1. Introduction
Evolutionary theorists have always been unanimous in the belief that behaviors are as important to natural selection as physical attributes. These theorists have to a large extent been successful in accounting for how much of the behavior observed throughout the animal kingdom evolved by showing how patterns of observed behaviors confer survival advantages onto those who practice them. Despite such successes, there are certain behaviors for which evolutionary theorists, including both biologists and philosophers of science, have been at a loss to provide convincing explanations. Altruism is one such behavior.
As it is used in the evolutionary literature, ‘altruism’ refers to behavior that reduces its practitioner's fitness while raising the fitness of others.Footnote 1 Once altruism is understood in this way, it is easy to see the difficulties it poses for any attempt to provide an evolutionary account of it. As Elliott Sober and David Sloan Wilson put it, “Altruism is the very opposite of the survival of the fittest” (Reference Sober and Wilson1998, 19). Though many ingenious and sophisticated theories have been offered to explain how altruism could evolve despite the fitness disadvantages it appears to bestow, they have been met by equally sophisticated rebuttals. While these criticisms shed light on the factors that appear to impede the evolution of altruism, they are often based on certain strong assumptions that have stacked the cards, unfairly I believe, against the likelihood that altruism can evolve. That these strong assumptions have become pervasive in contemporary discussions on the evolution of altruism is due partly to how such assumptions are as likely to be employed by those who argue for the evolution of altruism as by those who argue against it. By working under these assumptions, proponents of evolutionary altruism have often done more to perpetuate the controversies surrounding evolutionary altruism than to eliminate them. In this paper, I advocate a particular methodological approach for investigating how altruism could have evolved. By rejecting assumptions that are unwarranted, it is possible to establish that altruism can evolve under conditions that are less strict than commonly supposed.
2. Historical Analysis of the ‘Altruism Problem’
For the better part of a century beginning with Charles Darwin's Reference Darwin1871 work, The Descent of Man, attempts to explain altruistic behavior relied on the idea that group selection can influence adaptations. Group selection is often seen as being at odds with individual selection, which is the familiar cutthroat view of natural selection whereby one's reproductive fitness is determined solely by how one measures up to other individuals. The basic idea behind group selection is that certain characteristics that are selected against by individual selection can become prevalent in the gene pool because they confer advantages onto the groups consisting of individuals possessing these characteristics.Footnote 2 Even if one accepts that the basic theory behind group selection is sound, a satisfactory group selection account of altruism needs to explain how altruism can prosper, given that it appears to be selected against at the level of individual selection.
2.1. The Haystack Model
The view that groups can function as the unit of selection and that group selection can be a significant force in evolution has seen its share of detractors over the years. Some of the harshest and most influential criticism has come from John Maynard Smith, whose Reference Maynard Smith1964 article, “Group Selection and Kin Selection,” served, according to Sober and Wilson, as an “apparent nail in the coffin of group selection” (Reference Sober and Wilson2000, 188). In this article, Maynard Smith set out to understand the conditions under which group selection could occur. He listed three conditions that must be met in order for group selection to be possible:
1. Groups must be reproductively isolated;
2. Groups must be started by one or a few founders;
3. Groups must be separated in such a way that one group's resources do not stand to be threatened by encroachment from another group(s).
Maynard Smith illustrates how these conditions could be satisfied by constructing his famous model in which groups of mice live in haystacks. In his example, each haystack houses a group of mice that are initially populated by a single fertilized female. He stipulates that the mice are divided into two types: altruists and nonaltruists. The single females eventually give birth to a group of offspring that breeds among themselves for a number of generations before dispersing from the haystack at the end of one year. After leaving the haystack, the group mixes into the overall population of mice, mates at random, and goes back into haystacks, repeating the cycle.
Having constructed a model in which altruism might thrive, Maynard Smith evaluates the likelihood that conditions conducive to the evolution of altruism might appear in nature. He concludes that while it may be possible for altruism to flourish via group selection, the circumstances under which we should expect this to happen would be very rare. This view stems partly from his belief that “it is unlikely that species are often divided into a large number of small and completely isolated groups” (1146). Considerations like this lead him to draw a skeptical conclusion from the Haystack Model:
The model is too artificial to be worth pursuing further. It is concluded that if the admittedly severe conditions listed here are satisfied, then it is possible that [group selection could allow altruism to evolve]. Whether this is regarded as an argument for or against the evolution of altruistic behavior by group selection will depend on a judgment of how often the necessary conditions are likely to be satisfied. (1147)
Maynard Smith's arguments—especially when combined with the forceful antigroup selection arguments that appeared in George Williams’ Reference Williams1966 book, Adaptation and Natural Selection—were heavily influential during the widespread rejection of group selection in the latter half of the twentieth century. The conclusions of Maynard Smith and Williams were supported in the 1970s by a series of mathematical models (Levins Reference Levins and Gerstenbaher1970; Boorman and Levitt Reference Boorman and Levitt1973; Levin and Kilmer Reference Levin and Kilmer1974), suggesting that group selection could generate significant effects only under very strict conditions. The culmination of all of this was that group selection would have to wait nearly thirty years after Maynard Smith made his initial arguments against it before it would begin making strides towards widespread credibility.
