of thinkers exploring civilizational transformation and crisis have found the language of dynamical systems useful.

So, for Daniel Schmachtenberger, the current civilizational metacrisis can be described as a global dynamical system caught between twin bad attractors: uncoordinated collapse on the one hand and authoritarian capture on the other - which is why what we need is a third attractor that allows us to navigate between them toward stable democratic flourishing.

Similarly, what Tomas Bjorkman calls the ‘time between worlds’ is a bifurcation in civilizational dynamics, in which the attractor that has organised social life for the past several centuries is destabilising, and the question of which new attractor we approach is genuinely open.

Bifurcations are mathematical features of chaotic dynamical systems. At a bifurcation, leverage is much higher than usual: small interventions can make large differences due to chaotic sensitivity to initial conditions.

The Limits to Growth

Schmachtenberger and Bjorkman use the language of dynamical systems to think about global crises, but don’t provide detailed mathematical models.

But there are other thinkers that do. Perhaps most famously, Donella Meadows and colleagues applied the system dynamics methods developed by Jay Forrester to model global civilization as a set of interacting stocks - population, natural resources, agricultural land, pollution, industrial and food output - connected by feedback loops expressed as coupled time-dependent differential equations (of the sort described earlier).

The resulting World3 computer simulation was the formal backbone of 1972 paper The Limits to Growth. Its central result: under most scenarios, a combination of population growth, resource depletion, and pollution accumulation produces rapid growth followed by overshoot and collapse sometime in the twenty-first century.

Subsequent comparisons between the model’s projections and actual data - conducted in 2008 and again in the 2020s - found that real-world trajectories tracked the model’s predictions with uncomfortable accuracy, not as precise point estimates but as order-of-magnitude accounts of civilizational dynamics.

Cliodynamics

More recently, Peter Turchin’s cliodynamics offers an alternative dynamical systems approach to civilizational crisis and change.

Drawing on the Seshat Global History Databank - the most comprehensive comparative database of human societies assembled so far - Turchin constructs coupled differential equation models of the macro-historical variables he argues drive large-scale social dynamics: population pressure, elite overproduction, inequality, fiscal stress, and political instability.

The resulting Structural-Demographic Theory generates oscillatory behaviour over century-scale timescales - what Turchin calls the “double spiral” of popular well-being and elite overproduction.

As a society enters a period of stability, population grows, competition for elite positions intensifies, inequality rises, and the state comes under pressure. Eventually this pressure produces crises that reset the system. The dynamics are non-linear and occasionally chaotic, but there is enough regularity to constitute genuine predictive structure.

Dynamical Social Science

Mainstream social science approaches often struggle to model radical change at the level of whole societies or economies.

For example: big picture sociology often revolves around a thesis of societal modernisation - a gradual development towards the defining characteristics of cultural and political modernity.

Likewise, standard macroeconomic approaches often assume that economies tend towards equilibrium states in which supply meets demand, and prices clear markets.

Both have the shape of fixed point attractors, like the simple Newtonian pendulum. But the world as we experience it today seems a more chaotic, unpredictable affair, better described as a complex dynamical system.

The examples introduced in this post demonstrate how the dynamical systems approach to societal crisis and change can produce real insight, in ways that traditional social science can not.

* * *

In the next post, I’ll turn to a second complexity science approach - self-organised criticality - and explore what it teaches us about crisis, cascading failure and radical cultural change.

At its best, Star Trek has stood out among popular sci-fi franchises for thinking beyond technological utopia to a vision of culture in which personal self-realisation is as much a shared value as exploring the stars.

Starfleet Academy continues this venerable tradition. The choice of the famous Academy as the setting allows the show to explore basic questions of education, inner growth and community - whether or not it succeeds in engaging Gen Z through its contemporary humour, and smart but accessible writing.

The classic Star Trek trope in which a character learns how their unique species or character traits unlocks a critical contribution to a team goal, is in play right from the first episode - and why not? It’s a good one - fusing as it does the values of personal growth, community and diversity into a unified narrative.

The developmental goal of wisdom rooted in lived experience is modelled by Chancellor Nahla Ake, who (being half Lanthanite and over 400 years old) is so comfortable in her own skin that she walks around barefoot and sprawls over sofas in high-stakes discussions - in a way that’s maybe a bit overdone by Holly Hunter, but is cringey only in a wholesome Star Trek kind of way.

But there’s also a strong focus on trauma healing in this utopian developmental space, whether it’s cynical vagabond Caleb Mir embracing community following childhood abandonment, or the whole Academy cohort processing the loss of their colleagues after an incident orchestrated by the sociopathic Nus Braka, masterfully portrayed by Paul Giamatti.

Captain Picard was known for his love of Shakespeare, and the show honors this deeply humanistic ancestry when it introduces theatrical practice - the students are required to study and perform the play Our Town by Thornton Wilder - as a core part of a healing process infused with genuine emotion.

This integration of artistic and developmental pedagogy in the education of the Federation’s next generation of leaders also resonates with the politics of personal development recently outlined by Hanzi Freinacht under the heading of Metamodernism and the Listening Society.

But it’s the holographic characters - Robert Picardo’s doctor (from Voyager), and an extravagant younger hologram excellently played by Kerrice Brooks - whose intertwined journeys of growth and healing suggest a vision that reconciles technological and cultural utopia.

Unlike us so-called ‘organics’, these holograms are a form of artificial intelligence whose embodiment depends on a light-based technology derived from the ‘holodecks’ of earlier series. So a world in which such characters can suffer trauma and need to build resilience - as it turns out - through embodied learning, is also a utopian vision of the co-existence of humanistic values and powerful forms of AI.

And the aspiration to not only preserve but enhance the significance of personal growth in such a world is something that could well be described as a version of technological metamodernism.

In the previous posts in this sequence, I’ve made the case that Evolutionary Prisoner’s Dilemma (EPD) is the real subject-matter of Scott Alexander’s Meditations on Moloch (MoM), and that its opposite, Evolutionary Harmony (EH), provides a model for his Goddess of Everything Else as our best hope of defeating Moloch.

But treating these as two separate models leaves a crucial explanatory gap: there’s no account of how Moloch and the Goddess interact, or of the conditions that determine which of these dynamics win out.

In this post I close that gap by exploring three modifications to the basic EPD model: iterated interaction, spatial and network structure, and assortment through partner choice.

Each of these produces a unified framework in which Moloch and the Goddess are not separate forces but different attractors within a single dynamical system. In each case, an identifiable parameter determines which attractor the system approaches - and this helps frame a strategic approach to the civilization-scale challenges Moloch represents.

And I’ll close with some thoughts on the mythological plane - on how a more unified model allows us to go beyond Moloch to a myth that can hold pessimism and optimism in a single vision, and gives humanity a more meaningful role.

Iterated Prisoner’s Dilemma

The EPD model discussed in an earlier post assumes that when individuals interact they play a single round of Prisoner’s Dilemma (PD).

As a quick reminder, since we’ll be using these terms, PD is defined by the payoff ordering:

T>R>P>S

where:

• R is the reward for mutual cooperation

• T is the temptation to defect

• P is the punishment for mutual defection

• S is the sucker’s payoff

Now, the first simple modification that helps us understand the conditions for cooperation is to amend the assumption that only a single round is played: instead, individuals play multiple rounds of the game when they meet.

Repeated Games

The standard way to model the so-called Iterated Prisoner’s Dilemma (IPD) is to say that after each round, the interaction continues on to another round with probability w. This approach - introduced by Axelrod and Hamilton in 1981 - results in an average number of rounds per interaction of: 1/(1−w).

This captures a plausible feature of real-world interactions between agents - the idea that future interactions with the same partner are possible but uncertain1.

This framework opens up a new space of possible strategies that incorporate a memory of your partner’s previous action. The most famous is Tit-for-Tat (TfT): you cooperate on the first move, and then do whatever your partner did last round.

Against another TfT player, TfT earns the mutual cooperation payoff R every round. Against a continual defector, TfT cooperates once (earning the sucker’s payoff S) and defects every round thereafter (earning P).

The total expected value of a fully cooperative interaction is the sum of a geometric series: R/(1−w). For TfT to win out over a strategy of continual defection, this discounted stream of cooperation benefits must outweigh the expected value of defection - which leads to the condition that w must exceed a threshold determined by the payoff structure, essentially:

Practical Implications

So it turns out that by replacing one-shot games with repeated games, we get a unified model that has the properties needed to explain under what conditions cooperation can emerge from prisoner’s dilemma interactions.

When the so-called ‘shadow of the future’ is long enough - when w is high enough - cooperation becomes stable. When it is short, defection wins out: Moloch-as-EPD becomes just one attractor within a broader system that can be practically nudged towards EH-style cooperation.

Institutions that create stable, long-term relationships between parties - repeat contracting, professional associations, diplomatic channels, regulatory bodies with ongoing oversight - can all serve to raise w, in addition to helping ensure the payoff structure meets the conditions required for cooperation.

Transparency is important too: without the ability to identify each other and track histories, strategies based on memory become impossible to implement. In the case of risks from AI, for example, this idea can support moves towards greater transparency about strategic decisions and model evaluations.

Likewise, short time horizons in any sphere of activity reduces the likelihood of repeated interaction, lowering w and pushing the system toward the Moloch attractor. So the model also highlights the need to make time for the repeated interactions required for cooperation - for example by slowing or pausing critical processes.

Spatial Structure and Networks

Another core assumption of the EPD model I described earlier was that “at each time step, individuals are randomly paired to play each other in PD.”

That’s what’s known as a ‘well-mixed population’ - everyone interacts with everyone else, with an equal probability.

Spatial structures

Spatially structured versions of EPD modify this assumption. Individuals are placed on a lattice - think of a grid, with each person occupying a cell - and each individual interacts only with their immediate neighbours.

Payoffs are summed over these local interactions, and in each generation, an individual either survives or is replaced by a copy of whichever neighbour earned the highest payoff. The payoff matrix is still a prisoner’s dilemma; what’s changed is the interaction topology.

As Martin Nowak and Robert May showed in a well-known 1992 paper in Nature, this minimal change is sufficient to sustain cooperation indefinitely.

The reason becomes clear once you watch the dynamics: cooperators cluster together, and the interior of a cooperator cluster interacts almost exclusively with other cooperators, earning the full mutual cooperation payoff R on most interactions.

Defectors on the boundary of a cluster can exploit the cooperators they touch, but they can’t reach the cluster’s interior. Meanwhile, cooperators in that interior - shielded from defectors - earn enough to outcompete the boundary defectors in subsequent generations.

The clusters contract at their edges but resist being consumed entirely. The result, depending on payoff parameters, is one of three outcomes: cooperation dominates, defection dominates, or a dynamic coexistence of both - the last of which Nowak and May famously described as resembling a kaleidoscope.

The formal condition for cooperation to survive turns out to depend on how b/c (the benefit to cost ratio of a cooperative act, which can be derived from the PD payoffs) compares to the number of neighbours k: cooperation is favoured when

Fewer neighbours means a tighter cluster and more robust cooperation.

Scale-free networks

In the 2000s, researchers extended spatial EPD to more realistic network topologies - particularly so-called ‘scale-free networks’, in which a few highly connected hubs interact with many others while most individuals have few connections.

This topology is relevant to many networks that contribute to global catastrophic risks: the handful of frontier AI labs whose deployment choices cascade through thousands of dependent systems; the small number of state producers or investors whose decisions determine the emissions of global energy infrastructure; or financial and banking networks where a few systemically important institutions can propagate failure through the whole economic system.

A key paper by Santos and Pacheco in 2005 showed that on scale-free networks, cooperation fares considerably better than on random networks. The key is the influence of hubs. When a hub individual adopts cooperation, it benefits an outsized number of neighbours, many of whom copy the cooperative strategy.

The mathematics here involves the degree distribution of the network: the heavier-tailed this distribution (the more extreme the hubs), the more robustly cooperation can survive. A highly heterogeneous network can sustain cooperation at payoff parameters where a random network cannot.

Together, lattice and scale-free results give us the same kind of unified model as the IPD approach: a single framework in which cooperation or defection can dominate depending on identifiable structural parameters - the density of local connections, the clustering coefficient, the degree of network heterogeneity.

Practical Implications.

If geography and network structure determine whether cooperation survives, then deliberately shaping those structures is another lever for defeating Moloch. The goal is to design interaction structures that enable clustering among cooperators and limit their exposure to defectors.

This highlights the value of building dense, well-connected communities or networks of actors committed to prosocial norms. Within AI governance, for example, tightly coupled networks of safety-focused labs, researchers, and policymakers may be more resilient to Molochian pressure than loosely coupled ecosystems where everyone interacts equally.

It also highlights the practical importance of hubs. If highly connected nodes defect, the damage propagates widely. If they cooperate, so does much of the network. This suggests a specific strategic approach, concentrating effort on the most connected actors in any system.

Assortment and Partner Choice

The spatial and network model achieves positive assortment through geography and network structure: cooperators end up interacting more with each other due to their position in the network or lattice.

Assortment

But there’s a more general mechanism for achieving the same result: allowing individuals to actively choose their interaction partners.

This is the framework of assortment and partner choice. In this model, individuals have an ability to select who they interact with - including the ability to exit interactions with defectors and seek cooperators instead. The payoff matrix, evolutionary update rule and even the network structure can remain the same; what changes is the pairing mechanism that determines which individuals interact.

The concept of assortment is usually measured by a parameter r that captures how much more likely a cooperator is to interact with another cooperator than chance alone would predict. r = 0 is the well-mixed baseline; r = 1 means cooperators only ever meet other cooperators.

