Tag Archives: Cooperation


Thinking Outside the Box: The Power of Apologies in Cooperative Agreements

TWO kids I know—let’s call them Jeff and Mimi–wanted a cat, so they begged their reluctant parents for months. Eventually the parents gave in, but they forced the kids into an agreement: “The cat box will need to be cleaned every day. We expect you to alternate days. If you miss your day, we’ll take fifty cents out of your allowance and give it to your sister/brother. If you like, you can think of the fifty-cent transfer as a ‘fine’ that we pay to your sister/brother for your failure to hold up your end of the agreement.” The kids agreed and Corbin the cat was purchased (or, rather, obtained from the Humane Society). The litter box-cleaning arrangement went well for three whole days, but compliance started to wane on day four. Hostilities began to simmer. Jeff became reluctant to clean the litter box on “his” days because of Mimi’s failures to keep up her end of the bargain, and vice versa.

“Recriminations” began to pile up.

After two weeks, the agreement was declared dead. The parents became the chief cleaners of the litter box. The father continues to wonder whether he should have read more game theory before even entertaining the idea of getting a cat.

Cooperative agreements like these tend to be dicey propositions: What’s Mimi supposed to do if Jeff fails to clean the box on his appointed day? Should he view it as a sign that Jeff no longer intends to honor the agreement (in which case she should stop honoring it herself, notwithstanding the $.50 fine that Jeff had to pay to her), or should she view it as a one-off aberration (in which case she might want to continue honoring the agreement)? It’s not clear, and that lack of clarity can make problems for the stability of such agreements.

Luis Martinez-Vaquero and his colleagues addressed this issue in a recent article that caught my eye. The paper is complex, but it’s full of interesting results (some quite counter-intuitive) about when strategic agents should be expected to make commitments, honor commitments, retaliate when those commitments are broken, and so forth. I suggest you give it a read if you are at all interested in these issues. But what really grabbed my interest was the authors’ exploration of the idea that the key to getting such agreements to “work” (by which I mean, “become evolutionarily stable*”) was to build in an apology-forgiveness system that causes agreement-violators to pay an additional cost (over and above the fine specified in the agreement itself) after a failure to cooperate, which might cause the defected-against partner to persist in the agreement despite the fact that it has been violated.

The researchers’ results enabled them to be surprisingly precise about the conditions under which highly cooperative strategies that used apologies and forgiveness in this way would evolve*: The costs of cooperating (cleaning the litter box) must be lower than the cost of the apology (the amount of money the deal-breaker voluntarily passes to his/her sibling), which in turn must be lower than the fine for non-compliance that is specified within the agreement itself (fifty cents). When those conditions are in place, you can get the evolution of actors who like to make agreements, accept agreements, honor agreements, and forgive breaches of agreements so that cooperation can be maintained even when those agreements are occasionally violated due to cello lessons that run late, or unscheduled trips to the emergency room, or geometry exams that simply must be studied for.

I’ve written here and there (and here) about the value of apologies and compensation in promoting forgiveness, but the results of Martinez-Vaquero and colleagues suggest (to me, anyway) that forgiveness-inducing gestures such as apologies and offers of compensation can come to possess a sort of fractal quality: People often overcome defections in their cooperative relationships through costly apologies, which promote forgiveness. Throughout the history of Western Civilization, various Leviathans have capitalized on the conflict-minimizing, cooperation-preserving power of costly apologies by institutionalizing these sorts of innovations within contracts and other commitment devices that specify fines and other sanctions if one party or the other fails to perform. But after the fine has been paid for failure to perform, what’s to keep the parties motivated to continue on with their agreement? Martinez-Vaquero et al.’s paper suggests that a little “apology payment” added on top of the fine might just do the trick. Apologies within apologies.

By the way, Jeff and Mimi’s parents are reviewing the terms of the old agreement later this week. Perhaps it can be made to work after all.


Martinez-Vaquero, L. A., Han, T. A., Pereira, L. M., & Lennaerts, T. (2015). Apology and forgiveness evolve to resolve failures in cooperative agreements. Scientific Reports, 5: 10639.  doi:10.1038/srep10639.


*By which I mean “come to characterize the behavior of individuals in the population via a cultural learning mechanism that causes individuals to adopt the strategies of their most successful neighbors.”


