David BolesThe Cooperator’s Dilemma: How Martin Nowak’s Mathematics of Kindness Became a Blueprint for Control
Martin Nowak wanted to prove that cooperation is the animating force of evolution. He succeeded. His mathematical models, published across decades of work at Oxford, Princeton, and Harvard, demonstrate with formal rigor that cooperation is not an anomaly in a competitive world but a fundamental mechanism by which biological complexity arises. Genomes cooperate. Cells cooperate. Organisms cooperate. Societies cooperate. Without cooperation, there are no multicellular bodies, no ant colonies, no languages, no civilizations. This is not sentiment. It is mathematics. And it is precisely because the mathematics are correct that they are dangerous.
Nowak is Professor of Mathematics and Biology at Harvard University, an Austrian-born scientist trained in biochemistry and mathematics at the University of Vienna, where he worked under Peter Schuster on quasispecies theory and with Karl Sigmund on evolutionary game theory. He earned his doctorate sub auspiciis praesidentis, the highest academic honor Austria can bestow on a graduating student. He moved to Oxford, where he collaborated with Robert May (later Lord May of Oxford) on spatial evolutionary dynamics and virus population models. He established the first center for theoretical biology at the Institute for Advanced Study in Princeton in 1998. In 2003, he came to Harvard to found the Program for Evolutionary Dynamics (PED), where he would spend two decades formalizing the mathematics of cooperation, cancer evolution, language emergence, and infection dynamics.
His landmark 2006 paper in Science, “Five Rules for the Evolution of Cooperation,” laid out the theoretical architecture that his 2011 book SuperCooperators: Why We Need Each Other to Succeed (co-written with science journalist Roger Highfield) would translate for a general audience. The core argument is elegant and, on its face, optimistic: natural selection, left alone, favors defectors over cooperators, but five distinct mechanisms can reverse this tendency and allow cooperation to evolve. Those mechanisms are kin selection, direct reciprocity, indirect reciprocity, network reciprocity, and group selection. Each mechanism can be reduced to a simple mathematical rule specifying the conditions under which cooperation becomes the favored strategy. Each rule expresses a threshold: when the benefit-to-cost ratio of a cooperative act exceeds a critical value determined by the mechanism’s structure, cooperation wins.
The book is earnest. It is hopeful. It tells a story about vampire bats sharing blood meals, about cancer as a failure of cellular cooperation, about human language as the greatest cooperative innovation since the gene. Nowak calls humans “supercooperators” because we are the only species that deploys all five mechanisms simultaneously. The implication is that our capacity for cooperation is not just biologically real but biologically supreme. We are, in his framework, evolution’s greatest collaborative achievement.
This is all true. And none of it prevents the mathematics from being turned inside out.
The Five Mechanisms as Five Exploits
What Nowak mapped are not merely descriptions of how cooperation arises. They are, read from the other direction, specifications for how cooperation can be manufactured, directed, and harvested by any actor with sufficient control over the relevant variables. Each mechanism contains its own vulnerability. Each rule that tells you how to promote cooperation also tells you how to engineer compliance that feels like cooperation to the people inside the system.
Direct Reciprocity: The Obligation Engine
Direct reciprocity is the simplest mechanism: I help you now, you help me later, and we both benefit as long as we expect to interact again. Nowak’s mathematical condition is precise. Cooperation through direct reciprocity succeeds only when the probability of future interaction between the same two individuals exceeds the cost-to-benefit ratio of the cooperative act. If you and I will meet again many times, the cost of helping you today is offset by the expected return from your future help. The strategy that dominates in this environment is not pure tit-for-tat (which is too brittle, collapsing into mutual defection after a single error) but “win-stay, lose-shift,” a more forgiving strategy that sustains cooperation through noise.
The exploit is in the precondition. If you can engineer a situation where people believe they will interact with you repeatedly and indefinitely, you can extract cooperation from them even when the exchange is not mutual. Subscription services, employer-employee relationships with annual review cycles, government benefit programs tied to ongoing compliance, social media platforms that reward daily engagement: all of these create artificial conditions of repeated interaction. The person inside the system cooperates because their evolved psychology recognizes the pattern. They feel the pull of reciprocity. They return to the platform, they renew the subscription, they comply with the bureaucratic requirement, because the structure tells them the relationship will continue and defection carries a cost.
