225 | Michael Tomasello on The Social Origins of Cognition and Agency
What this covers
Michael Tomasello sits down with Sean Carroll to argue that human uniqueness rests not on superior individual intelligence but on a capacity for shared intentionality—the ability to coordinate minds with others toward common goals. The conversation traces how this social coordination ability became the foundation for everything humans do distinctively well: language, morality, agency, and culture itself. Tomasello draws on developmental psychology, evolutionary biology, and comparative studies of apes to build the case that a child isolated from human society would develop no language, mathematics, or institutions, regardless of raw cognitive capacity, because these are not products of individual brainpower but of assimilation into a cultural-social process.
The discussion moves through several interconnected domains. Tomasello explains how agency is not a mysterious inner essence but a control system—a goal-directed loop found in thermostats and lizards alike, which humans extended into shared intentionality and collaborative action. He argues that morality evolved specifically from the demands of collaborative foraging, where interdependence rather than genetic relatedness drove the evolution of fairness and obligation. The episode also takes up language as "the icing on the cake"—a late conventionalization built atop deeper capacities like common ground and recursive communication that preceded it. Throughout, Tomasello challenges modular accounts of human cognition (separate modules for language, theory of mind, institutions) in favor of a unified adaptive problem: collaborating to find food under conditions of partner choice. The conversation ventures into free will, the evolutionary timeline of hominin development, and what autism reveals about the biological substrate of social cognition.
Tomasello argues that what makes human cognition unique is not raw individual brainpower but a capacity for shared intentionality—coordinating minds with others—and that agency, morality, and even language are evolutionary outgrowths of cooperative, interdependent social life.
- A child raised alone on a desert island would not develop language, algebra, or social institutions, showing human cognition depends on assimilation into a social-cultural process.
- Morality evolved from interdependent cooperation in collaborative foraging, governed by a mathematics analogous to kin selection but substituting interdependence for genetic relatedness.
- Agency is a control-system organization (goal, perception, action to match) that scales from lizards to humans, with humans adding shared/plural agency.
Cooperation based on interdependence, not competition, drove the evolution of recursive theory of mind and moral obligation.
- Morality is the human psychology built for cooperation, and its key concept is interdependence: just as kin selection has a mathematics for how much to help genetic relatives, interdependence follows the same formula with interdependence substituted for genetic relatedness, so that to the degree I depend on you for mating, foraging, or other needs, it is in my long-term fitness interest to help you, share fairly, and treat you with respect—and computational cooperation models fail to capture this because they assume independent agents rather than interdependent social networks.
“The key concept is interdependence.”
- Recursive theory of mind likely evolved for cooperation rather than competition, because in cooperation I actively want my partner to read my mind—I make my plan perspicuous and facilitate mutual understanding—whereas competition gives no incentive to make one's intentions transparent.
“the thing about cooperating is I want you to read my mind. If we're cooperating, I'm trying to make it... If you and I are gonna go do something together, I try to tell you, Sean, here's what I'm thinking. Here's my plan.”
- Pure altruism that sacrifices the individual out of existence cannot evolve—because giving away all resources leaves no offspring to contribute to the gene pool—so Darwin was right that individuals must look out for themselves, but cooperation is not necessarily zero-sum and can be a win-win situation, which is the basis for the human version of cooperation.
“There's a famous quip that in evolutionary biology, the definition of altruism is, that which cannot evolve.”
- There are two distinct moralities: the morality of sympathy/helping—rooted in mammalian oxytocin-driven care for offspring and generalized to non-kin, widespread across mammals—and the morality of fairness and obligation, a more cognitive, rational process where equal effort generates a felt obligation to share equally; the sense of obligation is not the same as being nice and arises specifically from collaborative foraging.
“I distinguish... the morality of sharing and helping and sympathy, and the morality of fairness, which is a more cognitive rational kind of process.”
Humans uniquely leverage shared intentionality and collaborative agency to pool knowledge across individuals and generations.
- The uniquely human cognitive capacity is intimately bound up with a unique social capacity: humans don't operate on individual brainpower but are adapted to leverage what other people know, collaborating, communicating, socially learning, and teaching, which is why a chimp and a child raised in isolation would not end up that different, but a child raised in human culture develops language, mathematics, and institutions.
