Is culture what makes us cooperate?

This post by evolutionary biologist Jean-Baptiste André is part of a small series of posts on social learning and cooperation.

A recent series of papers by Laurent Lehmann and colleagues (including Marcus Feldman) shed new lights on the cultural evolution of social behaviors, at least on a theoretical perspective. Here, I mainly review the first paper of the series, published in The American Naturalist. This paper evaluates, with the tools of modern population genetics, the hypothesis of cultural group selection, originally formulated by Boyd and Richerson. The paper is very pedagogic and clear, and most admirably, it does not hide anything to the reader.  Overall, these results are not surprising, and they somehow correspond to what Boyd and Richerson originally proposed. However, they are quite helpful and refreshing. Over the years, and because of the careless way Boyd and Richerson’s work is referred to, it seemed sometimes that cultural group selection was a robust population process, favoring altruism automatically as soon as culture comes in. Well, it is not. It involves strong assumptions concerning the way culture is transmitted…

Blindly adopted social norms and inter-group competition: is that all there is to human altruism? V for Vendetta, by James McTeigue. Alan Moore wrote the Comics, the slogan is from Orwell (editor's note and choice of picture).

Lehmann et al. show that cultural transmission can strengthen the selection (or reduce the counter-selection) for altruism, when the horizontal transmission of cultural items does not depend on the items’ payoff consequences. This is the case when the source rather than the content of the item is the choice criterion (for instance, in the so-called “prestige-biased” or “conformist-biased” transmissions studied by Boyd and Richerson).
In this case, many individuals can easily adopt the cultural items from only a few individuals, independently of the content of these items. This tends to homogenize the phenotypes, or to increase the “cultural relatedness”, of fellow group members, which favors “altruistic” cultural items over “selfish” ones. And, in fact, in contrast with what is usually believed, prestige-biased transmission is more efficient than conformist-biased transmission at promoting altruism. Conformist transmission can be efficient at stabilizing altruism once fixed, but makes it impossible for altruism to rise in the first place (this is actually shown in the second paper of the series, published in Theoretical Populations Biology).
However, Lehmann et al. also show that, when cultural transmission is based on payoff comparison (e.g., imitating those with a larger payoff than the group average), an hypothesis that is at the core of many models in mathematical social sciences, then culturally transmitted traits are less likely than genetically transmitted ones to evolve toward altruism! Worst, they are more likely to evolve toward aggressiveness! And this is easy to understand intuitively, by considering an extreme scenario. Imagine that the individual with the largest payoff in the group is systematically imitated by all the others. In such a case, a cultural item that would make its bearer reduce the payoff of all its neighbors (e.g., by aggressive behaviors), should be very well transmitted!

My frustration, however, is still great after reading this great piece of population genetics. There is still, in the background of this work, the old idea that a single (or very few) learning rules operate for everything, and allow us to fill some little boxes in our mind (tool use, food habits, social behaviors…). And, in the precise case of altruism/selfishness, this error has important consequences. Altruistic behaviors are, by definition, individually costly. Natural selection (on genes) has very probably shaped our mind in such a way that we avoid, as much as possible, to express such costly behaviors. In any given instance, if we are able to detect that a behavior is costly, we should avoid doing it at all, rather than use some social learning rules (whatever they may be) to decide upon the behavior. If ever, we should use general learning rules only when we can understand nothing about the ins and outs of a behavior. And cultural group selection may have significant consequence only in these (probably very rare) cases! Are these few cases really worth fancy population genetics models?

Interestingly, with a little twist, it could be somehow one of the conclusions of the third paper of the series (a review in Trends in Ecology and Evolution). In this paper, Lehman et al. suggest that individual learning might play a more significant role in social behavior than most evolutionists believe. However, and this is again a disappointment to me, by individual learning they mean something like trials and errors, another simple and universal learning rule. The possibility that the ontogeny, and plasticity, of social behavior could be driven by a content-rich development program is still not even contemplated…

No comments yet