But the entity on the other side of the interaction is not bound by the same psychology. A corporation does not feel the tug of reciprocal obligation. A government agency does not experience guilt for failing to return a favor. The asymmetry is structural: the human cooperates because direct reciprocity is wired into primate social cognition; the institution extracts because it designed the interaction pattern to trigger exactly that response. The “repeated game” is real for the person and fictional for the institution, which can end the relationship, change the terms, or alter the benefit-to-cost ratio at any time without experiencing the psychological cost of defection.
Consider the modern employment relationship. An employee cooperates (works hard, stays late, defers complaints) because the structure of employment creates an expectation of continued interaction: there will be another paycheck, another review, another year. The employer benefits from this cooperation while retaining the unilateral power to terminate the relationship. The employee’s cooperation is genuine. The employer’s reciprocity is contingent. Nowak’s mathematics describe the employee’s behavior perfectly. They do not describe the employer’s, because the employer is not playing a repeated game. The employer is playing a series of one-shot games while the employee believes both parties are in a repeated game. This mismatch is not a bug in the model. It is the exploit.
Indirect Reciprocity: The Reputation Weapon
Nowak considers indirect reciprocity the most important mechanism for human cooperation, and he is probably right. Indirect reciprocity works through reputation: I help you not because I expect you to help me, but because others are watching, and my willingness to help builds a reputation that will cause others to help me in the future. The mathematical condition is that the probability of knowing someone’s reputation must exceed the cost-to-benefit ratio of cooperation. Language, Nowak argues, evolved in part to serve this mechanism. We gossip. We evaluate. We track who is trustworthy and who is not. This reputational calculus is what allows cooperation to scale beyond pairs of individuals who interact repeatedly.
The danger is obvious and immense. Whoever controls the reputation infrastructure controls the conditions for cooperation. And in the modern world, reputation infrastructure is not distributed among gossiping primates. It is centralized in databases.
Credit scoring systems (FICO in the United States, similar systems globally) are indirect reciprocity engines. They assign each person a reputation score based on their history of “cooperation” with financial institutions. A high score means you have reliably cooperated (paid debts, maintained accounts, avoided default). The score then determines whether others will cooperate with you (extend credit, offer favorable terms, rent you an apartment). The mathematics are identical to Nowak’s model. The probability of knowing your reputation is essentially 1.0 in a world of universal credit reporting. Therefore, the threshold for cooperation is easily met, and people cooperate.
But cooperate with what? With whom? The content of “cooperation” in a credit scoring system is defined by the institutions that build and maintain the scoring model. Cooperation means paying your bills. It means maintaining debt. It means participating in a financial system on its terms. The reputation system does not reward you for helping your neighbor move furniture or lending your car to a friend in need. It rewards you for being a reliable revenue source for financial institutions. The indirect reciprocity mechanism is operating exactly as Nowak describes. The mathematics are satisfied. But the cooperation is directed, not organic. It serves the architects of the reputation system, not the cooperators within it.
China’s social credit experiments take this further, attaching reputational scores to civic behavior, political speech, social associations, and consumption patterns. The mathematics are the same. The mechanism is the same. The outcome is that “cooperation” becomes indistinguishable from “compliance,” and the person inside the system cannot easily tell the difference, because the psychological experience of cooperating to maintain one’s reputation feels the same whether the reputation system is tracking genuine prosocial behavior or political obedience.
Social media platforms represent a third variant. Platforms like Instagram, TikTok, and formerly Twitter construct reputation systems (follower counts, likes, shares, verification badges) that trigger indirect reciprocity behavior. Users cooperate with the platform (producing content, engaging with others’ content, spending time on the platform) because the reputation system rewards them for doing so. The platform harvests this cooperation as engagement metrics, advertising revenue, and behavioral data. The user experiences the warm glow of reputational validation. The platform experiences profit. The mathematics of indirect reciprocity are perfectly satisfied in both directions. The exploitation is invisible precisely because it operates through a mechanism that evolution shaped to feel good.
