Blue is Glue!

Intro

A fascinating thought experiment went viral on Twitter recently (yet again — I remember it making the rounds a few years ago too). In case you were living under a rock, or more likely just not on Twitter, here it is: 

“Everyone in the world has to take a private vote by pressing a red or blue button. If more than 50% of people press the blue button, everyone survives. If less than 50% of people press the blue button, only people who pressed the red button survive. Which button would you press?”

Now, I just want to say that I love, love, love this thought experiment! It seems to be the ultimate scissor question — perfectly designed to split people into opposite camps, each fiercely convinced they're right and the other side is wrong. The combination of game theory, theory of mind, and morality makes it a perfect window into human psychology, and I think a person's answer - or at least their reasoning for how they arrived there - is deeply revealing of who they are.

As for me, I'm enthusiastically on Team Blue, and I want to document the reasons why here. But before I do, I want to show I fully understand the Reds — they're not all crazy, stupid, or evil — even though in the end I disagree with them quite strongly.

The Case for Red

Note: Here I am acting as Devil’s Advocate.

The strongest point in Red's favor is that it's the self-interested choice. Pick Red, and you're guaranteed to live. Pick Blue, and you might live, but you might also die, depending on the final vote tally. Why take a risk when the safe choice is right there in front of you? And while Red is self-interested, that doesn't make it selfish. This isn't analogous to the famous Prisoner's Dilemma, where defecting necessarily screws over the other participants for your own benefit. Red isn't like defecting, because the other voters are equally free to vote Red and walk away with the same outcome — survival. In other words, Red is both self-interested and universalizable: you can vote Red and unhypocritically tell everyone else to do the same. I'll vote Red to save myself, and I'm not harming you by doing so — you can just do the same. On this view, Red is the game-theoretically stable option. It doesn't require the uphill battle of coordinating people to vote against their own self-interest.

An alternate version of the thought experiment, the "Blender Game," underscores the case for Red:

"Everyone in the world has to decide whether to step into the blender. If at least 50% of the people do step into the blender, it will be unable to overcome their inertia to get started, and everyone survives. If less than 50% of the people step into the blender, then they all get blended up into paste and die. People who do not step into the blender suffer no adverse effects.

Would you step into the blender?

(Blue=step into the blender, Red=don't do that)"

Under this framing, which is logically equivalent to the original, Red becomes the obvious choice. Indeed, this reveals Blue to be utterly insane! Far from being the noble, empathetic, altruistic camp they claim to be, Blues are reckless and downright stupid. Why on Earth would you jump into the Blender of Death when you (and everyone else) could just…not do that? Like, duh?! And Blues have the audacity to call the Reds selfish for not bailing them out of their own idiotic decision. Blues created the problem in the first place by jumping into the blender — if they hadn't done that, Reds wouldn't need to bail them out! If Blues wish to foolishly and pointlessly put their lives in danger, they have only themselves to blame. Not only that, but urging others to vote Blue puts them in danger too — it's like signing people up for a suicide pact! Reds are under no obligation to take responsibility for Blues' recklessness. If Blues wish to cull themselves out of the gene pool, then so be it, and honestly, good riddance!¹

¹Not all Reds were this mean-spirited, but some were.

The Case for Blue

Ok, I'm done playing Devil's Advocate. Now for my actual position: Blue.

Blue's argument starts here: despite the logic of Red I described above, it is simply an empirical fact that many, many people will vote Blue. Why? Well, the initial² group of Blue voters — yes, the troublesome agents of chaos who create a conundrum for everyone else — might do so for totally innocent reasons. They might be children who picked Blue because it's their favorite color, or who just hit a button at random. The original wording specified "everyone in the world," though some on Twitter argued that including children was against the spirit of the question. But even restricting it to people of voting age, you still have those with dementia, mental disabilities, and so on, who have legitimate excuses for irrational behavior. And even if we stipulate they're excluded, voters who don't realize this will mistakenly assume they're in the pool anyway. Beyond that, plenty of people will read the question too quickly and misinterpret it. Others won't think too hard and will naively pattern-match Blue with cooperation in the Prisoner's Dilemma, even though it's not perfectly analogous. Or people who understand perfectly and intend to vote Red, but whose hand slips at the wrong moment and presses Blue. Oops. A silly mistake — but out of billions of voters, guaranteed to happen.

