Why You Cooperate Even When It Costs You

Think about the last time you chose to be generous with someone who hadn't earned it yet. Maybe you covered for a new coworker, stayed soft with a partner who'd been distant, kept showing up for a friendship that felt one-sided. And some part of you knew—logically, clearly—that you were taking a risk. That the math didn't necessarily favor you.

You did it anyway.

That instinct is worth taking seriously. Not as a character flaw, not as naivety, and definitely not as irrationality. It's actually one of the most sophisticated things a human being can do.

I've been sitting with a lecture from MIT's Economic Applications of Game Theory course (Fall 2025), taught by Ian Ball, and what keeps pulling at me isn't the math—it's the moment in the classroom when Ball asks his students to play the Prisoner's Dilemma against each other, face to face, three times in a row. And most of them cooperate.

The setup, briefly, because it matters

The Prisoner's Dilemma is a two-choice scenario: you either cooperate with the other player or you defect. In Ball's course version, the payoffs look like this—mutual cooperation gets you both 2 points; if you defect while they cooperate, you get 3 and they get -1; if you both defect, you both get 0. The uncomfortable truth is that no matter what the other person does, you personally score higher by defecting. Defecting is what game theorists call a "strictly dominant strategy."

So the rational move is to betray. Every time. The math is unambiguous.

Ball's students didn't do that.

To make the game playable in a classroom without screens, Ball had students use a hand signal system—essentially a two-option simultaneous reveal, like a simplified version of rock-paper-scissors but with only two choices, not three. (Worth being precise here: the Prisoner's Dilemma is a two-option game—cooperate or defect—not a three-outcome game. Ball used the rock/paper distinction purely as a physical signaling mechanism to make simultaneous choices visible.) What happened when MIT students looked each other in the eye and played? Cooperation broke out everywhere. Ball's reaction is wry: "I would argue maybe someone was a little irrational there, but we'll see."

Spoiler: the math does eventually vindicate the defectors. In a finitely repeated game—one where both players know exactly when it ends—the logic collapses backward from the final round. In the last round, there's no future relationship to protect, so you defect. Both players know this, so the second-to-last round is now also effectively the last meaningful round. So you defect there too. And so on, all the way back to round one. The unique "subgame perfect Nash equilibrium"—the game theorist's gold standard for rational play—is to defect every single time.

The students who cooperated were, by that measure, playing the wrong game.

Except I'm not sure they were.

What your nervous system already knows

Here's what the game theory model doesn't include, and what your body has been computing since childhood: most relationships don't have a known endpoint.

James Coan's social baseline theory—developed through decades of neuroimaging research—finds that the human brain treats other trusted people as literal cognitive and physiological resources. When we're in secure relationships, our nervous systems don't have to work as hard. Threat detection quiets down. Metabolic load decreases. We're not just emotionally better off; we're neurologically more efficient. Cooperation isn't a soft choice. It's a survival strategy that got baked into our biology because, for most of human history, being alone was the most dangerous thing that could happen to you.

The students who cooperated in Ball's classroom weren't being naive. They were responding to millions of years of adaptive pressure that says: if there's any chance this relationship continues, protect it. Their nervous systems were running a different calculation than the one on the chalkboard—and for most real-world situations, that calculation is actually correct.

This is exactly what Ball gestures toward when he mentions that the theory of infinitely repeated games looks completely different from finitely repeated games. He quotes economist Paul Milgrom—a Nobel laureate whose work Ball cites in the lecture—as saying: "The game is always larger than you think."

I've been turning that line over since I encountered it. From clinical observation, I hear it everywhere.

The game is always larger than you think

When sitting with clients who grew up in unpredictable households—parents whose moods were volatile, whose affection was conditional—what often emerges is a nervous system that learned to defect first. To armor up before the other person can hurt you. To protect yourself in round one because you know from experience that round one is all you get before things go sideways.

That's not irrationality. That's a learned strategy that once made complete sense. If your early relationships had knowable endpoints—if the warmth always dried up eventually, if the safety never lasted—then your system learned to treat every interaction like the final round. Defect now. Don't leave yourself exposed.

The Prisoner's Dilemma, run through an attachment lens, stops being an abstract economics puzzle and starts sounding like a clinical intake form.

And here's what makes the MIT classroom moment genuinely moving: these are students who knew the theory. Ball's class had already covered the math. They understood, at least intellectually, that the rational move was to defect. And most of them still looked someone in the eye and chose to trust.

Robert Axelrod's famous computer tournaments in the 1980s bear this out. When he invited game theorists to submit strategies for a repeated Prisoner's Dilemma and ran thousands of simulated games, the strategy that consistently outperformed every other wasn't the most sophisticated or defensive one. It was "tit-for-tat"—cooperate on round one, then mirror whatever your partner did. Start with trust. Punish betrayal. Forgive quickly. Cooperation isn't just emotionally appealing; in repeated, relationship-like conditions, it's strategically dominant.

The key word is repeated. And the key question is whether you believe the game is ongoing.

What this means for actual humans in actual relationships

Ball makes a point in the lecture that I keep coming back to: the conditions that enable cooperation are repetition and observability. You need a future, and you need to be seen. If either of those is missing, the incentive structure collapses.

That maps, with uncomfortable precision, onto what breaks relationships.

When someone doesn't believe the relationship has a future—because they're exhausted, because trust has been repeatedly violated, because their nervous system has catalogued too many final rounds—they stop cooperating. Not because they're selfish. Because the architecture that makes cooperation rational is no longer intact. No future, no observability, no cooperation. The math follows the emotional reality, not the other way around.

This is also why isolation is so clinically dangerous. It's not just that loneliness feels bad. It's that without the relational context that makes cooperation meaningful—without someone to be seen by, without a future to protect—people begin to make choices that look, from the outside, like self-destruction. But internally, those choices are often the only ones that make sense given the game they believe they're playing.

The students who cooperated in Ball's classroom were doing something that looked irrational on a whiteboard and deeply human everywhere else. They were betting on continuity. They were choosing to treat the relationship as larger than any single round.

Ball's theoretical framework ultimately confirms what they intuited: if the game is genuinely finite and both players know it, defection wins. But the deeper insight—the one Ball points toward and Milgrom apparently lived by—is that we almost never actually know when the game ends.

So the real clinical question isn't whether you cooperate or defect. It's what kind of game you believe you're in—and whether that belief reflects your present reality or a past you haven't finished grieving.

If you're navigating trust issues or relationship patterns that feel stuck, the American Psychological Association's therapist locator (apa.org/helpcenter) and Psychology Today's directory (psychologytoday.com/us/therapists) are good starting points. If you're in crisis, the 988 Suicide and Crisis Lifeline is available by call or text.

By Samir Patel, Mental Health & Wellness Correspondent