164 | Herbert Gintis on Game Theory, Evolution, and Social Rationality
What this covers
Herb Gintis argues for a unified behavioral science built on game theory and an expanded rational actor model. The core complaint is that economics, biology, psychology, and sociology offer incompatible accounts of human behavior without serious effort to reconcile them. Gintis proposes that rationality need not mean self-interest—it means having consistent, transitive preferences that include social and moral motives alongside material ones. Using this framework, he shows how game theory, grounded in human evolutionary history, can bridge disciplines and explain real human conduct far better than narrower models.
The conversation moves through several converging lines of evidence. Laboratory games like the ultimatum game reveal that people reject unfair offers even at cost to themselves, and that this holds across cultures with different social values—behavior strong reciprocity, not self-interest, explains. Gintis traces this capacity to human evolution after the advent of accurate throwing weapons made physical dominance unsustainable, shifting societies toward reverse hierarchy and cooperation. That same period saw gene-culture coevolution favoring language, cumulative culture, and what he calls social rationality—preferences for fairness, punishment of cheaters, and moral consistency. The discussion includes Gintis's criticism of standard economics for being insufficiently empirical, his critique of inclusive fitness as inadequate for explaining social behavior, and his defense of group selection against common misreadings. Throughout, the thread is that human behavior becomes intelligible only when you grant that our preferences are genuinely social and moral, not merely acquisitive, and that game theory is the proper language for modeling that fact across all behavioral sciences.
Gintis argues that the behavioral sciences should form one unified, mutually-consistent science of human behavior built on an expanded rational actor model—where 'rationality' means consistent preferences that include social and moral motives, not just self-interest—grounded in game theory and gene-culture co-evolution.
- Economics, biology, psychology, sociology and anthropology give contradictory accounts of behavior and rarely reconcile them, which is unscientific.
- The rational actor model (formal rationality = transitive preferences) is the core tool, but preferences are social and moral, not purely self-interested.
- Human social behavior evolved via gene-culture co-evolution after weapons made physical hierarchy unsustainable, favoring cooperation, language, and strong reciprocity.
Social sciences contradict each other because disciplinary isolation prevents fields from reconciling incompatible frameworks and core theories.
- Unlike the natural sciences, where overlapping fields agree or fight to find the truth, the social sciences overlap but say totally different things—economics assumes self-interested incentive-driven behavior with no morality while sociology talks only of norms and proper behavior—and they can't both be right; in fact both are wrong.
“But in the social sciences, they overlap, but they say totally different things... Well, they can't both be right. And in fact, they're both wrong.”
- The mutual neglect among social sciences stems from early-20th-century academic structures in which each field hires, fires, publishes and conferences only its own, so disciplines develop in strange, isolated directions without caring what other fields say.
“you have these developed... In the early 20th century, these feudal fiefdoms where if you're an economist, you publish in certain journals and you talk to other economists and you don't care what sociologists say.”
- Genuine interdisciplinarity does not mean combining the wisdom of different disciplines, because their wisdoms contradict rather than add up; real progress requires asking what each discipline must change to become compatible with the others—e.g., what economics must change to fit sociology and psychology, and vice versa.
“people think inter-disciplinary means, well, you just combine the wisdom of the different disciplines, but that's false because the wisdoms don't agree with each other. They don't add up. They are contradictory.”
- It is a strange attitude that scientists claim to care about truth yet, when they disagree with someone outside their own field, simply don't care and forget about it—a double standard that perpetuates contradictory cross-disciplinary accounts.
“scientists talk about being interested in getting the truth, but when they disagree with each other, if they're not in the same field, who cares? Forget about it. I think that's a very strange attitude.”
Humans are not self-interested; they cooperate, care about fairness, and value social norms even when doing so costs them materially.
- In complete markets behavior approximates self-interest, but in face-to-face and social interactions—even within firms and consumption choices—people are not self-interested: workplace cultures of cooperation are real economic behaviors, and consumers care about values like not buying goods made by slave labor.