Network Reciprocity: Whoever Designs the Graph Wins
Nowak’s third mechanism, developed in his landmark 1992 Nature paper with Robert May, showed that the spatial structure of interactions matters enormously. In a well-mixed population (where everyone interacts with everyone equally), defectors always win. But when interactions are local, restricted to neighbors on a network, cooperators can form clusters that protect themselves from exploitation. Cooperators surrounded by other cooperators thrive; defectors on the edge of cooperative clusters can invade, but the cluster structure slows the invasion and allows cooperation to persist.
The strategic implication is that whoever controls network topology controls the conditions for cooperation. This is not a metaphor. It is a direct application of the mathematics.
Social media algorithms determine who sees whose content, who appears in whose feed, who gets recommended as a connection. These algorithms are network reciprocity engines. They construct the “spatial structure” of online social interaction. A platform that clusters users into engagement-optimized groups is, in Nowak’s terms, constructing a network topology. If the topology is designed to maximize engagement (which is to say, to maximize the platform’s extraction of attention), then the cooperative clusters that form will be optimized for engagement, not for the welfare of the cooperators.
Corporate organizational design is another application. Who reports to whom, who collaborates with whom, who has access to information and who does not: these are network topology decisions. A company that understands network reciprocity can design org charts that promote exactly the cooperative behaviors it wants (cross-functional collaboration, knowledge sharing, collective problem-solving) while preventing the formation of cooperative clusters that might oppose management (unions, whistleblower networks, collective bargaining groups). The mathematics tell you which structures promote cooperation and which fragment it. The application is straightforward.
Gerrymandering is network reciprocity applied to democratic geography. By controlling which voters are grouped into which districts, political actors control the spatial structure of electoral cooperation. Voters who might form cooperative clusters around shared interests are separated. Voters whose “cooperation” (voting behavior) serves the redistricting party are grouped together. The mathematics of spatial evolutionary dynamics describe exactly why this works and how to optimize it.
Group Selection: The Factory of Tribes
Nowak’s treatment of group selection (which he and others now call multilevel selection) demonstrates that groups of cooperators outcompete groups of defectors, even when defectors dominate within groups. The mechanism requires that groups compete, that there is variation in the level of cooperation between groups, and that groups with more cooperators produce more offspring groups. Under these conditions, cooperation at the group level is favored even though individual defectors within groups do better than individual cooperators.
The exploit is the deliberate manufacture of group identity and intergroup competition. Nowak’s own reviewers noted the problem clearly: group selection favors within-group niceness and between-group nastiness. This is the mathematical basis of tribalism. It is also, historically, the most reliable mechanism by which authoritarian movements generate internal cohesion.
If you want a population to cooperate internally (pay taxes, report dissent, sacrifice personal interests for collective goals), you manufacture an external threat. The perceived competition between groups raises the benefit-to-cost ratio of within-group cooperation. Nationalists understand this intuitively. So do corporate culture architects who position their company against competitors while demanding employee loyalty. So do political parties that define themselves primarily through opposition. The mathematics of multilevel selection explain why “rally around the flag” effects work, why wartime economies produce extraordinary domestic cooperation, and why authoritarian regimes invest so heavily in identifying and publicizing external enemies.
The earnestness of Nowak’s presentation (he sees group selection as enabling the great cooperative achievements of human civilization, from agriculture to the United Nations) obscures how perfectly the same mathematics describe the cooperative achievements of fascism. The cooperative group that builds a hospital and the cooperative group that builds an internment camp are both satisfying the mathematical conditions for multilevel selection. The model does not distinguish between them. It cannot. The variables are the same.
Kin Selection: Manufacturing Family Where None Exists
Kin selection, formalized by W.D. Hamilton in 1964, is the oldest and most biologically grounded of the five mechanisms. Organisms cooperate with genetic relatives in proportion to their degree of relatedness, because helping a relative who shares your genes indirectly promotes the survival of those shared genes. Hamilton’s rule states that altruism is favored when the coefficient of relatedness between donor and recipient exceeds the cost-to-benefit ratio of the altruistic act. Nowak has a complicated relationship with Hamilton’s rule (his 2010 Nature paper with E.O. Wilson and Corina Tarnita argued that kin selection is less explanatory than previously thought, provoking a famous counterresponse signed by over 130 biologists), but the mechanism remains one of his five pillars.
