In this blog post, I will examine the explanation of reciprocal altruism presented in ‘The Selfish Gene’ from the perspectives of the theory of the selfish gene and game theory, and I will critically review the implications and limitations of applying evolutionary game theory to biological behavior in nature.

Explaining Reciprocal Altruism in Animals Through the Theory of the Selfish Gene and Game Theory

In ‘The Selfish Gene’, Dawkins ultimately describes all living organisms—or at least most living organisms except humans—as survival machines designed to pass on genes. Of course, since this view is difficult to accept based solely on the conclusion, Dawkins systematically unpacks his theory of how genes have been selected and spread over the course of several chapters. In this process, Dawkins posits the premise that genes are selfish. Here, the concept of “selfish behavior” refers not to a moral sense but to behavior that is advantageous for passing on oneself and one’s replicators to future generations. For example, if a gene—regardless of the process involved—is advantageous for transmitting its replicators to future generations, that is selfish behavior. Conversely, genes that fail to do so will be outcompeted by more advantageous genes and gradually disappear through the process of natural selection. Similarly, organisms influenced by such genes may tend to behave in ways that favor the gene’s continued transmission, while genetic traits that do not do so may decline over successive generations. Viewed from this perspective, it seems that all species we know of would act strictly selfishly. However, countless species we know exhibit altruistic behavior. Some species emit warning calls to protect their group from predators; others share food with individuals who have been unable to find it; and parents of countless species make sacrifices for their offspring. At first glance, these behaviors appear to contradict Dawkins’ theory; however, Dawkins explains altruism—and, in particular, the reciprocal altruism of animals that help one another—by examining these phenomena from the perspective of the selfish gene theory and game theory, which originated in economics. The concept of reciprocal altruism itself was not first proposed by Dawkins; rather, it was systematically introduced by Robert Trivers in a 1971 study. It is a theory that seeks to explain the conditions under which behaviors involving the exchange of help between unrelated individuals can be maintained through natural selection.
However, how to interpret game theory in this context becomes a critical issue. Traditional game theory has primarily been used to analyze situations—such as economic decision-making or strategic interactions—where each actor acts by considering the benefits gained from their own choices while anticipating the choices of others. Consequently, applying game theory directly to biological behavior carries the risk of interpreting it as though organisms actually calculate the behavior of others, predict future outcomes, and select strategies accordingly. However, modern evolutionary game theory does not presuppose such conscious and rational calculation. Evolutionary game theory has evolved into a theory that analyzes how an individual’s behavioral strategy changes and is maintained within a population through natural selection, and the concept of an evolutionarily stable strategy (ESS) is also used within this context. Therefore, rather than concluding that applying game theory to nature is inherently wrong, it is necessary to distinguish between classical, human-centered rational choice models and evolutionary game theory, and to examine in what sense Dawkins is using game theory.
Suppose a flock of birds is infested with ticks. If it is difficult for any single bird to eliminate the ticks on its own, it would be beneficial for all the birds to cooperate with one another to remove them. This type of mutually supportive relationship is called reciprocal altruism. According to Dawkins’ explanation, from the perspective of the selfish gene theory, the reason why organisms created by thoroughly self-interested genes exhibit such behavior is simple. It is because the genetic traits that induce such behavior are better suited for survival and reproduction, and thus are more likely to be passed on as a result of natural selection. Returning to the example of the flock of birds, we can assume that within the same group, there are individuals with genetic traits that induce the behavior of removing ticks from one another, and others without such traits. Individuals of the second type are more likely to suffer harm from ticks, which could put them at a disadvantage in terms of survival and reproduction. In other words, if the behavior of removing ticks from one another is ultimately advantageous for survival and reproduction, the genetic traits associated with that behavior are likely to occur more frequently. Furthermore, birds within the same group are likely to be closely related. This means they share many genes; the concept that indicates how genetically close two individuals are is called relatedness. For example, since each parent passes on half of their genes to their offspring, the relatedness between a parent and a child is 1/2. Therefore, if relatedness is high among individuals in the same group, behaviors that benefit members of that group may ultimately contribute to the transmission of the genetic traits they share. This explanation ties into the logic of kin selection and demonstrates that even though altruistic behavior may appear to be a sacrifice for another individual, from a genetic perspective, it can help ensure that one’s own genetic traits are passed on to the next generation.
