The "selfish gene" fully accounts for altruism

But what about the hard case of anonymous, one-shot cooperation?

By now, everyone knows or should know that charging Dawkins and his book The Selfish Gene, alongside the gene’s-eye view of evolution, with failing to at least partially account for human altruism is false.¹ (Read the footnote if you’re going to get upset about me using the word “altruism” in this post.)

The book, in fact, summarizes the two most well-known ultimate explanations for helping behavior among family members and non-related people with whom a person is exchanging favors. They are kin selection and reciprocal altruism (or direct reciprocity).

Kin selection

Just to quickly remind you, so we’re on the same page, kin selection explains why organisms help genetic relatives, even if genes only “care” about selfish survival and reproduction.

Suppose “a gene”² arises randomly (due to mutation) that inclines its bearer to help siblings at some cost to the bearer. Your sibling shares, on average, 50% of your genes by common descent. So when you help your sibling reproduce, there’s a 50% chance you’re helping to propagate a copy of the very gene that made you help in the first place. If the benefit to your sibling’s reproduction, discounted by that 0.5 relatedness coefficient, exceeds the cost to your own reproduction, then copies of the helping gene will tend to increase in frequency across generations.

The “helping gene” spreads not because it helps the individual organism (who is now actually worse off for having helped another organism), but because it helps copies of itself, wherever they happen to reside. The altruism is real at the level of the organism/your body, but the underlying genetic logic is quite self-interested.

Reciprocal altruism (direct reciprocity)

Of course, it’s not like only genetically related individuals help each other. Especially with humans (but, interestingly, not only humans), cooperation extends much farther than only family. How do we explain that?

Here, reciprocal altruism comes into play. The logic is again fairly simple. If I help you today and you help me tomorrow, and if the benefit to the recipient exceeds the cost to the helper, both parties can come out ahead over time, with these weird mutually-helping genes spreading across population. The key requirements are repeated interactions, the ability to recognize individuals, and the capacity to remember who cooperated and who cheated – things humans have no issue with, cognitively.

Consider two strategies in a population, and keep in mind that they don’t have to be conscious. It’s just genes in action, programming humans to act in certain ways given certain environmental cues. The first strategy is: cooperate with those who cooperate with you. The second strategy is: always defect and take without giving.

In one-shot encounters, non-helping defectors win. The genetically selfish thing to do is to not help. But in a world of repeated interactions, cooperators find each other and generate mutual gains while defectors stay on their own/get avoided or even punished. A gene that produces such mutually-helping psychology will spread because its bearers, on average, enjoy more cooperative partnerships and thus greater reproductive success than defectors who end up isolated.

For this to work, the expected future benefit of being helped (discounted by the probability of future interaction and successful recognition) must exceed the current cost of helping; otherwise there’s no net gain. For example, if person A gives a starving person B an extra fish and person B later gives person A an extra fish when A is starving, both have converted a temporary surplus into a kind of insurance against devastating results (in the extreme case, death). The gene for this kind of “strategic” (again, not necessarily conscious) helping spreads because it increases the lifetime reproductive output of its bearers via net direct fitness benefits over time.

(The post would get too long if I did a whole detailed section on proximate and ultimate explanations, but keep in mind that genes code for psychological traits, such as feeling gratitude when someone helped you. So, proximally, person A and B are helping each other because, being weird helping mutants, they have the genes that make them feel gratitude when helped – and warmth and happiness when they see a potential friend in need that they can help. In other words, the people in question are oblivious to the genetic fitness logic working in the background, ultimately accounting for why evolutionarily this type of behavior works. They’re just following their feelings and emotions.)

Kin selection and direct reciprocity are not enough to account for humans’ incredible amount of cooperation

So, case closed? No. There are two additional complications, a weaker and a stronger one. Take the weaker one first.

You need additional gene’s-eye-view-based explanations to account for important helping phenomena in the social world. One key mechanism people tend to forget about is indirect reciprocity. That might be because it’s not as intuitive as the other two and because it wasn’t covered by Dawkins (the development post-dates The Selfish Gene).