We can debate the size of this initial group — it depends on who exactly is in the voting pool — but the point is, they exist. You might think they're too small a group to risk your life for, or simply not worth saving. But other people will disagree, and they will press Blue for their sake. Whether that's noble heroism or further reckless endangerment is open to interpretation. But again: they also exist.

And now things get interesting. As the number of Blue voters grows, the case for joining them strengthens — both because there are more people to save and because crossing the critical 50% threshold becomes easier. So the case for voting Blue snowballs, and in the end, billions of people will vote Blue.

²I don’t mean initial as in chronologically first, since ostensibly everyone is voting simultaneously, but rather as in the “first-order” thinkers, who reason about their choice at a level below the “higher-order” thinkers who are considering the bigger picture.

Everyone Will Not Just

At this point, Reds are screeching, “NOOO! If everyone would just vote Red, then everyone lives!” Sure, that’s true, but here’s the problem: everyone is NOT going to just.

 

*Taps the sign*

Sorry, but it ain't gonna happen. A huge number of people will vote Blue. It is inevitable. It is undeniable. It is so. The very fact that this question is so hotly debated proves it. Reds can lament that it shouldn't be this way, but too bad: we must deal with human nature as it is, not as we wish it to be. And the snarky replies from (some) Reds — that they don't care because Blues deserve their fate — are pure bluff. Once they realize who's actually among the Blues — their family, lifelong friends, neighbors — their apathy will vanish, and the gravity of the situation will become crystal clear. Sure, Red ensures your own survival, but unless you're a complete psychopath, your utility function includes the survival of other people too.

The Schelling Point is Blue’s Selling Point

I've already explained Red's main advantage: it's self-interested and universalizable. Blue is also universalizable (if everyone votes Blue, everyone lives), but it isn't self-interested, because it asks people to accept some risk of death if Blue loses the vote. This is a hurdle for Blue, but not an insurmountable one. Blue has its own advantage: a wide net in which to land the optimal outcome.

 

What makes this thought experiment so diabolical is that the worst possible outcome (~50% dead) sits on a razor-thin edge right next to the best possible outcome (nobody dies), as the chart above shows. For outcomes on the left side of the possibility space, voting Blue (and urging others to do the same) just adds to the pile of dead bodies. But anywhere on the right side — where Blue gets >50% of the vote — the optimal outcome is achieved. So whereas Red requires 100% to get the best result, Blue can get there with anywhere from 50% to 100% of voters. I've already explained why 100% of people will never vote Red. This means if Red wins, billions of people are guaranteed to die — the largest disaster in human history. Call me crazy, but I think we should try really, really hard to avoid that scenario! And Blue is our only path away from it, because Blue doesn't need 100% — nowhere close. Everyone is not going to just, but a slight majority of people just might just, and realizing this advantage sways even more people into the Blue camp.

Framing Really Matters

The "Blender Game" framing I mentioned earlier shows how easily people's perceptions shift. In that version, a Twitter poll came out 81% Red. But we can also shift the Schelling Point back to Blue with other framings. Consider this: 

“Everyone in the world has to vote in an election. If the Red party wins, they’ll kill everyone who voted for the Blue party. The Blue party won’t kill anyone. Who are you voting for?”

In this version, Blue wins with about 76%. So even though the underlying structure is identical across each framing, it matters enormously whether the wording emphasizes "Blue is suicidal" or "Red is homicidal."

Another variable that moves the results is action vs. inaction. Consider the following two polls: 

This reveals a bias toward whatever option is framed as "doing nothing," regardless of the consequences. Inaction is always preferred as the default — whether it's functionally equivalent to Red or Blue in the original thought experiment. 

Text above reads: "The river is going to flood. Come down to press the blue button and wait in the flood path. If at least half of you press the button, we'll give you all life vests."

Other framings can make either option seem reasonable or ridiculous.

So, is it irrational to vote differently under each of these framings? I think not! Your choice should depend on what you predict most other people will do, and if other people take the framing into account, so should you. Other people are doing their own prediction of others' behavior, creating a recursive loop of modelling everyone else that makes the outcome extremely sensitive to framing. If the framing makes the Schelling Point obviously Red, then voting Red is perfectly sensible — you're avoiding certain death. In the Blender Game, it's safe to assume very few people will jump in. Avoiding a visible threat like that is instinctive even to those incapable of logical thought. And it would be impossible to assemble a winning coalition willing to jump in just to save that handful of lunatics.

