Evolution of Cooperation[4]: Origins of Cooperative Motivation

Reading notes on Michael Tomasello’s work. Translated from Chinese and lightly polished with Claude.

Anyone who has read the second and third essays will surely ask: if the main difference between apes and humans is the motivation to cooperate, how did human cooperative motivation evolve? This essay focuses on that question. In addition to the book, it also draws on a paper by Tomasello et al.¹¹ Tomasello, Michael, et al. “Two key steps in the evolution of human cooperation: The interdependence hypothesis.” Current Anthropology 53.6 (2012): 673–692. (hereafter, “the paper”).

1. The nature of cooperation and how it can evolve

Two issues first: what cooperation is, and how it can evolve at all.

1.1 The nature of cooperation

A clarification up front: cooperative behavior is not the same as a cooperative motive. As the third essay made clear, the cooperation in question here is a recursive belief, even a social norm.

The paper points out that many works find the evolution of a “cooperative motive” puzzling — why would a highly competitive evolutionary process produce altruism? Tomasello argues that cooperative motivation must have arisen in environments where “not cooperating means dying.” He describes cooperation through the Stag Hunt:

In Stag Hunt, the stag can only be caught if two hunters cooperate. If one chooses hare and the other chooses stag, the hare-hunter gets a hare and the stag-hunter gets nothing. Here, for a single hunter, a hare is worth 1 and a stag is worth 8 (4 after splitting).

If the energy required for survival is less than 1, then both players can survive without ever cooperating, and no environment of this kind will breed cooperation. To produce positive selection for cooperative behavior, the survival threshold must be above 1. I would push the low-payoff cells further down — far enough that the organism cannot survive:

So at the very beginning, cooperation was not altruistic — it was self-interested.

1.2 How can evolution work at all?

A behavior must first appear before it can be positively (or negatively) selected (you have to emit a behavior before you can reinforce it). So, showing that a trait was positively selected first requires showing that the trait could “possibly” appear in the population before the selecting environmental pressure took hold. The next two sections look at how the points in 1.1 and 1.2 actually manifest in apes — and so reveal the evolutionary roots of cooperative behavior.

2. Apes “cooperate,” but apes do not want to cooperate

2.1 Apes maximize individual gain yet produce group-beneficial outcomes

Apes cooperate in hunting monkeys. Anyone who has seen a wildlife documentary knows the pattern: a group of apes chases a monkey while others position themselves along the monkey’s likely escape routes.

This looks like a clear division of labor, even like coordinated action. But it is not, in fact, cooperative behavior.

Analyses of group-hunting in apes show that the sole motive of each participating ape is to be the one who finally catches the monkey, because the captor gets the most meat. So when a monkey appears, each ape watches the others’ positions and intentions and picks its own position to maximize its individual chance of catching the prey. Every ape is optimizing its own objective function, and the group probability of catching the monkey happens to go up as a side effect — much like Adam Smith’s “invisible hand,” where each rational individual maximizes their own gain and the market becomes maximally “efficient.”

Three further observations show that apes do not “want” to cooperate:

• Apes do not share their prey. Others mostly rely on begging or snatching meat from the captor. By contrast, human hunters even bring prey back to the group for those who did not participate.
• In experimental settings, apes simply do not care whether a conspecific gets food. Human infants will even split a windfall evenly.
• Apes are more likely to hunt monkeys in food-rich seasons. That is because, when food is abundant, the cost of a failed hunt is small. Going back to Stag Hunt: if hares are everywhere, you can afford to try for a stag, since you do not really lose the hare:

2.2 Free-riding is avoided “for free”

A classic problem in cooperation is free-riding. In the monkey-hunting behavior of 2.1, however, the self-interested tendency partially removes free-riding: since the meat reward decreases with distance from the kill, the best move is to participate in the kill as much as possible — to wait, say, along the monkey’s escape path.

2.3 Apes do not coordinate, yet are optimistic about the actions of others

Section 2.1 implies that ape group behavior is not coordinated — there is no top-down shared goal or role assignment, only a bottom-up self-interested tendency.

The “environment” in 2.1 has a structural feature: the cost of a failed monkey hunt is small. That is, the cost of failed cooperation is small.

