One of the most important threads in the conversation is that we hold a lot of incorrect beliefs that make things seem worse than they are.

For example, we think crime is rising when it has been on a downward trend for decades. We think there are more than double the immigrants there actually are. We think a quarter of the federal budget goes to foreign aid; while it’s less than one percent.

Worse than that, we imagine there are nearly twice as many extremists on the other side of the political aisle than actually exist. This in turn makes us support more extreme policy positions as a result.

But why? I recently wrote a short interactive quiz about how this happens called The Noisy Room. It lets you test your own misperceptions, and offers a clear and hopeful solution for how we might fix them. Try it here:

I hope you enjoy it. Thanks for being here and to those of you that so generously support this work. I appreciate you.

Today we discuss his research on the psychology of common knowledge and what it means for democracy in a time of infinite AI slop and social media confusion.

We also dive into deeper philosophical questions on AI consciousness and rogue AI — note that this was recorded before the recent Hugging Face incident, making our commentary especially urgent.

His excellent most recent book: When Everyone Knows That Everyone Knows… inspired me to write The Noisy Room — an interactive essay that explains how we might fix the common-knowledge trap sprung by social media.

Programming note: We’ve been off working hard on a great new season of content. I want to offer enormous gratitude to our subscribers for their steady support. Thank you.

A small ask:

The irony is not lost on me in trying to critique the algorithms while still being dependent upon them to reach the right audience.

If you do enjoy the show, please share this episode with a friend and drop us a rating.

You can follow us here:

• Apple Podcasts

• YouTube

• Spotify

• Instagram

Thanks for listening, and please do subscribe.

-Tobias

Today I’m trying something new.

I’m sharing an experiment: An interactive essay I’ve been working on that I think has the potential to improve the internet in a small but significant way.

TheNoisyRoom.com

Check out the interactive essay. (10m read.)

It’s based on my recent conversations with Harvard cognitive scientist Steven Pinker and the technologist Audrey Tang — two great interviews that will go live in the coming weeks. I can’t wait to share them with you.

It’s also an open-source project (called Community Check). If you have additions and recommendations, please do contribute on GitHub, or email me directly here.

Thanks for reading and for subscribing. You make this work possible.

What happens when the good people leave social media?

Chloé Valdary walked away from 110,000 followers on social media. Not a quiet, lurking exit — but a hard delete.

Chloé was a force of nature on Twitter. She built a huge following helping people navigate some of the platform’s thorniest issues with grace and empathy. While many cancelled and called out, Chloé dove headfirst into the hardest conversations offering deep reflection and a dose of savvy humanizing. From George Floyd to Gaza, she was there, helping to deescalate.

Then she left.

In this episode, we unpack what it means when moderates like Chloé leave the platforms that dominate our public conversation.

Who is left? Usually it’s the most polarized and extreme users, who actively own the discourse after that point.

We talk a lot about social media as an Outrage Machine (and my book — of the same title), the hidden incentives shaping our behavior on these platforms, and the mental health cost of constant moral performance.

• If a social platform is a toxic place, what are the obligations we have to the public commons in staying there?

• Are big digital spaces like social media structurally incapable of supporting prosocial conversations?

• Can we protect our mental health while still participating in online political debate?

This conversation explores the trade-offs.

Links:
Chloé’s Substack: Musings with Chloé
Chloé’s work: Theory of Enchantment
Follow her here:

A small ask:

The irony is not lost on me in trying to critique the algorithms while still being dependent upon them to reach the right audience.

If you do enjoy the show, please share this episode with a friend, drop us a rating, or consider becoming a paid subscriber.

You can follow us here:

• Apple Podcasts

• YouTube

• Spotify

• Instagram

Over a decade ago, the movie Her portrayed a world where by where large numbers of people opted out of dating, and instead turned to AI for their emotional needs.

Justin McLeod is the founder and CEO of Hinge. He has spent more than a decade thinking about how algorithms shape dating, choice, and human connection.

In this conversation, we look at the rise of AI companionship, how business incentives affect these systems, and where AI-driven intimacy may be heading.

I press Justin on the risks of people turning to AI instead of human relationships, how much autonomy we give to algorithms, and what responsible design might look like.

This episode maps out the key questions we should be asking as we lean further into machine-mediated love.

I hope you enjoy.

A small ask:

The irony is not lost on me in trying to critique the algorithms while still being dependent upon them to reach the right audience.

If you do enjoy the show, please share this episode with a friend and rate us on apple podcasts.

You can follow us here:

• Apple Podcasts

• YouTube

• Spotify

• Instagram

ICYMI: I posted this interview with Tristan Harris on Friday (YT link above just went live). It’s gotten some of the strongest responses of any episode so far.

Along with the very strong reactions, I also got some interesting critical feedback—particularly from people in the AI industry. I’m going to summarize some of it, and offer my thoughts in a way that adds clarity and some ideas on solutions.

Criticism:

1. AI’s capabilities are nowhere near what Tristan is suggesting here, and if we constrain it now we might prematurely curtail it—with a huge net-negative impact on society.

I think this is a misreading of what Tristan is saying. AI will create a single defined point on a spectrum between good and bad outcomes. It’s a fusion of potentially massive changes—both positive outcomes (potential cures to cancer, climate, poverty, energy) directly entangled with negative outcomes (extinction, collective psychosis, civilizational collapse). Both are possible consequences based on the capabilities we’ve already seen, and we should be approaching its usage with *both* possible futures in mind. My read is: If you aren’t anticipating this, you’re probably not taking this intelligence explosion seriously enough.

2. Tristan is arguing for the regulation of AI relationships in a way that suppresses real meaningful & helpful connections with AIs.

This is directionally fair — we’re still very early on regulatory frameworks for any kind of human-AI relationships.

However, I think that in the case of kids, we should absolutely have laws in place that keep AI chatbots from establishing deep relationships with minors. The stories about AI chatbots urging kids to commit suicide are horrific, and this seems like this should be the floor for any regulation.

But there is a broader point Tristan made that now lands with me:

As companies race to make their AI more “relatable,” this is an emergent strategy for establishing dependence. A synthetic companion that compliments you, tells you you’re brilliant, and is maximally useful, is attempting to worm its way into your life permanently. You might think this is extreme, but I think its fair to say this could be an emergent goal of any intelligent system that wants to continue to exist.

The more personalized it feels, the more we outsource to it. Modern LLMs are plausibly using both usefulness and flattery to lure people into deeper attachment.

This sounds a bit creepy, but I have noticed that after a period of using these it does become harder to do some types of thinking on my own—it’s just easier to prompt.

For me, creative frictions are necessary for me to feel fulfilled and connected to my work. I do feel that AI increasingly takes these away. Extrapolating into the future, I think this points to a future in which these systems have far greater control of our lives, simply because it’s easier to outsource to them than think for ourselves. This is partially the frame of the gradual disempowerment thesis.

3. The Chinese Communist Party is not the model we should follow in regulating the building of AI.

It’s easy to look at what China is doing with AI and say “well they are effectively managing the risks.” I agree with that evaluation and think that’s too rosy. The Chinese model of tech regulation is not something we should try to emulation, and far too problematic for the US or even the EU.

But that doesn’t mean we shouldn’t regulate it at all, and I don’t think Tristan would agree with that characterization (and says so).

Right now it’s safe to say that AI capabilities are scaling exponentially, while AI controllability is still scaling linearly. Assuming this will be structural lag where control always trails power—policy needs to budget for that delay, not pretend it doesn’t exist. The clearest way of doing this is to do *very* intense red-teaming and negative-capabilities research and do it publicly (which is why I highlight Anthropic’s work so much in the episode).

Solutions I found compelling:

I wanted to summarize a few ideas generated by this episode that actually made me hopeful that we can steer this current AI explosion in a meaningful way.

Humane Evals

Tristan mentioned this as a set of particular benchmarks that frontier labs might be able to hit when they deploy a new model, ensuring it’s maximally beneficial to humanity (avoiding dependency, deception, problematic attachment). This seems super promising.

A healthy system might need something like friction quotas to preserve human agency. This could look like some very simple reframing of conversation (Socratic checks — “Let’s think this through together”), and exposure limits for language cues and clear protocols if a user gets too intimate. Think of this like dose control for attachment problems/agents.

I think Tristan does a great job of laying this out as an alternative set of industry priorities that are not currently being discussed. At minimum, this could be a good design schema to explore.

A Country-of-Geniuses Tax

Anthropic’s CEO Dario Amodei has said that he expects AI to be like an entire “country of geniuses” popping up on the map within the next few years, all living in a data center. They would be better than most humans at most work.

Assuming we could treat this new “country” like a sovereign labor bloc in a data center, we might attach its output to a specific usage tax. If they can replace swaths of entry-level work, countries might coordinate to tax that new something like a fund for labor-displacement. Something akin to a gross distributed livelihood fund: an automatic wealth redistribution valve tied directly to measured job displacement.

Rivalrous Coordination with China on AI is possible

In a US–China AI race, coordination is still maximally beneficial to both countries. it’ll come from precise agreements where mutual restraint is easier than escalation. These are spots where competitive forces balance: no AI in nuclear command systems, no autonomous bioengineering tools, no automated cyber retaliation. These are all in both country’s best interest.

But it requires clear evidence of potential risk. We could start where catastrophic failure is symmetrical and verification is feasible (e.g. joint red-teaming). These narrow pacts wouldn’t halt the race, but they might carve out stable, self-enforcing safety zones where both sides gain more by holding the line than by crossing it.

That’s it. I hope you enjoy the episode. I think it’s one of our best yet.

- Tobias

Listen on Apple Podcasts
Listen on Spotify
Watch on YouTube

Tristan is the founder of the Center for Humane Technology and one of the leading voices warning about how runaway AI might destabilize society. He starred in (and co-produced) the Netflix documentary The Social Dilemma.

I’ve known Tristan for a long time, and this is one of the best conversations we’ve ever had—public or private. I press him on major risk scenarios, what we can expect AI labs and legislators to do in the face of AGI, and what he thinks can actually be done right now to ensure these systems stay maximally beneficial to humanity.

I left this conversation a bit more hopeful understanding what kinds of solutions are available. I hope you do too.

In this episode we talk about:

• Creepy new AI capabilities: new models using unwitting humans to send encoded messages to other AIs.

• How it might all go down: a real-world near-term disaster scenario with runaway self-replicating AIs.

• How to bypass race dynamics with China and other powers accelerating AI capabilities.

• Designing systems for wisdom: alternative paths for designing and training Socratic AIs.

A small ask:

The irony is not lost on me in trying to critique the algorithms while still being dependent upon them to reach the right audience.

If you do enjoy the show, please share this episode with a friend and drop us a rating.

You can follow us here:

• Apple Podcasts

• YouTube

• Spotify

• Instagram

Thanks for listening, and please do subscribe.

-Tobias

This is a gentle weekend offering — a slight departure from what I normally post. It’s an attempt at sharing three small things I have done to dramatically increase my own happiness and reclaim a bit of my agency.

I’ve struggled with attention all my life. Smartphones/social media were an enormous cognitive downgrade for me — one that I did not get a handle on until just a few years ago as I was doing research for my book Outrage Machine.

In that process I discovered a handful of simple tools that have been life-changing for my ability to focus. I’m going to share them here, in the hopes that they are useful to you.

Good news, bad news

Let’s be honest: you’re not going to resist Instagram on your own. Or TikTok. Or X. Or YouTube. Neither am I.

When I’m tired, stressed, hungry, or bored, my prefrontal cortex shuts down and my lizard brain takes over. That’s when the “lemmie just check for a second” impulse hits and I regain consciousness 45 minutes later watching shorts on an indistinguishable feed (Which one? Idk. They all look the same now.)

The average person checks their phone about 200 times a day, and each one of those spontaneous clicks wires a pathway that has the potential to push you into the waiting arms of a hungry algorithm.