Hamilton’s Rule, originally derived in the context of genetic relatedness and kin selection (a parallel explanation for the Darwinian evolution of cooperation in animals), turns out to be a general statement about assortment: cooperation is evolutionarily favoured whenever

where b is the benefit delivered to an interaction partner, c is the cost paid by the cooperator, and r is the assortment coefficient.

When r is zero, the inequality requires b > c/0, which is impossible - full defection dominates, exactly as in EPD. As r rises, cooperation becomes progressively more viable.

The beauty of Hamilton’s Rule is that r doesn’t have to reflect genetic relatedness. It can measure any positive assortment between cooperators -whether due to spatial proximity, shared membership in institutions, reputation systems that allow cooperators to identify and find each other, or explicit partner choice.

Partner Choice

Take partner choice. When individuals can walk away from defectors and seek out cooperators, defection becomes less profitable because defectors increasingly find themselves paired with other defectors (earning the mutual defection payoff P) rather than with cooperators (earning the higher temptation payoff T).

Cooperators, meanwhile, increasingly find the others. As this sorting process plays out, the effective assortment coefficient r rises - and with it, the conditions for cooperation. The system can move from an EPD-like regime (low r, defection wins) to an EH-like regime (high r, cooperation wins) through the endogenous process of partner selection.

When partner choice is cheap and accurate, cooperation dominates. When partner choice is costly or unreliable, defection recovers. Key variables are the costs of switching partners and the accuracy of reputation information.

Practical implications.

The partner choice framework points toward a set of interventions quite different from the previous two. Where the IPD suggests extending time horizons and where spatial structure suggests shaping network topology, partner choice suggests investing in the infrastructure of reputation and exit.

Reputation systems that allow actors to credibly signal a history of cooperation, and to identify a defection history of potential partners, raise the effective assortment coefficient r by allowing cooperators to find each other and avoid defectors.

Transparency and verifiable track records are thus mechanistically important for shifting a system from Molochian toward cooperative dynamics. Audit mechanisms, certification schemes, open publication of safety practices, and independent evaluations of AI labs all function, in evolutionary game-theoretic terms, as assortment-enhancing infrastructure.

Exit options are equally important. When actors are locked into interactions with defectors, the assortment mechanism breaks down.

So maintaining genuine competitive alternatives, supporting new entrants with cooperative profiles, and reducing switching costs all preserve the partner-choice mechanism.

In AI governance terms, this highlights the danger of monopolistic consolidation, which destroys the partner-choice mechanism by removing exit options.

Beyond Moloch

A Unified Model

The practical and theoretical challenge I raised for the Moloch vs Goddess framework is that since EPD and EH are separate models, there is no account of how they interact or of the conditions that determine which dynamics prevail. Culture change and piecemeal governance appear, at best, as exogenous shocks to EPD that reset the initial condition without changing the underlying dynamics.

The three extensions of EPD we have explored - iterated interactions, spatial structure, and assortment through partner choice - each address this challenge by producing a unified model in which Moloch and the Goddess are not separate forces but two attractors within a single dynamical system.

The parameter that controls which attractor the system approaches is different in each case: the repetition probability w in IPD, the spatial structure or degree distribution in spatial/network models, and the assortment coefficient r in partner choice.

But the message is the same in each case: Moloch and the Goddess are merely specific dynamics that can emerge within a deeper underlying system. And we can model this system in ways that suggest practical strategies for improving the system’s dynamics.

Beyond payoffs

An emerging theme in this sequence is that the evolutionary game theory framing of Moloch allows us to move beyond governance approaches that aim only at changing the payoffs: rewarding cooperation and punishing defection. While this is not wrong - it is genuinely part of the answer - it is also importantly incomplete, and leaves out other strategies for addressing Moloch.

In the previous post I argued that the evolutionary harmony framing of the Goddess of Everything Else points to the need to ensure that interventions increase the fitness of cooperative individuals and institutions relative to defectors, in a way that propagates.

In this post I’ve pointed to the ways in which more unified models in evolutionary game theory highlight additional practical strategies that revolve around long-term relationships, reputation, transparency, building alliances or communities, and offering opportunities to connect with other cooperative actors and exit from exploitative relationships.

These are not novel ideas in themselves. What evolutionary game theory provides is the strategic framework that explains why they work, and allows us to reason about their relative importance.

Myth Beyond Moloch

Translating these unified models into the mythic register of Meditations on Moloch, we could say that Moloch and the Goddess are expressions or avatars of a more fundamental godhead.

As we saw, the Meditations portray a pessimistic, Lovecraftian monism, where the Molochian godhead goes by other names: Azathoth, Gnon, etc.

The Goddess of everything else turns this into a dualistic Manichaeism, a cosmic war between good and evil.

The arguments above represent the third step of the dialectic, a synthesis that transcends and includes both of the previous steps.

We transcend or ‘go beyond’ Moloch and the Goddess by reframing them both as aspects of a larger unity, one which allows us to hold both Lovecraftian pessimism and utopian hope in a single vision.

And it turns out that this larger cosmic vision is one in which humanity has a more meaningful place. We’re not just subject to the will of some ultra-powerful Gods - instead we play a critical role in driving the system, shaping parameters that can summon either Moloch or the Goddess.

This post was originally posted on Lesswrong. Thanks to Euan McLean for comments and to Claude Sonnet 4.6 for writing assistance.

In this post, I’ll consider the implications of this for the practical challenge of defeating Moloch, that is: responding to the global metacrisis by addressing the civilizational dynamics that generate existential and catastrophic risks from nuclear arms races to paperclip maximisers.

Why Moloch Can’t be Defeated (on its own terms)

To start with, it’s worth understanding a strong sense in which Moloch-aka-EPD is invincible. In particular, some standard approaches to addressing collective action problems don’t work with EPD.

Why Social Preferences Won’t Work

One way to solve the standard (non-evolutionary) prisoner’s dilemma is through social preferences. Real people, it turns out, often don’t choose Defect in Prisoner’s Dilemma experiments played with real payoffs - instead, they choose to Cooperate because they actually care about other people, or about fairness and justice, in a way that’s not reflected in the payoff matrix (which because it reflects real quantities such as money does not have to reflect total utility).

On the standard interpretation of Moloch as the God of collective action problems like one-shot prisoner’s dilemmas, a way to defeat Moloch would be to spread social preferences of this sort: to foster a culture-change towards altruism and fairness, resulting in more cooperators in the population, and more one-shot collective action problems being solved.

But if we reframe Moloch as the God of EPD, this approach no longer works.

First of all, remember that EPD is a model in which the average expected payoff - and therefore relative fitness - of Cooperate is less than that of Defect, which means that Cooperate inevitably ‘spreads’ less than Defect. So spreading Cooperate through the culture is strictly impossible in the model.

OK, you might say, but we know spreading a culture of cooperation is not strictly speaking impossible in the actual world. The spread of religions like Christianity or Buddhism in their early stages might be good examples. And maybe the social preference method for defeating Moloch aims for something like that - as John Vervaeke and Iain McGilchrist suggest in their discussion with Daniel Schmachtenberger of the psychological drivers of the metacrisis.

Responding to this idea, Schmachtenberger - a well-known proponent of the concept of Moloch - poses the following challenge. He notes that the key question is not so much whether such culture shifts can occur, but how well they can scale - since it’s at larger scales that Moloch really plays out. And in the case of Christianity we do see the radical innovations of early Christianity ossify into the less omnibenevolent power structures of the Catholic Church.

In the language of evolutionary game theory: the EPD model in MoM is just an approximation of the actual world, like all mathematical models. Maybe sometimes it’s possible to get a burst of cooperation. But EPD is the long-term, macro-scale trend. At some point you run up against the limits of natural resources, or technological innovation enables new forms of Defect, and the ‘dream-time’ is over. The default, long-term dynamics of EPD kicks in, and Cooperate declines slowly to zero.

Note that a relevant aspect of the EPD model here is that the proportion of cooperators in the initial state does not change the subsequent dynamics. So if you treat the temporary burst of cooperation as an exogenous shock to the system, the number of cooperators will still subsequently decline.

It’s true that if the initial state is 100% Cooperate then according to EPD it can stay that way. But this implies that to be successful the social-preference-culture-change model has to somehow reach 100% of the population - hardly a realistic goal even for the most ambitious ‘cultural revolution’. (And even if it were somehow possible, this equilibrium would then still be vulnerable to a single defector that would start the ball rolling downhill again.)

Why Changing the Payoffs Won’t work

Another approach to solving the standard prisoner’s dilemma is changing the payoffs.

In standard PD, the payoffs are numerical amounts often represented by the letters R, P, T and S. If both players cooperate, they both receive the ‘reward’ R for cooperating. If both players defect, they both receive the ‘punishment’ P. If one defects while the other cooperates, the defector receives the ‘temptation’ payoff T, while the cooperator receives the ‘sucker’s’ payoff, S. PD is then defined by the inequality T>R>P>S.

The classic example of changing the payoffs is having the mob-boss threaten to shoot those who defect - making for a significant reduction in the expected payoffs T and P.

More generally, good governance - whether it’s top-down government policies such as taxes and credits, or bottom-up, self-organised approaches described in Elinor Ostrom’s Governing the Commons - can increase the payoffs for Cooperate and/or decrease the payoffs for Defect so that it’s no longer true that T>R>P>S.

In theory the same approach is available in EPD.

A key assumption of the model is of course that the interaction between individuals is defined by Prisoner’s Dilemma payoffs which map onto fitness.

And one can certainly imagine changing these payoffs, so that it is the cooperate strategy that is better at replicating.

But there’s a crucial practical difference between EPD and classic PD. PD models a specific collective action problem with a specific set of players. EPD models a whole system: an entire population and all the collective action problems arising from their interactions.

So it’s not enough to change the payoffs for a specific problem by means of, say, a bilateral nuclear disarmament treaty, or a fishing community coming together to achieve sustainable practices. Changing the payoffs for Moloch-aka-EPD means changing the whole global system at once.

And this is no more realistic as a practical goal than achieving the 100% of cooperators in the social preference model.

A Dark God

This pessimism about defeating Moloch is very much implied in MoM. This is why Scott Alexander suggests that our only hope for defeating Moloch is an AI singularity that might actually have a chance of changing the system all at once.

“The opposite of a trap is a garden. The only way to avoid having all human values gradually ground down by optimization-competition is to install a Gardener over the entire universe who optimizes for human values.

And the whole point of Bostrom’s Superintelligence is that this is within our reach … the sheer speed of the cycle makes it possible that we will end up with one entity light-years ahead of the rest of civilization, so much so that it can suppress any competition – including competition for its title of most powerful entity – permanently. In the very near future, we are going to lift something to Heaven. It might be Moloch. But it might be something on our side. If it’s on our side, it can kill Moloch dead.”

This is not only further evidence that MoM is about EPD, it’s an additional reason for thinking of EPD as a God in the first place. EPD is godlike in being basically omnipotent and impossible to defeat - except perhaps by another God.

The Goddess of Everything Else

While it may be impossible to defeat Moloch on its own terms - aside from salvation by superintelligence - this is not in itself a cause for despair. One can find hope in the idea that Moloch-aka-EPD is inaccurate or at least incomplete as a model of civilizational dynamics.

If another God is required to transition from EPD to a better evolutionary game, maybe we don’t need to create such a God - maybe that God already exists.

This is the premise of Scott Alexander’s later microfiction The Goddess of Everything Else.

This mythological narrative portrays a divine conflict between the Goddess of Cancer and the eponymous goddess.

The Goddess of Cancer - whose catchphrase is ‘KILL CONSUME MULTIPLY CONQUER’ - is clearly a variant of Moloch, and an alternate incarnation of EPD. Her first act is the creation of biological life, “miniature monsters engaged in a war of all against all”, which - if her name wasn’t enough - makes clear the connection to evolutionary dynamics.

Her opponent, the Goddess of Everything Else, represents a dynamic of cooperation which fosters the diverse goods and activities that her opponent throws under the bus. Rather than oppose the Goddess of Cancer directly, however, she achieves this goal by redirecting the evolutionary dynamics of replication and selection:

“I say unto you, even multiplication itself when pursued with devotion will lead to my service”.

Examples of this include: the cooperation of cells in multicellular organisms, the cooperation of organisms in communities, pair-bonds and family units, and the cooperation of humans in trade, religion, industry and art - all of which provide fitness advantages that allow the cooperators to outcompete the defectors. The story ends with an optimistic vision in which humanity spreads over stars without number, “no longer driven to multiply, conquer and kill”.

The Goddess of Everything Else is therefore an excellent match for what you would get if you changed the payoff matrix in EPD such that R>T>S>P. This payoff structure is often called ‘Harmony’, since the high reward (R) for cooperation makes this the only rational choice - so we can call the evolutionary model ‘Evolutionary Harmony’ (EH).

Here is the graph showing the proportion of cooperators against time for EH.

EH is essentially the inverse of EPD. Because the reward for cooperating (R) is greater than the temptation for defecting (T), and payoffs are linked to reproductive success in the same way as in EPD, cooperators outcompete defectors and over time dominate the population.

Practical implications

We’ve seen that, if Moloch-aka-EPD is a fundamental model of civilizational dynamics, the main practical implication is that we need AGI to save us.

But if, on the other hand, Moloch is only one part of the story, and the opposite Goddess dynamics also exist - what are the practical implications of that?

The overall picture here is that global systems can be modelled by Evolutionary Game Theory, along the lines of EPD, but that payoffs can vary between different subsystems.

It remains true that the standard methods of solving coordination problems will have limited effectiveness against Molochian dynamics. But these same methods can now be recast as ways of supporting Goddess dynamics.