This produces, on the surface, an optimistic expectation about whether others will participate in “joint action.” In a real Stag Hunt, the cost of uncoordinated action is large. But in the real world: (1) the two parties may not be making decisions independently — there can be leader–follower strategies, persuasion strategies, and so on; (2) the cost of failure is low.

In one experiment, apes face a real Stag Hunt — they can either take food A, or together take food B (which is much better than A), but if an ape moves toward food B, food A disappears.

The experimenters found no sign of prior communication — even when apes could see each other, they did not even bother to watch the other’s action. Instead, apes showed a leader–follower or persuasion pattern. But this does not mean that the first mover is a “cooperative-minded” ape — recall that in Evolution of Cooperation [2], apes exploit humans to achieve their own goals, and treat others as tools.

The conclusion: in a real Stag Hunt, apes still retain the behavioral pattern of an abundant-food world — they do not coordinate in advance; they remain optimistic about others’ participation, since defection has no real cost anyway.

3. Apes that better tolerate co-feeding are preferred as partners

In another experiment, an ape and a partner have to pull a rope together to get food. When there are two portions of food, the apes have some probability of pulling together; when there is only one portion, joint pulling almost never happens. In the two-portion case, if the two apes can eat together peacefully — rather than fight to monopolize the food — they are more likely to act jointly in the future.

This shows that apes who are more tolerant of cofeeding are more likely to form cooperative partnerships — even when each is purely self-interested. Three elements matter in this experiment. First, “no cooperation, no food.” Second, the cooperative payoff has to be large enough to feed both apes (food is split, each gets a share). Third, an ape is more willing to cooperate with a conspecific that is more tolerant. The third point echoes the difference seen in Evolution of Cooperation [2] between an ape’s behavior toward humans versus toward conspecifics: when an ape meets a significantly more cooperative individual, new behaviors emerge — even if those behaviors are still self-interested and imperative, without any cooperative or altruistic motive.

4. The environmental features that positively select for cooperative individuals

So far I’ve described (1) the way a cooperative individual can elicit new behaviors from conspecifics; and (2) the requirement that the environment make non-cooperation lethal (a large cost of failed cooperation). Now let’s see how nature actually shaped such an environment.

4.1 Scavenging large carcasses as mutualism

Two million years ago, the global climate became colder and drier. Terrestrial monkeys multiplied; plant-based food became increasingly scarce. Apes had to rely more and more on the large carcasses left by big carnivores.

This situation has three features. First, cooperation is necessary, because other large scavengers compete for the carcass. Second, “no cooperation, no food,” because there is not enough plant food left to fall back on. Third, the cooperative payoff is large enough for all participants, and cofeeding is required, because the kill is large.

In this environment, counter-dominance emerged: any conspecific who hogged food or refused to cofeed was driven out of the group, and so eliminated by selection (it could not forage alone). Over time, the group came to consist of more tolerant members.

4.2 The “reputation” mechanism

The natural environment gave early humans a built-in shared goal. Through nothing more sophisticated than associative learning, an early human individual could come to understand that “only those who participate in the group hunt get food.” Those who grabbed food or stood around could not join the hunt and so got no food.

Under cooperative pressure, obeying others’ instructions (imperative gestures) became adaptive. The group is now full of more cooperative individuals — and we saw in Evolution of Cooperation [2] that apes do use imperative gestures (e.g. pointing) toward such individuals. In this setting, understanding another’s intent under a shared goal — and following it — became the more adaptive behavior.

The paper frames this whole set of behaviors as reputation, since apes already have the ability to recognize individual identity. To participate in more hunts, an individual has to maintain a good “image” in the group; those better at coordinating and at sharing get positively selected, and so do the behaviors themselves.

4.3 Interdependence breeds altruistic tendencies

The two preceding sections only describe ape cooperative behavior in a cooperative context. But human infants will act altruistically even in non-cooperative contexts — e.g., spontaneously sharing food with a conspecific (click “read the original article” to see a video example).

How could this have evolved?

The paper argues that it stems from a sense of interdependence — my own interest can only be realized through the group. So, in a non-cooperative context, pure individual altruism is still, fundamentally, self-interested (so long as the cost of “everyday” altruism is smaller than the benefit of being integrated into the group): it is used to establish one’s reputation within the group. This “self-interested origin” of altruism is what enabled the behavior to be emitted in the first place — at the dawn of evolution, no organism would behave altruistically for no reason at all.