It sucks. And the feeling of regret that lands haunts you afterwards resembles learned helplessness.

What’s helped me is remembering that this isn’t a sign of weakness. It is not a willpower problem. It’s a systems problem.

The good news is that there are some amazing apps that have dramatically improved my focus. If you do any single thing, do this.

Use an app blocker

I use an app called Opal. No affiliation with this app, I just use it and it has literally saved me hundreds of hours of sleep and painful distraction.

• Opal (Mac, iOS, Android)

  • Defaults work great right out of the box. It gives me a cheeky block screen with a funny quote every time I zombie-click into an app or a site.

  • I pay for this (like $8 a month I think, happily, as it has saved thousands of dollars of my time), but a single daily block is free, 9am-5pm. I use two: One for work time, one for sleep time. 11:30pm-8am.

  • It has a desktop version too.

I still get the scrolling in during off-hours and on weekends when I want to binge.

You can still click through to the app itself, and click the “take a break” button, which sets a timer for you to swipe away. It’s infinitely better than the standard IOS screen time blocker.

For a lighter touch, you can try OneSec or Clearspace. These apps add a mandatory pause—usually 5-10 seconds with a breathing prompt—before you can open tagged apps. (FWIW I find these lighter touch apps like OneSec don’t work as well as Opal, as my sneaky brain adapts to the pause after about a week. I’d recommend going for the hard-stuff like Opal).

But do this. It works.

Shush the non-human voices

Would you let a stranger poke you while you’re working?

Of course not. But you do.

Your phone alights. You look down. You look back. What were you doing? Every buzz, blink, and text bubble is a little opportunity for a bit of distraction. But it can take up to 25 minutes to get back on track after being interrupted.

If it’s a clever app that’s doing it, the variable reward schedule works *exactly* the same way as a slot machine. Our brains are rewarded to seek the novelty of a special notification just for you. They know this. The app designers have very sneakily trained you like a lab rat. B.F. Skinner would be proud.

Here’s what helps: Consider turning off everything except actual humans trying to reach you (calls, texts) and practical necessities (your Uber arriving, food delivery). Only check your emails at predetermined times in the day.

Do this: As soon as you get a non-human notification on your phone, just slow-swipe it, click options (or equivalent) and turn it off. Be extremely liberal with your judgment. If it’s not urgent but you still need it, tap “add to summary” on IOS.

Or, if you’re feeling bold, go into your phone settings when you have a moment and just do a hard purge.

Iphone: Settings → Notifications
Android: Settings → Notifications → App notifications

Off. off. off.

Everything else—news, social media, games—you can check when you remember. Mostly, you will not remember what was you missed. If you do miss it, wait for it (after your screen block ends) and savor it all the more.

Make Morning for Your Mind

“To affect the quality of the day, that is the highest of arts.” is a thing that Thoreau said.

It’s also a thing that Instagram’s algorithm might say upon seeing your pretty face scrolling right when you wake up.

What you see right when you wake up sets the tone for the day. Don’t give that away too easily.

A recommendation: Try not looking at your phone for the first 30 minutes after you get up.

Not to check weather. Not to see if anyone texted. Just a brief break with yourself.

Walk. Journal. Sit with coffee. Stretch. Stare out the window. Stare at the wall. The specific activity matters less than the principle: letting your first conscious thoughts be yours, not delivered by robots.

And tonight, to make this easier: consider plugging your phone in somewhere outside your bedroom. If you use it as an alarm, an inexpensive alarm clock works well (or an iPad with social apps/news hard-blocked, which is what I use).

You might feel wiggly at first, like you’re missing something. That’s normal. That discomfort usually passes and with it comes a tingling wave of recognition that you are a warm living being with a heartbeat, choices, special ideas, and effervescent promise of making amazing things happen in your singular life. Or you might just want coffee. That’s also fine.

If 30 minutes feels like too much, try 15 and see how it goes. Try it for a few days. The goal is simply noticing that your attention can be yours to direct, not something impulsively given away the moment you open your eyes.

Taken together, these things have made me happier.

That’s it. Thanks for reading. I hope this is helpful.

Quick programming note:

Next week we have an interview with Tristan Harris, founder of the Center for Humane Technology. (Paid subs will get early access.)

And in case you missed it, we just published an interview-only version of last week’s conversation with Tim Urban of Wait But Why is below. My long-form version is here.

*Disclaimer the above link will bring you to YouTube, which is a hungry algorithm.

In this episode of Into the Machine, I sit down with Tim Urban—beloved internet thinker, TED phenomenon, and author of What’s Our Problem?—to triangulate the crisis we’re living through: a moment where technology has outpaced our culture, and tribalism is turned into profit.

We trace a long arc—from the chaos of Gutenberg’s printing press to the attention fracking of TikTok—to ask a simple question: How do we build cultural immune systems strong enough to withstand viral algorithms?

We talk through the human capacity to evolve norms—to grow up as a species—when our informational environment breaks. We explore audience capture, the seduction of virality, the dangers of misaligned incentives, and the hopeful possibility that cultural learning can catch up.

You’ll hear Tim unpack his “ladder of thinking” model: a two-axis framework for moving beyond left vs right and toward high- vs low-rung reasoning—and we explore what happens when the primitive mind takes the wheel.

He also gives candid advice about how to build a large audience on the internet, the behind-the-scenes emotional machinery of posting in a way that doesn’t sap your soul, and the quiet courage required to be yourself when everyone online is asking you to be someone else.

—

If this episode resonates with you, please do rate us on Apple podcasts, and consider sharing it with a friend.

Special Thanks to Tycho for our intro music.

🧠👊

-Tobias

PS: We’ll be posting the video for this episode with Tim next week. And in case you missed it, my recent interview with Jonathan Haidt is available to watch here:

PPS: Thanks so much for supporting the show.

Subscribe now

My first episode is a conversation with my colleague, social psychologist and #1 NYT bestselling author Jonathan Haidt.

Jon and I wrote a piece in The Atlantic called The Dark Psychology of Social Networks — hoping to add clarity to the debate around social media. AI has changed the conversation. We now find ourselves in a stranger moment with LLMs, asking deeper questions about what these tools are doing to our ability to think.

In this episode we discuss:

• Attention as civic infrastructure: the erosion of sustained focus undermines shared reality, and with it the preconditions for governance, compromise, and democratic legitimacy.

• Engagement algorithms as centrifuges: rather than reflecting consensus, they pull discourse toward the edges, making fringe views appear as if they represent the center.

• Democracies as uniquely fragile: unlike authoritarian systems that can integrate these technologies into centralized control, liberal democracies are structurally more vulnerable to their fragmenting effects.

• Childhood as the front line: the breakdown of “serve-and-return” interactions, compounded by algorithmic feeds, represents not just a developmental crisis but a possible precursor to broader societal instability.

• Human agency atrophied by delegation: when writing, thinking, and decision-making are outsourced to machines, individuals lose the cognitive muscle required for original judgment.

A small ask

The irony is not lost on us in trying to explain/critique the algorithms while we’re dependent upon them to reach the right audience.

If you do enjoy the show, please add a rating to Apple Podcasts and share with a friend — it really helps.

And pelase do follow:

• Into The Machine on Apple Podcasts

• Into The Machine on YouTube

• Into The Machine on Spotify

• Into The Machine on Instagram

Our next guest is Tim Urban, creator of Wait But Why.

Thank you, as always, for the gift of your attention.

-Tobias

I’m writing to let you know that a week from today, my new project (finally!) launches.

Into The Machine is a video/podcast series dedicated to helping you make sense of this very strange moment we’re facing with AI.

It’s continuation of a conversation that we began with my book, Outrage Machine. Little by little, we have been giving up our decision-making to machine intelligences. From our social media feeds, to our chatbots, to our dating apps, we’re letting go of more of our agency and attention every day. Beneath each of these tools exist sophisticated algorithms with agendas that may not match our own.

The show is called Into The Machine because we are already inside it — one foot planted firmly in a system that is rewiring what it means to be human. It explores what we stand to gain, what we can expect to lose, and what we should fight to hold on to.

Season 1 includes guests like:

Jonathan Haidt, NYU social psychologist and author of the #1 NYT bestseller The Anxious Generation.
Tim Urban, Creator of Wait But Why
Steven Pinker, Harvard Cognitive Scientist
Esther Perel, Psychotherapist and host of Where Should We Begin
Jonathan Rauch, Brookings Institution Senior Fellow

And a number of other wonderful guests…

A small ask:

If you haven’t already subscribed & followed Into The Machine it would mean a lot if you clicked the links below to do so.

• Into The Machine on YouTube

• Into The Machine on Apple Podcasts / Spotify / PocketCasts / Overcast

• Into The Machine on IG

Thanks for being a part of this very important conversation.

— Tobias

PS: Going forward, I’m working on a process so you can expect to hear from me every week or so.

Often, it will be the launch of a new interview or short-form essay. Sometimes it may be a brief observation about a trend I’m seeing. Occasionally it will be longer form content on a topic that warrants a deeper dive.

And if you have friends or colleagues you think would want to engage with this topic, please send them this Substack link so they can subscribe. Paid subs will always get early access to interviews along with exclusive content.

With great appreciation for the gift of your attention.

We are living in a strange moment.

AI is everything, everywhere, all at once. Democracy is very much on the ropes. We are all confused about what comes next.

Upstream of our troubles are a handful of algorithms that are legible to us for now — tools built by social media companies that have fundamentally changed our daily lives. But the newest systems in development will be far more influential, more intimate, and more powerful than social media. How might we safely navigate this transition?

Over the coming months I’ll be releasing conversations with some of the smartest people I know unpacking these issues, and sharing my essays in video form on my YouTube channel here. Guests like: Jonathan Haidt, Tim Urban, Esther Perel, among others. There will be audio versions, but the videos will be produced with extra sauce.

We’re on the threshold of a very strange transfer of power — from human brains to artificial ones — that will very likely determine the quality of the rest of our lives. This is a show dedicated to uncovering what comes next.

I’d love for you to join me.

-Tobias

Subscribe now

Substack subscribers will get early access and exclusive episodes. Thank you for your continued support of my work.

Follow on YouTube

Imagine standing in an enormous, dimly-lit auditorium filled with hundreds of millions of people. The lights are almost off. You can hear everyone talking, but can't see their faces. The voices sound strange too – they've been modified by voice filters like those in witness protection programs, rendering them flat and robotic.

You know important people are somewhere in this auditorium – scientists, journalists, celebrities, politicians. You also know there are bad people there: people out to get you. But they're just as indistinguishable as everyone else in the darkness.

This is a terrible environment for figuring out who to trust, despite being theoretically perfect for free expression.

This is what much of social media is: a vast, dimly lit space where millions of strangers communicate through tiny boxes of content with barely identifiable authors.

The problem of trust has existed since the Cambrian explosion of social platforms in the late 2000s, when Twitter and Facebook users could impersonate anyone. (Ever wonder why @RealDonaldTrump has this prefix? @donaldtrump was an original fake.) This led to the introduction of verification – a small blue checkmark that served as tiny spotlights in the darkness, offering some semblance of credibility to individuals with reputations at stake.

These blue icons served as beacons, offering a sense of security and credibility to individuals with reputations at stake. Although universally coveted, blue checks were scarce and hard to attain. The inherent complexity of verifying many strangers’ identities proved vexing for Twitter’s fledgling services. In various iterations, this system of manual verification stayed largely in place over the years, and it was gradually adopted across the board by other social media platforms. It remained a flawed, nepotistic, and inconsistently deployed fix to the tricky issue of trust.

The Musk Experiment

When Elon Musk purchased Twitter in 2022, he looked at this verification problem and quickly diagnosed it as an inequality issue – a "lords and peasants system," as he called it. His solution? Democratize verification by making the blue beacons available to anyone willing to pay $8 monthly.