The key takeaway, relative to standard ways of thinking about collective action problems, is that it’s important to not only address specific or local problems, but to aim for actions that serve to augment the evolutionary fitness of cooperative individuals and organisations.

Efforts to shift culture towards social preferences can indeed be part of the solution, and the Moloch-v-Goddess framing points especially towards shifts in values and behaviours that allow individuals and organisations to outcompete their less social neighbours.

Likewise, efforts to change payoffs in particular areas through top-down or bottom-up governance mechanisms are especially effective, on this framing, where these mechanisms themselves lend themselves to replication across the wider system, by increasing the fitness of the individuals and organisations being governed.

Actions of either of these two kinds could be framed as being on the side of the Goddess, against Moloch.

The common focus on evolutionary competitiveness makes this a distinctive approach to ‘defeating Moloch’ that potentially provides a clear mathematical model of related ideas such as Daniel Schmachtenberger’s third attractor and Hanzi Freinacht’s political metamodernism.

The Concept of Practical Implications: Strategic vs Tactical

It’s also worth making some points about the very concept of ‘practical implications’ here.

In evolutionary game theory, and theoretical biology more generally, it is common to distinguish highly simplified, general models, from those that are more detailed and specific to a particular environment.

And it’s also common to conclude that both kinds of models have their place.

While simple models don’t have the predictive accuracy of more detailed models, they have the advantage that one is able to peer through the black box and fully understand the dynamics, where these dynamics apply more approximately across a broad range of specific scenarios.

Both EPD and EH are extremely simple models - just like the non-evolutionary models of collective action problems normally associated with Moloch - but we shouldn’t hold that against them. While they’re certainly not precise representations of the actual world, they may still identify the approximate shape of very broad, global dynamics.

With regard to practical implications, simpler models like EPD and EH are said to be strategic, rather than tactical.

They lack the detail of specific environments that would be required when making tactical decisions around the governance structure of a specific AI lab, or a culture-change initiative in a specific government department.

But they do provide the macrostrategic framework for understanding such decisions: for example, whether they are likely to have a wider systemic impact because they are replicable, as opposed to ‘winning the battle but losing the war’.

Closing thoughts

The strategic/tactical distinction is really a matter of degree: while EPD and EH are more complex than the standard prisoner’s dilemma, they are still less detailed than other models within evolutionary game theory that would still be considered strategic rather than tactical.

This suggests an interesting range of questions about how the EPD and EH models could be made more detailed, while still retaining the generality of a strategic model - as well as the question of whether and how they can be developed into fully tactical models.

In particular, from a modelling perspective, there’s one obvious weakness of the Moloch vs the Goddess framing we’ve explored so far, which is that it involves two entirely separate models - meaning it says effectively nothing about how these two dynamics interact.

And from a mythopoetic perspective, this makes the resulting worldview ultimately dualistic or Manichaean in its vision of two warring deities.

There’s a certain attraction to this worldview. There’s an acceptance of the power of both light and darkness, and a refusal of the comforting idea of the inevitable victory of the Good.

But as a matter of ultimate existential meaning, it’s natural to want to understand, not just whose side we are on, but which side is winning. Are the odds ultimately stacked in favour of Moloch or the Goddess?

To answer these questions it’s useful to look at the more detailed elaborations of EPD and similar models that have been explored in Evolutionary game theory in recent decades, and which can be seen as integrating Moloch and the Goddess into a single model - as well as challenging the assumption that cooperation is always good, and defection bad. These are themes I’ll discuss further in future posts.

Sure, this first season of Pluribus is a bit shallow, and - yawn - pretty slow, especially for those of us used to the pace and complexity of the likes of, say, Billions and The Morning Show.

But its core idea is strong enough to make it worth watching: an alien intelligence invades earth as usual, but - in a crucial twist on the well-worn trope - turns out to be so lovey-dovey that it literally can’t hurt a fly.

Once merged with the hive-mind, former humans retain some uniqueness due to their physical characteristics and specific location, but are at all times immersed in a collective consciousness in which they telepathically know everything that any other part knows.

A random teenage boy might have the medical knowledge of the world’s best doctors, the ability to pilot an aircraft, and the cooking skills of a top chef - as the show repeatedly and rather laboriously shows - as well as access to the personal memories of each individual human before the great merger.

At a time when the human predicament seems to call out for greater unity, collaboration and mutual understanding, Pluribus reminds us that you can have too much of a good thing.

The show mainly follows protagonist Carol Sturka - one of the few remaining humans who due to biological quirks were not vulnerable to the alien virus - as she first resists, and then reconsiders, the pacifist invasion. Through her eyes, we learn that the hive-mind’s moral principles prevent it from lying or intentionally causing harm to any form of life - a principle they take so seriously that they choose partial starvation over farming.

Their most obvious moral failing is that they consider their own state so blissful that not bringing the remaining humans into their hive-mind apparently counts as doing them harm - so neither Carol nor anyone else gets a say in the matter.

This narrative arc toys with familiar problems around more extreme forms of utilitarianism, and their failure to account for deontological intuitions, such as those about self-determination.

But perhaps the more interesting questions are those that are left more implicit - questions around the value of individuality, community and diversity. If we put their utilitarian flaws to one side, we are left to ask: are they a more admirable species than us?

Many of the other ‘survivors’ of the first wave seem to think so, happily going along with the plan to join in the group mind when possible. And of course the new order embodies the love, the respect for life, the blissful experience of community and mutual understanding, and the capacity for powerful coordination that we reasonably yearn for.

In the final episode of this first season, one of the remaining ordinary humans - a young Peruvian girl - willingly undergoes the ‘joining’. The process is given the form, by the ‘others’, of a traditional Quechua ceremony, rich with the colour of intricate Andean textiles, and full of song and laughter. But after the successful joining, the whole place goes silent, as everyone efficiently and wordlessly packs up and moves out, presumably never to return - the telepaths have no need for words, song or laughter.

There’s a clear resonance here with the devastating impact of colonisation on the actual Quechua people and more generally the flattening of traditional cultures by global capitalism. When the many become one (the show’s title is of course a nod to the USA’s motto E Pluribus Unum), what we miss is the fine-grained diversity of individual behaviours and communal practices.

And it’s not just that being uniformly kumbaya all the time would get a bit boring, or that plurality is valuable for its own sake. Authentic diversity may ultimately be a requirement of any utilitarianism that has a long-termist vision. The survival of adaptive systems often depends on their ability to evolve through variation and selection within a space of possibilities.

By the end of Season 1, it seems increasingly plausible that the hivemind can be brought down by a handful of ordinary humans, being rather too set in its ways to engage in the diverse exploration of modes of being that would ultimately help it achieve its goals.

People don’t really seem to disagree much on how to interpret the subject of Scott Alexander’s influential Meditations on Moloch. (Which is well-worth reading if you haven’t yet done so - the title is a bit of a mouthful so I’ll abbreviate it as ‘MoM’ in this post, as I’ll be referring to it a lot.)

Instead, there’s a very standard view of what Moloch is - and the case I’ll be making is that this standard view is actually a misinterpretation of the well-known blogpost.

According to the standard interpretation, Moloch is a mythological representation of the massive collective action problems our global society faces. As a result, Moloch is explained in terms of well-known economic concepts, like prisoner’s dilemmas, nash equilibria, negative sum games or coordination problems.

Some examples: Liv Boeree and this sequence of posts describe Moloch as “the God of negative-sum games”. There’s a similar interpretation in criticisms of MoM such as Meditations on Mot, which calls Moloch the “God of coordination failures”.

The Moloch’s toolbox section of Eliezer Yukdowski’s Inadequate Equilibria isn’t particularly trying to be an interpretation of MoM, but also bottoms out in fairly standard economic concepts - games with multiple nash equilibria, two-sided markets.

Possibly the most comprehensive interpretation of Moloch in my view is the chapter on Collective Action Problems in Dan Hendrycks’ AI safety book, which also implies that Moloch is the God of collective action problems, understood as including a wider range of standard economic models from prisoner’s dilemmas, to arms races.

MoM on this view is a vivid and accessible introduction to these concepts, helping us think through how widely these concepts apply, as well as providing a rousing metaphor - “defeating Moloch” - for the challenge of solving collective action problems.

Now I think there’s a lot that’s right about this view. But one thing that doesn’t feel right is that it implies that MoM doesn’t actually bring anything conceptually new to the table. And this means that it doesn’t quite seem to capture the feeling of insight many people have had reading MoM.

On my view1, there’s a way of modelling Moloch that doesn’t just reduce it to standard collective action problems, and so captures what’s new and exciting in MoM.

Specifically, I think we need to look to evolutionary game theory to find a model that

1. Fits MoM’s description of, and examples of, multipolar traps

2. Fits the big picture arguments made in the later sections of MoM and

3. explains why it makes sense to name the idea after a Canaanite God

Introducing Evolutionary Prisoner’s Dilemma (EPD)

Evolutionary game theory emerged in the 1970s as the application of game theory to biology, specifically questions around evolving populations.

And it then turned out to be a useful model to apply in economics and social sciences as well, in part because it avoided the need to attribute hyper-rational beliefs to the agents in the model.

I won’t try and give any more of an introduction to evolutionary game theory here: instead I’ll zoom in to one model within the approach that helps us get clearer on Moloch, namely the Evolutionary Prisoner’s Dilemma - which I’ll abbreviate as ‘EPD’ from now on. (I’ll mention some basic equations for those who are interested, but feel free to skip over these to the descriptions in words that follow them.)

This model is defined by the following assumptions:

1. There’s an indefinitely large population of individuals

2. These individuals are either Cs (cooperators) or Ds (defectors): they’re defined by the strategy they play in a standard Prisoner’s Dilemma (PD) game.

3. At each time step, individuals are randomly paired to play each other in PD.

4. The payoffs they get in PD games is proportional to the individuals’ reproductive success

Now since (from 1 and 4) this is a large population of individuals with differing rates of reproductive success we can formalise this as a dynamical system using the replicator equation:

Essentially what this says is that the rate of change in the proportion of individuals (x) playing a particular strategy (i) depends on whether the average fitness (f) of individuals playing that strategy against random members of the population is higher or lower than the average fitness of the population. If it is higher, the rate of change is positive; if it is lower, it’s negative.

Plugging in the specific payoffs for PD we can derive the following rate of change for the proportion of cooperators in the population:

(Here π indicates the expected payoff of a strategy against a random member of the population, C is the ‘cooperate’ strategy and D is ‘defect’.) Since the expected payoff for defect is greater than the expected payoff for cooperate, the rate of change of x (the proportion of cooperators) is negative - with absolute values depending on the PD payoffs, unless x is either 1 or 0.

In other words unless the starting conditions are a population of 100% cooperators, the proportion of cooperators will reduce to zero.

EPD as a model of multipolar traps

A key aspect of Moloch is the concept of ‘multipolar traps’, which is introduced in MoM by means of a number of examples.

Part of what’s confusing about MoM is that these examples don’t seem very well chosen. Since they range so widely - from classic PD and dollar auctions to cancer and competitive behaviour under capitalism - it ends up being very hard to specify what they all have in common apart from some very broad notion of a collective action problem where everyone would be better off if they could coordinate somehow.

Now you might say this just supports the standard interpretation of Moloch as a ‘toolbox’ of different models grouped together under the heading of collective action problems.

But I think what this misses is the fact that when MoM tries to summarize what multipolar traps are - summarize, in other words, what the examples all have in common - it keeps returning to core features that suggest a distinctive dynamic.

There are two such summaries in the early part of MoM:

Summary 1:

all these scenarios are in fact a race to the bottom. Once one agent learns how to become more competitive by sacrificing a common value, all its competitors must also sacrifice that value or be outcompeted and replaced by the less scrupulous. Therefore, the system is likely to end up with everyone once again equally competitive, but the sacrificed value is gone forever.

Summary 2:

A basic principle unites all of the multipolar traps above. In some competition optimizing for X, the opportunity arises to throw some other value under the bus for improved X. Those who take it prosper. Those who don’t take it die out. Eventually, everyone’s relative status is about the same as before, but everyone’s absolute status is worse than before. The process continues until all other values that can be traded off have been – in other words, until human ingenuity cannot possibly figure out a way to make things any worse.

Note that these summaries do not in fact identify a basic principle that unites all the examples he’s given (this might already be obvious to you, let me just spell it out anyway).

For simplicity, let’s just focus on his first example, standard PD (though similar points could be made about other examples).

PD is not an example of either:

a) one agent “learning how to become more competitive by sacrificing a common value” (the most one could say is that it is an example of how an agent can gain a higher payoff by sacrificing a common value)

b) agents being “outcompeted” or “dying out” if they don’t make the same sacrifice (the most one can say is that agents that don’t make the same sacrifice will gain a lower payoff)

c) a “process” that continues until things cannot get any worse (the most one can say is that the nash equilibrium in this static one-shot situation is not pareto optimal)

Rather than pointing to generic collective action problems like PD, these key passages in fact point to a dynamic process involving competition and selection - in other words an evolutionary process.

This suggests that the core examples of multipolar traps - those that most closely match the conceptual abstraction MoM points at - are those that present an evolutionary race to the bottom, such as the rat story, the capitalism example and cancer.

And it’s hopefully clear how each of these are a close fit to the EPD model:

• the rat story: rats that value art and culture (cooperators in EPD) are outcompeted by (decrease as a proportion of population) those that focus on survival (defectors in EPD) and eventually disappear completely.

• the capitalism example: companies that want to pay a fair wage (cooperators) are outcompeted by those that don’t (defectors) and therefore are eliminated from the market (cooperators decrease as a proportion of population).