The logic of indirect reciprocity explains why you might help someone who can never repay you directly. At first blush, that seems like pure altruism, which simply cannot be accounted for with the “selfish” logic of replicating genes. But it can be, if others are watching. Indirect reciprocity explains cooperation based on reputation. In the simplest version, you help someone not because they helped you, but because they’re known to help others, or because helping them will enhance your own reputation in the eyes of third parties. More formally put, A helps B; C observes this interaction; and later C (or someone else) is more willing to help A because A has earned a good reputation. This explains some instances of charitable donations and why we care so much about what people think of us.

If you’re like me, this isn’t completely convincing. That is, it wasn’t always clear to me from this explanation why indirect reciprocity would be evolutionarily stable/fitness enhancing. For instance, to account for A’s help (to B), person C is now incurring an unaccounted-for cost.

But if you just add two additional steps, the whole thing becomes more understandable. First, going back to the example with persons A, B, and C, because C helped, he/she is also (if observed) more likely to be helped by yet other people (D or E) in the future. So even though C helped A and thus incurred a cost (one that A isn’t repaying), if that very fact also helps C build a reputation for being helping, which nudges other people to help C, then the logic goes through.

Second, people want to associate or trade with, and marry, reliable cooperators. Why? Well, suppose you’re choosing a partner for some joint venture like hunting or raising children. You want a partner who will actually cooperate with you rather than exploit you, and past behavior is your best predictor of future behavior. So a mutant gene that inclines you to preferentially associate with people who have demonstrated cooperation with others will tend to land you in more beneficial partnerships (even if, again, you’re oblivious to this fact). That increases your chances of reproduction and survival, meaning that the gene for this kind of prosocial behavior has a higher probability of spreading through the population.

In any case, with indirect reciprocity, both sides of the equation are genetically self-interested. Namely, the donor gives because being seen giving attracts help or better partners, and the observer chooses cooperators because cooperators make better partners.

The multifaceted genetic basis for cooperation. Figure from West et al. (2011).

What about the big issue: anonymous one-shot altruism with strangers?

But wait, there’s a stronger problem here. We’ve discovered in lab experiments, and have anecdotally known for a while longer, that people cooperate with others even in one-shot interactions (so no direct reciprocity) with non-related players (so no kin selection) and without the possibility of building up a reputation (so no indirect reciprocity). People also altruistically punish non-cooperators in these settings, even when punishment is costly and yields no future benefit.

Some have explicitly proposed that these facts show the gene’s-eye view to be inadequate for the task. Apparently, you need the concept of “strong reciprocators,” who have an evolved tendency to cooperate with others and punish defectors even when no possible (direct or indirect) benefits were likely. The idea, associated with researchers like Ernst Fehr and Herbert Gintis, is that humans evolved strong other-regarding preferences, going beyond what the gene’s-eye view alone can explain. As Fehr and colleagues put it, human behavior in these games “cannot be rationalized as an adaptive trait by the leading evolutionary theories.”

I think it’s actually more complicated than that.³

Hidden, unaccounted-for benefits

First, the theoretical models of “strong reciprocity” don’t necessarily show what their authors claim they show. When you work through the mathematics, these models often rely on standard direct and indirect fitness benefits. For instance, models by Bowles and Gintis assume something called limited dispersal between groups, which generates significant relatedness among group members. Yet the authors claim that “there are many unrelated individuals, so altruism cannot be explained by inclusive fitness.” That doesn’t seem fair.

These models also assume that cooperation provides benefits to the whole group, including the cooperator, through mechanisms like reduced group extinction or success in between-group competition. That’s a direct fitness benefit for the individual (even if not only for him or herself). As West and colleagues put it, formal analysis reveals that “the earlier verbal arguments were incorrect.”