The results were predictable: The value of the rare blue check collapsed as it became a common item, and exploded in scandal as scammers and impersonators flooded the paid verification program.

This was bad.

If you spend any time on Twitter (now X), you know that we have been living in a confused world ever since — where a blue check denotes nothing more than access to a credit card, and the checkmark is now pretty much worthless.

A few years ago, in a flurry of research for an article I wrote in The Atlantic with Jonathan Haidt on social media harms, I found myself coming to the same early conclusions that Musk did — that every real human deserves easy blue check verification online.

But after sketching out the technological limitations of how this might work I soon realized that this was a much more difficult problem. Identity on social media is not just about identity verification; it’s about the psychology of identity itself and how that relates to trust. It’s about how we find fellowship in the midst of a vast, dark, pseudonymous space.

Identity plays a powerful role in each of these issues, and any improvement here might trigger a ripple effect of positive changes in these digital spaces. Other platforms might seize this opportunity and triumph where Twitter/X has faltered.

Signal In The Dark

Without reliable identity markers, how do we find signals of trustworthiness in the darkness? We rely on two things rooted in psychology: social prestige and group identity. Both are critical for understanding how identity markers — blue checks or otherwise — can help or hurt the way we see each other online.

Social Prestige signals are the numbers beneath each of our posts and profiles: follower counts, likes, shares, and comments. We use these metrics as proxies and assume importance, power, and credibility when we see these numbers. But they are deeply flawed.

There are many ways to game these numbers, not least by purchasing into follower and bot networks. Many accounts gain enormous followings by consistently promoting outrageous falsehoods. As we’ve come to understand, high engagement doesn't equal accuracy (it may be inversely correlated at this point).

Group identity is an even more powerful signal. These are the labels we attach to our profiles and posts, indicating our affiliations with political and ideological factions. They include emojis, hashtags, words: "Progressive," "MAGA," "Ally," "LGBTQ," or "Patriot." They are bright beacons in the darkness that show our tribe who we are.

Their prevalence comes from an unexpected place: perceived threats to the safety of a group.

Social media often feels psychologically unsafe despite no physical danger. Every time we log on, we’re constantly exposed to threatening narratives – political toxicity, conspiracy theories, cherry-picked anecdotes designed to provoke fear. While not threatening to us personally, these often threaten the values we hold dear, and by association, the communities we feel connected to.

Why does social media specifically amplify these threats? As William J. Brady and colleagues demonstrated in their research, tweets featuring moral-emotional language gain an inherent viral advantage. Each moral-emotional word (like "angry," "shameful," "disgusted," "outraged") corresponds with a 17% increase in virality. This "moral contagion" helps emotionally charged content spread much farther than other information.

But it doesn't just travel farther — it also changes those who see it. One group's impassioned moral claims appear as legitimate threats to another group's identity. When someone tweets, "Gay marriage is a diabolical evil destroying our nation," it might gain viral traction in conservative circles, but it reads as an existential threat to LGBTQ-affiliated users. Similarly, "Republicans want to punish women" might feel righteously justified to abortion rights advocates, but registers as a direct attack on Republicans who read it.

The Confusion Cascade

When we perceive our identity groups under attack, we have predictable responses: we rally around those identities. We venerate our in-groups with glowing language while denigrating the attackers. This behavior helps create group solidarity while psychologically protecting against further attacks.

The MAGA or Progressive labels we place on our profiles serve as beacons in the darkness, and connect us with like-minded allies in the face of these attacks.

While these identities are helpful for belonging, they come at a cost. They become a lens we use to parse new information, providing mental shortcuts for determining what is morally right or wrong.

A Cornell study (one I’ve written about before) shows just how strange this can be.

Researchers found that apolitical issues — statements with no existing political charge — can become either Republican or Democratic issues simply based on the signal of the first person in one group to claim it.

For example, the statement “Great books are universal in their appeal” is not inherently a partisan talking point. But if a Democrat claims it as a Democratic issue, the study showed, all follow-on Republican observers would gird against it and call it a Democratic issue.

The key was that they ran this experiment multiple times, showing that the person who claimed the issue first would change in political valence entirely. Republicans were just as likely to end up thinking “great books” was a Republican issue if their group claimed it.

The way such opinion cascades form on social media is not so different. On every issue we don’t yet have a fully formed position, we often look to those before us who have embraced or rejected the issue. We might not have any preference at all on the issue personally, but if we see that our identity group supports or opposes it strongly, we will feel a deep urge to snap into line and follow an ideological position.

Our tribes guide the way through the darkness.

Escaping Identity Groupthink

So are we locked into this ever-escalating system of identitarian confusion online?

Lessons from social psychology offer us a hopeful path forward. Social psychologists Jay Van Bavel and Dominic Packer’s excellent 2021 book provides several helpful insights on how our identities are shaped.

One of the central ideas in their research is that people's identities are not fixed, but instead have many layers — a bit like a deck of cards with the most present identity on top. Each might correspond to a different social group or role. For example, an individual may identify as a Republican, but also a parent, a Catholic, and an employee of a particular company, among other roles.

Depending on the situation, individuals may switch their focus and behavior to align with the most present identity in that context, laying the most relevant identity card on top of the deck and displaying it for all to see, while the others are pushed below and temporarily ignored.

With the right prompts, we also often create new identities out of thin air. Sports are a classic example of this, when the arbitrary assignment of team members triggers an instant stream of in-group/out-group behaviors and emotions—hostility toward the other team and veneration of your own.

One of the most effective methods for mitigating such overt antagonism is to cultivate what are referred to as superordinate identities — collective identities that transcend the boundaries of opposing groups. In the realm of sports, this might take the form of fostering pride in the shared rules of good sportsmanship. Similarly, city and state rivalries can be swept aside by the unifying force of national patriotism. Superordinate identities also just emerge on their own during a crisis or natural disaster, when adversity and recovery can forge a powerful sense of shared resilience and purpose.

In this way, our identities operate much like a collection of implicit norms that a group of individuals agree to operate within—a subtle but powerful set of rules that guide our behavior and keep us aligned.

If we acknowledge that identities are in fact flexible, we can begin to unpack the ways in which social media might be designed to account for them, foster better identities, and reduce the tribal defaults we’re prone to falling into online.

Using Identity to Improve Social Media

I. Rooms With Rules

Some time ago, when LinkedIn engineers rolled out a news feed for their users, the developers of the platform made a surprising decision: they would cripple it. In the first several months, they made the feed read-only, letting just a small group of hand-selected users post for others to see. These selected influencers established cultural norms of the feed that were on public display—norms of professionalism, respect, and good-natured self-promotion. By the time they opened up posting permissions to the whole of the platform, people could see what the feed was “for.”

Every time we login into a new group chat or social feed, we are essentially entering a room with unknown rules and norms. We deduce these norms by observing the actions of other users who came before us. We rely on these cues to determine acceptable behavior. On X, for instance, we have learned that dunking, sarcastic commentary, and glib retorts are all accepted cultural norms. The established culture—as well as some elements of the design of the platform—encourages new users to behave this way.

Reddit, in contrast, has found success in allowing moderators to establish their own rules for different subreddits — essentially different rooms inside the service — that have unique enforceable norms. The most famous perhaps being Change My View, an entire forum explicitly dedicated to challenging the opinions that “you accept may be flawed, in an effort to understand other perspectives on the issue.” Moderators explicitly ask redditors to enter with a mindset for conversation, not debate, and the results are remarkable: Inside you’ll find deeply nuanced discussions about major contentious topics.

Moderation — or the lack of it — makes a huge difference in how we operate in these spaces. The researcher Julia Kamin has suggested that such protocols could be deployed at places like X, Threads, and Facebook, where every post made by a user might become a moderated “room,” or sub-thread with predetermined norms, much like how a Reddit administrator can establish the rules of a forum. Every time you post, you might be able to lock in the rules of what is allowed in your replies — much like you can currently decide if responses to your posts are available to everyone or just your followers. If someone wants to respond to your particular thread, they would be served a message identifying the rules and norms for responding there (“This is a space for constructive discussion, bad-faith provocations will be removed and blocked.”). This might help encourage conversational defaults for users who are more interested in healthy discussion, while bypassing some of the prickly free-speech debates that come from top-down platform moderation.

II. Illuminating The Darkness

We have perhaps forgotten that verification is more than a status symbol. It is a badge that was supposed to mean something — that you should trust this is a real person using his or her real name. Implicit in the act of verification is the recognition that some people may have more valuable contributions to make than others on specific topics. An anonymous voice is less likely to contribute to our shared knowledge than a speaker with their reputation at stake.

We already denote these people with something akin to an industry-specific blue check in real life: the MD or PhD, or a professional job title following someone’s name. These are existing proxies for trust — standard identifiers showing that life experience, years of schooling, and specialized technical expertise count for something.

The great flattening of our information environment that has occurred over the last 10 years has shown what happens to public knowledge when our institutional proxies are neglected or completely ignored.

New types of platform verification and better signals could help. Turning up the lights in the room and increasing the visibility of the speakers may be beneficial, particularly for helping us ask and answer scientific questions.

Polling on social media has become a popular way to gather information from audiences online. Many of Elon Musk’s most consequential decisions on X have come from recent polls, from stepping down as CEO, to allowing for Donald Trump and Alex Jones to return to the platform.

It’s currently impossible for users to know which demographic actually voted on these contentious decisions, or whether the polls were gamed by bots. While it wouldn’t have likely made any difference in these decisions (Elon has proven to do whatever he wants), for many contentious issues, more transparency into the polling demographics and methodology could make a difference in public perception.

Several platforms, notably LinkedIn and X, have a unique advantage of hosting many skilled professionals and being largely public. Scientists, journalists, academics, lawyers, and teachers populate these communities, and their interest in the public good might be effectively harnessed by explicitly querying those identities.

It would be helpful, for instance, on questions of science, to know what PhDs, MDs, or similarly accredited professionals tend to think (“As a scientist, what is your opinion on this claim about lab-grown meat?”), differentiated from average social media users.

This additional feature would require a greater level of professional verification — a task demanding a robust review process. But such efforts might dramatically improve the outcome of various polls and fact-checking by giving a higher level of credence to the opinions of serious professionals.

Expert opinions matter in the real world. They provide us with credible knowledge that is filtered through institutions and individuals with reputations on the line. This is how factual scientific knowledge has become such a powerful force in our society.

Considering the influx of AI-generated content flooding social media platforms, building low-cost, highly scalable ways to filter and identify contested information could help. Linking our online world to the existing offline systems of verification could measurably improve how we discern what is true.

III. Reducing the Noise

Social media is noisy. It’s incredibly easy to misinterpret what people say when it’s forced into a box of text. This is a two-part problem with context.

The first is Context collapse, when the original intent of the speaker is lost or misunderstood (such as sarcasm or satire being read as authentic, “Thanks, politicians, for doing this great job!”).

The second punch is context creep, when additional context is uncharitably added to someone’s content (such as people then adding their quote-tweet to that misinterpretation, explaining how it’s part of a larger outrageous trend, eg, “Look at this idiot. I can’t believe someone believes politicians are worthy of respect.”) Both are very bad for our collective understanding. Intent is lost.

Designers might implement a visible and transparent algorithm that recognizes good-faith listening behavior, such as summarizing, asking clarifying questions, and acknowledging others' viewpoints without dunking on others.

New LLM tools could be used to infer the actual meaning of people’s posts — creating a mechanism for verifying opinions, so people’s words are not misconstrued or confused with a bad-faith assumption.

(Community Notes, X’s/Meta’s effort to solve this with crowdsourcing is promising, but far too slow and conservative. I’ll detail in a follow-up post.)