The Trouble with Oxytocin, Part III: The Noose Tightens for The Oxytocin–>Trust Hypothesis

https://i1.wp.com/media-cache-ak0.pinimg.com/736x/2b/1f/9b/2b1f9b4e930d47f31b1f7f3aecd0b0cf.jpgMight be time to see about having that Oxytocin tattoo removed…

When I started blogging six months ago, I kicked off Social Science Evolving with a guided tour of the evidence for the hypothesis that oxytocin increases trusting behavior in the trust game (a laboratory workhorse of experimental economics). The first study on this topic, authored by Michael Kosfeld and his colleagues, created a big splash, but most of the studies in its wake failed to replicate the original finding. I summarized all of the replications in a box score format (I know, I know: Crude. So sue me.) like so:

Box Score_Dec2013By my rough-and-ready calculations, at the end of 2013 there were about 1.25 studies’ worth of successful replications of the original Kosfeld results, but about 3.75 studies’ worth of failed replications (see the original post for details). Even six months ago, the empirical support for the hypothesis that oxytocin increases trust in the trust game was not looking so healthy.

I promised that I’d update my box score as I became aware of new data on the topic, and a brand new study has just surfaced. Shuxia Yao and colleagues had 104 healthy young men and women play the trust game with four anonymous trustees. One of those four trustees (the “fair” trustee) returned enough of the subject’s investment to cause the subject and the trustee to end up with equal amounts of money; the other three trustees (designated as the “unfair players”) declined to return any money to the subject at all.

Next, subjects were randomly assigned to receive either the standard dose of intranasal oxytocin, or a placebo. Forty-five minutes later, participants were told that they would receive an instant message from the four players to whom they had entrusted money during the earlier round of the trust game. The “fair” player from the earlier round, and one of the “unfair” players, sent no message at all. The second unfair player sent a cheap-talk sort of apology, and the third unfair player offered to make a compensatory monetary transfer to the subject that would make their payoffs equal.

Finally, study participants took part in a “surprise” round of the trust game with the same four strangers. The researchers’ key question was whether the subjects who had received oxytocin would behave in a more trusting fashion toward the four players from Round 1 than the participants who received a placebo instead.

They didn’t.

In fact, the only hint that oxytocin did anything at all to participants’ trust behaviors was a faint statistical signal that oxytocin caused female participants (but not male participants) to treat the players from Round 1 in a less trusting way. If anything, oxytocin reduced women’s trust. I should note, however, that this females-only effect for oxytocin was obtained using a statistically questionable procedure: The researchers did not find a statistical signal of an interaction between oxytocin and subjects’ sex, and without such a signal, their separation of the men’s and the women’s data for further analyses really wasn’t licensed. But regardless, the Yao data fail to support the idea that oxytocin increases trusting behavior in the trust game.

It’s time to update the box score:


In the wake of the original Kosfeld findings, 1.25 studies worth of results have accumulated to suggest that oxytocin does increase trust in the trust game, but 4.75 studies worth of results have accumulated to suggest that it doesn’t.

It seems to me that the noose is getting tight for the hypothesis that intransasal oxytocin increases trusting behavior in the trust game. But let’s stay open-minded a while longer. As ever, if you know of some data out there that I should be including in my box score, please send me the details. I’ll continue updating from time to time.

Of Crackers and Quackers: Human-Duck Social Interaction is Regulated by Indirect Reciprocity (A Satire)

1280px-221_Mallard_DuckWatching the ducks on a neighborhood pond can be an entertaining and rewarding pastime. I myself, along with my nine-year-old co-investigator, have taken daily opportunities to feed some ducks on a nearby pond over the past several months. In doing so, we not only had fun but also managed to conduct some urban science that led us to a new scientific discovery: Mallards (Anas platyrhynchos L.) engage in indirect reciprocity with humans. Scientists have known for decades, of course, that indirect reciprocity was critical to the evolution of human social interaction in large-scale societies, but we believe we are the first to identify indirect reciprocity at work in human-duck social interaction.

Here’s how we made this discovery.