2.2. Strong Altruism Revealed
While Maynard Smith's arguments—which have become something of ‘the conventional wisdom’ in evolutionary theory—have led many theorists to deny that group selection is a potent force in evolution, recent defenses of group selection have called into question the methodology he employed. Sober and Wilson, two of the more prominent latter-day champions of group selection, have criticized Maynard Smith's model the following way:
Maynard Smith simply stipulated that the within-haystack, individual selection part of his process was as powerful as it could possibly be. He assumes without argument that altruism is driven to extinction in all haystacks in which it is mixed with selfishness; the only way that altruism can survive in a haystack is by being in a haystack that is 100 per cent altruistic. (Reference Sober and Wilson2000, 193)
Sober and Wilson's objection to Maynard Smith's argument against group selection is persuasive. Though it may be true that altruists will necessarily decline in mixed populations consisting of both altruists and nonaltruists, Maynard Smith overstates the case against group selection by adopting the very strong assumption that altruism would decline to zero within a year in all haystacks with mixed populations. The burden is on him to provide a plausible explanation as to why this must occur.
Though Sober and Wilson deserve credit for pointing out one of the ways in which the Haystack Model employs an unnecessarily strong assumption, they fail to identify another illicit assumption that is built into this model. Namely, that altruists will behave altruistically in every given circumstance. From this point on, I will use the terms ‘strong altruism’ and ‘unconditional altruism’ in referring to this type of altruism. The impact that this assumption has had on evolutionary debates about altruism is hard to overstate. For one thing, it has been adopted by the vast majority of accounts that seek to explain how altruism might evolve. It is part of virtually all models that employ game theory to explore the origins of altruism. Indeed, this should not come as a surprise since the person who first began using game theory to investigate the evolutionary origins of altruism is none other than the person who built the assumption of the unconditional altruist into the Haystack Model—John Maynard Smith.
Game theorists use simple models simulating the factors relevant to evolution to explain how altruism can evolve. Game theoretical accounts of altruism often appeal to a game called the ‘Prisoner's Dilemma’. In this game, a player will either cooperate (the altruistic act) or defect (the selfish act). The options open to a player in this game constitute a hierarchy of optimal circumstances that is structured as follows (this hierarchy is looked at from the point of view of player A):
1. A defects/B cooperates (best-case scenario for A);
2. A cooperates/B cooperates;
3. A defects/B defects;
4. A cooperates/B defects (worst-case scenario for A).
Given the parameters of this game, there is always a higher payoff attached to behaving selfishly than there is to behaving altruistically. Note how game-theoretic models like the Prisoner's Dilemma typically stipulate a narrow set of players—those who are wholly altruistic and those who are wholly selfish. This stipulation, which was initially used to keep the mathematics of these models tractable, has become a common element in virtually all contemporary discussions about the evolution of altruism. Even Sober and Wilson's group selection-based evolutionary account of altruism adopts this practice.Footnote 3
3. The Explanatory Benefits of Weak Altruism
What needs to be acknowledged is that despite the fact that attempting to explain the evolution of altruism in terms of how a strong form of altruism could be selected for is the predominant approach taken by contemporary evolutionary theorists, this methodology is problematic in that it makes the task of providing a plausible evolutionary account of altruism more difficult than it needs to be and more difficult than it should be. Given that the forces driving selection at the individual level are heavily opposed to strong altruism, it is not surprising that theorists have run into difficulties providing a persuasive account of how strong altruism can evolve. A reasonable question to ask at this point is why we should bother taking up the project of explaining how unconditional altruism can evolve. Might it not be possible to develop a satisfactory evolutionary account of altruism by explaining how nature can select for a type of weak altruism? I believe that this question warrants an affirmative answer, and I will describe an evolutionary model for a weaker type of altruism that foregoes the extremely strong assumptions that were included in models like Maynard Smith's.
What does it mean to say ‘weaker form of altruism’? By this, I mean a kind of altruism whereby one's propensity towards altruistic behavior is less than it is in the strong or unconditional form of altruism that dominates discussions in the evolutionary literature. To illustrate, imagine an individual whose motivation to engage in altruistic behavior corresponds to the degree to which she has secured her own fitness needs. While this individual would not generally be motivated to act altruistically while these needs were unsatisfied, she would become motivated to behave altruistically once these needs were met. In other words, this individual's propensity to behave altruistically is correlated with her fitness. I will refer to this kind of individual as a fitness-relative altruist. By appealing to the kind of weak altruism exhibited by the fitness-relative altruist, I believe that it is possible to develop an evolutionary account of altruism that avoids many of the difficulties that traditional accounts have faced.