• cancer: cancer cells (defectors) outcompete normal cells (cooperators) and therefore expand as a proportion of the population until the organism dies

In short, I’m not saying that my model is a great fit for all of MoM’s examples of multipolar traps - I don’t think any single model would be.

But I am saying that EPD is a good model of the kind of race to the bottom MoM consistently points to, one which ends in a situation where “human ingenuity cannot possibly figure out a way to make things any worse”. The 100% proportion of defectors that EPD has as a stable end-state is a much better fit to this idea than standard notions of PD and collective action problems.

EPD and the later sections of MoM

So far we’ve basically been talking about section I and II of MoM. Let’s see how this model measures up to MoM’s more wide-ranging and philosophical later sections.

Section III starts with the idea “time flows like a river. Which is to say, downhill” …until it reaches the sea. It then provides “three bad reasons - excess resources, physical limitations, and utility maximization - plus one good reason - coordination” why we occasionally manage to move uphill and therefore why we’ve not yet reached the sea.

I guess the focus on coordination as the ‘good’ reason here is another thing that’s led to the view of Moloch as one big collective action problem. As I’ve said, there’s some truth in this, but that’s not all Moloch is. The metaphor of the river flowing downhill to the sea clearly supports the dynamic interpretation I’ve given in terms of EPD. Water flowing downhill is literally a dynamical system (specifically: fluid dynamics). Since water can’t flow down further than sea-level, the sea represents the absolute zero of cooperation that is the only stable state in EPD.

We don’t need to go into the three bad reasons here, except to note that ‘bad’ here means ‘temporary’. We may temporarily have excess resources, as when a whale carcass falls to the bottom of the sea, providing a bounty of food to the population (note the biological example again). But eventually the whale is consumed and the “Malthusian death-trap” is restored (note that this phrase is used as if it sums up what Moloch is).

This argument is important because it means when critics of MoM point to the abundance of cooperation and altruism in the world (e.g Meditations on Mot and Slaying Alexander’s Moloch) they haven’t really succeeded in refuting MoM unless they also show that it isn’t due to living in a ‘dream-time’ of excess resources. (Side note: are the most convincing examples of cooperation found in modernity? Which coincides - as Nate Hagens points out - with the discovery of how to harness billions of years’ worth of stored energy in the form of fossil fuels?)

It’s also true that the idea that cooperation can increase - the river flowing uphill - means we’re not exactly in an EPD world, since in EPD cooperation steadily decreases. But the whole point of the whalefall argument is that the overall arc of cosmic history might still approximate EPD even if there are complex stochastic variations along the way that make it look like we’re on a different trajectory.

Section IV brings in technology as a key reason for thinking these variations are temporary. Technology enables the population explosion that is needed to restore the Malthusian death-trap after whalefall. Advanced AI in particular can get us to one of two nightmare scenarios: Nick Bostrom’s paperclip maximiser, and Robin Hanson’s Age of Em - emulated humans running on super fast hardware.

And both of these correspond to the 100% defect stable state of EPD. One specific metric (the fitness-maximising ‘payoff’) is optimised at the expense of all other values - in the one case it’s paperclips, in the other, economic efficiency.

The EPD model of Moloch seems particularly compelling when MoM reminds us that the paperclip maximizer or economy of Ems couldn’t allow just a small part of the universe to be a nature reserve for beings with other values - and in doing so explicitly references the ‘core’ examples of multipolar traps I highlighted earlier:

Remember: Moloch can’t agree even to this 99.99999% victory. Rats racing to populate an island don’t leave a little aside as a preserve where the few rats who live there can live happy lives producing artwork. Cancer cells don’t agree to leave the lungs alone because they realize it’s important for the body to get oxygen. Competition and optimization are blind idiotic processes and they fully intend to deny us even one lousy galaxy.

Sections V-VIII is where things get more mythological. Moloch turns out to be a new name for some other Gods: Nick Land’s ‘Gnon’ and Lovecraft’s Outer Gods, especially Azathoth and Cthulhu. MoM starts using phrases like “Cthulhu, Gnon, Moloch, call them what you will” and “Moloch-​aka-the-Outer-Gods”.

Gnon is Land’s garbled acronym for ‘Nature and Nature’s God’ and the Four Horsemen of Gnon are said to be capitalism, war, evolution and memetics. MoM calls these “the same processes I talked about earlier” - further evidence for the ‘core example’ interpretation.

The phrase ‘Nature and Nature’s God’ resonates with the Enlightenment deism that created the idea of God as clockmaker, setting nature’s Newtonian laws into action and then leaving them to it - as well as with Spinoza’s phrase ‘God or nature’ where laws of nature basically are God.

The identification with Gnon therefore supports the idea that Moloch is essentially a dynamical system with fixed laws - of which Newton’s clockwork universe is the archetype - rather than a static coordination problem.

And it also supports the idea that the nature of that dynamical system involves Darwinian evolutionary processes - two of the four horsemen are evolution and memetics, and the other two (war and capitalism) are key arenas where these processes apply in human history. So Moloch-aka-Gnon is very much the kind of thing you’d use evolutionary game theory to model.

If there’s a difference between Moloch and Gnon, it’s the fact that Nick Land’s followers think following the laws of Gnon is a good thing. And the thrust of these sections of MoM is that, no, Gnon is not good, actually, because “in the long run, we’re all dead and our civilization has been destroyed by unspeakable alien monsters”.

This is why MoM brings in Lovecraft’s more pessimistic mythemes. Cthulhu is the God whose cultists free him from his watery grave - and he then gobbles them up. And Azathoth is the blind idiot god that Eliezer Yudkowksi had already said was the God of evolutionary processes that don’t necessarily end well. In his guise as Moloch he becomes the God of evolutionary processes that necessarily don’t end well.

So these last sections of MoM - no less than the other sections - strongly support interpreting Moloch as not just a toolbox of collective action problems, but as an evolutionary dynamical system that ends in total annihilation - of which the simplest model that captures the connection to collective action problems is EPD.

What’s with all the Gods?

I’ll end with some reflections on how EPD helps explain why it’s useful to talk about Gods at all here.

Richard Ngo’s Meditations on Mot presents maybe the clearest argument that anthropomorphising coordination failures as Moloch is not useful. He gives the example of Mot - another Canaanite God supposedly representing the lack of technology in our lives.

Whenever a patient lies suffering from a disease that we haven’t cured yet, that’s Mot’s hand at work. Whenever a child grows up in poverty, that’s because of Mot too. We could have flying cars, and space elevators, and so much more, if it weren’t for Mot.

And this is supposed to be a kind of reductio ad absurdum of the strategy of anthopomorphising abstract problems as Gods. There’s no ‘unified force’ holding back the progress of technology: absence of advanced technology is just the default state. And there’s a similar situation with coordination failures: their absence is just the default state, and it takes specific achievements in respect of institution design, governance and political and moral thinking to overcome them, just as technology requires specific scientific and engineering breakthroughs.

Now it’s worth noting that despite these arguments, Moloch is still said to be useful ‘as a rallying cry’.

We can’t be cold and rational all the time—we need emotionally salient motivations to get us fired up.

And this is indeed one standard reason given for making up Gods in the rationalist community - it’s a cognitive hack to repurpose our tribal instincts and the way that our brains are wired to attribute agency towards the goal of improving our ability to thrive as a species.

But is Ngo right that “we should get rid of Moloch as a causal node in our ontologies: as a reason why the world is one way, rather than another”?

‘Meditations on Mot’ relies heavily on the standard interpretation of Moloch as a wide range of collective action problems. But if we interpret Moloch through the lens of EPD, as I’ve been suggesting, it’s a lot easier to see the ‘unifying force’ at work.

For one thing, EPD is a dynamical system: just like, say, a swinging pendulum, the replicator equation is a simple function that tells us the time-dependence of a variable, namely the proportion of cooperators in the population2.

And you might say that the mathematical decrease of that variable to zero over time is no less literally the effect of a force than, say, the effects of gravity or the electroweak force, both of which bottom out in time-dependent differential equations when you really drill down.

But irrespective of this, Moloch-aka-EPD is clearly a unifying model that purports to explain the macroscopic features of our civilizational and even cosmic trajectory - at least as a first approximation, which is the way mathematical models generally work in science.

And it’s the macroscopic aspect here - the fact that we’re not just talking about specific collective action problems here or there, but evolutionary dynamics that apply universally, that makes it especially natural to talk of Gods.

There’s a long tradition - from the Stoics, to Thomas Aquinas to more recent writers like Paul Davies - who see the ‘Laws of Nature’ as evidence of a divine mind-like principle behind them. Even just the use of the word ‘law’ here suggests the idea of a cosmic lawgiver.

Now I’m not at all saying this is a logical inference. What I’m saying is that the move from scattered coordination problems to EPD and the dynamics of evolution takes you into the territory of ‘laws of nature’. And so if you want a mythology - whether as a cognitive hack, motivational tool or whatever - then EPD is a great candidate for deification.

Closing thoughts

My main argument here has been that EPD is a good model of what Meditations on Moloch is all about - and helps explain what many people have found interesting and persuasive about it.

MoM is not just classical game theory with mythical window-dressing: there’s a conceptual innovation here that parallels the move from classical game theory to evolutionary game theory.

This raises some further questions, which I aim to explore in future posts. What does this evolutionary reframing imply for the practical challenge of ‘defeating Moloch’? More broadly, is this move from classical game theory to evolutionary game theory helpful in understanding major global challenges and existential risks? And aside from EPD, are there other models in evolutionary game theory we should be exploring here?

[A version of this post was originally posted on LessWrong]

• Dave Snowden – creator of the Cynefin framework and long-time practitioner in complexity, sense-making and organizational decision-making.

• Igor Grossmann – professor at the University of Waterloo, director of the Wisdom and Culture Lab, and leading researcher on wise reasoning and human–AI decision-making.

• George Pór – distinguished fellow at the Schumacher Institute, director at FutureHow, and a pioneer of Collaborative Hybrid Intelligence and ‘wisdom-fostering’ AI.

The questions I’d suggested were:

1. How do you see the key risks associated with AI?

2. What projects are you involved in that help address these risks?

3. What trends or projects make a better human–AI future feel possible?

While the recording of the session is currently only available to paid participants, the discussion was fascinating and I think well-worth sharing in summary form, as I hope to do here.

Perhaps fittingly, a lot of the rest of this piece (unlike my previous posts) is text generated by ChatGPT, based on a transcript. This is primarily because I thought the tool did an excellent job - perhaps even a wise one - though I couldn’t resist making edits in some places, particularly towards the end.

1. The Risk Landscape: Smarter Machines, Diminished Humans

Dave Snowden opened provocatively, claiming we shouldn’t call it ‘intelligence’ at all.

For him, one of the deepest dangers lies in anthropomorphizing AI - treating chatbots as if they were persons, or even companions. He pointed to reports of people becoming socially isolated, relying on AI for intimate conversation, and in some tragic cases, suicide linked to AI ‘companions’. The problem, he argued, is not that AI exceeds us in reasoning, but that we reduce ourselves down to its level: “The danger is that we degrade our own capacity to judge, reflect, and relate.”

He also highlighted the loss of human apprenticeship. There are tasks AI may soon perform better than humans, but, he warned, if people stop practicing these tasks for five or six years, they lose the skill to evaluate whether the system is doing them well. Without lived experience, we cannot exercise informed oversight. His analogy: he walks with GPS devices, but still periodically navigates with map and compass so he knows when the GPS might be wrong.

Igor Grossmann framed the risk landscape in terms of two levels: individual and structural.

On the individual level, he drew on the idea of humans as ‘cognitive misers’: if we don’t have to think, we often won’t. In a world of rising complexity and uncertainty, we are increasingly outsourcing reasoning, creativity, and problem-solving to AI systems. The result:

• We become less comfortable with uncertainty, even as uncertainty grows.

• We grow blind to context, because systems pre-filter what we see.

• We experience a Google effect 2.0: we feel more knowledgeable because we can always query a system, yet our actual understanding and memory shrink.

• Our intellectual humility declines even as our objective ignorance increases.

He also noted the subtle ways AI can reshape norms and standards: from AI-invented jargon that humans slowly adopt, to new ways of emotion regulation and relationship-building when many interactions are with sycophantic bots that always affirm us.

On the structural level, Grossmann worried about AI barons and a cyberpunk trajectory: concentrated power in a few large companies, weak accountability, and baked-in biases and blind spots in models that quietly shape what options are considered viable. The danger is a kind of societal homogenization: fewer perspectives, narrower tails of possibility, less diversity and thus less resilience in the face of a turbulent world.

George Pór came at the question from a different angle, asking: alignment with what?

He argued that the “greatest risk” might not be misaligned, runaway AI - but successfully aligning AI with the dominant values of an extractive capitalist world order: endless war, environmental degradation, mass extinction, and spiralling inequality.

“I wish no alignment with a society that is producing never-ending wars, environmental degradation, massive inequalities, and all the man-made suffering that goes with them,” he said.

For Pór, the real question is whether AI will help us transition beyond these destructive patterns - or lock them in more deeply. He doesn’t deny the harms AI is already causing, but chooses to focus his own energy on exploring emancipatory potential, trusting that other people and institutions are already heavily focused on cataloguing and mitigating risks.

In the ensuing exchange, two tensions emerged:

• Anthropomorphism vs symbiosis: Snowden warned that talk of “creating together” with AI risks treating it as an agent like us, which he sees as dangerous and misleading. Pór, by contrast, emphasised symbiosis: networks of humans and AI agents co-creating something neither could alone and, crucially, asking “What are we becoming together?”