“people do not behave in the economy as though they are self-interested. Now when you have very complete markets you get almost self-interested behaviour but when you get interactions among people, face-to-face interactions or other social interactions even in the economy people do not behave self-interestedly”
- In the honesty game, where player A truthfully or deceptively tells player B which box holds more money, the game is logically indeterminate, yet almost always player A tells the truth (at a cost to himself) and player B assumes truthfulness—as long as the split is not too uneven—demonstrating that people dislike being dishonest.
“when you actually play this game, almost always the player A tells the truth. As long as the splitting of the money is not too uneven, player A will tell the truth, and player B will assume the player A is telling the truth.”
- Formal rationality does not mean caring only about yourself; the only people who care exclusively about themselves are sociopaths (about 10% of the population), while everyone else cares about a wide variety of things beyond themselves.
“Rational means you only care about yourself. Well, that's of course ridiculous... The only people who only care about themselves are sociopaths.”
- In the ultimatum game, self-interest predicts the proposer offers one dollar and the responder accepts, but across societies worldwide (Boston, Palo Alto, Peruvian jungle, Mongolian highlands, African jungle) the most common offer is half, and offers of $3 or less are commonly rejected because responders derive more satisfaction from punishing unfairness than from the cash—demonstrating humans are not self-interested and like to hurt those who treat them unfairly.
“When you play this game that has never happened... The most common offer is half. I'll give you $5. And if you offer $3 or less in most societies, people will reject it.”
Humans evolved cooperation through selection for group-level traits, throwing weapons eliminating dominance hierarchies, and cumulative culture enabling complex social coordination.
- Group selection has been wrongly reduced to groups competing with one another; properly understood, saying a trait is selected does not require competition (escaping a fox enhances fitness without fighting a conspecific), so group selection means only that the evolutionary dynamic favors the evolution of groups with certain social interactions, regardless of intergroup competition.
“the notion of group selection, when people talk about it, it's... They've reduced it to groups competing with each other... but that's not right... group selection in general, means that the evolutionary dynamic is favourable to the evolution of groups with certain social interactions, that's all.”
- Only humans really have cumulative culture, where a body of knowledge is maintained and passed from one generation to the next; animals have culture but not cumulative culture—when birds learn to open milk bottles, the knowledge is later forgotten and lost rather than accumulating.
“humans only... Only humans really have cumulative culture that is where from one generation to the next, you maintain a body of knowledge and pass it on, and animals, animals have culture, but they don't have very much cumulative culture”
- Wars are won by foot soldiers rather than by atom bombs, bombers, or tanks, as shown by the United States losing in Vietnam, Iraq, and Afghanistan; the rise of small-caliber 'democratic' weapons over cavalry helped drive the push toward democracy after the world wars.
“foot soldiers win wars... it's not the atom bombs and the big bombers, and the tanks, it's the foot soldier”
- Chimpanzees cannot speak not because they are stupid but because they lack the tongue muscles, cheek muscles, and a low larynx; human language arose through gene-culture co-evolution in which a little communication ability was valued and rewarded with more offspring, producing genetic changes that enabled more communication and increased cultural dependency in a self-reinforcing circle.
“when people say language is because people have big great brains, that's just not right... People have language because of gene culture evolution.”
- Unlike chimpanzees who cannot throw accurately and fight only with hands, early humans became skilled with throwing weapons, which meant an alpha male could be killed in his sleep; this made power-based hierarchy unsustainable and pushed human societies toward cooperative, 'reverse hierarchy' leadership where people choose leaders who promote the group's values and fitness.
“once you get an accurate weapon, you can kill a guy in his sleep. So you can't maintain a hierarchy based on power. And human societies move towards cooperative leadership. It's called reverse hierarchy by Chris Boehm”
- Human cooperation evolved partly because hunter-gatherer groups made war on each other, which required cooperation; non-social species do not make war, and only social species such as ants or humans can, so evolution explains not only conflict but also cooperation.
“cooperation evolves because hunter-gatherer groups make war against each other, and to make war, you have to cooperate. Non-social species do not make war, only social species can make war, like ants or humans.”
- Because dominance could no longer be maintained by force after the advent of accurate weapons, the way to win leadership became having a good mind and being able to speak it; this selected for language and the human vocal apparatus, so the invention of weapons led to the evolution of language.