Working within a model of weaker altruism like the one I have put forth yields important advantages for those seeking to provide a satisfactory account of how altruism can evolve. If the fitness-relative altruist's propensity towards altruism is correlated with her fitness, the negative impact that her fitness will sustain on the level of individual selection in virtue of her altruistic tendencies will be minimal. The selection forces working against this weaker form of altruism would be less than those working against unconditional altruism. This can be illustrated by referring back to the Prisoner's Dilemma. Suppose that there are three types of individuals playing in repeated versions (i.e., games with multiple rounds) of the Prisoner's Dilemma—an egoist (who always defects), an unconditional altruist (who always cooperates) and a fitness-relative altruist. A fitness-relative altruist's move in any given round will be determined by how her fitness fares in comparison to the other players in her group (whom she may or may not be interacting with in a current game). To keep things simple, stipulate that if her fitness is equal to or above the median fitness level of all individuals in her group, she will cooperate. Otherwise, she will defect.
It is easy to see how the fitness-relative altruist will make out as well as or better than the unconditional altruist given that they play multiple games against the same opponents (including at least one defector). The reason is that whereas the unconditional altruist will always be hurt in every round in which he is playing against a defector, the fitness-relative altruist may not. If her fitness should drop below a threshold level, she will begin to defect. Upon defecting, she will hold an advantage over the unconditional altruist on every move. Thus, the fact that it is open to the fitness-relative altruist to defect renders her less apt to be selected against than the unconditional altruist. These considerations indicate that once we abandon the project of explaining how nature can select for a strong type of altruism in favor of explaining how a weaker type of altruism can evolve, the task of providing a plausible evolutionary account of altruism becomes less difficult.
3.1. Fitness-Relative Altruism vs. Tit-for-Tat Altruism
I should point out that employing conditional strategies in game theoretic models of evolution is by no means a new concept. My goal in this paper is to stress the importance of taking a certain methodological approach in addressing the question of how altruism could evolve, rather than to offer a revolutionary new way of dealing with this issue. This being said, the particular weak form of altruism that I offer holds advantages over other weak forms of altruism that have been discussed in the evolutionary literature. In particular, the fitness-relative altruism that I employ in my admittedly brief sketch of an evolutionary model has some explanatory advantages over the most prominent example of weak altruism discussed in evolutionary theory—the so-called ‘Tit-for-Tat’ strategy. Tit-for-Tat is a reciprocal strategy that has gained notoriety for having outperformed more selfish strategies in repeated versions of the Prisoner's Dilemma. Tit-for-Tat employs the following tactics:
1. Cooperate on the first move.
2. On each subsequent move, do whatever the other player did on his previous move.
Despite the premature celebrations of some theorists who took Tit-for-Tat's successes as conclusively demonstrating that nature could select for altruism, there are reasons for doubting that this strategy by itself could constitute the basis for a plausible evolutionary account of altruism. For one thing, Tit-for-Tat altruism seems ill-equipped to address what is known in evolutionary theory as the ‘problem of origination’. A good synopsis of this difficulty is given by Sober and Wilson:
Altruism can easily evolve after the discriminating altruists exceed a threshold frequency, but a model that requires altruists to exist at a frequency of 20 percent (for example) before they can be favored by natural selection fails to address a fundamental problem about how altruism can evolve. (Reference Sober and Wilson1998, 136)
The main difficulty for a player using the Tit-for-Tat strategy is that he will do worse than an opponent whenever he is pitted against a purely selfish strategy in the Prisoner's Dilemma. Tit-for-Tat's success in repeated versions of the Prisoner's Dilemma is due to the fact that individuals adopting this strategy do very well when paired with each other, whereas more selfish strategies do poorly when paired together. While Tit-for-Tat may have an advantage over selfish strategies if the overall population has a requisite number of players using similar kinds of reciprocal altruistic strategies, we need an explanation for how a population could come to have the number of reciprocal altruists necessary for making Tit-for-Tat a profitable strategy. Attempting to explain the evolution of altruism by employing a model that stipulates that altruistic strategies are initially common, bypasses the very question that needs an answer in the first place. Namely, how could an altruistic strategy get a starting foothold in a population?