Dawkins also explains reciprocal altruism under the assumption that individuals are not related by blood. Suppose the members of a certain group frequently fail to remove ticks from one another. In that case, this group may be at a disadvantage in terms of future survival compared to a group whose members do remove ticks from one another. Ticks can lead to serious consequences, such as spreading infectious diseases; even if, by sheer luck, the damage caused by ticks is not particularly severe, this difference may be difficult to ignore. The key point here is that this behavioral strategy can be understood not as a process in which individuals consciously modify their strategy based on repeated experiences, but rather as a behavioral trait that is inherited and whose frequency within a population changes through natural selection. For example, if an individual possesses the behavioral tendency of a “free-rider”—one who only takes help but never gives it—that tendency may not easily change over the course of that individual’s lifetime. Dawkins assumes a scenario in which the “altruist” strategy (unconditional help), the “free-rider” strategy (taking help without giving it), and the “revenge strategy” (withholding help only from those who have betrayed the individual) are mixed. If we simulate a scenario where these three strategies are moderately mixed and compete for survival, the “helpers” initially appear to be at a disadvantage compared to the “free-riders.” Over time, the helpers nearly disappear, but the free-riders are put at a disadvantage by the “revenge-seekers,” who refuse to help them, and eventually only the helpers and revenge-seekers remain. Since the resentful individuals do not betray first, a tendency not to betray one another can be maintained in a situation where the “bees” and the “resentful” coexist. Ultimately, this demonstrates that altruistic behavior can lead to higher survival and reproductive success than selfish behavior in certain environments. Therefore, from the perspective of genes, genetic traits that induce altruistic behavior may persist at a higher frequency than those that induce only selfish behavior.
Dawkins explains this phenomenon once more through game theory. Although he actually cites the example of vampire bats in this case, let’s use the case of birds and ticks as an example since the situation is logically similar. First, to better understand game theory, let’s look at the Prisoner’s Dilemma, one of the most famous examples in game theory. In the Prisoner’s Dilemma, two prisoners, A and B, play a game. Suppose that if A cooperates and B also cooperates, both receive $300, but if B defies, B receives $500. Although the additional gain from defection may seem relatively small, it is ultimately more profitable, creating an incentive to choose defection over cooperation. Furthermore, suppose that if A defects and B cooperates, B loses $100, but if A defects and B also defects, B loses only $10. Therefore, defection is advantageous for B in any scenario; the problem is that this fact applies equally to A as well. Ultimately, both players end up defecting and suffering a loss. This is a disadvantageous outcome for both parties compared to cooperating and earning $300 together. While the outcome of this dilemma itself is straightforward, the situation changes slightly if the game is repeated. This is because, in repeated interactions, players can adopt strategies such as retaliating when betrayed or resuming cooperation if the other player cooperates again. If the other player continues to defect, one can apply pressure by retaliating with continued defection, making it clear that both parties will suffer losses. Furthermore, while the gain from defecting when the other player cooperates is relatively small, the loss incurred if the other player defects can be significant. Ultimately, one must consider not only how much benefit is gained or how much loss is incurred, but also how many times these interactions are repeated. Therefore, in repeated interactions, long-term outcomes become more important than the benefit gained from a single action, and the likelihood of maintaining a cooperative relationship increases. This repeated Prisoner’s Dilemma later became one of the key models used in the study of evolutionary game theory and the evolution of cooperation.
One of the concepts developed using this theory is ESS. ESS stands for Evolutionarily Stable Strategy and is generally translated as “evolutionarily stable strategy.” ESS is a concept that describes a state in which, when a particular strategy is widespread within a population, it is difficult for other strategies to infiltrate and displace it. In other words, it does not simply mean that a particular strategy is the best strategy, but rather refers to the stability of that strategy—meaning that even if mutations or other strategies emerge in a population where it is widespread, it is not easily replaced by natural selection. The concept of ESS developed into a core concept of evolutionary game theory through the research of John Maynard Smith and George Price, and today it has established itself as a representative theoretical tool for analyzing biological behavior.