If the framing makes the Schelling Point obviously Blue, some Reds will still argue that voting Red is sensible because everyone survives anyway — "No Harm, No Foul." I disagree, because that meme is unstable: if it spreads too widely, it could backfire and ironically end up causing real damage. Better to nip it in the bud. Spreading the Red mindset even when Blue is poised to win easily feels like pointlessly wandering toward the abyss. Even if we have some space between here and the cliff's edge, why encourage people to walk up to it? But whatever — I won't push the point too hard here. Suffice it to say that voting Blue in this case is at least not harmful, so go ahead and do it.

The most interesting case is when the Schelling Point isn't obvious — so let's return to the original framing.

Who Would Really Win?

I think the reason the Twitter poll was so contentious is that it's framed about as neutrally as possible. Both choices require the same action — pressing a button — and neither is explicitly framed as suicidal or homicidal. That leaves it entirely to the reader to insert their own interpretation and values. So, in the end, who actually won the poll? With just shy of 100k votes, Blue won with 58%. A separate poll by the YouTuber MrBeast yielded similar results: 56% Blue with over 300k votes.

But wait — a Twitter poll probably isn't representative of the general population. Near the height of the Twitter fiasco, the question was also posed to a nationally representative sample of Americans, and here were the results:

So: a more lopsided victory for Blue at 63%, and if we exclude the 15% who voted "Not sure," that jumps to 74%. Interestingly, the results were also broken down by how people voted in the last two US Presidential Elections. Not surprisingly, Democrats were more likely to press Blue, but even among Republicans, Blue was the majority. It's unclear how much these results were confounded by the real-world association of blue and red with the Democratic and Republican Parties. I said earlier that the question is framed neutrally, but I think it could have been better still if the buttons were, say, Orange and Yellow — to avoid this confound.

Still, these results were only for the United States. What about the rest of the world? Unfortunately, I couldn't find any international polling on the question — someone should get on that! Some Reds on Twitter argued that the Blue majority in the US would be overwhelmed by Reds elsewhere, but I'm not convinced. If anything, the US has a strong reputation for individualism over collectivism — and Blue still won here. In East Asian societies, most notably Japan, far greater emphasis is placed on advancing the good of the whole community over individual interests. And even in places that don't share this ethos, an intermediate unit between the individual and the wider society is highly valued: the family. Those with deep concern for their family may vote Blue for their sake even if they care little about strangers.

So the limited empirical evidence we have points toward a Blue victory. But Reds have a legitimate objection: the polls aren't realistic, because nothing was actually at stake. No lives were on the line. It's easy to vote Blue on a Twitter poll, but much harder to truly put your life at risk. Perhaps some Blue voters — even enthusiastic and vocal ones — would cowardly switch to Red at the last second. Publicly supporting Blue is useful because it signals virtue to others, but in the privacy of the voting booth, that incentive disappears. In the end, self-preservation would likely pull additional votes toward Red beyond what the no-stakes polls suggest. Whether that's enough for Red to achieve a majority is anyone's guess. But what I hope is clear to everyone is this: we should all be rooting for a Blue victory. No matter what, billions of lives are on the line, and they will only survive if Blue wins. We can reasonably expect the results to be close, and given that, I do view voting Blue as cooperating and voting Red as defecting — even understanding the differences from the Prisoner's Dilemma. So I encourage everyone to join Team Blue!

But My Blue Vote Doesn’t Matter!

Hold on a minute, though. Voting Blue only saves people at a single point: when yours is the pivotal vote. If the rest of humanity votes majority Blue, an incremental Blue vote doesn't improve the outcome (a slight majority is already enough), and an incremental Red vote doesn't hurt it. If the rest of humanity votes majority Red, an incremental Blue vote is harmful because it adds to the death toll, and an incremental Red vote helps you by sparing you from the doomed.

Blues argue you should vote Blue to save other people, but the only scenario where an individual Blue vote actually does that is if the rest of humanity votes exactly 50/50 and your vote pushes Blue over the edge.³ With over 8 billion voters, this is astronomically unlikely! So why not vote Red as the only guaranteed way to save yourself, and let everyone else vote Blue?

The insignificance of a single vote is a problem even for normal elections, not just bizarre hypotheticals. When voting for President, any individual voter is almost certainly irrelevant to the outcome. Your one vote is but a drop in the ocean, so why bother? Indeed, many people refrain from voting for this very reason.