“the way you defeat a person who has a good mind and can speak it, is to have a good mind and speak it... The group won't elect you their leader because you bashed the leader over the head, only if you're having better ideas.”
Strong reciprocity—rewarding fairness and punishing unfairness without material return—is the fundamental basis of human social stability and cooperation.
- Beyond Trivers's reciprocal altruism ('you scratch my back'), humans exhibit strong reciprocity—spontaneously helping others and spontaneously hurting those who hurt them, with no expected return; this love of rewarding nice behavior and punishing unfair behavior, not governments, is the major source of social stability, since hunter-gatherer societies had no jails, judges, or police, only praise and ostracism.
“strong reciprocity is what kept things going, people help those that they thought were being nice and hurt those that are not being nice... this is a major reason you have social stability, it's not governments”
- The negative side of strong reciprocity—vengeance and retaliation—is treated as abnormal in sociology textbooks but is actually a basic human behavior: people derive pleasure from hurting those who treated them unfairly, as in the ultimatum game and revenge movies, with no material gain in mind.
“people love to hurt people who hurt them. Go to your sociology book and see if you can find the word retaliation... it's treated as a... It's treated as an abnormal behaviour to want vengeance and retaliation, but it's one of the basic human behaviours”
- In small-scale hunter-gatherer societies cooperation pays off for individuals because cooperators are rewarded by other group members and non-cooperators are ostracized; these societies are run democratically rather than hierarchically, making human sociality singular among species.
“there's a huge benefit to cooperation, which pays off for the individual, because they're rewarded by other members of the group when they behave in a cooperative way, and they're ostracized when they behave in a non-cooperative way.”
Game theory across evolutionary and behavioral contexts provides the universal language for modeling decision-making across all disciplines.
- In an evolutionary game where players play repeatedly and successful strategies reproduce more, the only equilibria of that dynamical system are Nash equilibria of the underlying game; this justifies using game theory wherever social evolution has occurred, because evolution favors strategies that converge on Nash equilibria—for example fingers players evolving toward 50-50.
“In an evolutionary game, where people play the game over and over and the people who do well get to reproduce more like in the Darwinian evolutionary sense, then the only equilibria of that dynamical system is a Nash equilibrium of the underlying game.”
- Game theory is decision theory with more than one decision maker under partial information: players, strategy sets, and payoffs depending on combined choices; it is the basic language usable across all behavioral disciplines, whether modeling dung beetles on cow patties or humans voting in elections.
“it's really the basic language that can be used across all behavioural disciplines, whether you're talking about dung beetles on cow patties or you're talking about humans voting in elections”
- The major development in game theory came from biology when John Maynard Smith and Price built a game-theoretic model to explain fights among butterflies over mating space, which grew into evolutionary game theory, the basis for understanding the dynamics of animal and human behavior.
“the major development that happened in game theory, really in biology, when Maynard Smith, John Maynard Smith and Price, developed a game theoretic model to explain fights among butterflies for mating space.”
- The textbook claim that 'rational players choose Nash equilibria' is false; the prerequisites for it are extremely implausible, as shown by the matching-fingers game where if one player randomizes 50-50 the other is indifferent to all strategies, removing any reason for either to play the equilibrium.
“They say, "Rational players choose Nash equilibria." Well, that's just false.”
People act from moral principles about what is right rather than calculating consequences, exemplified by voting and Kantian equilibrium reasoning.
- A Kantian equilibrium is a moral equilibrium where each chooses the strategy that, if everyone chose it, would make everyone best off; voting exemplifies it because no rational self-interested or even altruistic actor should vote (no large election has ever been decided by one vote), yet people vote and care, obeying a deep moral principle—'behave so that if everybody behaved that way we'd all be better off'—without even knowing they obey it.
“If you're rational, according to the rational actor model you don't vote, you'll never vote... no vote in a large election an election more than 40,000 voters, no vote has ever been determined by one person.”
- Most people, especially about politics, are not consequentialist: they do not reason about the consequences of their actions but instead think in terms of what is right and what is wrong, wanting what's right to happen regardless of consequences.
“I think people are fundamentally in many ways, not consequentialist. They don't think in terms of what the consequences are. They think in terms of what's right and what's wrong.”