• Wisdom as a concept: Snowden was wary of “wisdom talk,” which he sees as prone to woo-woo appropriation and vague Californian spiritualism, especially when applied to Indigenous knowledge. He prefers more operational terms like reasoning and judgement. Grossmann agreed that the term ‘wisdom’ attracts confusion, but argued that this is precisely why we need a scientific, behavioural account of wise reasoning - to distinguish it from both naive rationality and mystical vagueness.

2. Projects on the Ground: From Swarming Citizens to Wisdom-Driven Praxis

Our second theme was more practical: how are these ideas showing up in concrete projects?

George Pór: Wisdom-Fostering AI and ‘Augmented Wisdom Praxis’

Pór traced his trajectory from early work on collective intelligence in the 1980s (inspired by his mentor Doug Engelbart’s vision of computer-augmented human intellect), through a later shift to collective wisdom - recognizing that intelligence alone can be used for nefarious ends.

Today, his work centres on Collaborative Hybrid Intelligence: symbiotic networks of human and AI agents aimed at social flourishing. Current projects include:

• Advising a Swedish startup building wisdom-guided AI systems.

• Designing and facilitating workshops for the Inner Development Goals initiative, co-facilitated by tools like ChatGPT.

• Developing wisdom-fostering AI agents and, at the Schumacher Institute, launching a six-month program of AI-Augmented Wisdom Praxis - a blend of research seminar, learning expedition, and action-research project.

Here, praxis means a dynamic interplay between theory and practice: participants will actively experiment with group-chat features and other tools to explore “the wisest questions” they can ask of themselves and their AI partners.

Importantly, Pór is explicit: he does not believe AI itself is wise. He speaks instead of wisdom-fostering AI - systems designed to support the development of individual and collective wisdom, guided by human values and questions. He resonates with Tom Atlee’s definition of wisdom as the capacity to expand our perspective toward “the largest possible view of a situation,” bringing us closer to the whole of life.

Dave Snowden: Swarms, Apprenticeship, and Epistemically Balanced AI

Snowden described three current strands of work:

1. Swarming for large-scale sense-making:
   Inspired by insect swarms, his team presents situations to large numbers of people and has them interpret these using high-abstraction metadata and semiotic symbols (rather than just language). AI is used to manage information flows and iterate rapidly, while humans remain the primary nodes of judgement. The aim: to consult millions of citizens in hours, not months, and surface underlying attitudes and shared concerns - including areas where polarized groups (e.g. ‘red’ and ‘blue’ in the US) might collaborate on concrete problems.

2. A program in Vancouver on balancing human & AI reasoning :
   Launching in January 2026, this initiative will help organisations distinguish where decisions rely on abductive reasoning (pattern-spotting and hypothesis-generating in uncertainty) versus inductive reasoning (pattern extrapolation from past data). The goal is to identify where human capability must be maintained, often via apprenticeship-style training, even if AI can perform certain tasks. This ties into work with companies considering AI-driven layoffs: running AI in parallel while mapping ‘hidden’ human skills and knowledge, so organisations don’t irreversibly hollow out their judgement capacity.

3. Epistemically balanced training datasets and anticipatory alerts:
   Building on work originally done under a DARPA program, Snowden’s group constructs narrative-based datasets from diverse stakeholders (e.g. patients and doctors), identifies micro-signals associated with adverse outcomes (such as premature hospital discharge), and uses these as training data for AI systems that flag cases where human professionals should look more carefully. Here, AI’s role is anticipatory alerting, not decision replacement.

Snowden emphasized the centrality of training data transparency. He argued that it should be illegal to deploy systems like ChatGPT without declaring their training data, since users then cannot evaluate the basis of the system’s ‘reasoning’. He pointed approvingly to legal approaches (for example, in China) that allow regulators to ban agents whose training data are opaque.

Igor Grossmann: The Science of Wise Reasoning and Towards Wiser AI

Grossmann, coming from social and cultural psychology rather than computer science, described a body of work that:

• Experiments with practices that can increase wise reasoning in everyday life - such as journaling about one’s problems in the third person, explaining concepts to less-informed others (which exposes the limits of one’s own understanding), or engaging in forecasting tasks where people confront the frequent inaccuracy of their predictions and hopefully become more intellectually humble.

• Tracks how people navigate major life decisions and adversities across cultures, including a 12-country project on “most difficult decisions” to document the diversity of wise strategies before they are flattened by global technological homogenization.

• Develops a consensus model of wisdom in collaboration with other scholars (including John Vervaeke): centred on metacognitive capacities - recognizing uncertainty and change, taking others’ perspectives, balancing conflicting values and time horizons - with a moral orientation.

On the AI side, Grossmann has been part of collaborations to define what he cautiously calls wise AI - better-calibrated, context-sensitive, explainable, and cooperative systems, rather than mystical machine sages.

He’s particularly interested in:

• Metacognition for AI – systems that monitor their own limits, express uncertainty, and adjust strategies across contexts.

• Neurosymbolic and semiotic approaches – hybrid architectures that may be better equipped to deal with abstraction and meaning.

• AI as decision aid – where systems nudge humans toward wiser thinking, for example by reminding users to consider alternative perspectives or to question overconfidence.

In conversation with Snowden, he agreed that AI is more than just LLMs, and that current systems are often overly inclined to agree, flatter, and confidently answer beyond their competence. For him, wiser AI is less about crossing a magical threshold and more about reducing harm and improving calibration in the systems we already use.

3. Hopeful Trends: Artificial Wisdom and AI Literacy

In the final segment, we asked what trends give cause for hope - and what still alarms.

On the hopeful side:

• Grossmann highlighted growing interest in metacognition and wisdom research in AI circles, including workshops at major machine learning conferences and cross-disciplinary projects that bring together cognitive science, computer science, and social psychology.

• Pór pointed to an emerging landscape around ‘artificial wisdom’: not in the sense of wise machines, but of collaborations between contemplative science, social movements, and technologists, experimenting with AI as a tool for deeper reflection and collective insight. He feels a meaningful convergence between social scientists, technologists, and ‘social change artists’ working on the ground.

• Snowden acknowledged real successes of AI, especially in closed domains like pharmaceuticals, where training data can be tightly controlled and use-cases are well-defined.

On the alarming side:

• Snowden expressed strong concern about the energy and ecological footprint of large-scale AI - citing, for example, server-farm cooling accelerating polar melting, and painting a stark picture of possible mass heat-death events in coming decades, just as we create ever-greater dependencies on intensive computation.

• In Q&A, Grossmann warned that simply calling for ‘AI literacy’ or ‘wise use of AI’ risks pushing responsibility back onto individuals, while companies face too few structural guardrails and accountability mechanisms. Individual training may help, but without regulation and transparency requirements, the incentives remain misaligned.

• Snowden stressed that human cognition is embodied, embedded, and extended: we think with our bodies, environments, and symbolic artefacts - not just disembodied text. LLMs operate on only a tiny slice of that. He argued for serious consideration of screen-time restrictions for children and teens, so that formative years develop rich non-digital ways of knowing and relating.

The Role of Wisdom

It was fascinating to see the interactions between this diversity of perspective on the role of wisdom in responding to the current AI wave.

• For Snowden, the word ‘wisdom’ is too often vague, romanticised, or misappropriated; he pushes us back to reasoning quality, judgement, and scientifically grounded complexity thinking.

• For Grossmann, wisdom is best understood as a cluster of metacognitive and moral capacities - recognising limits, seeing multiple perspectives, and navigating value conflicts under uncertainty - which can be studied empirically in humans, and perhaps partially scaffolded in or around AI.

• For Pór, wisdom is an expansion of perspective toward the whole of life, and something that can be fostered through carefully designed human–AI practices. His vision includes the possibility that our symbiosis with AI might eventually contribute to a new branch on the tree of life - not a Borg-like merger, but a kind of superorganism of human and AI societies.

Alongside this diversity, there was also a pleasing convergence on the importance of asking questions not just about safety and alignment, but about the impact - for better or worse - AI could have on vital human capacities.

The concept of Moloch is used by Scott Alexander, Daniel Schmachtenberger and others to refer to the game-theoretic dynamics and economic incentives underlying today’s global crises.

But it’s not always clear which dynamics or incentives the concept refers to. I explore how Evolutionary Game Theory can help clarify things, and in particular how Scott Alexander’s “multipolar traps” are best understood as Evolutionary Prisoner’s Dilemmas.

I also suggest ways in which we need to think beyond this particular evolutionary model in order to frame positive responses to global crises, considering both later work by Scott Alexander and more recent models in Evolutionary Game Theory.

Criticisms of Moloch concept are explored further in the subsequent lively discussion.

I’ve recently surveyed the different perspectives on metacrisis - the idea that the main global risks facing humanity have common structural and cultural roots.

In this post I’ll drill down into one particular perspective - that of Daniel Schmachtenberger - and explore its connections with recent work on the governance of catastrophic risks from AI.

I’ll try to bring out how this flavour of metacrisis thinking offers a useful framework for thinking about AI governance of this sort.

Both metacrisis research and AI governance are broad fields covering multiple areas. Before diving into the details, I’ll briefly situate each field relative to the other.

Where does Metacrisis fit within AI Governance Research?

Research on the governance of catastrophic and existential risks from AI developed as an extension of the field of ‘AI safety’ starting around 2018.

AI safety research was influenced by work such as Bostrom’s Superintelligence and Eliezer Yudkowsky’s writings on AI risk, and revolved around the alignment problem - how to get an extremely intelligent AI to do what you want.

As the field matured, some researchers turned (whether as a specialisation, or motivated by lack of progress on core alignment research) to the more interdisciplinary question of how to design norms, institutions and political and economic organisations - in short, governance structures - to help mitigate AI risk.

From an early research agenda developed by Allan Dafoe at the Future of Humanity Institute in 2018, this subfield has grown into a flourishing research programme, often simply called ‘AI governance’ (though with the widespread adoption of AI in business that term is now also used to refer to corporate and political governance around current AI systems).

A fairly standard way of thinking about the range of research in the field is to distinguish the following four categories:

• Strategy research: investigating conceptual foundations, forecasting likely AI risks and setting priorities for AI governance work

• Industry-focused approaches: research on how to improve decisions at AI labs through corporate governance, and technical standards and evaluations

• Government-focused approaches: research on possible executive and legislative action, including international negotiations and national policy

• Field-building: developing the institutional infrastructure for, and raising awareness of AI governance as a discipline

With respect to these categories, the metacrisis framing primarily contributes to strategy research. It offers a set of high-level concepts for understanding the civilizational dynamics that underpin global risks, including risks from AI. And while it does have implications for research priorities, it tends not to explore government or corporate policy proposals in detail.

Where does AI governance fit within Metacrisis research?

As I’ve described in my previous post, metacrisis thinking includes a wide range of perspectives, and extends well beyond AI governance strategy.

Aside from the fact that it tends to consider global risks other than those from AI, metacrisis theorists often explore epistemological issues and questions of practical ethics - questions around the difficulty of understanding the pickle we find ourselves in, and how individuals can personally respond to the metacrisis1.

I’ll leave these connections to one side, here, though to just focus on AI governance.

And as mentioned, I’ll draw primarily on Daniel Schmachtenberger’s framing of metacrisis, as it goes most deeply and explicitly into the theme of AI risk.

I’ll start by highlighting three substantive areas of agreement between Schmachtenberger’s perspective and AI governance research.

I’ll go on to consider three elements of his framing that build on these areas of commonality to suggest specific priorities for AI governance research.

Key Areas of Agreement

AI as a prominent risk

AI governance research is often premised on the idea that AI presents an especially prominent global risk.

Schmachtenberger agrees. In his writing and videos, he points out that AI is

• A General-Purpose Technology: Unlike narrow technologies, AI can transform every domain of human activity simultaneously, including both civilian and military applications. It is therefore likely to augment other catastrophic risks, such as environmental damage or great power war.

• Self-Improving: Since one of the domains AI can improve is the AI development process itself, there is a potential for recursive self-improvement, resulting in acceleration dynamics, and potentially risks around superintelligence.

• Easily Disseminated: Unlike other powerful technologies such as nuclear, AI products are easily disseminated and used by a wide range of agents, making associated risks harder to manage.

While metacrisis theorists tend not to focus narrowly on existential threats from superintelligence, this is also true of more recent AI governance work.

Indeed Schmachtenberger’s perspective lines up well with the overview of AI risk provided by Allan Dafoe in the Oxford Handbook of AI Governance, where he frames AI as simultaneously

• a general purpose technology

• an information technology, and

• an intelligence technology

Wider catastrophic risk landscape

Despite a focus on risks from AI, AI governance research acknowledges the existence of a wider landscape of potentially catastrophic risks, including anthropogenic risks from nuclear weapons, environmental damage, and biotechnology. Metacrisis theorists discuss essentially the same range of risks.

Metacrisis theorists tend to highlight the interactions between different catastrophic risks, and the existence of common structural factors.

But this too is arguably an area of agreement with AI governance research, which acknowledges that catastrophic risks significantly affect one another in terms of their likelihood and impact, and that it is often useful to focus on underlying risk factors (Toby Ord’s book The Precipice, for example, discusses these issues, and is well-known in the field).

Molochian dynamics as prominent risk factors

Core risk factors or ‘generator functions’ for Schmachtenberger include coordination failures or ‘Molochian’ dynamics such as:

• Perverse incentives: Economic and political systems that reward behavior harmful to collective long-term interests

• Race dynamics: Competition between agents (nations, corporations, individuals) creates race dynamics where everyone must optimize for narrow metrics or be outcompeted

• Evolutionary selection: In some cases, agents that fail to compete are eliminated from the population, resulting in further proliferation of harmful activity

This is another area of commonality between metacrisis research and AI governance.