The exploit is the simulation of kinship where none exists. “We are family.” “Band of brothers.” “Our company family.” “Fellow citizens.” “Children of God.” These are not merely sentimental phrases. They are invocations of kin selection psychology, designed to lower the threshold at which people will sacrifice personal interest for the group. When a military unit trains together, eats together, sleeps together, suffers together, and adopts shared rituals, symbols, and origin stories, it is manufacturing fictive kinship. The result is that soldiers will take risks for their unit-mates that they would not take for strangers, because their psychology has been calibrated to treat those unit-mates as kin.
Religious organizations, fraternities, political movements, cults, and nationalist ideologies all exploit this mechanism. The more completely an institution can simulate the markers of genetic relatedness (shared appearance through uniforms, shared language through jargon, shared history through founding myths, shared suffering through initiation rites), the more effectively it triggers kin selection psychology, and the more cooperation it can extract from its members. The cost of this cooperation is borne by the members. The benefit accrues to whoever designed the kinship simulation.
The Epstein Entanglement: A Case Study in the Exploitation of Cooperation Science
Any serious discussion of Martin Nowak’s work must confront the fact that the Program for Evolutionary Dynamics, the institutional home of his cooperation research, was founded with $6.5 million from Jeffrey Epstein. This was the largest single gift Epstein made to Harvard, part of a total of more than $9 million in donations to the university between 1998 and 2007. Epstein, a convicted sex offender who would later be charged with sex trafficking before his death in federal custody in 2019, did not merely donate money and walk away. He embedded himself in the program.
Harvard’s own 2020 review found that after Epstein’s 2008 conviction and release from prison, he continued to visit the PED offices more than 40 times between 2010 and 2018. He had a personal office in Nowak’s lab. He had a key card. He was typically accompanied by young women described as his assistants. His publicist requested that PED post information about Epstein on the harvard.edu domain because, as the publicist wrote, it would be helpful for Google search results. PED complied. Epstein’s foundation page was linked from the PED website under a tab labeled “Friends.” Epstein was the only “Friend” listed.
More recently, documents released by the U.S. Department of Justice in 2025 revealed that Epstein’s involvement went beyond access and reputation-laundering. He discussed research topics with Nowak and his graduate students. He suggested lines of inquiry, including “commercial evolution” and “prelife.” He facilitated visa arrangements for at least one graduate student. He funneled scholarship money through a Ph.D. student to young female mathematicians in Romania. He reviewed page proofs of a Nature paper before publication and offered advice on handling criticism. In 2025, Nowak was placed on administrative leave a second time after his name appeared more than 8,000 times in the newly released DOJ Epstein files.
The irony is lacerating. Epstein was a man who built his entire social and financial empire on the exploitation of cooperation mechanisms. His method was indirect reciprocity: he cultivated relationships with scientists, politicians, and financiers by offering gifts, access, and introductions, building a reputation as a brilliant and generous patron of science. He used network reciprocity: he positioned himself as a hub connecting elite nodes (Harvard professors, tech billionaires, heads of state), making himself indispensable as a broker of social capital. He manufactured fictive kinship: his dinners, his island retreats, his intellectual salons created a sense of belonging and shared identity among participants. He exploited direct reciprocity: every gift came with an implicit expectation of return, whether it was a letter of reference, a favorable public statement, or simply continued access and association.
And he funded, specifically and deliberately, the research program that mathematically formalized every one of these strategies. He did not fund a chemistry lab or an engineering department. He funded the mathematics of cooperation. He then used the institutional affiliation (Harvard, the Program for Evolutionary Dynamics) as a reputational asset, laundering his public image through association with the most prestigious cooperative institution in American academia.
Nowak has not been charged with any crime related to Epstein’s offenses. The Harvard review found policy violations, not criminal conduct. But the structural relationship between Epstein and PED is itself a perfect illustration of the very dynamics Nowak’s research describes. Epstein was a defector masquerading as a cooperator, using the mechanisms of cooperation (reputation, network position, reciprocal obligation, fictive kinship) to extract value from a system whose participants genuinely believed they were cooperating for the advancement of knowledge. The mathematics predicted this possibility. The researchers did not see it, or chose not to.