On random days, we take a soda cracker along with us to feed to a single lucky duck. On the other days, we take our walks without a cracker. What my young co-investigator and I have noticed is that on cracker days, after we’ve fed the cracker to the first duck that approaches us (the “focal duck,” which we also call “the recipient”), other ducks (which we call “entertainment ducks,” or “indirect reciprocators”) appear to take notice of our generosity toward the recipient. Almost immediately, the indirect reciprocators start to perform all sorts of entertaining behaviors: They swim toward us eagerly, they waddle up to us enthusiastically, they stare at us with their dead, obsidian eyes, they quack imploringly. It’s all very amusing and my co-investigator and I have a great time. Take note of the fact that we always bring only a single cracker with us on cracker days. As a result, the indirect reciprocators have absolutely nothing to gain from the entertainment they provide. In fact, they actually incur costs (in the form of energy expended and lost foraging time) when they do so. Thus, their indirect reciprocity behavior is altruistic.

Our experience with the indirect reciprocators is very different on non-cracker days. If a focal duck comes up to us on a non-cracker day, there’s just no cracker to be had, no matter how charming or insistent the request. Dejected, the focal duck typically waddles or paddles away within a few seconds. Now, what do you suppose the entertainment ducks do after we refuse to feed the focal duck? That’s right. They withhold their entertainment behaviors. This pattern, of course, is exactly as one would expect if the entertainment ducks were regulating their entertainment behaviors according to the logic of indirect reciprocity.

Theorists typically assume that the computational demands for indirect reciprocity to evolve are quite extensive. For instance, indirect reciprocators need to possess computational machinery that enables them to acquire information about the actions of donors—either through direct sensory experience of donor-recipient interactions, or (more rarely) language-based gossip, or (even more rarely) social information stored in an external medium, such written records or the reputational information that’s often available in online markets. Indirect reciprocators also need be able to tag donors’ actions toward recipients as either “beneficial” or “non-beneficial,” store that social information in memory, and then feed that information to motivational systems that can produce the indirect reciprocity behaviors that will serve as rewards to donors. However, the indirect reciprocity we’ve identified in our mallards suggests that those computational requirements may be fulfilled in vertebrates more commonly than theorists originally thought.

Neither of us could figure out for sure whether the focal ducks were transmitting information about our generosity/non-generosity to the indirect reciprocators through verbal (or non-verbal) communication, but we think it is unlikely. Instead, we suspect that the indirect reciprocators were directly observing our behavior and then using that sensory information to regulate their indirect reciprocity behavior.

In support of this interpretation, we note that on several cracker days, it was not only other ducks that engaged us as indirect reciprocators, but individuals from two different species of turtles (which we believe to be Rachemys scripta and Apalone ferox) as well. The turtles’ indirect reciprocity behaviors, of course, were different from those of the ducks, due to differences in life history and evolutionary constraints: The turtles didn’t reward our generosity through waddle-based or quack-based rewarding, but rather, by (a) rooting around in the mud where the focal duck had received the cracker earlier, and (b) trying to grab the focal duck by the leg and drag it to a gruesome, watery death. The fact that turtles engaged in their own forms of indirect reciprocity suggests that they, at least, were obtaining information about our generosity via direct sensory experience, rather than through duck-turtle communication or written or electronic records: It is widely accepted, after all, that turtles don’t understand Mallardese or use eBay.

The involvement of turtles as indirect reciprocators also suggests that indirect reciprocity might be even more prevalent–and more complex–than even we originally suspected. Not only does indirect reciprocity evolve to regulate interactions within species (viz., Homo sapiens), and between species (viz., between Homo sapiens and Anas platyrhynchos L., as we have documented here), but also among species (Homo sapiens as donors, Anas platyrhynchos L. as recipients, and Rachemys scripta and Apalone ferox as indirect reciprocators).

Finally, we should point out that although our results are consistent with the indirect reciprocity interpretation that we have proffered here, other interpretations are possible as well. We look forward to new work that can arbitrate between these two accounts (and perhaps others). We also see excellent opportunities for simulation studies that can shed light on the evolution of indirect reciprocity involving interactions between two or even three different species, which my co-Investigator thinks she might pursue after she has mastered long division.

h/t Eric P.

I’m feeling Edge-y about Human Evolutionary Exceptionalism

Unless your Internet has been broken for the past few days, by now you’re probably aware that John Brockman, via his Edge.org web site, has published the responses to his Annual Edge Question of the Year. For more than 15 years, John has been inviting people who think and write about science and the science-culture interface to respond to a provocative question. The question for 2014 was “What scientific idea is ready for retirement?” Brockman explains:

Science advances by discovering new things and developing new ideas. Few truly new ideas are developed without abandoning old ones first. As theoretical physicist Max Planck (1858-1947) noted, “A new scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die, and a new generation grows up that is familiar with it.” In other words, science advances by a series of funerals. Why wait that long? What scientific idea is ready for retirement? Ideas change, and the times we live in change. Perhaps the biggest change today is the rate of change. What established scientific idea is ready to be moved aside so that science can advance?