While the problem of origination poses a difficulty for my evolutionary account of altruism as well, this difficulty is somewhat less formidable than that faced by accounts that rely on Tit-for-Tat altruism to make their case. The reason is that the fitness-relative altruist in my model faces less of a disadvantage against a defector in a Prisoner's Dilemma than does the Tit-for-Tat altruist. Remember that the fitness-relative altruist only cooperates if her fitness is equal to or above the median fitness level of all individuals in her group. Thus, if a mutant fitness-relative altruist were introduced into a population consisting only of defectors, she would cooperate only on her first move in the very first game that she plays.Footnote 4 Though this would put her at a disadvantage relative to all others in the population, this disadvantage would be minimal. Consider, on the other hand, the Tit-for-Tat altruist, who when placed in a similar population would cooperate on every first move against everyone she plays against. While the fitness-relative altruist's relative fitness in a population of defectors would remain unchanged after the first game that she plays, the Tit-for-Tat altruist would find her relative fitness decreasing in every game that is played. Hence, the fitness-relative altruism of my account is better able to cope with the problem of origination than Tit-for-Tat altruism.
3.2. Empirical Support for Fitness-Relative Altruism
Another virtue of my evolutionary model is that fitness-relative altruism matches up well with the actual kinds of cooperative behaviors that evolutionary theorists are attempting to explain. Surveying the empirical literature on human beings, for example, shows that cooperative behavior is significantly influenced by factors that impact individual fitness. For instance, several studies have shown that increasing or decreasing a subject's self-esteem (a key indicator of relative social standing, and, hence, of personal fitness) is correlated with a likewise increase or decrease in the propensity to behave cooperatively (see Isen, Horn, and Rosenhan Reference Isen, Horn and Rosenhan1973; Midlarsky Reference Midlarsky1971; Weyant Reference Weyant1978; Crandall Reference Crandall1978; Batson Reference Batson, Moore and Isen1990). Even events that would appear to have only a slight impact on a subject's fitness (such as finding a dime in a public telephone) have been shown to significantly impact the likelihood that subjects will cooperate (see Isen and Levin Reference Isen and Levin1972). These considerations recommend switching the focus of the evolution of altruism debate from explaining how strong altruism might evolve to explaining how a weaker form of altruism could evolve, since the kinds of cooperative behaviors that we observe do not resemble unconditional altruism. Furthermore, the research I cited on human cooperative behavior suggests that the fitness-relative altruism at work in my model is more empirically sound than other weaker types of altruism (e.g., Tit-for-Tat altruism) that have been discussed in the evolutionary literature.
4. Conclusion
Skeptics about the evolution of altruism usually do not claim that it is impossible for altruism to evolve. Rather, they claim that it is highly unlikely to evolve given that the conditions that would allow for it make up a very narrow window. My contention is that this alleged narrowness is due primarily to how evolutionary theorists (e.g., Maynard Smith) have tended to focus on a kind of altruism that is stronger than the discussion warrants. The main position I have argued for in this paper is that the difficulty of providing a credible evolutionary account of altruism is diminished if we turn our efforts towards explaining how nature can select for a weaker kind of altruism that is less prone to being extinguished by the forces driving individual selection. I have also suggested that the task of constructing a satisfactory evolutionary account of altruism is made easier by allowing for a type of altruism that is sensitive to considerations of relative fitness. This type of altruism is less apt to be selected against than other weak forms of altruism, like Tit-for-Tat altruism. Furthermore, fitness-relative altruism resembles much of the cooperative behavior that is actually found in nature. In stating this, I am not denying the explanatory significance of Tit-for-Tat altruism (or other forms of weak altruism) altogether. I see no reason to reject the possibility of developing an evolutionary model of altruism that employs a type of altruism that combines aspects of both a fitness-sensitive altruism and a reciprocal-type altruism like Tit-for-Tat. Indeed, such a model may work better than a model that employs either of these two types of altruism alone.
My primary concern in this paper has not been to offer an account of how altruism actually did evolve in human beings or other species. I have taken up the more modest goal of clarifying why traditional attempts to explain the evolution of altruism have been wrongheaded in taking up the project of explaining how unconditional altruism can evolve. The error of this approach is borne out by how the cooperative behaviors that evolutionary theorists want to explain do not match the types of behaviors that we would expect from unconditional altruists. By focusing their efforts on explaining how strong types of altruism can evolve, evolutionary theorists have made the task of providing a satisfactory evolutionary account of altruism appear more difficult than it is.
The particular model that I have put forward to explain how altruism might evolve (where one's tendency to act altruistically is correlated with one's fitness), is likely to be merely one of many possible models that can offer a theoretically plausible account of how a weak type of altruism can evolve. Once we acknowledge this, the task of explaining how altruism can evolve becomes much less forbidding. It seems reasonable that different kinds of altruism could evolve under a variety of conditions that could be explained by a variety of models.