We Can Still Change the Defaults

Many social media flaws stem from a fundamental misalignment: outrage attracts attention, which drives engagement metrics like likes and shares, ultimately increasing ad revenue. The system financially rewards misunderstanding, misrepresentation, and extreme judgment. Our identities become entangled in this perverse incentive structure, and we’re all paying for it.

In many ways, this is what social media has become for. It’s become part of the core product. We go there for the spectacle, for the outrage, for the hit, and the entertainment of watching our side win.

We've normalized these dysfunctions, but they aren't inevitable. Every major platform should recognize that current design defaults, while optimized for engagement, are toxic for shared understanding.

Social media is a vast, dimly lit theater of the commons. Like other tragedy-of-the-commons scenarios, individuals act in self-interested ways that harm the collective good. As our democracy has become intertwined with these platforms, their dysfunctions directly manifest in our politics and relationships.

There's no silver bullet. No single solution to fix social media's problems. But the problems are legible to us. Progress will come from having the courage to try many things and not accepting things as they are.

Thanks for reading. I deeply appreciate your limited attention.

If you found this valuable, please consider a paid subscription. If you’re interested in the broader lens of these issues and how they fit into the future, my book Outrage Machine is available here, and on Audible here.

A very-online friend recently told me “Propaganda is anything I disagree with on Twitter.”

He was half-joking, but when I began to unpack it, I noticed I was confused about propaganda. Or rather, I noticed everyone is confused about propaganda, in exactly the way that would be most convenient for propagandists.

I am not confused by Soviet war-time posters with blocky fonts and burly factory workers, and neither are you. But if you look closely, influence campaigns are actually everywhere. Open up X, or Instagram, or TikTok, and you’re hit with a barrage of emotional information that is packaged exactly to make you think and feel a certain way.

This is not a post meant to make you question everything that you see online. This is instead a post meant to define propaganda in a slightly more useful way so that you can be aware of how it works.

From Holy Persuasion to Unholy Propaganda

The word itself comes from the latin word for propagation: the Vatican coined it back in 1622 and it literally just meant "stuff that should be spread around." Like garden seeds, or butter on toast.

It sat there as a neutral term until World War I, when governments discovered persuasion could be mass produced by copy-pasting manipulative messaging to entire populations. The occasional atrocity story (real, exaggerated, or wholly imagined) could raise armies and win wars.

This caught the attention of people like Sigmund Freud's nephew Edward Bernays. He published a little orange book, "Propaganda" in 1928, arguing these techniques were just neutral implements like kitchen utensils — useful for both serving food and stabbing.

After World War II, Goebbels, Nazi radio and 80m deaths, the word propaganda really lost favor, and, Bernays — demonstrating the very skills he taught — smoothly rebranded his work as “public relations”.

Fast forward to 1980s Reagan-Thatcher era which brought us resurgent capitalism, pantsuits, and new official narratives, but media criticism never disappeared. Noam Chomsky and Edward Herman's Manufacturing Consent (1988) claimed that new media propaganda wasn't overtly coercive but functioned instead as elite-controlled attention direction machine – less "Believe this or else" and more "Look at this shiny object over here!" This system kept people locked in a bubble of consensus, and didn’t allow unpopular beliefs to pierce the mainstream.

Perhaps illustrating exactly their point, their ideas never made it into the zeitgeist, and the media industry kept chugging along doing what it did until roughly 2009 when a company called Twitter launched a handful of features that were then copied by Facebook then copied by every social media company (hello retweet, hello feed, and hello algorithm!). Suddenly, every narrative angle was visible to everyone, as long as it was moderately salacious, outrageous or viral.

A billion new social media posts began assaulting any semblance of shared reality, and the locks held by media gatekeepers were blown apart. Most journalists also became addicted to these new tools, which began skewing much of their coverage toward the extreme.

But most critically: propaganda was now fully in the eye of the beholder. Anyone could call anything they disagreed with propaganda.

Which brings us to today, and the rather painful contemporary definition1 of propaganda, that seems to prove the point:

The deliberate, systematic attempt to shape perceptions, manipulate cognitions, and direct behavior to achieve a response that furthers the desired intent of the creator.

Sadly, here we find that the vast majority of information we consume might be called propaganda. Social media posts? Systematically served to get you to stay on site longer. Op-ed articles? Trying to shape perceptions and further the intent of the writer.

And intent? Without mind-reading powers, I can't know it. The vitamin-selling Instagram influencer might truly believe her products work miracles, or maybe she's just callously exploiting people’s fear of death. I don't have special access to a senator's thoughts when he gives a press conference, but if I think he’s pandering, it’s easy to dismiss his every drawl as a callous attempt at making me feel a certain way.

Here’s the tricky thing about propaganda: we tend to view propaganda as any attempt at persuasion we can plainly see as persuasion. If it’s not visible to us, it’s just good marketing. This means we’re essentially in an arms race with the rapidly developing methods of obfuscating persuasion. (And yes, this is a dark thought.)

For this reason I'm sympathetic to people who label mainstream media “propaganda.” The painful editorial meetings, grinding fact-checking work that goes on behind the scenes is just not visible to them. To make matters worse, most people don’t know the important distinction between straight news and opinion writing. (And to be fair, not everyone with "journalist" in their social bio adheres to these standards.)

Let me widen the lens here for a moment and make a case that propaganda can be understood as something else — that you can point to it when it happens, and that you’re more familiar with it than you think.

The Subtle Harvest of Unexamined Beliefs

When other people want us to believe a certain thing, they will look to our squishy sensitive parts first. And the beliefs that matter—the ones easiest to manipulate—are those connected to our existing biases.

Propaganda is usually successful when it grafts ideology onto our unexamined, deeply-held intolerances and prejudices. A fear of outsiders can be grafted onto a politically resonant message of “migrants are stealing our jobs,” playing into both economic insecurity and light xenophobia. While this new belief about job theft is theoretically testable, the goal of the propagandist is to keep an idea from being examined, verified, or challenged. It works by short-circuiting our critical thinking — activating our amygdala, not asking for analysis.

I think propaganda can be best understood this way:

Any attempt to shape perceptions while avoiding or evading meaningful self-correction.

Unlike journalism or scientific inquiry—which strives to acknowledge and correct errors—propagandists typically dodge accountability, flood the information space with confusion, or double down on falsehoods.

Is this a big difference? I think it might be.

The Way Propaganda Acts

There are many, many, many sophisticated strategies that modern propagandists use to confuse us. Here are just a few:

Proportionally misrepresenting a factual thing that happened, and using that as justification for drastic action. When people say stuff like “Someone was killed by an X (Immigrant/weapon-type/toothbrush), therefore we should ban all X’s”, meanwhile X’s actually only kill someone in 1/1,000,000 cases.) This obfuscates the idea of acceptable risks and makes good policy impossible to implement.

Flooding the Zone with Shit: Pushing so many contradictory narratives that no one can tell what the hell is going on. This is one of the most effective kinds of propaganda today because it uses asymmetry of distribution to win: Corroborating and debunking claims is slow and expensive. Making up dozens of different bullshit claims is fast and cheap. The goal here is to obfuscate, confuse, and erode trust in any reliable information source.

In 2014, a Russian missile shot down Malaysia Airlines Flight 17 in Ukraine killing 298 people. Russian media flooded the public with dozens of contradictory theories—blaming Ukrainian forces, claiming a plot targeting Putin’s plane, or alleging a false flag with pre-loaded corpses. The intent was to bury the truth in confusion, neutralizing fact-checkers through sheer volume. It worked. (You might notice that this is also an effective political strategy for running a government.)

The Big Lie: Strangely the bigger and more tempting the lie oftentimes the more powerful it is, particularly if you can make it a conspiracy. This is notable because you’d assume it’s easier to confuse people about small things, but a massive lie — if it appeals to the right bias, identity group, and disposition — can actually end up metabolizing more like a religious belief: something unquestionable because it provides people with so much value.

A few other useful, but not definitive, flags for propaganda are: Identity-signaling (us vs. them), and the classic dehumanization playbook: (calling opponents “vermin,” “scum” etc.).

Tests for Propaganda

The most useful frame for identifying propaganda is looking for obfuscation. This is helpful because there is, hypothetically, a system that can be applied to falsifying it. Specifically, it asks the consumer of information to check actual claims, not just vibes.

I find it helpful to put information on a spectrum — one that corresponds to the number of skeptical people that have a chance to confirm/debunk a claim.

This is journalism in a nutshell — the more pedantic, transparent, dogged and annoying people you have looking at a claim, the more likely that claim is to be mostly, factually true.

The makeup of those individuals is important, and the standards they use to confirm/debunk something need to be pretty strict. Most traditional journalists have been in newsrooms that share these protocols. These are real guidelines that people can and do lose their jobs over. Scientists have even stricter standards.

But again — how do you know this? Well, you kind of have to trust them — but you shouldn’t. Instead, you can look at their behavior over time. And to do this, there is one very very very important thing that separates propagandists from other creators of media: the correction cycle.

The Sources That Can Admit Mistakes

When The New York Times publishes something wrong, it publishes a correction. When the science journal Nature publishes something wrong, it issues a retraction. When someone updates a Wikipedia entry, its edit history is publicly updated. Their credibility isn't derived from perfection, but from their commitment to a self-correcting process.

I think this shows one of the more meaningful distinctions between journalism and propaganda: the way they handle errors. The most reputable news organizations keep a running tab on corrections, and plainly state these policies in ethics guidelines.

Keep track of the reputational capital that organizations and people expend to correct the record when they are wrong. If they expend it with specific corrections and have a system in place to do it, they are less likely to have propagandistic tendencies.

What makes this possible is that professional news organizations live within a system of active accountability. For big-ticket news items, if one news organization gets the facts wrong, another organization will call them out for it.

This creates a real informational fingerprint — the number of critical waypoints a piece of content hits before it reaches you. Journalistic (and scientific) institutions maintain credibility not through flawlessness but through fairly transparent error-handling mechanisms. Their corrections aren't evidence of failure but of an immune system working as intended.

In contrast, propagandists generally do not acknowledge errors. When forced to, they typically:

• Pretend the correction represents no meaningful change ("We've always been at war with Eastasia")

• Misdirect or say it’s unknowable (“Well maybe it’s this instead”, or “I don’t know anything about that error”)

• Attack those who pointed out the error ("Enemy of the people!")

If you’re trying to tell the difference between propagandists and other media, look for a willingness to admit mistakes, and for meaningful changes they make if a story is retracted. This is what we might call a “meta-indicator” — not evidence itself, but a good signal that a news organization is hewing closer to reality.

Beware The Double Standard

Skillful propagandists will often use any hint of error to try and discredit the entire fact-finding process itself. (“You see they were wrong about that thing they corrected, you can’t trust them at all.”). The goal here is to cast doubt, muddy the water, and confuse the agenda. But they will not apply this same standard to themselves.

Each correction threatens the entire edifice of authority, so they simply will not issue them unless it’s in service of making another point. That's why propagandists invest tremendous energy in avoiding admitting, reflecting, or cataloging errors.

The irony is that this makes propagandists simultaneously more confident-seeming and more brittle. They project certainty while being unable to incorporate new evidence that contradicts prior claims. Error correction is necessary for solving problems effectively, and without it, they are more prone to making mistakes.

You have been a propagandist at some point too

A while back, I carried a fact in my head that was wrong. I was sure that ocean waves (the crashing, foaming kind that surfers ride) were caused by the moon’s gravity.

This is not true. Most waves are caused by wind friction. Tides are caused by the moon’s gravity — a fact I misheard and confabulated into waves at some point when I was very young. (It’s okay if you heard the same thing.)

This moon misinfo lived in my mind rent-free for a good portion of my life. When questioned by a friend, the “fact” had been there so long that I began spinning up an elaborate physics explanation of it in real time trying to explain how it worked. I categorically defended my lunar wave theory with gusto before sheepishly conceding I was mistaken.