For example, consider a situation involving a “hawk” and a “dove.” Suppose that when two individuals from the same animal population encounter each other, the winner receives 50 points and the loser 0 points; suffering serious injury results in -100 points; and the cost of a protracted conflict is -10 points. When two doves encounter each other, they do not attack but only engage in threatening behavior; therefore, they bear the cost of wasted time and are assumed to score 40 points and -10 points, respectively. When two hawks meet, one wins and scores 50 points, while the other sustains severe injuries and scores -100 points. When a dove and a hawk meet, the hawk wins and gains 50 points, while the dove gains 0 points. If all individuals are doves, the average score is (40–10)/2, or 15 points. However, if a hawk intervenes in a situation consisting solely of doves, the number of hawks will gradually increase, and eventually, the ratio of hawks to doves may converge to a steady state. If we apply the above payoff structure directly, the expected payoffs for both strategies become equal in a mixed state where the hawk ratio is 7/12 and the dove ratio is 5/12. Therefore, in this example, a strategy frequency of 7 hawks and 5 doves could constitute an evolutionarily stable mixed state. However, since this ratio depends on the game’s pay structure, it does not appear identically in all hawk-dove games. Furthermore, the mixed ratio does not necessarily mean that an individual chooses between hawk and dove strategies with a fixed probability at every moment; rather, it may refer to a state in which different strategies coexist at a certain frequency within the population.
Let’s explain the process by which birds cooperate from this game-theoretic perspective. Suppose there are two birds, A and B. Assuming A cooperates, if B also cooperates, both can eliminate the mites, though B expends a small amount of energy and time. On the other hand, if B defies, B can reap the benefits without bearing the cost of eliminating the mites. Although the gain is very small, there is still an incentive for the defector. Conversely, if A defects, B will expend energy only if B cooperates; if B defects, B can avoid that cost. Looking at a single interaction, defection appears advantageous in both cases, so it might seem that defection would prevail. However, since ticks do not attach to a bird just once in its lifetime, birds repeatedly experience interactions similar to the Prisoner’s Dilemma. Similar to the previous example, the benefit gained from defection is merely a small energy savings, whereas the loss incurred from failing to remove the tick due to the opponent’s defection can be quite significant. Ultimately, under these conditions, an evolutionary stable state in which cooperation is maintained at a certain frequency may emerge, and behavioral tendencies that induce cooperation from others or respond to defection may be selected within the population. For certain species, as in the example of blood-feeding bats presented by Dawkins, the ability to distinguish and remember the past behavior of other individuals can play a crucial role in maintaining cooperation. If such an ability exists, it enables a response to or retaliation against betrayal. Therefore, rather than betraying others simply to gain a minor benefit from a single interaction, individuals may be incentivized to cooperate with one another through repeated interactions. Ultimately, behavioral strategies that enable sustained cooperation can be selected in specific environments, and the genetic traits associated with those strategies can thrive within a population. It is through this very logic that Dawkins links repeated interactions with natural selection to explain cooperation and reciprocal altruism.

What are the issues to consider in Dawkins’ explanation?

Both of the above explanations describe the process by which reciprocal altruism arises and is maintained, and they reach similar conclusions in that behavioral tendencies that induce cooperation can be preserved through natural selection. However, there is a point in the second explanation that warrants careful consideration. As mentioned earlier, traditional game theory has been widely used to describe situations in which intelligent agents adjust their strategies based on the context. Therefore, if we accept this concept at face value, it is easy to interpret it as agents observing their opponents’ strategies, adjusting their own strategies accordingly, and acting with an eye toward long-term benefits rather than short-term gains. However, evolutionary game theory does not necessarily presuppose such conscious calculation. Rather, in biology, “strategy” is interpreted as a genetically influenced behavioral tendency or rule, and the process by which the frequency of strategies changes through natural selection can be analyzed. Therefore, applying game theory to nature cannot be said to be fundamentally wrong; the problem arises when the logic of the classical rational actor model is applied without distinguishing it from that of evolutionary game theory.