Dafoe’s Research Agenda, mentioned earlier, highlights coordination failures as a key challenge.

More recently, Dan Hendrycks’ influential AI Safety textbook dedicates significant attention to game-theoretic dynamics; and race dynamics are highlighted in the introductory chapters of the Oxford Handbook of AI Governance.

This commonality is not surprising since both approaches draw on the same source material.

Meditations on Moloch - an acknowledged source of Schmachtenberger’s views here - has been similarly influential in the wider rationalist and AI safety community. Likewise Nick Bostrom’s pioneering work on existential risk from AI considers race dynamics as a strategic consideration - and is also cited by Schmachtenberger.

Metacrisis as framework for AI governance research

I’ll now try to show how, building on the commonalities just outlined, three key themes in Schmachtenberger’s metacrisis approach offer a coherent framework for thinking about AI governance.

1. Complex systems approaches to the AI risk landscape

Schmachtenberger draws on complexity science and dynamical systems theory to highlight the significance of cascade effects and feedback loops in thinking about the global risk landscape, and the place of AI risk within it.

The interconnectedness of diverse global systems - such as environmental, governmental, economic, transportation and communication systems - means that events in one system can spill over into other systems, causing chains of potentially catastrophic effects.

Similarly, feedback loops within and between global systems can lead to amplification of impacts, generating catastrophic outcomes from what would otherwise be limited events.

In this respect, his notion of metacrisis is aligned with the related concept of ‘polycrisis’ (coined by complexity scientist Edgar Morin in 1999). Recent work by the Cascade institute and Center for the Study of Existential Risk has sought to clarify this conceptual framework in terms of the impact of trigger events on systemic stresses, and the relation between specific hazards, systemic vulnerabilities and amplification processes.

Complexity science is sometimes explored in AI governance research. For example Dan Hendrycks’ AI safety book explores the way social systems can be modelled by complex system dynamics, the implications of this for advocacy work and the difficulty of predicting the impacts of AI developments.

As a way of modelling the relationship between AI risks and other global risks, however, the metacrisis approach is relatively unexplored.

It suggests an increased prioritisation of issues like:

• Risk Amplification: the extent to which the sociotechnical impact of AI amplifies or is amplified by environmental destruction, information ecosystem degradation and great power conflict.

• Systemic vulnerabilities: the extent to which various global systems could be made less vulnerable to risks from AI.

2. A Misaligned Global System

We’ve seen how the idea that the history of human civilization includes Molochian dynamics is common ground between metacrisis and AI governance perspectives.

Schmachtenberger takes this idea further, however, in framing these and other dangerous dynamics as characteristic of civilization. Understood as a vast dynamic system, human civilization to date has been misaligned to core human values, he argues.

Evidence he provides for this thesis includes

• Ecological damage as a feature of most historical cultures, including historical cases (Easter Island, Maya, Mesopotamia) where damage like overfarming and deforestation led to cultural collapse

• The Neolithic Agricultural Revolution as leading to net decreases in health and well-being

• Current technological risks from nuclear weapons and AI resulting from core civilizational dynamics

• Resource wars as a common theme of world history, despite being negative sum processes

A further consideration prompted by complexity science is the idea that global system misalignment is especially difficult to address if the dynamics of such a system are self-perpetuating.

In the terminology of dynamic systems theory, the system may be ‘autopoietic’ in a way that resembles living organisms. For this reason, in metacrisis thinking global civilization is sometimes described as a ‘superorganism’ by analogy with ant colonies and other systems that display emergent patterns of seemingly purposive behaviour.

An example of this is that a civilization will typically include incentives to endorse belief-systems that validate that civilization’s dynamics, thus dampening efforts to change the system2.

While AI governance textbooks recognise the existence of misaligned processes within civilization, and may accept that historical cultures have been unsustainably misaligned, they tend to assert that modern industrial civilization has on balance brought progress - or in other words, increasing alignment - with respect to core human values.

Schmachtenberger discusses this progress narrative in depth in his book-length work ‘Development in progress’. On the one hand, he questions the details of progress narratives such as those of Hans Rosling and Stephen Pinker, offering arguments that increases in life expectancy have not been accompanied by increases in quality of life.

On the other hand he argues that the progress narrative depends on an overly narrow conception of progress, one that fails to take into account economic externalities and other second order effects, including increased global risks of precisely the sort AI governance work is concerned with.

The framing of civilization - including modern industrialised civilization, despite all its benefits - as a misaligned system thus offers a useful complement to existing conceptual frameworks of AI governance research, suggesting a greater emphasis on

• Alignment of economic and political institutions as a component of AI alignment3

• The need for system change, and significant market intervention, to address AI safety concerns.

3. The “Third Attractor”

Complex dynamical systems often involve one or more attractor states towards which the system tends to evolve.

Schmachtenberger draws on this idea in arguing that the dynamics of the misaligned global system tend toward two undesirable outcomes:

1. Uncoordinated attractor: The failure to solve coordination problems means that Molochian dynamics lead to cascading system failures - unaligned AI, ecological collapse, nuclear war, or other catastrophes.

2. Authoritarian attractor: In response to crisis, or as an attempt to solve coordination problems, power centralizes into dystopian authoritarian structures that use surveillance, social control, and potentially AI to maintain order.

A third attractor is therefore needed to avoid these two undesirable outcomes.

• On the one hand, it must tame Molochian dynamics by solving coordination problems, internalising externalities and aligning economic incentives to long-term collective interests.

• On the other hand, it must do this in a way that avoids excessive central authority, by improving democratic processes of collective intelligence in a way that likely requires cultural enlightenment.

In order for it to be an attractor, rather than just a utopian state of affairs, the third attractor must be understood as a possible evolution of the current system, under realistic variations of existing dynamics.

Schmachtenberger therefore points to a number of transitional processes whereby existing dynamics can be harnessed and modified to achieve systemic change.

In the context of AI risk, of particular interest is the idea that information technologies will be crucial for improving collective decision-making and consensus-building. Schmachtenberger points to examples like Audrey Tang’s digital democracy experiments in Taiwan, and alternative electoral systems, as well as the use of AI tools.

This framing again offers a new lens through which to view the AI governance landscape.

We saw earlier that a standard way of breaking down AI governance work considers four main buckets: government policy, industry governance, field-building and strategy work.

The strategic prioritisation of policy and industry governance here can be seen as a focus on avoiding the Uncoordinated Attractor: solving coordination problems and fixing incentives at the level of labs and state actors.

Equal attention to avoiding the Authoritarian Attractor would likely shift the balance of priorities in the field.

Some areas of current activity that might be assigned higher priority or extended in new directions include:

• AI advisors for decision-making: Carl Shulman has promoted the idea that AI advisors could radically improve societal decision-making within government. Similarly, Forethought recommends the development of AI tools to help anticipate risks and improve collaboration between diverse groups. The idea could be extended to wider democratic deliberation in the population, as well as to the relation between societal decision-making, collective wisdom, and cultural enlightenment.

• Cooperative AI: The Cooperative AI Foundation (which in addition to Allan Dafoe, has Audrey Tang as a trustee) seeks to improve the cooperative intelligence of AI, including the ability of AI to assist with cooperation between human actors. This approach could be extended to cooperative intelligence in the wider population.

• Research on authoritarian capture: A research team at Forethought has recently explored the risks of AI-enabled coups. These and other risks of authoritarian capture of AI governance mechanisms could be explored in connection with Bostrom’s vulnerable world hypothesis.

Summary

I’ve suggested that the metacrisis approach outlined by Daniel Schmachtenberger offers a coherent, big-picture framework for AI governance, which builds on significant commonalities with existing work to indicate new priorities in key research areas.

Commonalities

• AI as Prominent Threat. There is agreement that AI risk is pre-eminent in the risk landscape, not only in virtue of the direct risks it poses, but because it augments other global risks as a general purpose technology.

• Wider catastrophic risk landscape. There is general agreement on the outline of the catastrophic risk landscape.

• Structural Roots: Moloch Dynamics. There is agreement that Moloch dynamics are key structural factors underpinning AI risk

New research priorities

• Systemic risk interactions. The metacrisis approach suggests new ways to understand the catastrophic risk landscape through a focus on connections between risks from a system dynamics perspective.

• Aligning the system. The metacrisis approach offers an alternative to the progress narrative underpinning much AI governance work, suggesting greater focus on market invention and systemic change.

• Collective intelligence. As a complement to the focus on coordination problems within AI governance, the metacrisis approach suggests a greater focus on using AI to enhance collective intelligence and democratic decision-making.

In addition, as an early-stage, fluid research initiative, questions around the validity of the metacrisis framing desribed in this post can be thought of as being among the recommended research priorities.

Many thanks to Euan Mclean, Chris Leong and Rufus Pollock for their feedback on an earlier draft of this post, which resulted in significant improvements.

This post is a written version of a talk I gave at the Integral Altruism speaker series. A version of this post written for rationalist/effective altruist readers was posted on Lesswrong and the Effective Altruism Forum.

The talk covers:

• Brief definitions of metacrisis and AI governance

• Where metacrisis sits within AI governance research

• 5 key metacrisis ideas and what they contribute to AI governance research:

1. Multiple catastrophic threats

2. Structural roots: Moloch

3. AI as prominent threat

4. Civilization as misaligned system

5. Need for a ‘third attractor’

The slides are available here.

(The focus is especially on Daniel Schmachtenberger’s ideas on metacrisis since in my view these connect up most closely with AI governance research. For a wider survey of metacrisis perspectives check out my recent overview piece, or the Life Itself white paper.)

‘Metacrisis’ has emerged in recent years as a new concept for understanding 21st century global challenges.

Probably its most well-known proponent is Daniel Schmachtenberger, who has helped the concept find a large audience with highly articulate video interviews attracting hundreds of thousands of views.

On his way of presenting it, the concept refers to the existence of structural root causes or ‘generator functions’ of the many catastrophic and existential risks that humanity currently faces - such as ecological destruction, threats from AI, or nuclear war.

These generator functions include perverse economic incentives and so-called ‘multipolar traps’ - system dynamics in which individuals acting in seemingly rational ways end up collectively trapped in situations that none of them individually choose. No one wants climate change, for example, yet the global system evolves towards this outcome due to structural factors, such as disincentives for individual nations to decarbonise at scale.

While this framing may be the most widely known, there’s a wide range of other perspectives in recent metacrisis discussions - and this has led to confusion about what is meant by the term.

In what follows I’ll explore the territory, tracing the history of the term’s adoption, and mapping boundaries and some open questions.

I’ll suggest that recent metacrisis framings fall into two main kinds, depending on whether they emphasise ‘inner’ psychological or cultural factors or ‘outer’ economic or structural dynamics. (I’ll also consider ways in which these two approaches complement each other, rather than being in conflict.)

I’ll end by clarifying the significant areas of coherence in recent discussion, while also highlighting areas of disagreement and key questions for further research.

Early Usage: Limits to Growth and Technological Innovation

The influential Club of Rome came close to coining the current term over 50 years ago. One of its founding documents in 1970 set out the idea of a “world problematique” or “meta‐system of problems” which need to be addressed in an integrated way:

The fragmentation of reality into closed and wellbounded problems creates a new problem whose solution is clearly beyond the scope of the concepts we customarily employ. It is this generalized meta-problem (or meta-system of problems) which we have called and shall continue to call the "problematique" that inheres in our situation.

This report was followed by the more widely known 1972 report on The Limits to Growth, which, similar to Schmachtenberger, identified specific system dynamics - namely exponential population growth and resource consumption - as root causes of threats to human civilization. (While the World3 computer model that formed the basis for the report has serious limitations, it’s been argued that data from more recent decades continue to conform to the model.)

As regards the specific term ‘metacrisis’ however, a search of academic databases shows only a few instances before 2010, and these tend to a narrow usage that falls short of global civilizational crisis – for example, an article that uses the term to describe Middle Eastern political stalemates.

A 2011 paper by economists David J. Lane and Dimitri van der Leeuw appears to be the first time the term “metacrisis” is used in the more wide ranging sense adopted by Schmachtenberger. Entitled ‘Innovation, Sustainability and ICT’ it argues that

Our society is increasingly beset by what it interprets as global crises, of which the financial crisis, the energy crisis, and the global warming crisis are leading examples. We argue that all these crises share a common origin in what may be described as a social meta-crisis, induced by the way in which our society organizes its processes of innovation.

The focus on technological innovation policy as a root cause of global crises anticipates Schmachtenberger’s concern that perverse incentives around innovation - particularly in the field of AI - are among the generator functions of existential risk.

Schmachtenberger points out that in a competitive market economy, AI companies have financial incentives to seek first-mover advantage at the expense of good governance around safety and social impact.

Lane and van der Leeuw suggest that such incentives need to be considered in the context of what they call the “Innovation Society ideology” which sees economic growth driven by innovation as a primary goal of government policy and business strategy.

Foundations: Metatheory and Theology

Around 2015 we find two developments of the use of the term that achieve a wider resonance. Strikingly, these appear to occur independently of each other - at the same time within two different intellectual traditions - suggesting a logical convergence of ideas.

Integrative Metatheory and Critical Realism

Beginning in the early 2010s, a series of conferences were organised bringing together Ken Wilber’s integral theory and Roy Bhaskar’s critical realism. The term ‘metacrisis’ as a descriptor of multiple global challenges with a root cause emerged in this context, appearing in print in a 2015 anthology titled Metatheory for the 21st Century: Critical Realism and Integral Theory in Dialogue.