“the really unique cognitive part is intimately bound up with the unique social part, that what we do is put our heads together with others”
- Beyond the executive control systems of mammals that plan ahead and consider possibilities, humans evolved to be shared (plural) agents: when two people collaborate they form a joint agent with a shared goal, shared knowledge maintained through communication, and joint self-regulation, where shared agency is the behavioral collaboration and shared intentionality the underlying cognitive capacities enabling it.
“humans have evolved to be shared agents or social agents, or plural agents.”
- Thinking feels like a private act in one's own mind, but like a jazz musician playing alone in his apartment, the individual learned the genre from others, plays an instrument others built and perfected, and works within forms invented before him—so even solitary symbol manipulation is the product of an evolutionary and cultural process into which the individual is assimilated.
“it's like a jazz musician playing a jazz rift in his apartment by himself. Yes, he's doing it by himself, but it's he learned jazz from other people.”
- Humans have a far longer ontogeny than other apes—youngsters in hunter-gatherer groups don't produce more calories than they consume until around age 16—and this extended dependency is dangerous and costly for both children and investing parents, so it must serve a function: providing time to learn, socially coordinate, and become a member of a culture that transmits the accumulated knowledge enabling humans to live in environments like the Arctic where an unaided individual would quickly die.
“there are studies with hunter-gatherer groups that youngsters don't bring in more calories than they consume until they're 16-years-old”
Meaningful cooperation requires individual agents with genuine choice within evolved decision-making systems, not physics-level determinism.
- Denying free will amounts to taking a physicist's view of psychology and eliminating psychology altogether; while the decision-making apparatus is itself evolved and not chosen, given that apparatus the individual genuinely chooses within it, and the existence of agentive decision-makers requires keeping levels straight rather than reducing everything to deterministic physics—moreover, no one truly behaves as if free will is false, since they still get angry at others and worry over decisions.
“people who want to deny free will are essentially taking a kind of a physicist view of psychology”
- Children with autism tend to be missing exactly the social-cognitive capacities Tomasello identifies as uniquely human, indicating these capacities have a biological basis—though autism is a spectrum and individuals vary.
“you have children with autism, is the syndrome, they tend to be missing exactly the things that I'm zeroing in on.”
- Every species is by definition unique in some way, so merely noting that humans are different from other animals says nothing in itself.
“Every species by definition, every species is unique in some way.”
- A 'good negative' in comparative experiments is when chimps pass a control condition very similar to the experimental condition while children pass both, making the chimps' failure in the key condition a meaningful fact rather than an uninterpretable null result, since ordinary negative findings could be due to anything.
“what I call a good negative is that the chimps pass a control condition that's very similar to the experimental condition, that they're only a little bit different, and kids in the same as close to this identical methodology as possible, they pass. And that to me, sets the chimp failure in the key condition in... It makes it a meaningful fact.”
- Agency evolves precisely where the environment is unpredictable: nature cannot in principle specify what a squirrel should do with a particular nut in a particular landscape, so instead of determining the action it builds in a decision-making apparatus—the organism gathers information and makes the best decision—meaning the apparatus is evolved/determined but the particular decision on a given occasion is not.
“agency evolves when there are unpredictabilities and nature builds in an apparatus for the individual to deal with those unpredictabilities through what I call informed decision making.”
- Agency and behavior can be a leading edge in evolution: if insects on the ground disappear and only tree insects remain, lizards that choose to climb trees survive, and subsequently selection favors traits like longer claws that support tree-climbing—so the behavioral choice precedes and directs the genetic change, a process called genetic assimilation that Darwin himself recognized as the agency of organisms.
“behavior and agency can be a leading edge in the evolutionary process itself.”
- The current challenge for humans lies in the cooperation realm rather than the cognitive one—we risk blowing ourselves up or degrading the environment beyond livability—and progress on civilization-scale advances came historically when isolated knowledge-building cultures bumped into each other (e.g., the Greeks getting math from the East), so a more integrated world community would give the social-cultural process another boost, but only if the cooperation problem and nationalistic resistance can be solved first.
“right now the challenge for humans is more on the cooperation realm. We're gonna blow ourselves up or we're gonna degrade the environment to the degree that we can't live in it anymore”
Language presupposes and builds upon prior non-linguistic capacities like concept formation, communicative intention, and recursive thinking.
- Language is critically important for uniquely human cognition but is 'the icing on the cake'—a later conventionalization that presupposes prior capacities like concept formation, communicative intentions, recursive thinking, and non-linguistic communication; in principle you cannot start from nothing and simply assign words, so explaining humans by 'language' requires accounting for those enabling capacities first.