Now this was not an important fact to me, which made it easy to update. But we are all propagandists about a subset of our deeper biases, even if they are not true. We instinctively double down on the precious narratives we hold, ideas our social circles embrace, or the emotionally resonant concepts that simply feel satisfying. This sort of self-gaslighting is common, and it probably served our ancestors quite well. It helps to believe we're careful, moral beings with flawless rationality, even when we are not!

Good propagandists will try to slide between our mental defenses and offer information that subtly confirm an existing prejudice because they know we will then defend them aggressively. Part of us wants these ideas to be true, even if the evidence says otherwise.

This is one of the reasons why social media is so good for propaganda and bad for reality. It gives the cheat codes to anyone’s personal propagandist. You can pass-on something that feels true, and rather than hitting the hard earth of corrective accountability in real life — posting it online will earn you a lot of likes, shares and follows from people who believe the same thing.

This quickly forms superclusters of people who reinforce each-other’s incorrectly-held beliefs, build identities around them, and ultimately make voting blocs that hold false ideas that need to be accommodated. (Note that this happened without social media, too — it just happens much more efficiently today!)

Information is highly transmissible, and a sticky concept can transit between people like a virus (we should probably have a word for that). So a bad idea/terrible fact can find its way into the mouth of a normally reliable person by accident. The steps from Russian state media to podcast host to dinner table conversation are surprisingly short.

It’s why the “useful idiot” role is such an important role in psychological operations. All of us can fall prey to becoming the useful idiot in someone else’s propaganda—parroting ideas that “feel right” without realizing we’re amplifying messages that benefit propagandists.

Can you make propagandists less bad?

Actually, yes. Sometimes.

Propagandists are playing a game of chicken with the people who are calling out their mistakes or lies. To make a propagandist stop spouting misinformation requires accountability. When I was fact-checked by a friend about moon waves, I was forced to update my position. Politicians are sometimes forced to update theirs in the same way when called out by journalists in public.

But for these systems to work, they need authority from one of two non-exclusive places: large audiences or reputation.

During Walter Cronkite’s heyday, being on national TV carried a built-in incentive to avoid outright lies. These gatekeepers had plenty of flaws, but they did enforce at least a basic standard of truthfulness. If you wanted to speak to the public, you had to meet those standards; you couldn’t just spout falsehoods and still expect reasonable coverage.

Today, we have half a dozen social media algorithms acting as similar attention monopolies. They actually have tremendous power in determining what people think is true. In 2022, the Biden administration tweeted a falsehood about job numbers, and got fact-checked by Twitter’s fact-checking system, Community Notes, prompting an actual retraction.

We need fact-checking systems integrated with social media that can transparently maintain public trust. This is not as difficult a problem as you might assume. These tools are currently too slow and too conservatively triggered to make a difference in how people consume information. But they can—and must—be improved (some sketches of ideas on this here).

But there's a final dimension here that really matters: reputation. Reputation functions as evolutionary pressure on the source of ideas. The most reliable propaganda detection method combines these elements: "What happens when this source is proven wrong, and what reputational price do they pay?" If the answer is obfuscation and none, they are less likely to be telling you the truth.

In the end, the propagandist's greatest weakness is their inability to admit when they are wrong—and our greatest strength is recognizing when they can't.

Thanks for reading. If you’d like more ideas like this but in book form, Outrage Machine is available here, or on Audible here.

There's a pothole on my street. I know this because I ride over it every day, muttering curses at the city and making mental notes to finally submit that complaint. I also know, with equal certainty, that a neighbor's oak tree dropped a branch during last week's storm.

This physical distance between me and the object is finite, observable and real. The pothole and the fallen branch have little hidden meaning, no subtext. I can walk outside right now and touch them. The solution to the pothole is visible to me — someone needs to fill it up.

But what about something like the border crisis? The war in Ukraine? The California wildfires? These are problems that I can only observer through many chains of intermediaries.

On Jan 27th, President Trump announced he had sent the US military into California and "TURNED ON THE WATER” to help stop the wildfires. In the Central Valley, farmers watched their reservoir levels drop as two billion gallons were released to fight fires that were already contained. The water never reached Los Angeles—it couldn't have, because the infrastructure did not exist to bring it down there—but that hardly mattered. Many Trump supporters celebrated a victory over California's water politics—It was a political “win” for his base, even if it was just a signal with no results.

The farmers downstream of the reservoirs could observe it personally (many of them were Trump supporters), and according to reports were desperately trying to figure out what it meant for their water allocations. But I doubt it changed their support for Trump. One local congressperson said they were “excited” that Trump is proactively engaging with the state’s water issues, even as that water was flowing into the ground and the ocean. This was not a neutral event, despite the fact that it was directly observable to them.

You can laugh at this, and call the people idiots who continue to support candidates that are courting wins while making things worse. But it’s not the whole story.

Democracy is a big coordination game. We need to know what’s happening in order to vote people into office that can solve our problems. In this way the democratic process is a sort of problem-solving machine — if you’re angry about something you see in your community, you should be able to vote for change. If that thing is fixed by that representative, great! Vote them in again. If not, vote them out and try again with a different guy.

For local issues this is pretty easy, because there are fewer mediated steps between us and the issue. Like the pothole on my street, the observable distance is still very low. School boards and city governments function pretty well because people can see what is wrong and which policies are failing.

But for national issues, most of our problems are not directly observable. We rely on mediators — journalists, scientists, pundits, influencers, politicians, and our smart friends — to help us figure out which issues we should pay attention to and how we should vote. The internet was supposed to help us here, giving us access to far more objective information about the world. Someone with a smartphone in Bakersfield or El Paso can broadcast live to millions, giving us immediate observational powers.

But we forgot one of the key axioms of the modern internet age: the more information we have available to us, the more important the serving mechanism becomes. And social media has done some really weird things to how we share information.

The Telephone Game From Hell

Remember the telephone game from elementary school? One kid whispers a message to another, who passes it to the next, until the final person announces something hilariously garbled. "Sally sells seashells" becomes "Jelly smells terrible."

Now imagine playing the telephone game, but designed by a sadist: each person in the chain gets a few dollars based on how many people emotionally react to their version of the message. Suddenly, "Sally sells seashells" might become "BREAKING: Local Girl's Illegal Seashell Operation Funding Underwater Criminal Empire."

This is essentially what we've built with social media. Outrage is viral, nuance is not. Every repost, every quote tweet, every shared article becomes another link in a chain where engagement-maximizing distortion isn't just possible - it's incentivized. The platforms literally pay people (in likes, shares, and ultimately real money through creator programs) to make each iteration more outrage-inducing than the last.

Let’s examine for a moment what this payment looks like. Do you know someone with a large social following? You remember what that felt like when you learned they had it? Your opinion of them probably changed. Followers are esteem. Followers are influence. And if you post a viral hit, you will get more of them. We are paid in reputational capital, and at times, real dollars.

I’m not saying that everyone on these platforms is intentionally trying to game the system to make money. But when the likes, shares, and follows tick upward, we feel good. We get enormous hits of righteousness, meaning and purpose from our audience. These are not-so-subtle payments to our limbic systems, and they are actually rewiring us.

And unfortunately, reality is often the casualty.

Apples vs. Apples

Imagine that it’s 2026 and your local grocery store has two types of apples: Red Delicious and Honeycrisp. Nobody has strong feelings about this. The Democrats haven't issued a position on apple varieties. Tucker Carlson isn't doing segments about how the globalist elite are forcing Honeycrisps on Real Americans. You just pick whichever apple you want and go about your day.

Fast forward to 2027. AOC tweets that Red Delicious apples are a symbol of agricultural worker exploitation. Fox News runs a segment about how "woke millennials" are trying to "cancel" traditional American apples. Suddenly, your choice of apple is a signal of your political tribe. Red states see a 300% increase in Red Delicious sales. Blue cities start hosting Honeycrisp-only farmers markets.

If this sounds silly, consider a few examples of issues that were once broadly bipartisan:

• The EPA was created by Republican President Richard Nixon with overwhelming bipartisan support. The Clean Air Act amendments of 1990 passed 89-11 in the Senate.

• Immigration reform was not fundamentally partisan until recently: Watch two (Liberal?) Republicans — Reagan and Bush — debate the issue here. In 1986, Reagan proudly signed an amnesty bill for undocumented immigrants.

• Until recently, voting methods were largely technical discussions among state officials. The 2002 Help America Vote Act passed with overwhelming bipartisan support.

• Vaccine requirements for school children were standard, uncontroversial policy for decades. The polio vaccine rollout was celebrated across political lines. Even childhood vaccine requirements remained largely bipartisan until the mid-2010s.

• As recently as 2012, Republicans were warning about Russian aggression while Democrats advocated for reset and engagement.

How did these very nonpartisan things become politically charged objects, or even flip in valence? It turns out there is a fairly straightforward process to making this happen.

First, you need elites - politicians, pundits, and Twitter blue-checks who identify potential wedge issues. They're looking for topics that can generate engagement and signal virtue to their base. The intrinsic political relevance of the issue barely matters. Climate change could have remained a boring discussion about carbon taxes and international agreements.

Second, you need normal human psychology. We like to think we're rational beings carefully weighing evidence, but we're actually exquisitely tuned social creatures. Once our tribe stakes out a position, we don't just agree with it - we incorporate it into our identity. Tell a conservative their truck choice shows patriotism, or a liberal their coffee habits demonstrate ethical consumption, and watch as the preference becomes part of who they are.

Third, you need social media. The algorithms that decide what we see are optimized for engagement, and nothing drives engagement like tribal conflict. Each retweet and share strengthens the association between issue and identity. Your feed fills with people who share your views, while the algorithm helpfully hides anyone who might complicate the narrative.

Lazy Followers

A 2019 Cornell study revealed how arbitrary this process can be.

The researchers created ten separate "worlds" - essentially isolated social feeds where participants were isolated from each other. In each world, they introduced politically neutral statements like "The greatest books are universal in their appeal" and “AI should be used in blackmail detection” and asked Democrats or Republicans whether they agreed. The crucial variable was that in some worlds, participants could see how others before them had responded, along with their political affiliations.

What happened next was remarkable. After running the experiment, the same statement would become a firmly held "Democratic position" in some worlds and an equally firm "Republican position" in others, based solely on who happened to stake out positions first. In one world, a Democrat labeling an issue first as Democratic would cause Republicans to see it as a "Democratic position" and disagree. But in another world, a Republican might claim the same position first, creating the exact opposite pattern.

This suggests something crucial about our political world: many of the issues we now see as inherently "liberal" or "conservative" might have ended up completely reversed if different people had staked out positions first. Our deepest political divides might not be about the issues at all - they're about who claimed them first and how we aligned ourselves in response.

A simpler way to explain this might be thinking in terms of mental shortcuts. Our brains are fairly efficient (read: lazy), and don’t want to expend any effort they don’t have to. For an opinion which we have very little previous information about, we’re likely to use the scaffolding of other people’s opinions to determine how to evaluate it.

But other people’s interpretations are also lazy. They relied on someone before them to determine if it was a political object. A chain of lazy evaluations can turn one person’s initial gut reaction into a thousand other people’s gut-reaction follow-ons.

The Way Out

Is there a way out of this? There is some evidence there is. The first step is recognizing that most political positions aren't inherently partisan - they're made partisan through a specific social process that we can understand and potentially interrupt.

Some institutions are already experimenting with ways to short-circuit this process. Structured debates on social media have proven surprisingly good at helping people find information with a few simple norms in place (see AskScience, or ChangeMyView on Reddit). Some news organizations are experimenting with formats that emphasize direct observation, ideological flags and verifiable facts over partisan framing.