While this is an understandable reaction, there are good counterarguments. Although a single vote is extremely unlikely to sway the outcome, the impact if it does is massive. Multiply some miniscule number (the low probability of being the pivotal vote) by an enormous one (the impact of the election going one way versus the other), and you can end up with some moderate in-between number — the "expected value" of your lonely, individual vote. Just because your vote is one among many doesn't mean it doesn't matter in expectation. Depending on the precise estimates, your vote could be quite important, or at the very least, not trivially unimportant.

In the Red vs Blue button experiment, being the pivotal vote means saving or killing half of humanity — about 4 billion people. That's a pretty big deal if you ask me, something worth taking even a tiny chance of affecting. So what's that chance? Under a uniform prior — where we assume the vote tally is equally likely to be anywhere from 0% to 100% Blue — the probability of being the pivotal vote works out to roughly 1/N, or 1 in 8 billion. In reality, though, we have decent evidence to expect the tally would be close — Blue wins in the polling, but Red would probably gain ground if the stakes were real. At the other extreme from the uniform prior, if we assume everyone's vote is literally a 50/50 coin toss, the probability of breaking a tie is much higher — about 1 in 112,000. This isn't a realistic assumption either; an accurate model would land somewhere between these two extremes. It's very debatable where in this range it should land, and reasonable people can disagree. But what can't be denied is that the vote is likely to be close — and a tight race is precisely where voting Blue has the most leverage for good.

³This assumes the total number of voters, including oneself, is odd. What if the total number of voters is even, and the final vote tally is exactly 50/50? The wording of the original question leaves this scenario undefined — I guess in that case there'd have to be a re-vote!

Ok, But I Still Don’t Care!

The concept of expected value is useful, but relying on it absolutely leads to a thorny philosophical issue: fanaticism. Fanaticism refers to the disposition to take any action, no matter how costly or risky, if it offers even a near-infinitesimally small probability of an arbitrarily large payoff. In expected-value terms, the logic runs: if the reward is infinite (or sufficiently enormous), it multiplies out to dominate every other consideration, even with near-zero probability. The classic example is Pascal's Mugging. Suppose you're accosted by a mugger who proposes a deal: give him your wallet, and in exchange he'll reward you with 1,000 quadrillion happy days of life. The mugger is almost certainly lying, but even so, the expected-value calculation may favor surrendering your wallet simply because the upside is so enormous — and you can never truly be 100% certain he's lying, because you're not infallible. Still, most people share the intuition that giving in is crazy — which shows that pure expected-value reasoning can lead us astray.

So back to Red vs Blue. Perhaps you reject fanaticism and argue, on that basis, that the probability of swinging the outcome is too low to factor into your decision at all. Below a certain threshold, you just don't care. Fair enough. If you're in that camp, I've got a few more arguments up my sleeve.

Red Is Selfish

Let's add another twist. Suppose that if Red wins, instead of only the Blue voters dying, a randomly selected set of people die, regardless of how they voted. The number of people in this death group would equal the number of Blue voters. If Blue wins, the result is still the same: everyone survives.

It's revealing that in this version — with self-preservation removed, since voting Red no longer ensures your survival — Blue becomes the obvious, intuitive choice. This suggests that selfishness, not merely self-interestedness, is at least one of the motivating factors for Red voters — and that Blue is the impartial choice.

Hopefully I don't need to convince you that selfishness is at least a moral problem, but I won't pretend this is uncontested — ethical egoists argue self-interest is the only rational basis for action, and many people hold strong partiality intuitions, especially toward family. But here's the basic case against selfishness: from any view that takes morality to be impartial — that the suffering of a stranger counts the same as the suffering of someone you know — selfishness is just the refusal to weigh other people's interests at the rate they actually have. Your survival isn't worth more than someone else's just because it's yours; it only feels that way from the inside. Almost every major ethical tradition — utilitarian, Kantian, most religious frameworks, even most virtue ethics — converges on some version of this. They disagree about almost everything else, but they agree that "I matter more because I'm me" isn't a real reason. And even most partialists don't extend partiality all the way down to just yourself — they make room for loved ones, communities, sometimes nations, which means they've already accepted the principle that concern extends past the self. The disagreement is about the radius, not the existence, of impartial concern. The Red vs Blue button stress-tests how far that radius reaches. If it stops at you, Red is consistent. If it extends past you at all, Red gets harder to defend.