However, even if we acknowledge this distinction, there remains a problem with equating the behavior of genes and individuals with the choice of strategies in game theory. Genes are not agents that actively modify their strategies by analyzing situations based on their genetic information throughout an individual’s lifetime. While an individual’s behavioral tendencies may be influenced by its developmental process and environment, the fundamental process by which natural selection operates is the differential survival and reproduction of genetic variants within a population across generations. Therefore, the “change in strategy” referred to in evolutionary game theory does not necessarily mean that an individual calculates and changes its strategy during the course of a game. Rather, it should be understood as a process in which the survival and reproductive success of individuals with different behavioral strategies vary, resulting in changes in the frequency of strategies within the population over successive generations. Considering this, rather than simply concluding that “it does not satisfy the premises of game theory,” it is more accurate to examine which form of game theory is being applied and to what extent the model’s assumptions align with the actual process of natural selection.
When using the theory of the selfish gene to explain reciprocal altruism, the explanation is relatively straightforward. Genetic traits with various variations will exist through processes such as random mutation and genetic recombination, and some of these will be more advantageous for survival and reproduction in specific environments. Among these genetic traits, there may be some associated with altruistic behavioral tendencies. As Dawkins emphasizes, what matters is not whether such traits emerged with a specific purpose from the outset, but rather what outcomes they produce through the process of natural selection once they have emerged. A tendency toward altruistic behavior may contribute to the transmission of one’s genetic traits by aiding closely related individuals, or it may be selected even among unrelated individuals if mutual aid is advantageous for long-term survival and reproduction. On the other hand, explanations that apply game theory to the evolution of genes and populations analyze the process by which different behavioral strategies interact, and how the frequency of a particular strategy increases or decreases depending on the outcome. In both cases, the result is the same: reciprocal altruism can be sustained under specific conditions. They also share the commonality that, among various possibilities, relatively more successful behavioral strategies persist through natural selection. However, the former focuses on biological processes such as genetic variation and natural selection, while the latter employs mathematical models involving interactions and payoffs between strategies, as well as changes in strategy frequencies. In other words, just because the outcomes may be the same does not mean that the processes and implications of the two explanations are identical.
Some might counter as follows: The process of rationally trying out various options is, to some extent, similar to the process by which, among individuals with different genetic traits, those with traits closer to the optimum survive in greater numbers through natural selection; likewise, the process by which high-performing strategies are selected is similar to the process by which genetic traits that yield high reproductive success are retained among individuals. Therefore, one could argue that since the process by which the frequency of a specific strategy increases through natural selection resembles the process of rationally selecting the optimal strategy, the application of game theory is justified. In fact, evolutionary game theory is a theory that has developed these similarities into mathematical models, rather than treating them as mere analogies. Therefore, it is difficult to criticize the application of game theory to nature as inherently wrong.
However, it is also an excessive leap to claim that all assumptions of a specific game theory model apply directly to nature simply because the two are fundamentally similar. In particular, it is necessary to verify how closely the strategies, payoffs, frequency of interactions, memory capacity, and group structure defined in the model align with actual biological conditions. Furthermore, even if the two models yield the same conclusion, the dynamics and timescales leading to that conclusion may differ. In the case of repeated games, the process by which individuals respond to their opponents’ previous actions is central to the analysis, whereas in natural selection, changes in strategy frequencies across multiple generations are key. In fact, even considering just the examples of the bee, the cheater, and the grudge-holder, it may take time to approach a specific equilibrium as the frequencies of each strategy change. Furthermore, since nature is not limited to a single species but involves the simultaneous interaction of multiple species, the rates of change in other species and environmental changes also become important variables. In the case of birds and ticks, if the ticks have not yet spread widely within the flock and the frequencies of the three behavioral strategies change rapidly, the situation could shift before the damage caused by the ticks increases significantly. Conversely, if the changes in behavioral strategies are slow and the ticks spread rapidly, the outcome could be different. In other words, the outcome may vary depending on the temporal and ecological differences between the structure of interactions defined by game theory and the selection processes that actually occur in nature. Of course, this discussion is based on highly simplified assumptions. However, the mere fact that a model’s assumptions may differ from the actual natural environment is sufficient to highlight that, when applying game theory, one must also explain under what conditions the model is valid.