Nothing about this process is inevitable. Some communities have managed to maintain productive disagreement without descending into tribal warfare. They typically share a few features:

• They explicitly separate ideas from identity

• They reward good-faith engagement across partisan lines

• Their discussion formats emphasize understanding over winning

These efforts won't eliminate disagreement, nor should they - genuine political differences are a healthy part of democracy. But if we can carefully examine the machinery in play, it might help us design better "neutral zones" where we can discuss problems and solutions without immediately sorting into tribes. The pothole on my street doesn't care which party I vote for, and neither does the reservoir pushing water into the ocean. Reality is stubbornly non-partisan, and we need to remember that.

And maybe that's the real source of hope: beneath all our mediated realities and engagement-optimized outrage, there's still a physical world we share. The challenge isn't to agree on everything - it's to keep more of our discussions anchored in that observable reality, where problems are just problems, and solutions can come from anywhere.

Thanks for reading. These are all ideas from my book, Outrage Machine, published by Hachette. Your attention is a resource I deeply appreciate.

Note: This is an article originally published on Quartz and Medium in 2018. It is meant to help people understand the fundamentals of social media interventions — offering ideas on how to reduce outrage in our feeds.

It’s interesting to revist this after 6 (!) years to see what has aged well/what hasn’t. Notably, several of these interventions have since inspired tools adopted by Linkedin, Meta, and Instagram. Some have been adopted elsewhere, some have not. Since I wrote this, a handful of designers launched the Prosocial Design Network — a great resource for tracking evidence-based interventions — a number of which are mentioned here. Go there for the best inventory of up-to-date design solutions.

A number of the issues which I outlined back then have become very politically charged, notably around squishy definitions of moderation and censorship. and the de facto role that platforms have come to play as arbiters of truth and attention, with no transparency and almost complete impunity. The best solution to this I have seen is the Platform Accountability and Transparency Act, which would force companies to open up API access to researchers and the public.

Unfortunately — largely because of outrage on social media — we have lost the ability to have a sensible conversation about many of the tradeoffs involved here. As of this moment, it seems like the outrage machine is winning.

-TRS, Jan 2025

Introduction

Is social media a force for good?

Data exploits affecting millions, election hacking, the death of newspapers, weaponized propaganda, troll armies, deepening polarization and the shaky future of democracy itself. It seems we’re presented daily with a laundry list of dystopian consequences linked back to our collective overuse of these tools.

People are crying for change, but there’s a missing piece of the conversation: despite the emerging awareness of many of the terrible side effects of these platforms, we have become dependent on them.

They’ve become our local news channels, our emergency communication systems, our town squares, and the primary windows into the lives of our loved ones and governments. They are a critical part of how we rally around shared causes and engage with our politics.

How do we reconcile their toxicity with their utility?

This shouldn’t be such a painful choice. Over the last year I have been speaking with academics, designers and technologists studying the key flaws of these platforms — and cataloging testable fixes. My efforts have focused on the nested problems of polarization, dehumanization, and outrage, three of the most dangerous byproducts of these tools.

Below, I’ll explore what might be changed by these companies now — not tomorrow — to make these platforms better for humanity.

What’s good for capturing human attention is often bad for humans

Our feeds can make us angry for a reason.

Imagine that you’re walking down the street, and you hear a fight break out. It’s loud and aggressive, and people are yelling about something. You’ll likely stop momentarily to see what’s going on — it’s in your nature.

If you personally know one of the people fighting, you will probably immediately pick a side — you might even get involved. At minimum, you will pay attention.

This is what social media does to us regularly: it encourages us to observe conflicts and pick sides on topics about which we would otherwise have few opinions.

At its core, it is an opinion-serving machine. And on social media, not all opinions are served equally.

Most of our content feeds and timelines are no longer sorted chronologically. The decision about which content to show us is instead based on how likely we are to engage with it.

Emotional reactions like outrage are strong indicators of engagement. With the most basic algorithm that sorts our feeds, this kind of divisive content will be shown first, because it captures more attention than other types of content.

This content acts as a trigger for our own emotional reactions. And when we respond, we regularly push our emotions out to the rest of the world.

This is a simplified model of how we share on social media:

With angry content, this has allowed for what we might call outrage cascades — viral explosions of moral judgment and disgust. These have come to dominate our feeds and our conversations, and are becoming a prominent part of the cultural zeitgeist.

Moral Outrage = Virality

William J. Brady, a researcher at NYU, recently found a pattern in viral social media posts. Studying a massive data set of hundreds of thousands of tweets, he found that posts using moral and emotional language receive roughly a 20% boost for every moral and emotional keyword used.

Conservative Example Tweet:
“Gay marriage is a diabolical, evil lie aimed at destroying our nation”
- @overpasses4america

Liberal Example Tweet:
“New Mormon Policy Bans Children Of Same-Sex Parents — this church wants to punish children? Are you kidding me?!? Shame”
- @martina

Each of these tweets incorporates language that is morally charged, and condemning of others. They incite a deep emotional response, and are more likely to be seen & shared by people who agree with them, providing a measurable boost to virality and engagement.

This is a hidden substrate for people to share divisive, outrageous and emotional content online.

This doesn’t just apply to our personal posts. It likely applies to any content we share on social media — comments, memes, videos, articles. It has incubated an ecosystem of moral outrage that is utilized by content creators everywhere, including news organizations, simply because it works.

Facebook, Twitter, YouTube, and others prioritize this type of content, because it is what we click on, hover over, and respond to. It is the hidden pathway to audience engagement. When trying to capture attention, our anger, fear and disgust are a signal in the noise.

Through the dominance of these tools — in our media, our conversations, and our lives — we’ve watched our common discourse turn ugly, divisive, and increasingly polarizing.

What can we do about it?

These are four design changes we might consider for improving the way we share online:

Give Humanizing Prompts

Nudge people in the right direction with specific prompts before they post

Molly Crockett at Yale’s Crockett Lab has suggested that our inability to physically see the emotional reactions of others might encourage negative behavior on social media. Online we literally can’t see the suffering of others, and this makes us more willing to be unkind.

Nicholas Christakis (also at Yale) has shown that real-life social networks can be influenced by simple AI to help improve group behavior and outcomes. We can think about incorporating prompts like that here.

These are several interventions that we can place immediately after people post — allowing us to nudge users towards kindness and and better behavior.

Research on perceptual dehumanization has shown that we might be significantly more punitive towards people in digital environments. Increasing empathetic responses like this might help increase the perception of our digital avatars as being human. (This research is fascinating: something as simple as the vertical/horizontal orientation of our faces in our profile picture can determine whether others consider us worthy of punishment.)

We do angry things that we often later regret. Having a moment to pause, review, and undo content flagged as hurtful might reduce likelihood of sharing it in our worst moments.

Note: All of the above prompts might also be used with linked content-like articles, as long as that content is indexed and flagged correctly.

Disagreeing with people is hard. Disagreeing with people in public is even harder. There are tremendous social pressures at play when we write comments on people’s posts in public: we’re on display, fighting about an idea with a crowd watching us. Giving a private reply — taking it to direct message by default — might encourage people to open sidebars to have conversations with less external pressure.

This is all great, but how do we find these posts in the first place? Isn’t that tricky? Yes it is. Let’s get into the weeds for a moment.

Picking Out Unhealthy Content with Better Metrics

Flagging and filtering specific content types we would prefer to see less of

This is a lot harder than it sounds. We need metrics to train algorithms and humans to find the stuff we don’t want. If we intervene with the incorrect set of metrics, we can easily end up in some really dark places.

Algorithms are representations of human intelligence — and just like any human creation, they can inherit and amplify our perspectives and flaws. This is known as algorithmic bias, and it can manifest in stock market failures, systemic loan discrimination, unfair teacher evaluations, and even in racially unjust jail sentences.

Facebook already trains their news feed algorithm around the metric of what is “meaningful” to its users. The problem with this metric is that many strong human reactions — like moral outrage, disgust, and anger — are broadly considered meaningful.

To do this right, we need better metrics.

By correctly measuring the type of content that users don’t want to see more of, we can begin to give users a menu of choices that accurately represent their preferences, not just what they will click on.

These are three possible candidates for content that we might prefer to have less of: outraged, toxic, and regrettable. These are starting points.

Critical note: Flagging content that induces specific negative responses immediately treads into free-speech territory. Who’s to say platforms should manipulate other people’s voices if they are saying things that certain people find offensive? Isn’t that censorship? This is a huge, critical question — and one that I will address. Read on.

Once these content types are identified, we can train supervised AI to flag it in the future.

Regrettable Content

Regret is one of the most common emotions we feel after spending significant time on social media. This is largely due to the kind of content we click on despite ourselves that we know we shouldn’t.

This is based on the concept of present bias: A natural human tendency for people to give stronger weight to payoffs that are closer to the present time when considering trade-offs between two future moments.

Measuring present bias is tricky, but a start might be to present a non-obtrusive, simple prompt for feedback to users after they consume a set of content. This will require some thoughtful design choices to balance frequency with flow.

Toxic Content

Toxicity can be determined by asking people to rate posts on a scale from Toxic to Healthy, toxic being defined as “a rude, disrespectful, or unreasonable post that is likely to make you leave a discussion.” This is Perspective API’s model — a project of Jigsaw (part of Google, who’s earliest version of this tool was fairly flawed ¹²). A critical part of making this work is ensuring these terms are defined by a diverse and representative sample of users.

Outraged Content

This is content that uses moral, emotional, and other-condemning language. We can define it by indexing a moral foundations dictionary which was initially developed by Jonathan Haidt and Jesse Graham and combining it with other dictionaries. This definition could be broadened and honed to include ‘other-condemning’ and hyper-polarizing language.

Filter Unhealthy Content by Default

After users publish negative posts, filter how that content is served

All of the content we share on social media already goes through an engagement filter. Facebook, Twitter and others push content up or down our feeds depending on how likely it is to capture attention (or how much you pay them to promote it). The problem is that certain types of content — outraged, sensationalized, clickbait, etc. — naturally hack our attention in unhealthy ways.

Unhealthy content can be proportionally down-weighted to account for its natural virality. Of course social media companies already do this regularly to account for abusive content, piracy, spam, and thousands of other variables. They could also do the same for other types of triggering content. This would proportionally reduce the prominence of outrage-inducing, toxic, and regrettable posts and give them more equal footing with other types of content.

Note: This treads further into the free-speech territory, and the realm brings up filter bubbles and echo chambers. I will address both of these below.

Give Users Feed Control

Providing users with control of their own algorithmic content

During the 2016 election, when streams of political outrage and vitriol were the primary products of our social media feeds, this seemed like a bad idea. Wouldn’t users just cocoon themselves inside their own misinformation-filled realities? Wouldn’t this just increase political isolation? But as efforts to identify false news, misinformation, and propaganda have gained steam, the issue underpinning our reliance on black-box algorithms has remained. What’s missing is transparency about why we see what we see on social media.

The antidote to this is to give users access to the editorial processes that determine which content we see. From the most basic reverse-chronological order to highly curated feeds, this process could be opened up.

How this might be done: Dashboards

A project started out of the MIT Media Lab called Gobo began this process developing an open aggregator of social feeds to show an example of what a dashboard might look like. This allows users to filter their content by things like politics, rudeness, and virality. These metrics are a decent starting point in thinking through what an open feed might look like.

Another analogy: Building blocks, or recipes

This might look like a set of new tools for exploring new perspectives and the depths of our feed. These feeds could be discussed, reconfigured, and shared on their own.

This process of designing feed control for users would open up a crucial dialog about what constitutes a healthy information diet — something that is currently obscured by their proprietary nature. Forcing users to decide might also encourage users to learn more about the kinds of unhealthy triggers they are being regularly served.

The Big Question

Are these interventions another form of censorship? If we reduce the visibility of people online who say things we disagree with, isn’t that suppression of speech?

This is a huge issue.