Beyond that, there's a stronger argument that selfishness might actually be irrational too: see this post for the case. In a nutshell, "the boundaries between people are nebulous and a matter of degree, and caring much more about yourself is arbitrary."

Blue is the Glue that Holds Society Together

 

Imagine a narrow Red victory. 4 billion Blue voters are suddenly dead — the largest catastrophe in human history. Chaos ensues, economies collapse, governments are in shambles. The Red survivors face the aftermath — half of their loved ones have perished, and many are racked with guilt over their votes.

And this, I argue, is just the first disaster. Even once the initial shock subsides, the dust settles, and the logistical rearrangements are made — what comes next? What kind of society will the Reds rebuild?

The Blue voters did something extraordinary. They put their lives at risk. Whether that was a noble sacrifice or a foolish self-own is something of a Rorschach test. But they did it because they had something the Red voters lacked: trust. The Blues trusted their fellow world citizens to do the right thing in sufficient numbers. And they were themselves trustworthy — they held up their end of the bargain. The Reds were fundamentally distrustful — they didn't believe a majority would pull through for Team Blue and therefore chose to just look after themselves. I'm not saying there's never any wisdom in that. Being too trusting can itself be a vice — it allows you to be exploited. There's a balancing act between trust and warranted suspicion, and it's clear that Reds and Blues tip that scale differently. Trust is the fundamental fault line between the Reds and the Blues.

The importance of trust is well-known in economics, political science, sociology, and anthropology. The difference between high-trust and low-trust societies has been studied extensively, and it's stark. Countries like Denmark, Sweden, Norway, Finland, the Netherlands, and New Zealand consistently score highest on social trust measures. In these places, most people assume strangers will generally follow rules, keep their word, and not exploit you. This baseline expectation shapes everything:

• Institutions work better — bureaucracies are less corrupt, contracts are honored, tax compliance is high without heavy enforcement
• Transaction costs fall — less need for lawyers, guarantors, middlemen, and verification at every step
• Public goods thrive — people pay into systems (universal healthcare, pensions) because they trust others won't free-ride
• Risk-taking increases — entrepreneurs start businesses more readily when they trust courts, partners, and customers

In contrast, many parts of sub-Saharan Africa, Latin America, the former Soviet states, and parts of the Middle East score lower. Trust tends to be particularized — strong within family or ethnic group, weak toward outsiders. Consequences include:

• Nepotism becomes rational (you can only trust kin, so you hire kin)
• Corruption is the workaround for unreliable institutions
• Economic activity stays small-scale and informal
• Social capital concentrates within tight networks rather than spreading broadly

Economists have shown that trust functions almost like a tax rate in reverse — high-trust societies get economic activity for free that low-trust societies have to pay enormous costs to replicate through contracts, enforcement, and monitoring. Some estimates suggest moving from low to high trust has a larger effect on per capita income than most policy interventions.

So, back to my question: what kind of society would the Reds rebuild? The most trusting half of the population — the Blues — are now gone, and this would be common knowledge among the Red survivors. This is a recipe for disaster — the second disaster, following the wiping out of half of humanity.

The Reds will inevitably struggle to recreate the old world. A society of Reds is one where people are constantly looking over their shoulders, always on guard against scammers, hackers, and thieves. It's a society where children cannot roam freely outside, library books and shopping carts don't get returned, and deception is the norm.

If you think I'm being overly dramatic, the survey results below should ground the picture. A polling organization asked people the Red/Blue button question alongside commonly used psychometric questions, and the four items that most strongly predicted a Red answer all cluster on the same axis: lower sympathy for others' feelings, less truth-telling, greater appetite for dangerous situations, and more willingness to cut corners to get ahead. These weren't cherry-picked to embarrass anyone — they were the most predictive items in the battery. I'm not arguing that all Red voters are bad people; you can vote Red for principled reasons. But the correlations confirm what the trust framing predicts: Red and Blue track real differences in how people orient toward others, not just differences in game-theoretic reasoning. Concentrate one half of that distribution into the entire surviving population, and the society you get looks meaningfully different from the one we have.

Source

Conclusion

Blue is glue! It is the glue that holds society together. Blue is what builds civilization. Even if you yourself are a Red, even purely out of self-interest, you should want to live among the Blues — the blueness of a society is correlated with its excellence.

So be pro-social, cooperate, and vote Blue!