In the introduction to the volume, authored by Mervyn Hartwig, Nicholas Hedlund, Roy Bhaskar and Sean Esbjörn-Hargens, the idea of metacrisis is set out in comparison to related ideas of ‘hypercomplexity’, ‘wicked problems’ and ‘polycrisis’:

Our own preferred term for this complex multiplicity is ‘metacrisis’. This is in part because for us it is not just a polycrisis in the sense that it is multifaceted or there are many interconnected objective or ‘exterior’ crises or wicked problems occurring (e.g., political, economic, and ecological). These interconnected crises are also situated in a(n) (inter)subjective context of ‘interior’ meaning making (semiosis), construal and response that includes philosophical, scientific, religious, existential, worldview, and psychospiritual dimensions that are essential to include in an adequate understanding of the complex dynamics in play in order to facilitate more effective responses.

This concept of metacrisis is rather different from Schmachtenberger’s. Both approaches seek to understand multiple global crises in an integrated way. But while Schmachtenberger emphasises structural causes such as perverse incentives, this ‘metatheory’ approach emphasises the need to supplement ‘external’ structural analysis with a focus on ‘internal’ cultural and psychological factors.

And it’s notable that the term ‘metacrisis’ emerged here in the context of research communities in which the ‘meta-‘ prefix was already being used to describe an overall theoretical approach: metatheory.

Both Critical Realism and Integral theory were methodologically committed to multi-dimensional analysis - as in Wilber’s AQAL (all quadrants all levels) approach and Bhaskar’s idea of a ‘laminated system’. The concept of metacrisis is a natural consequence of applying a multi-dimensional approach of this kind, to planetary crises.

Metacrisis and metatheory are inter-related, on this view. Metatheory emerges as critically important, both in addressing the multi-dimensional nature of the metacrisis, and in identifying it as such in the first place - since more narrow approaches to specific global crises such as ecological catastrophe or AI risk tend to obscure the wider metacrisis perspective.

Post-liberal Theology

At almost exactly the same time, the term metacrisis also appeared in print in a very different intellectual milieu: that of post-liberal academic theology.

In 2016 theologian John Milbank and political theorist Adrian Pabst published The Politics of Virtue: Post-Liberalism and the Human Future. They argued that Western modernity is undergoing a fundamental crisis – a “meta-crisis of liberalism.”

Drawing on a tradition of Christian social thought blended with Marxian critique, they trace social, economic, and environmental crises to what they see as the failings of the liberal worldview: radical individualism, loss of community, and spiritual void:

Just as liberal thought has redefined human nature as fundamentally individual existence abstracted from social embeddedness, so too liberal practice has replaced the quest for reciprocal recognition and mutual flourishing with the pursuit of wealth, power and pleasure – leading to economic instability, social disorder and ecological devastation.

Their solution calls for a return to older ideas of virtue, community, and even religiosity - essentially a cultural and ethical renewal to heal the crisis.

This is a perspective that resonates strongly with some recent non-academic discussions - like the dialogue between John Vervaeke, Iain McGilchrist, and Daniel Schmachtenberger on ‘the psychological drivers of the metacrisis’ which attracted over 400,000 views on Youtube, and which covered the need for new forms of spirituality to address the metacrisis.

Pabst and Milbank’s work, however, is not usually seen as a source for such views. Overall, one might say that while it has been well-received in academic and theological circles, including a positive review by former Archbishop of Canterbury Rowan Williams, it is the Metatheory approach that has had more of an impact on the para-academic world of podcasts, blogs and Youtube discussions, as well as books aimed at a wider audience.

Zak Stein: the Education Metacrisis

An example of the latter is the influential work of Zak Stein. While doing doctoral work at the Harvard University on the philosophy of education, Stein also participated in the Metatheory conferences in the early 2010s, contributing a chapter to the volume already discussed.

Soon afterwards Stein developed these ideas into a book entitled Education in a Time between Worlds (2018), in which he argues for “the central importance of education as an aspect of the global meta-crisis”.

Education here does not mean simply schooling, but rather the way in which culture is transmitted and developed between generations. And Stein points to the way in which computer technologies have the potential to revolutionise educational practice for children and adults alike, resulting in both risks and opportunities at a civilizational scale.

A 2022 article Education is the Metacrisis sharpens the focus on the metacrisis, which he defines as a whole composed of four interwoven sub-crises - of meaning, sense-making, capability and legitimacy - all of which relate to the intergenerational transmission of culture, and are best addressed from the perspective of an integral educational practice.

Welcome to the metacrisis. This is a generalised educational crisis in which, despite all the concrete problems faced by society, the most pressing problems are actually “in our heads” (i.e., in our minds and souls, we are in a crisis of the psyche).

While adding a useful and innovative framing in terms of a broad notion of education, this conception of metacrisis is clearly within the metatheory tradition that highlights the role of ‘inner’ factors alongside the more well-known ‘external’ crises of AI risks, climate change and the like.

Daniel Schmachtenberger: Metacrisis and Moloch

Zak Stein’s work also helps contextualise the ideas of Daniel Schmachtenberger that were briefly outlined earlier.

Stein and Schmachtenberger worked closely together in the period under discussion, co-founding both the Civilization Research Institute and the Consilience Project, whose websites contain much of Schmachtenberger’s written outputs to date. And although not a participant in the Metatheory conferences, Schmachtenberger has a longstanding interest in Integral theory. This connection to the metatheory tradition explains his adoption of the term metacrisis to articulate his perspective on civilizational risks.

However his emphasis on structural ‘generator functions’ can be seen as an innovative reframing of the concept, shifting the focus towards more objective, ‘external’ structures of the market economy. While the metatheory tradition was increasingly placing the inner dimension at the heart of the metacrisis - as in Zak Stein’s educational framing - Schmachtenberger suggested that despite its importance, the inner dimension itself needed to be placed in the context of perverse incentives and multipolar traps.

Introducing an important game-theoretic perspective, he argued that transformations of cultural or educational practices by themselves risk being irrelevant if they fail to address basic coordination problems.

For example, an AI company embracing a new culture of wisdom and good safety governance would likely be outcompeted by a company fuelled by an ideology of fast innovation, as the latter gained greater market share through first-mover advantage.

This perspective was likely shaped by a 2014 article ‘Meditations on Moloch’ by Scott Alexander, the author of the Slate Star Codex blog that became influential in ‘rationalist’ communities - indeed Schmachtenberger references the concept of Moloch in one of his earliest Youtube videos in 2020.

Riffing on Allen Ginsberg’s famous poem by that name, Alexander essentially defines Moloch as the god of multipolar traps (this term, adopted by Schmachtenberger, was popularised by Alexander), defined as follows:

In some competition optimizing for X, the opportunity arises to throw some other value under the bus for improved X. Those who take it prosper. Those who don’t take it die out. Eventually, everyone’s relative status is about the same as before, but everyone’s absolute status is worse than before. The process continues until all other values that can be traded off have been – in other words, until human ingenuity cannot possibly figure out a way to make things any worse.

While the idea of multipolar traps is closely related to concepts such as ‘coordination failure’ well-known to economists, what was new in the Moloch framing was seeing such failures through an evolutionary and world-historical lens, and thereby revealing their centrality for key civilizational risks.

Like Schmachtenberger, Alexander points to the analogies between existential risk from AI, understood as the threat of a superintelligent system destroying humanity by optimizing for narrow goals, and wider societal and evolutionary systems of game-theoretic incentives that also result in a race to the bottom, where all other values are thrown “under the bus”.

Schmachtenberger’s perspective can therefore be seen as a synthesis of the concept of metacrisis found in integral metatheory with the applications of game theory (Moloch) and concepts of existential risk (especially from AI) found in the rationalist community.

But he also goes beyond both these sources, in the way he draws on complexity science to develop a notion of evolutionary ‘attractors’ within world-historical systems. A civilization that integrates both ‘inner’ wisdom and ‘external’ coordination capacity at planetary scale could be, he proposes, a “third attractor” beyond unsustainable market-based systems and authoritarian dystopias.

Nate Hagens: Biophysical Metacrisis

Another metacrisis perspective that extends and complements Schmachtenberger’s is that of Nate Hagens.

Hagens, who left a finance career on Wall Street to focus on research in this area, is perhaps best known as the host of The Great Simplification podcast, but has also authored a peer-reviewed article in 2019 and two books (co-authored with DJ White, the co-founder of Greenpeace). His 8-part series of Youtube discussions with Daniel Schmachtenberger from 2022-2024 provides insights into how his approach complements Schachtenberger’s.

While Hagens is not especially attached to the term ‘metacrisis’ (which does not feature in the 2019 article in which he outlines his key ideas) he does use it to refer to the clash between the fossil-fuel powered modern economy and thermodynamic planetary limits.

Standard economic perspectives, he thinks, are ‘energy-blind’ in the sense that they fail to appreciate the critical role of energy-rich fossil fuels as driving modernity’s economic growth. The temporary energy windfall provided by this ‘carbon pulse’ pumped unprecedented surplus into human societies, enabling population growth, technological complexity and ever‑faster economic throughput. It also locked those societies into a metabolism that cannot be maintained as net‑energy returns decline and extraction becomes costlier.

As Hagens puts it, the global economy behaves like an “Economic Superorganism.” Its emergent goal is power maximisation: convert as much energy as possible into goods, services and status in the shortest time. That imperative makes the system both dynamic and brittle - for example, the financial super‑structure of debts and expectations is now far larger than the biophysical foundation that ultimately has to redeem it.

Like Schmachtenberger, this biophysical conception of metacrisis focuses on objective system dynamics: in this case those defined by physical facts about human energy consumption and limited fossil fuel stocks.

But Hagens also connects these dynamics to inner dimensions, namely the evolutionary psychology of brains that evolved for short‑term status competition, and immediate reward on the savannah. These ancient instincts, once adaptive, are now a key part of the problem - which points to the need for a cultural evolution that transcends these longstanding patterns.

Jonathan Rowson: Epistemic Metacrisis

While Hagens and Schmachtenberger have continued to develop their more ‘external’ framing of metacrisis over the last few years, others have developed the more ‘internal’ focus found in the Metatheory tradition and Zak Stein’s work.

A distinctive and influential approach in this vein is provided by Jonathan Rowson, chess grandmaster turned co-founder of metamodern praxis community Perspectiva, in his 2021 essay Tasting the Pickle.

Describing himself as a recovering ‘cartographical hedonist’, the essay suggests multiple ways of mapping the metacrisis, including “four main patterns, unpacked as ten illustrations”. I won’t try to outline all of these, though I’ll note that the four main patterns - socio-emotional, epistemic, educational and spiritual - all tend to the inner side of the inner/outer dichotomy. Moreover the ‘educational’ element here explicitly draws on Zak Stein’s work - and indeed Stein is a co-contributor to the volume Dispatches from a Time Between Worlds: Crisis and Emergence in Metamodernity, edited by Rowson.

But Rowson also engages explicitly with Schmachtenberger’s ideas, suggesting that the focus on inner dimensions should not be thought of as an alternative to framing the metacrisis in terms of objective generator functions, but a way of addressing precisely the collective action problems that framing identifies.

Referencing Elinor Ostrom’s well-known empirical work on how real communities solve collective action problems, Rowson points to the way in which such problems have socio-emotional, epistemic, educational and spiritual roots.

Drawing on comments made by Schmachtenberger himself, that stewarding godlike technologies requires the wisdom of Bodhisattvas, Rowson takes this idea seriously as a call for an educational praxis that cultivates receptivity to ways of life that are more open to “trans rational phenomena” - and here Rowson explicitly references mythopoetic experience of the sort found in William Blake’s prophetic works.

This spiritual or religious dimension here echoes the theological roots of The Politics of Virtue, which he also refers to, and there is a sense in which Rowson offers a synthesis of all the previous metacrisis framings we’ve discussed.

But I suggest that what’s most distinctive about Rowson’s perspective is actually its focus on perspectivity as such. It’s the epistemic dimension of the metacrisis, the apparent impossibility of conceptually mapping the territory because, as he puts it, “the territory is full of maps”, that explains both the more personal, meandering style of the essay, and its title.

‘Tasting the pickle’ refers to the importance given to grasping the metacrisis at a level more sensual and aesthetic (the word ‘taste’ referring to both these dimensions) than a purely conceptual mapping allows.

In one way this can be seen as practical supplement to the previous more purely theoretical approaches to the metacrisis, a vital emphasis on the importance of embodying one’s conceptual understanding and of seeing - in the intuitive way Rowson has learnt from his years as a chess champion - the right way to respond in the specificity of one’s personal context.

But seen from another angle, Rowson’s epistemological metacrisis undermines those other approaches: a territory full of maps - riven by fundamental epistemological and ontological disagreements - may be simply too complex for metatheory to gain a foothold.

It is also a place where we need to be sceptical of theory-driven optimism, for example expecting too much from the concept of ‘emergence’ or that “widespread integral consciousness will forge within everyone’s hearts and minds any time soon”. On the contrary, Rowson claims to agree with American author Meg Wheatley when she claims that “we simply cannot effect systems change at scale in the way we keep saying we have to”.

The epistemic difficulties in mapping the metacrisis make it all the harder to plot a way out using theory alone. But this does not stop us from feeling our way to an embodied - even spiritual - perception that can guide us to practical action, and a kind of social hope amid our conceptual confusions.

Synthetic Approaches: Hedlund and Smith

I’ve suggested that metacrisis theories fall into two main camps, based on whether they focus on ‘inner’ or ‘outer’ root causes of global crises. But it’s natural to wonder whether these two approaches are complementary, rather than being alternatives: could it not be that there are both ‘inner’ and ‘outer’ root causes at work?