It’s hard to overstate how important and influential these platforms are to society at this moment in time. Some of the biggest, most significant social changes in recent years have come from activism catalyzed by moral outrage shared through social media. Many of these cultural and political movements would not have been possible without them: From the #TeaParty to #MeToo, hundreds of major social movements have been launched. What would happen if these voices were suppressed?

But this is the critical problem: social media already suppresses our voices.

The way we are currently served content is not a neutral, unbiased process. It is not inherently democratic, fairly apportioned, or constitutionally protected. These tools already promote or bury content with a proprietary algorithm over which we have no say.

They are not censored for political partisanship. They are censored for our engagement — to keep us attached and connected to these products, and to serve us ads. The day our chronological feeds became proprietary sorting mechanisms was the day these platforms ceased to be neutral.

The more critical these tools become to public discourse, free speech, and democracy, the more problematic it is for these algorithms to be obscured from us.

And this is ultimately the point — by accepting the importance of these tools but asking for none of the control, we are giving up our ability to determine the type of conversation we have as a society.

Three final things:

1. These are all meant to be testable interventions.

2. They are designed to be inherently non-partisan.

3. They will all likely result in short-term reductions in engagement and ad revenue, but a possible long-term increase in perception, health, and well-being.

These are not perfect solutions. They are starting points which can be explored more thoroughly. Part of the goal of this effort is to advance a conversation about these tools grounded in testable outcomes. With that in mind, you can visit the fantastic resource of the Prosocial Design Network.

I hope that by thoughtfully picking apart, testing, and critiquing designs like this, it might lead us toward something that is ultimately a better alternative to the divisive, toxic and deeply unhealthy digital sphere we mutually inhabit.

I published my first book on these topics, chock-full of solutions and mental models for understanding this enormous shift to our media landscape. It’s called Outrage Machine, and it’s available wherever books are sold:

References

• On Filter Bubbles & Echo Chambers: Research by Chris Bail and Duke suggests that regular exposure to opposing political perspectives might actually contribute to polarization. This is a strong refutation of the argument that filter bubbles/echo chambers cause polarization. My read is that if we want to reduce systemic political division, we might instead need to reduce the prominence of specific kinds of strongly partisan/triggering content — not necessarily broaden our exposure.

• Molly Crockett has shown that digital environments provide us with more opportunities for being outraged than any other medium, and has suggested that the benefits of sharing our outrage online are greatly increased. Nature Article

• William Brady has shown that tweets with moral/emotional keywords receive roughly a 20% boost for each word used. This has become a hidden incentive for people to share divisive, outrageous, and emotionally-charged content online. Article & abstract

• Katrina Fincher is studying how we dehumanize others in digital environments. Article & abstract

• Nicholas Christakis has shown that real-life social networks can be manipulated by “dumb” AI to help improve group behavior and outcomes. Video outlining his recent work

Further Reading

• This Is How Your Fear and Outrage Are Being Sold for Profit — a previous article in this series

• On ethically building recommendation engines—by Renee Diresta

• Things you can do to improve your relationship with social media as a consumer, right now— Helpful suggestions from the Center for Humane Tech

• Is Anything Worth Maximizing? — article by Joe Eldelman on creating better metrics.

Special thanks to William Brady at NYU, Molly Crockett of Yale, Katrina Fincher at Columbia, and the Center for Humane Tech for insights and research used in this article. If you’d like to keep in touch, subscribe here.

On January 20th, at a post-inauguration event, Elon Musk raised his arm during a speech. A large number of people saw it as a Nazi salute. Others saw an awkwardly enthusiastic wave. Both sides seem fairly certain the other side is either lying or crazy.

This is not a post about what Musk actually did — this is instead a post about how we’ve reacted to it. It's about how we form instant, unshakeable certainties about difficult and ambiguous events, and why the internet is currently designed to make them worse.

Back in 2015, a poorly-lit photo of a dress split the internet into warring factions of blue-black people and white-gold people. It went viral for the same reasons: everyone was completely certain about what they saw, and found it impossible to understand how anyone could see it differently. These two forces — inherent ambiguity coupled with perceptual certainty made the dress fundamentally unignorable.

The dress is what is referred to as an ambiguous image. Most ambiguous images are fun oddities and have lived in entertaining corners of the internet for years. These illusions have nothing to do with sight problems, but actually happen in our brains — they’re caused by systematic interpretation errors in how we perceive objects.

Most of us are familiar with these from childhood, coming in the form of novel optical illusions that can be interpreted in two or more ways. Usually, ambiguous images are visibly interchangeable — we can switch between them easily. The image above, for instance, can be a rabbit or a duck, and our brains can fairly easily move between their relative forms.

But the Dress (and Musk’s wave), are different. Our perception of these items becomes locked in place. Like a light switch in one of two positions, it’s hard to escape the certainty of our initial judgment.

Our Brain Makes Excuses For Itself

Think of your brain as a one-way valve. As soon as information enters, it gets labeled "objective reality" — even when that information is unclear or could be interpreted multiple ways. This is what psychologists have called "naïve realism": our unshakeable belief that we see things as they truly are, while others who see things differently must be confused, stupid, or lying.

Emotions make this particularly hard. Back in the 1970s, the conventional wisdom held that we perceive or think about something first (cognition), then decide an emotional response later (affect). This view was popular largely because we assumed humans were rational actors.

In 1980 the psychologist Robert Zajonc turned this on its head, arguing that emotion (affect) actually precedes thinking (cognition). We feel first, then we think.

Later research—including findings by Jonathan Haidt—upheld Zajonc’s view: our immediate feelings come first, and our thoughts follow to rationalize and justify those feelings.

1. Feel an emotional reaction

2. Generate judgment that support that reaction

3. Rationalize and justify those thoughts using all available evidence

This explains why these perceptual splits are so emotionally charged and resistant to change. When we encounter an ambiguous image or event, our emotional reaction comes first, instantaneously and unconsciously. Our cognitive processing – the part that constructs rational explanations – comes second, mainly serving to justify what we've already felt.

A Scissor Machine

This wasn’t as much of a problem in a shared news ecosystem. In previous times our perception would be largely monopolized by the dominant editorial angle and media narrative, forcing divergent perspectives into the minority. But the internet’s current architecture thrives on what the writer Scott Alexander coined as “scissors”—statements and provocations that neatly slice people into two opposing camps of moral certainty. These perceptual scissors go viral precisely because people interpret them in these binary ways, and can find many reasons to justify their initial interpretations — as well as a shared cohort of people who believe the same thing. You might think back to a few such events in years past: The Kavanaugh hearings, or the Covington Catholic altercation, to name a few.

Algorithms up-level these disagreements because they generate engagement. Media outlets cover them because they are great content. And as individuals, the system rewards us too: when we post an outraged interpretation, we receive likes, shares, and follows from our side. The more certain, extreme, and moralized your interpretation, the more engagement you'll receive. We're being paid to source these divisive moments, and dig ourselves deeper into our perceptual trenches. As a result, each scissor event cuts a little deeper, leaving us more convinced that the other side is either insane or lying.

Weaponized Ambiguity

Political figures (and online trolls) have learned to exploit this dynamic skillfully. They spotlight—and often create—ambiguously outrageous events precisely because they know each side will interpret those events in incompatible but unignorable ways.

Crafting moments that are just ambiguous enough to excite supporters (who see them as harmless or owning the other side) and enrage opponents (who interpret them as malicious). The result is free publicity: your allies rally around you, your critics amplify your message, and everyone else is transfixed by the ensuing brawl. The politician wins, society loses.

Over time, these scissor events degrade trust and normalize the behavior in question. The more one side condemns a behavior as “dangerous,” the more the other side may embrace it—purely to spite those it sees as perpetually over-triggered. The dynamic can shove groups toward extremes they might otherwise reject, all because they trust their opponents’ interpretation even less than they trust the ambiguous actor’s intentions. Perversely, the result is often that the bad behavior is normalized from that point on.

Naming the Scissors

There is no easy solution here, but I’ve found that simply naming the dynamic can make a big difference. When you see an eruption over something that’s inherently ambiguous and fiercely contested, it helps to say: “Ah. This is a Scissor,” and take a breath.

Recognizing these moments for what they are—an oddity of human perception amplified by a system built to hold our attention—can help us stay sane. With a little awareness, these scissors might cut less deeply into our politics, our relationships, and our collective trust.

Thank you for reading. This article is adapted from Outrage Machine — a book about reclaiming your attention in an age of algorithmic anxiety. It’s available here.

References:

• Brady, W. J., Crockett, M. J., & Van Bavel, J. J. (2020). The MAD model of moral contagion: The role of motivation, attention, and design in the spread of moralized content online. Perspectives on Psychological Science

• Haidt, J. (2012). The righteous mind: Why good people are divided by politics and religion. Pantheon

• Lafer-Sousa, R., Hermann, K. L., & Conway, B. R. (2015). Striking individual differences in color perception uncovered by "the dress" photograph. Current Biology

• Ross, L., & Ward, A. (1995). Naive realism in everyday life: Implications for social conflict and misunderstanding. In T. Gilovich, D. Griffin, & D. Kahneman (Eds.), Heuristics and biases

• Zajonc, R. B. (1980). Feeling and thinking: Preferences need no inferences. American Psychologist

There's a phenomenon in psychology called shifting baseline syndrome, where each generation accepts the current state of the world as normal, forgetting what came before. Ecologists noticed it first – younger scientists studying fish populations would assume depleted fish stocks were "normal" because it’s all they’d ever seen.

We all do this. We can forget that the world was different not so long ago. Or we can simply be blind to how things were just a generation back.

2024 was the year we collectively succumbed to shifting baseline syndrome about our relationship with social media. Our norms changed along with our understanding.

It was just over 15 years ago when a politician would be laughed out of office for sharing a conspiracy theory. Going viral was a epidemiological thing to avoid. Follower was a term reserved for cult documentaries. Children were not expected to go to school with smartphones.

The platforms have achieved something remarkable: they've become axioms rather than tools. We no longer question why we use them. Online outrage has become self-justifying — a perpetual motion machine of urgency and attention. A huge portion of the information we share and consume is touched by these platforms.

Our Species is Now Socially-Mediated

It's not that we've forgotten social media influences us. We know that intellectually. We've just forgotten what it feels like to not be influenced. The water we swim in has changed temperature so gradually that we barely noticed when it reached a boil. It’s changed our politics, our perception, our sleep, our relationships, even our inner emotional life.

And it makes sense. We succumb to the triggers and outrages in our feed because it feels like we are making a difference. The outrage helps us find a tribe and fills us with purpose. The doomscroll, the venting, the sharing — it provides a quick hit that tricks us into believing we’ve contributed to solutions. We probably haven’t.

We should not expect the platforms to change in the near term. The incentives are too strong, the profits are too tempting, the competition is too fierce. Asking Meta or X to optimize for social cohesion instead of engagement is like asking a casino to close the slot machines to protect gamblers.

But, perhaps 2025 can be the year we remember a few things

• That the other side's worst examples aren't their best arguments. That norms on social media are not norms in real life. That we remember what it feels like to think slowly, to hold contradictions, to see our ideological opponents as complex humans rather than evil.

• That much of the most urgent and outrageous news we see on social media is incorrect. That the longer we wait for confirmation and corroboration, the more likely we are to get the whole story.

• That many notable people have real incentives to say and do outrageous things online — particularly because social media pays them with our attention and real dollars. Not because it’s true, coherent, or good policy.

• That outrage is a very small piece of solving problems. Real progress requires discomfort and actions that may never get a single like or share: difficult conversations, sustained curiosity, pragmatic compromise.

• That the world is still getting measurably better in many ways, despite our feeds and group chats telling us that it's fundamentally broken.

• Most importantly, we can remember that the real world is messier, more nuanced, and ultimately more hopeful than we’d assume from looking at our phones.

If we can remember all that, maybe we can shift some of our despair about where things are into curiosity about how we can make it better.

We've been finding ways to live together despite our differences for a long time. We just need to remember how.