- Treating voters as self-interested is inconsistent because a self-interested person would not vote at all; voting for your social group—e.g., a union member voting the union line—is a form of social rationality, not selfishness.
“the whole idea that people are self-interested if they vote in an election is just inconsistent. You may vote for your social group, if you're a union member, you may say, "Okay, I vote the union," that doesn't make me selfish, if I were selfish, I wouldn't vote at all.”
People hold culturally specific values that shape their behavior in laboratory games and real-world interactions regardless of anonymity or context.
- In a 16-country MacArthur-funded study, people took the values of their own society into laboratory games even under complete anonymity, behaving the same way they do in their societies; cooperative whale-hunting cultures like Lamalera even made 'hyper-fair' offers of $8 out of $10 and rejected offers that were too generous, proving that values genuinely matter.
“people take the... They take their values from their society, whatever they are, and they take them into the laboratory, even in conditions of complete anonymity”
- Humans uniquely evolved both a private persona—going about daily business without evaluating the impact on the wider world—and a public persona that enters a Kantian frame to judge what is right, who to support, and what values to accept, supporting social rationality rather than individual rationality; people switch between these, and crucially humans can act on the rules of the game, not just play within them.
“the public persona, you automatically enter a different frame in which you're thinking as a Kantian kind of categorical person... one that supports what I called social rationality as opposed to individual rationality, and people do both all the time.”
Inclusive fitness models fail for social species because multi-locus genomes create conflicting evolutionary interests that simpler relatedness assumptions cannot explain.
- Hamilton's inclusive fitness rests on a single-locus, additive model assuming genes don't affect each other, but in real multi-locus genomes different loci have conflicting interests (one gene's poison provokes suppressor mutations elsewhere), creating a complex evolutionary dynamic in which inclusive fitness no longer works; it explains non-social species but cannot explain the incredibly complex cooperation of social species.
“The model that he built to show this is a single-locus model... his theory is what happens exactly at one locus... when they affect each other, inclusive fitness, no longer works.”
- The idea that social bees are one queen's highly related offspring is a myth—queens die often, some species have six or seven queens and many males, so measured worker relatedness in some bee and wasp species is near zero—yet they show incredible cooperation, demonstrating that inclusive fitness cannot explain social species.
“bees are incredibly social, they're not highly related by the way, this is a myth, one queen, and they're all the offspring, no way... if you measure the relatedness of workers in some species of bees and wasps, they're actually quite low.”
- When 127 biologists signed a letter to Nature rejecting the Nowak-Wilson-Tarnita critique of inclusive fitness, the appeal to numbers echoes the '100 authors against Einstein' episode, to which Einstein replied that if he were wrong, one would be enough—truth is not settled by the number of objectors.
“the response was, 127 biologists sent a letter to Nature saying, "This is wrong."... after Einstein developed his special theory, there was a book written in German called a hundred authors against Einstein... And he said, Well, if I'm wrong, one would be enough.”
The rational actor model is a mathematical framework enabling differential equations, not a literal description of psychology or behavior.
- The fundamental theorem of the rational actor model is that any consistent (transitive) set of preferences can always be represented as maximizing a utility function; this is a formal mathematical convenience that brings in calculus and differential equations, not a literal description of what people are doing.
“That's fundamental theorem of rational actor model. They talk about utility and maximizing something, but it's all really... That's not what's really going on... for mathematical purposes, it's really nice to have a function you can maximize.”
- Behaviors that look dumb, like smoking, are not irrational in the formal sense; people often act against their own interest, make performance errors, or are momentarily out of their minds, but no one has ever developed a systematic alternative to the rational actor model for explaining behavior.
“if you smoke cigarettes, that's a pretty stupid thing to do these days, but it's not irrational... The point is that no one's ever developed anything alternative to the rational actor model that you can use systematically to explain behavior.”
- Rationality has two meanings: instrumental rationality (choosing efficient means to a single goal) and formal rationality (having transitive preferences—if you prefer A to B and B to C you prefer A to C); because people have multiple competing goals, instrumental rationality is often irrelevant, and formal rationality with transitive preferences is what lets us model behavior with mathematical decision theory.