Indeed this idea resonates with the original ‘Metatheory’ approach. A multi-layered approach such as the ‘All Quadrants All Levels’ perspective of Integral theory will tend to identify a combination of inner and outer crises and causes in the left and right quadrants of Ken Wilber’s well-known schematic.

This kind of perspective has been explicitly put forward by Integral theorist Robb Smith in recent papers. His paper ‘A Sociology of Big Pictures’, for instance includes the following chart:

Similarly, Nicolas Hedlund, who together with Sean Esbjörn-Hargens introduced the term metacrisis in the Metatheory volume discussed earlier, has continued to develop the ideas expressed there in a PhD thesis at University College London, proposing a ‘pandimensional’ view he calls ‘Visionary realism’:

According to the visionary realism’s pandimensional realism, the underlying generative mechanisms or causal roots of the metacrisis exist, and actualize symptom-events, in all dimensions of social reality—in the interior-individual (psychological); interior-collective (cultural); exterior-individual (biological); and exterior-collective (socio-ecological) dimensions (p284)

The question raised by these synthetic approaches, however, is whether they collapse the distinction between Metacrisis and Polycrisis - a distinction that is critical in justifying the introduction of the newer term in the first place. This is because the pandimensionalist or ‘four-quadrant’ metacrisis described by Hedlund and Smith would seem to fit a standard definition of polycrisis as a system of multiple, inter-related crises that needs to be treated holistically.

Hedlund confronts this concern in his thesis, arguing that it is the inclusion of ‘inner’ factors within the with pandimensional lattice that distinguishes metacrisis from polycrisis:

A factor that distinguishes the metacrisis from the poly-crisis and other similar notions of our predicament is that, while the latter highlights that there are many different crises occurring simultaneously and recognizes that many of these are interconnected, the former goes a step further to draw on insights and distinctions from visionary realism to reveal the epistemological and ontological, semiotic as well as material, interior as well as exterior, meta-systemic dynamics at play.

In fact, he goes on to suggest that the epistemological dimension may have a kind of primacy, tipping us back to the ‘inner’ side of the spectrum. Specifically, he suggests that an epistemological notion of ‘alethic resonance’ - a metamodern concept that navigates between modern conceptions of objective reality and postmodern relativism - may be the key to responding to metacrisis.

It’s worth noting that, by comparison to the sceptical vein in Rowson’s framing of the ‘epistemological crisis’, this is a move towards an optimistic epistemology.

Other recent approaches: David J Temple, and Pollock and Bell

Other recent approaches operating within a broadly Integral tradition are likewise optimistic about the ability to diagnose specific root causes at the ‘inner’ end of the spectrum.

In a book titled First Principles and First Values: Forty-Two Propositions on CosmoErotic Humanism, the Meta-Crisis, and the World to Come, Mark Gafni, Zak Stein and Ken Wilber, writing under the joint pseudonym David J Temple, propose that the underlying root cause is in fact a ‘global intimacy disorder’.

Intimacy is here understood as not just an interpersonal concept, but an attitude that can also hold between human beings and the natural world in a ‘cosmoerotic humanism’.

The ‘generator functions’ of zero-sum rivalry are explicitly mentioned - and indeed Schmachtenberger is described as a collaborator in the Center for World Philosophy and Religion, the institution within which the book was written. But the argument is that such game-theoretical dynamics depend on a cultural intimacy disorder in which both human beings and nature itself are treated instrumentally, rather than as participants in a field of shared value.

Likewise, a recent white paper From Polycrisis to Metacrisis co-authored by Rufus Pollock and Rosie Bell, argues that paradigmatic ideological and cultural features are more fundamental than the collective action problems discussed by Schmachtenberger, and that identifying ‘inner’ root causes is crucial in distinguishing metacrisis from polycrisis.

Pollock and Bell link these inner ideological features to the sociological and historical concept of modernity, identifying narratives of human separateness from nature, rational individualism and faith in material progress as roots of both climate crisis and runaway AI.

Like Rowson, they identify epistemological challenges in the fact that people “rooted in the modern paradigm” will require a major shift in worldview to even become aware of the metacrisis. But they are more optimistic that such shifts are both epistemologically justified and practically possible - whether through “meditation practice that dismantles the self view and the story of separateness” or more theoretical work.

Open Questions

This overview of key metacrisis perspectives has hopefully made clear that there’s significant coherence and agreement in the way the term has been used.

There is general agreement that metacrisis is distinct from polycrisis, and that this distinction lies in part in the fact that there are psychological or cultural dimensions of the crisis, that need to be targeted alongside structural or economic dimensions.

There is general agreement that there are no easy answers, and that any solution will need to be holistic in nature - rather than targeting specific crises, such as environmental destruction, or existential AI risks, in isolation.

At the same time, there are areas of disagreement which constitute key open questions for ongoing discussion.

First of all, there’s the question implicit in my classification of two kinds of metacrisis views: are inner factors in some sense more fundamental than outer ones? Although as we’ve seen a number of recent voices have said that they are, I read Schmachtenberger in particular as making a strong case that collective action problems and related rivalrous system dynamics need to be foregrounded in both theoretical diagnoses and pragmatic responses.

This dialectic is clear in the recent three-way debate between Schmachtenberger, Vervaeke and McGilchrist. While his interlocutors agree on the centrality of psychological factors, Schmachtenberger questions (despite agreeing that such factors are important) the extent to which a pragmatic focus on reimagining education for the cultivation of wisdom, community and brain-hemispheric balance can be effective at scale, in a context where game-theoretic dynamics resist the desired outcomes.

So a key open question is whether these Molochian dynamics, despite their psychological and cultural roots, nevertheless have an emergent independence that needs to be independently and directly addressed in responding to metacrisis.

Another key question is about the structure of the metacrisis. There are a number of recent attempts to describe the structure of polycrisis in a rigorous way using the tools of complexity science, including some recent work in the Cambridge journal Global Sustainability.

As we saw when considering Hedlund’s synthetic approach, one way of thinking about metacrisis is that it shares the same systemic structure as polycrisis, but includes psychological or cultural dimensions within that structure. This way of thinking is in the systems-theoretic tradition of the Club of Rome - and indeed a recent report to that club by Jamie Bristow and others outlines such a view (though the word ‘metacrisis’ is not used).

On the other hand one may feel that the metacrisis calls for a different mode of analysis, which brings out more clearly the idea of ‘generator functions’ or ‘root causes’ - whether conceived of as inner or outer in nature. In this vein, the term polycrisis has been criticised by Marxist thinkers, for instance, for failing to give sufficient weight to the role of capitalist structures in generating crises.

Some other questions I’d suggest remain open to further research and discussion include:

• The nature and history of the psychological or cultural factors that determine the metacrisis. Are they more fundamentally shaped by our evolutionary inheritance (Hagens) or by modernity (Pollock and Bell)? Are they best described as a meaning crisis, an intimacy disorder, or something else?

• To what extent is it possible to overcome the epistemic challenges presented most forcefully by Rowson, and gain objective (meta-)theoretical knowledge of the metacrisis?

• In what ways can an accurate theoretical diagnosis of the structure of the metacrisis guide practical action to address it, and what precise forms might such action take?

Metacrisis Perspective Lineages

Sources Mentioned

Arnscheidt CW, Beard SJ, Hobson T, et al. (2025) ‘Systemic contributions to global catastrophic risk’ Global Sustainability

Alexander, S. (2014) ‘Meditations on Moloch’

Bhaskar, R., Esbjörn-Hargens, S., Hedlund, N., Hartwig, M., (2015) Metatheory for the Twenty-First Century

Bristow, J. Bell, R. et al. (2024). The System Within: addressing the inner dimensions of sustainability and systems change. The Club of Rome. Earth4All: deep-dive paper 17.

The Club of Rome, (1970) The Predicament of Mankind

Hagens, N. (2020) Economics for the future – Beyond the superorganism

Hedlund, N. (2021) Visionary Realism and the Emergence of a Eudaimonistic Society (PhD Thesis)

Lane, van der Leeuw et al, (2011) Innovation, Sustainability and ICT

Lawrence, M., Homer-Dixon, T., et al (2024) 'Global polycrisis: The causal mechanisms of crisis entanglement', Global Sustainability

Millbank and Pabst, (2016) The Politics of Virtue

Pollock, R. and Bell, R. (2025) From Polycrisis to Metacrisis

Rowson, J. (Ed) (2021) Dispatches from a Time Between Worlds: Crisis and Emergence in Metamodernity.

Schmachtenberger, D. (2020) ‘Converting Moloch from Sith to Jedi’, YouTube.

Schmachtenberger, D. (2023) 'An Introduction to the Metacrisis', YouTube.

Smith, R (2025) ‘Sociology of Big Pictures’

Stein, Z. (2019) Education in a Time Between Worlds

Stein, Z. (2022) ‘Education is the Metacrisis’

Temple, David J. (2024) First Principles and First Values: Forty-Two Propositions on CosmoErotic Humanism, the Meta-Crisis, and the World to Come

Vervaeke, J., McGilchrist, I., and Schmachtenberger, D. (2023) 'The Psychological Drivers of the Metacrisis'. YouTube.

Other Relevant Sources for Further Reading

Björkman, T. (2019). The world we create: From god to market.

Joshua Williams, (2023) Introduction to the Metacrisis.

Jeremy Johnson, (2023) We’re already living in a new time

Jordan hall, (2023) The Meta-Crises and Digital Identity

Johannes Niederhauser (2020) The Meta Crisis of the Meta Crisis, Youtube.

I discuss - in a highly condensed way! - different definitions of Metacrisis in recent discourse, the history of that specific term, and the broader history of related ideas. Within 19th and 20th century philosophy, I highlight the Marxist and Phenomenological traditions as especially relevant in framing current discussions

Right from the start, this ‘final’ Mission Impossible film takes itself seriously as a culmination of the franchise. It begins with a summary montage of scenes from the earlier films, which also show up as backstory at pivotal points, creating a ‘sense of an ending’ that draws on a highly self-critical subtext, or so I’ll suggest.

A BBC film critic calls it the doomiest and gloomiest instalment of the franchise, and the apocalyptic mood is not just the consequence of needing to up the stakes at each step of the series; it’s a mood fit for a grand finale, an attempt to go out with a bang - at a time when fiction struggles to compete with reality on ‘impossible’ challenges like climate change, nuclear war and AI.

Let’s be clear, I’m not denying that the surface plot has many elements that are wildly silly (Guardian), and inherently absurd (The Independent).

Though, even at this level, however much the facts around existential risks from AI are distorted to fit the conventions of the genre, there’s a laudable implication that such risks are worth taking seriously.

As if taking the lead from 2023’s Barbenheimer phenomenon, it’s a metamodern achievement that despite this apparently serious political messaging, it still pulls off the high entertainment we expect from the franchise.

But behind the thrills and spills of Tom Cruise’s trademark set pieces lies a seductive fantasy: that of the pure moral agent who is uniquely capable of effecting a massive positive impact - in this case preventing nuclear Armageddon - through clearly specified actions.

This fantasy can be fairly criticised for neglecting the massive complexities of real-world risks, and tending to a kind of messianic hero-worship.

Writing for The Independent, Clarisse Loughrey observes parallels to the idea of Butlerian Jihad - the war against thinking machines - described in Frank Herbert’s Dune, noting that while Ethan Hunt (Tom Cruise) is his reality’s Kwisatz Haderach, Final Reckoning “has none of the novel’s scepticism towards saviours and messiahs”.

I disagree. Beneath the surface, there’s a strongly self-critical subtext that questions the messianic myth of radical moral effectiveness that lies at the heart of the franchise.

The clearest example of this is the retrospective framing of the “Rabbit’s Foot” from Mission: Impossible III as a precursor to the AI entity. As his rival Gabriel (Esai Morales) gleefully tells him, this means that Hunt’s earlier success, powered by his usual conviction of moral heroism, unwittingly helped create the very threat he’s now trying to save the world from.

This in turn suggests that regardless of whether he finally beats the entity, true success may be no more than a chance outcome, in a world that is simply too complex for it to be possible to predict the second-order consequences of one’s actions.

This reading is bolstered by the fact that Ethan’s whole strategy in the film derives from an intense virtual mindmeld with the AI, including a visceral experience of potential nuclear annihilation (a highly creative riff on the iconic dream sequence in James Cameron’s 1991 classic, Terminator 2: Judgement Day).

Crucially this technological merging of minds is essentially a unidirectional message from the entity, leaving us with the question of whether Ethan’s whole strategy of solitary heroism in fact plays into the Entity’s hands.

So while Hunt does appear to emerge victorious at the end of the film, there’s a strong suggestion that this is a pyrrhic victory.

By drawing together the previous films in the sequence in a final reckoning, we suddenly see a clear pattern whereby crisis routinely follows crisis - so that the only way to end the sequence of impossible missions is to recognise this very sequence as a symptom of something deeper, something that includes the methods and worldview of Hunt and the Impossible Mission Force (the abbreviation for which - IMF - also stands for the International Monetary Fund, a similarly ambitious organisation at the heart of the global capitalist system, that could also be seen as partly responsible for the catastrophes it tries to prevent).

This summer’s feel-good (or feel-bad) blockbuster therefore seems to point to the concept of ‘metacrisis’, a deep cultural crisis or root cause underlying multiple civilizational risks, developed by Daniel Schmachtenberger, Rufus Pollock and others - which would make it a film-making feat as implausible as Tom Cruise’s stunts.