The Cyclical Trap
The deepest and most troubling insight in Nowak’s work is the finding that cooperation is inherently cyclical. Cooperators increase in number and trust, building successful clusters and institutions. Then a minority of defectors, positioned to exploit the high-trust environment, invade. Cooperation collapses. Eventually, cooperators rebuild. The cycle repeats. Nowak frames this as a feature of evolutionary dynamics, a permanent oscillation that can be modulated but never eliminated.
For a government or corporation seeking to exploit cooperation, this cyclical pattern is not a problem. It is the business model. The cycle describes exactly what extraction looks like over time. During the cooperative phase, the institution harvests trust, labor, engagement, compliance, and revenue. During the collapse phase, the institution restructures, rebrands, and resets the conditions for a new cooperative phase. The people inside the system experience the collapse as a betrayal. The institution experiences it as a cost of doing business.
Platform companies cycle through this pattern visibly. A new platform launches, cultivates a cooperative community of early adopters, builds network effects, then monetizes by degrading the experience for users while extracting more value from advertisers. Users eventually defect (leave the platform), but by then the platform has captured enough network position to survive, or a new platform launches and the cycle restarts. This is Nowak’s evolutionary dynamic playing out in real time, at internet speed.
Political cycles follow the same pattern. A new administration or movement builds cooperative coalitions around shared goals. Trust increases. Policy achievements accumulate. Then insiders begin to extract (corruption, patronage, self-dealing), trust erodes, the coalition fragments, and a new movement arises to rebuild cooperation on different terms. The cycle is so regular that political scientists have formalized it independently of evolutionary biology, but Nowak’s mathematics show that it is not unique to politics. It is a property of any system in which cooperation and defection coexist.
What Nowak Missed, or Chose Not to Say
SuperCooperators is a book about the conditions that produce cooperation. It is not a book about the conditions that produce just cooperation. This is not a minor omission. It is the central weakness of the work.
Nowak’s five mechanisms are content-neutral. They describe the structural conditions under which organisms will choose to cooperate, but they are silent on what the cooperation is for. Cooperation to build a hospital and cooperation to build a surveillance state satisfy the same mathematical conditions. Within-group cooperation that produces a democratic parliament and within-group cooperation that produces a paramilitary organization both emerge from the same multilevel selection dynamics. A reputation system that tracks genuine generosity and a reputation system that tracks political loyalty both promote cooperation through indirect reciprocity.
The book occasionally gestures toward this problem. Nowak acknowledges that defectors can invade cooperative groups, that cooperation cycles, that punishment mechanisms can themselves become exploitative. But these acknowledgments are treated as complications within a fundamentally optimistic narrative, not as structural features of the mathematics that demand equal weight. The title is SuperCooperators, not SuperExploiters. The framing celebrates cooperation’s triumphs without adequately confronting the fact that the same mathematics, applied with different intent, describe cooperation’s capture.
This omission is not unique to Nowak. It is endemic to a certain strain of evolutionary optimism that mistakes the existence of cooperation for its benevolence. Cooperation is not inherently good. It is a strategy. It can be deployed in service of any goal. The mathematics do not care. A reader who absorbs Nowak’s five rules as a celebration of human goodness will be poorly prepared to recognize those same rules operating as mechanisms of control in their workplace, their government, their social media feed, and their financial system.
The Responsibility of the Mapmaker
Nowak drew a map. The map is accurate. The territory it describes is real. But a map can be read by anyone, and the same map that helps a traveler find water helps an army find the traveler. The five rules for the evolution of cooperation are, simultaneously, five rules for the engineering of compliance. The mathematics are identical. Only the intent differs.
The question is not whether Nowak should have refrained from publishing his research. Suppressing accurate mathematics is never the answer. The question is whether the scientific community and the reading public have a responsibility to read the map with both eyes open: to see not only the beautiful cooperative structures it reveals but also the exploitative architectures it enables. The answer, given what we now know about who funded the map’s creation and what they used its institutional credibility to accomplish, should be self-evident.
Cooperation is real. It is mathematically demonstrable. It is essential to every level of biological and social organization. It is also the single most exploitable feature of human psychology, and anyone who tells you otherwise is either not paying attention or is the one doing the exploiting.
#consumerValue #cooperation #engineering #epsteinFiles #fiveMechanisms #harvard #martinNowak #needingEachOther #oxford #politics #princeton #research #rogerHighfield
TheEdinburghBookshop