“there's another called formal rationality, an individual is formally rational, if his preferences are transitive... if you prefer A to B and B to C, you'd prefer A to C”
Behavioral game theory empirically observes actual choices and infers values, resolving the unfalsifiability objection to non-selfish preferences.
- The old objection that allowing non-selfish utility lets you fit anything (epicycles) is answered by behavioral game theory, which takes subjects into the laboratory, observes how they actually behave, and infers what they value from that behavior, making the standard objection to non-self-interested preferences obsolete.
“we now have a whole discipline called Behavioural game theory in which we take subjects in the laboratory and see how in fact they behave and infer what they value from their behaviour and that is very advanced now.”
- Objections that ultimatum-game results are artifacts of trivial stakes are wrong: when poor farmers earning about $300 a month play for $900-1000, nothing changes—they still reject offers worth roughly a month's wages.
“get farmers who make about $300 a month and play the ultimatum game for $900 or a $1000. There, you'll see whether they accept it. You know what happens? Nothing is different. They still reject three... They'll reject a month's wages.”
Interdisciplinary training requires mastery of single disciplines first, then common foundations in statistics, methods, and theory across all behavioral sciences.
- High levels of political ideology taint judgment across the behavioral sciences—sociology is extremely liberal, anthropology has embraced postmodernism's denial of science and resists experimenting on simple societies—which made interdisciplinary, experimental work like Gintis's hard to fund through anthropologist-run NSF review and hard to publish.
“I have found very high levels of political ideology tainting people's judgment in almost all of this field... anthropologists have gone into what's called post-modernism, which is a denial of the importance of science”
- Only economics and biology have core theories that everyone worldwide learns the same way, and both use the rational actor model; psychology and anthropology lack cumulative core theory and remain voguish and scattered precisely because they avoid the rational actor model, which requires mathematical modeling that is genuinely hard to do.
“there are two of the behavioural sciences that have core theories... that's economics and biology, and these are the two disciplines that use the rational actor model. Psychology is all over the place”
- The internet has transformed the ability to do interdisciplinary work: where learning a new subject once required weeks of hauling journals from the library, Gintis can now learn a new subject in about a year and communicate with distant experts in other fields, facilitating transdisciplinary research.
“Now with the internet, I can learn a new subject in a year without any problem... I think the internet has made all the difference in the world in the ability to gather information from all over, and to talk to experts in fields that are not your own”
- Young researchers should first go deep into a single discipline and master it before branching out, because without working out the intricacies of one particular field you cannot deal with the intricacies of any discipline; an ideal training would give behavioral students a common first year of statistics, mathematical modeling, scientific method, and the cores of psychology, sociology, political theory, economics, and anthropology, then specialize.
“for young people, when you start out, you have to go into a particular discipline and you have to really learn them very well before you can branch out”
- Economics is not as empirical as it should be: a thousand-page graduate economics textbook contained not a single empirical fact, unlike a quantum mechanics text full of empirical effects, because economists try to derive human behavior from the concept of rationality alone—which cannot determine choices among the huge number of relevant variables.
“In the economics book there wasn't a single empirical fact, not one. 1000 pages no facts and the fact is what the economists are trying to do is derive human behaviour from the concept of rationality”
Entangled minds cause people to form beliefs against social networks rather than evidence, and humans operate private and public personas governed by different rationality principles.
- The rational actor model's serious defect for humans is that it assumes individual subjective priors, whereas people actually have 'entangled minds': they form and validate beliefs not against empirical evidence but against the beliefs of their social network, becoming adamant about truths simply because everyone around them believes them.
“people have what we call entangled minds, that is what you believe and what you think it's a function not of you, yourself, but of your network of entangled brains around you”
- Universal cooperation can be supported as a Nash equilibrium only if there are no errors, via trigger strategies ('cooperate, but if anyone defects, never cooperate again'); but once errors exist—someone gets sick and misses work—cooperation collapses, models become far more sophisticated, and cooperation is never universal.
“That's called a trigger strategy. And that's fine until someone gets sick and they don't come to work one day... So if there are errors, then you have big complications and the models become much more sophisticated and cooperation is never universal.”