Conclusion

In ‘The Selfish Gene’, Dawkins gradually persuades readers to accept concepts that may sound radical to the general public. As a result, he explains relatively easily that we are part of a biological system that operates for the survival and replication of genes, and that even altruistic behavior can, under certain conditions, lead to outcomes favorable for the transmission of genes. However, the way he combines the theory of the selfish gene with game theory to explain reciprocal altruism warrants closer scrutiny. In particular, it cannot be assumed that game theory is a theory that deals exclusively with situations—such as human-to-human relationships—where rational agents calculate their own interests and anticipate the behavior of others. On the contrary, modern evolutionary game theory has applied game theory to biological behavior to explain natural selection and changes in strategy frequencies, and the concept of the ESS (Evolutionally Stable Strategy) has also developed within this evolutionary context. Therefore, it is difficult to view the fact that Dawkins applied game theory to nature as an error in and of itself.
Nevertheless, there is ample room to critically examine the explanation of reciprocal altruism using game theory presented in this book. While the mechanical process of natural selection and the process of game theory—which analyzes strategic interactions—can be linked, the two processes are not entirely identical. In particular, we must distinguish between an individual actually calculating and selecting a strategy and the increase in the frequency of that strategy as the reproductive success of individuals with a specific behavioral tendency rises over multiple generations. Since evolutionary game theory analyzes the evolutionary stability of strategies while acknowledging this distinction, the application of game theory should be interpreted not to mean that “organisms make rational calculations,” but rather that “the frequency of strategies changes through natural selection.” Without making this distinction, explanations based on game theory may be interpreted in a way that differs from actual biological processes.
In practice, the process by which numerous variants compete within a population, and in which individuals with specific behavioral tendencies achieve higher reproductive success, ultimately leading to the persistence of a particular strategy, can be linked to the stability of strategies predicted by game theory. Therefore, rather than viewing the two explanations as contradictory, they can be seen as describing the same evolutionary phenomenon at different levels.
In fact, evolutionary game theory has developed with the aim of linking natural selection and strategic interactions, and it has been widely used to analyze the evolution of reciprocal altruism and cooperation. Since Trivers proposed the possibility of natural selection for reciprocal altruism, topics such as repeated interactions, responses to cheaters, and the stability of cooperative strategies have become representative areas of research at the intersection of evolutionary biology and game theory.
Therefore, the criticism in this paper is not that applying game theory to nature is inherently wrong, but rather that it is necessary to clarify which type of game theory is being applied and under what assumptions. Just because an explanation using game theory yields the same conclusions as the selfish gene theory does not mean that the logical processes of the two theories are identical. Furthermore, just because cooperation emerges as a stable strategy in a specific model does not mean that the same result will necessarily occur in the natural world. This is because the behavior of actual organisms is influenced not only by genetic factors but also by a variety of elements, including the environment, developmental processes, relationships between individuals, the frequency of interactions, population structure, and relationships with other species. Therefore, to apply the results of game theory to natural phenomena, one must first verify the conditions under which the model holds.Of course, from an academic standpoint, the computational results may ultimately be the same, and one might argue that the difference between the two explanations is insignificant as long as the results are identical. However, ‘The Selfish Gene’ is not a mathematical model intended solely for expert researchers; it is a book written to explain how evolution works to a general audience. From this perspective, it is necessary to consider the possibility that readers might interpret the concept of “strategy” in game theory as behavior involving conscious calculation and choice, similar to that of humans. Conversely, if Dawkins’ actual intention was to present the perspective of evolutionary game theory, the theory would have been all the more persuasive had he clearly distinguished and explained that point. After all, applying game theory to nature is not logically impossible. In fact, evolutionary game theory now exists as a distinct field of research and is utilized as a key method for analyzing natural selection and strategic interactions. However, when translating the mathematical results of game theory into biological reality, one must explain the differences between the model’s assumptions and the actual process of natural selection. Treating the two processes as identical merely because the results are the same—without making such distinctions—can undermine the logical precision of the explanation. Therefore, rather than assessing Dawkins’ game-theoretic explanation of reciprocal altruism as entirely incorrect, it is necessary to critically examine it with a view to sufficiently distinguishing between classical game theory and evolutionary game theory, between strategies and genetic behavioral tendencies, and between models and the actual natural environment.