“language is the icing on the cake. It's the jet rocket at the end, it's not the starting point.”
- While humans are biologically evolved to use language—prepared with capacities for concept formation, communication, imitation, and internalization—the specific generative grammar/syntax that Chomsky proposed as innate is actually the least likely aspect of language to be innate, because grammatical structures vary substantially across the world's languages even as concepts and pragmatic principles show commonalities.
“the actual grammatical structures are quite different in different languages. So I actually think that's the least likely thing to be innate out of all the aspects of language.”
- Uniquely human gestures like pointing and iconic pantomime communicate by relying on shared common ground: the same point can mean a piece of spear-wood or some berries depending on what partners mutually know, and the recipient searches only the space of what both know in common—a capacity chimps lack and which serves as a non-linguistic halfway house to language.
“with the same pointing, I can point over there and mean there are some berries... I can mean anything with the pointing, but it depends on our common ground. It depends on what we know in common.”
A single adaptive challenge of collaborative foraging under partner choice explains multiple human capacities hierarchically organized together.
- Rather than many independent evolved modules, humans face one organizing problem—collaborating to get food under partner choice—which brings with it about fifty hierarchically organized sub-problems (cognitive, communicative, socio-moral, reputational), making it more plausible that a single big adaptive challenge in the last million years organizes many smaller capacities than that we evolved separate adaptations for language, theory of mind, institutions, and mathematics each.
“We have a problem and the problem is collaborating to get food. It's just that that problem has about 50 sub-problems that it brings with it, and they are both cognitive problems, communicative problems, socio-moral problems”
- For roughly the first four million of six million years hominins were essentially bipedal apes with ape-sized brains; stone tools appear around 2-2.5 million years ago, collaborative foraging less than a million (perhaps half a million) years ago, and culturally distinct groups with conventional language only about 100,000-200,000 years ago—so the genuinely distinctive human capacities emerged fairly late.
“for the first four million years, some people might say more like three or three and a half, but somewhere more than half, we were basically bipedal apes”
- Evolution does not select for organisms simply 'to be smarter' or 'have bigger brains'; selection requires a specific concrete problem, and since big brains are metabolically expensive there must be a specific benefit driving their evolution.
“evolution doesn't work like that, you have to have a specific problem.”
Each organism occupies an experiential niche shaped by its agentive needs; humans uniquely include the normative world of how things ought to be.
- Just as each species occupies an ecological niche, each occupies an 'experiential niche'—the world as it perceives and understands it, shaped by its agentive needs, scanning only for what is relevant to its decisions—so humans do not enjoy privileged access to reality as it really is; uniquely, however, part of the human experiential niche includes the mental worlds of others, culture, and the normative world of how things ought to be, since no other creature plans its actions with respect to what ought to be the case.
“I use the term experiential niche. The experiential niche is the world as you perceive and understand it, which evolved to support your agentive actions in it.”
- Great apes share an understanding of the perception and goals of other individuals—a subordinate chimp behaves as if it knows a dominant cannot see food, and apes understand what others are trying to do even when the goal isn't achieved—but they do not understand mental states more narrowly defined, such as sharing thoughts or disagreeing about opinions.
“This is sort of ground zero that I think the great apes all share, is an understanding of the perception and goals of other individuals and not the mental states more narrowly defined.”
- An agent is not a homunculus but an organism with a specific control-system organization: it has a goal state it wants in the environment, perceives the environment to check whether it matches the goal, and acts to make it match—the same circular/cybernetic causality as a thermostat or bodily homeostasis—so behavior should be understood as goal-directed control rather than stimulus-response, just as 20th-century biology replaced vitalism not with a vital substance but with a different organization of ordinary matter.
“it has to have a goal state, that it wants to be the case in the environment, it has to perceive the environment and see if it matches with the goal state. And if it doesn't, it has to be able to do something to make it match.”
- Goal-directed agency likely requires a nervous system: single-celled organisms and bacteria probably do not have goals, while vertebrates from fish and especially lizard-like land animals onward clearly pursue goals as shown in experimental research—though invertebrate lineages like octopi and insects are a separate, uncertain case.
“single-celled organisms, bacteria, single-celled organisms and that stuff do not have, they don't... Organisms without a nervous system, probably don't.”