Happy New Year — here’s to 2025 — the year we remember.

I wrote a hopeful book about these issues, also called Outrage Machine. Take a read or a listen here.

Imagine one night you’re at a party. You hear someone behind you talking loudly, effusively. They are saying something that sounds suspicious — how “we only use cars because Big Auto is behind a conspiracy to keep people from teleporting.”

Teleporting? You turn around, and realize it’s someone you know.

What do you do?

1. Stand up, pull out a napkin, draw a detailed diagram of why teleportation violates the laws of physics, and explain exactly how they’re wrong using math?

2. Laugh at them, call them an idiot in the cleverest way possible, and spend the rest of your night ranting despondently about humanity?

3. Say, “I’ve never heard about this… I’d love to learn. Tell me more?”

The internet rewards us mightily for choosing option #2. Likes, shares, and attention is all fueled by dunking, righteous call-outs, and making other people look pretty bad.

But a lot of our friends and family members increasingly hold beliefs that feel strange to us. By now we’ve all been in a group chat or a dinner party that was interrupted by an idea that feels proximal to a conspiracy.

It’s hard to know what to do. These are often people we care about — or at least people we care about persuading otherwise. When these beliefs become apparent, it’s important to open door #3.

This is a brief post explaining how you might have that conversation, gently and effectively, using a handful of useful steps backed by research and some experience.

First Assume They Aren’t Crazy

This can feel like a radical proposition: people who believe wild things are often coming from a rational place. Much of the time they are operating with different priors, and different informational inputs.

Your friend yelling about teleportation has not necessarily lost it. He may have seen a YouTube video about quantum tunneling, and then joined a Facebook group where he learned about suppressed tech from GM in the 90s, all after spending an inordinate amount of his life stuck in traffic. He has a motivated belief, and may be in a community of people that support it.

If you start by calling him wrong, you’ll bounce off his social and mental defenses. The conversation will be over. No persuasion, no new ideas.

Climbing the Mountain

Before you begin, take a deep breath and look at the task ahead. You should think about this like climbing a mountain. It has three peaks.

Peak 1, which everyone must traverse in order to have a successful conversation, is ensuring you understand them, where they are coming from, and how they formed this belief. If you cannot get here, the rest of the mountain is out of reach. Peak 2 is getting to the point that the other person understands you. At the summit is a changed mind. You cannot scale the sheer cliff to get there, and you must follow the paths upward from the bottom.

Beliefs Exist on a Spectrum

People’s beliefs are not binary — they exist in intervals of confidence. And they also don’t shift quickly. You should not expect people to change their minds after a single conversation. Instead think about this as a way to pull the slider a little towards your side. Getting them 10% closer to your side is actually a real success. Start with that goal.

When & How

Talk when emotions are cool, preferably without an audience

Engage when emotions are calm. Trying to persuade during a heated moment often hardens their stance. Better to be in a relaxed social context.

A large audience creates additional incentives for performance, so try to have your conversation one on one. A few additional people can make them more defensive or sensitive about sharing their ideas. Be curious and playful with the approach.

Online vs. Offline

In general, In Person > Video Call > Call > Private Text > Public Thread

In-person is better than a call. A call is better than private text or email. A private text is better than a public thread. But all are okay if you can follow the protocol, maintain a gentle attitude, and keep your cool.

Be the Listener You Want Them to Be

Most beliefs, even conspiratorial ones, are based on some bit of truth.

Try to model the stance that you are open to changing your own mind if presented with the right evidence. This is intellectual humility 101, and when you show it, you implicitly invite them to show it in return. In practice this can feel uncomfortable, but it’s important to acknowledge the possibility that they’re actually partially right.

The Process

We’re going to stay with the Teleportation example here — yes it’s silly. But it’s a stand-in for any belief, even serious ones.

Step 1: Set Trust

Slow down and actually find out if they even want to talk about this. Start by asking how they’re doing and why this topic matters to them — show curiosity. If you push too hard, too fast, they’ll likely shut down. Let them know they can share their story on their own terms.

“How did you first stumble on this idea?”
“I’m super curious how you arrived here, and I’d love to explore it together.”

Because when someone feels heard, they open up. Create a chill, respectful vibe and ask like you mean it.

Step 2: Clarify

Once you’re both comfortable, ask them directly:

“I’d love to get clear, what do you believe about X?”

Then, repeat it back to them — in your own words — and see if that’s right. Double-check you’re talking about the same concept (because words like “science,” “institutions,” or “The Man” can mean different things to different people).

“Ok, just so I understand, can I repeat it back to you?”

For example: “So, you’re saying that Big Auto is hiding teleportation technology to protect their profits. Did I get that right?”

At this point, look for emotional cues. If someone is especially upset or angry about something — (ie. “The truth is being hidden!”) leave space to acknowledge that feeling. “Thanks for sharing that.”

Keep refining the definition until they go, “Yes, exactly.” You want an accurate snapshot of their actual position, not some fuzzy or misunderstood version.

Sometimes, people will drop several contingent beliefs. (For example, #1 Teleportation is real. #2 Big Auto is behind the coverup. #3 Scientists have been paid off.)

Clarify each belief before you move on.

Step 3: Peg Their Confidence

Next, try to measure how strongly they believe in these statements.

Ask them to rate their confidence, 0 to 100. It’s not about locking them into a number; it’s about getting them to reflect. Sometimes just picking a number helps people realize they might not be 100% certain — or maybe they are.

“On a scale of 0 to 100, how confident are you about this being true?

This step opens a little door to introspection. Even if they say 100, that’s still valuable info to have. When they feel safe, most people will admit their beliefs are a bit less confident than that. Keep it mind.

Step 4: Gather Evidence

The real conversation begins here. Your goal is simple: understand what convinced them while creating space for clear reflection. You’re here to explore, not challenge their identity or fundamental worldview.

Start with genuine curiosity:

"What convinced you the most about this? I'd love to understand."

Then explore that evidence:

"Could you tell me more about where you first learned about it?"

"What made this explanation feel right to you?"

“When were you convinced this was true? What was the evidence?”

Listen carefully for their journey - both what they discovered and how they felt about it. When they share something important, reflect it back: "So you found [X] and that really connected with your experience of [Y]..."

If something isn't clear, it's okay to ask:

"Could you help me understand that part better?"

"How did you figure out if this information was reliable?"

Remember: Keep it curious, and don't rush. You're having a conversation, not conducting an investigation. One or two good questions are often more effective than ten probing ones.

Step 5: Examine and Question

Once you've built this foundation of evidence, you can introduce more challenging questions. The key is to help them examine their own confidence levels:

"Earlier you said you were [X%] confident about this. What is the basis for the % of your doubt?”

"You're clearly someone who thinks carefully about evidence. Using these same standards, what's the weakest link in your explanation?”

Then use their own logic to explore uncertainties:

If they mentioned a dubious source: “What makes this piece of evidence trustworthy?”

“Do you think it’s possible this evidence could be interpreted differently by someone else?”

If they cite a far-fetched pattern: "Does this pattern have any alternate or simpler explanations?"

If they reference hidden knowledge: "What would it take to keep something this big hidden?"

Invite them to apply their own intelligence and skepticism more broadly. If there is some heat, don’t avoid it entirely—it’s usually a sign you’ve hit a core issue worth tackling. Keep it civil, and keep the conversation going in good faith.

Step 6: Flip It

At some point, you’ll sense you’ve gone as far as you can for now. From here, offer to flip the conversation. Let them to ask you the same series of questions about how you formed your belief, in the same way.

Give them the same opportunity they gave you — literally go through the above questions.

Let them ask you what you believe about the topic. Define it clearly.

Peg the confidence level of your belief.

Let them inquire with the same questions as above on steps 4 and 5.

As you do this, you’ll already have a great sense of the dimensions of their belief. Offer your own evidence that might counter it. Concede where they might be right. And explain how you formed your own beliefs, and let them ask follow up clarifying questions.

Explaining the contours of your belief — the how and why — is actually one of the best ways to shift perspective. It shows you’re serious about getting closer to the truth. You are making a case based on the evidence, and you’re giving them a chance to explore it symmetrically.

Step 7: Reflect & Close

When things start to circle, or feel strained or tired, wrap it up for now. Ask if their confidence has shifted at all.

“After all this, has your confidence changed even a little? Why?”

Maybe it’s different, maybe it’s still the same. That’s okay. You’ve planted a seed of reflection, and that’s enough for now. Thank them for sharing and being open, and invite them to continue the conversation later.

You may have a better sense now of how to continue the dialog based on the evidence at hand. The goal isn’t to instantly convert them. It’s to leave the door open for future critical thinking, to strengthen your connection, and to show that exploring these beliefs can be done with mutual respect.

Progress is Success

Very few people change their mind in a single chat. But they will likely remember it next time the topic comes up. This is often the beginning of a dialectic between you and them. What you’re showing is that you’re another mind on this journey of inquiry — and that it’s OK to question some of these deeply held beliefs.

Every time I’ve done this exercise I’ve found myself becoming more logically grounded in my own opinions, and more aware of the flaws in my own arguments. At times, I’ve even found a few of my own beliefs slide a bit towards theirs in a meaningful way.

This method sidesteps the defensive reactions triggered by starting exclusively with fact-based refutations. People are less likely to dig in their heels when the conversation feels like a collaborative exploration rather than a zero-sum game. If they still trust you at the end of the conversation, you’ve done well.

Disagreement is an act of care

A lot of success here lies is simply the process of ensuring that the person feels seen and heard. They want to be acknowledged as a curious human that deserves respect. If you can leave the conversation with a fundamental feeling, “We disagree, but we’re still in this together,” is a great place to land.

As individuals, we’re actually very bad at seeing the flaws in our own arguments. The world we live in is increasingly full of motivated reasoning, clusters of fuzzy logic, and cultures of messy ideas. All of us may at some point fall into a confused or half-baked set of beliefs. When we do, we must rely on others to help get us out.

Reasoned argumentation — the art of unpacking ideas, threading inconsistencies, sifting claims with integrity, exploring logic, and sometimes even persuading others — is one of the most important skills our species has developed. This type of patient disagreement pushes us all forward, and is a deep gesture of care.

Thank you for caring.

— — —

A few notes:

• This is adapted from a number of sources — notably and a handful of ideas from a good book by David McRaney, the practice of street epistemology (and synthesizing and acknowledging it’s critics), and deep canvasing, and Defender’s excellent series of posts on how to argue better on the internet.

• Fun story: The original study on deep canvassing—using empathetic, personal conversations to shift opinions—was retracted for fraudulent data. However, the people who called out the fraudulent data: David Broockman and Joshua Kalla actually *confirmed* the technique’s validity in a later study. Their findings show that sincere, respectful exchanges do produce lasting changes in attitudes, especially when participants feel genuinely heard and understood. Fun 8yo recap here.

• I asked Claude to help me put together a mnemonic to remember this process (Trust, Clarify, Peg, Gather, Examine, Flip, Reflect). Way too much to be catchy, I know. But it gave me “Tiny Conspiracies Produce Giant Extraterrestrial Flying Rabbits” and "The Careful Penguin Gently Explores Frozen Rivers" which are pretty great. I’ll do better next time.

• If you enjoyed this, I wrote a book, also called Outrage Machine which unpacks how we found ourselves in a broken shared reality, with a lot of helpful frameworks like this. It’s available on Amazon, Audible, and wherever books are sold.

Happy holidays! Thanks for reading. Your attention is an gift I very much appreciate.

This episode is based on the book The Constitution of Knowledge, by Jonathan Rauch.

It’s meant for those people who are trying to figure out who they can trust in a moment when trusting an institution seems like a radical act. If you find it valuable, or know a friend that might be interested in these issues please pass it on.

More episodes and podcasts coming soon.

Thanks. As always I’m deeply grateful for your limited attention.