So the strong reciprocity models don’t seem to provide a novel solution to the problem of cooperation. They rediscover the same solutions we already had, like kin selection via limited dispersal or direct benefits via group augmentation, while just verbally claiming to have transcended them.

The possibility of misfiring

Second, set the model aside for a bit. There exists a fairly straightforward gene’s-eye view interpretation of the seemingly peculiar lab results themselves. It’s sometimes called the “misfiring” or “big mistake” hypothesis. The argument is that our minds evolved in small-scale ancestral environments where true anonymity was essentially impossible. In a band of 50-150 people where everyone knew everyone, where gossip was rampant, and so on, helping behavior almost always had reputational consequences. There was no such thing as a genuinely one-shot anonymous interaction, which the lab experiments are now artificially generating.

In this view, our cooperative psychology is calibrated to the ancestral world, not the modern lab. When you place someone in an artificial situation where their behavior truly has no consequences for their reputation or future interactions, their evolved psychology doesn’t know how to respond to this unprecedented context. It misfires, or defaults to cooperation because that’s what would have been adaptive in the environments that shaped our minds.

There’s some (though relatively weak) evidence to support this interpretation. Players in experiments sometimes adjust their level of cooperation in response to artificial cues like eye-spot pictures on computer desktops, or even just the presence of other individuals who have no influence on the game. If our psychology were well-calibrated to these artificial environments and attuned to switching between anonymous and non-anonymous contexts, such irrelevant cues wouldn’t matter. But they do, which tells us the brain is processing the lab as if it were the real social world where being observed matters.

No gene’s-eye view imperialism here

I definitely don’t want to be misconstrued as saying or implying that every single instance of human altruism (or behavior in general) has to have a direct adaptive explanation – a specific story about how that behavior increased genetic fitness. Not at all. As should be clear from what I just said above, many behaviors are byproducts of adaptations, enabled by our psychological machinery that evolved for other reasons/in a different context. There’s also simply noise and individual idiosyncrasies.

I’m relatedly in complete agreement with people like Gintis and Joseph Henrich that cultural evolution is an important additional layer on top of (or in parallel with) genetic evolution. While genetic group selection doesn’t usually work with humans (unless it’s simply redescribing kin selection), cultural group selection is a perfectly reasonable concept that not only works in principle but has been successfully used to explain some interesting and important social phenomena. I’m convinced it also has something to do with the scale of human cooperation, although I don’t at this moment know how much, so I’ll save that for a future post on cultural evolution.

So, a more accurate title of my piece would be “The ‘selfish gene’ can fully (or virtually fully) account for altruism.” It can do so, but where we want it to is a different matter. We might need to bring in cultural evolution for the fully convincing and truly exhaustive explanation, though that doesn’t contradict the genetic account and is compatible with it (as the name “gene-culture coevolution” clearly suggests).

In any case, there’s more to the explanatory power of the gene’s-eye view – even when it comes to cooperation/altruism – than meets the proverbial eye. You should take it very seriously.

1. Oh boy, how do we define altruism? In everyday language, altruism means something like helping others at a cost to yourself, perhaps with the connotation of genuine other-regarding motivation. In evolutionary biology, the term has a narrower technical meaning. Altruism is a behavior that reduces the actor’s direct genetic reproductive success while increasing the recipient’s. By this definition, reciprocal helping isn’t actually altruistic – if you get paid back later, your direct fitness goes up, not down. It’s mutually beneficial cooperation. True genetic altruism, where the actor takes a net fitness hit, can only be explained by kin selection (or something called the greenbeard effect, which I’ll skip here). Throughout this post, I’m mostly talking about cooperation in the broad sense, which includes both altruism and mutual benefit.

2. It’s actually not that simple, as there are usually many genes involved with a single trait, not just “a gene.” But I’ll use that word for pedagogical purposes, for the sake of easier understanding.

3. I’m strongly leaning here on a wonderful paper by Stuart West et al. from Evolution and Human Behavior (2011) titled “Sixteen common misconceptions about the evolution of cooperation in humans.”