I received an invitation to participate this year, and it didn’t take me long to settle on a topic. Something has been bugging me about the application of evolutionary thinking to human behavior for a while, and the Question of the Year was the perfect place to condense my thoughts into a 1000-word essay. What scientific idea, in my opinion, is ready for retirement? I nominated Human Evolutionary Exceptionalism. Here’s how I framed the problem:

Humans are biologically exceptional. We’re exceptionally long-lived and exceptionally cooperative with non-kin. We have exceptionally small guts and exceptionally large brains. We have an exceptional communication system and an exceptional ability to learn from other members of our species. Scientists love to study biologically exceptional human traits such as these, and that’s a perfectly reasonable research strategy. Human evolutionary exceptionalism, however—the tendency to assume that biologically exceptional human traits come into the world through exceptional processes of biological evolution—is a bad habit we need to break. Human evolutionary exceptionalism has sown misunderstanding in every area it has touched.

In my essay, I went on to describe examples of how human evolutionary exceptionalism has muddled the scientific literatures on niche construction, major evolutionary transitions, and cooperation. You can read my entire essay over here, but for this blog I’m reproducing what I had to say about the major evolutionary transitions—for reasons that I will make clear presently. The most critical part below is bolded and italicized. Here’s what I wrote:

Major Evolutionary Transitions. Over the past three billion years, natural selection has yielded several pivotal innovations in how genetic information gets assembled, packaged, and transmitted across generations. These so-called major evolutionary transitions have included the transition from RNA to DNA; the union of genes into chromosomes; the evolution of eukaryotic cells; the advent of sexual reproduction; the evolution of multicellular organisms; and the appearance of eusociality (notably, among ants, bees, and wasps) in which only a few individuals reproduce and the others work as servants, soldiers, or babysitters. The major evolutionary transitions concept, when properly applied, is useful and clarifying.

It is therefore regrettable that the concept’s originators made category mistakes by characterizing two distinctly human traits as outcomes of major evolutionary transitions. Their first category mistake was to liken human societies (which are exceptional among the primates for their nested levels of organization, their mating systems, and a hundred other features) to those of the eusocial insects because the individuals in both kinds of societies “can survive and transmit genes . . . only as part of a social group.”…

Their second category mistake was to hold up human language as the outcome of major evolutionary transition. To be sure, human language, as the only communication system with unlimited expressive potential that natural selection ever devised, is biologically exceptional. However, the information that language conveys is contained in our minds, not in our chromosomes. We don’t yet know precisely where or when human language evolved, but we can be reasonably confident about how it evolved: via the gene-by-gene design process called natural selection. No major evolutionary transition was involved.

This past Monday morning, right as I was about to go to the Edge web site to check out some of the other essays, someone e-mailed me to let me know about an uncanny coincidence. Just hours before my Edge essay came out—in which I was calling for the retirement of the misconception (among others) that human language was the outcome of a major evolutionary transition, Martin Nowak had published an essay on the Templeton Big Questions web site in which he was pushing in exactly the opposite direction. Here’s what Nowak had to say (my emphasis in boldface and italics):

I would consider these to be the five major steps in evolution: (i) the origin of life; (ii) the origin of bacteria; (iii) the origin of higher cells; (iv) the origin of complex multi-cellularity and (v) the origin of human language. Bacteria discovered most of biochemistry, higher cells discovered unlimited genetics; complex multicellularity discovered intricate developmental processes and animals with a nervous system. Humans discovered language.

Human language gave rise to a new mode of evolution, which we call cultural or linguistic evolution. The enormous speed of human discovery and invention is driven by this new mode of evolution. An idea or concept that originates in one brain can quickly spread to others. Structural changes (memories) are imprinted from one brain to another. Prior to human language the most crucial information transfer of evolution was mostly in terms of genetic information. Now we have genetic and linguistic evolution. The latter is much faster.  Presumably the collective information in human brains evolves at a much faster rate than any previous evolutionary system on earth. The growing world wide connectivity speeds up this linguistic evolutionary process.

Now, Nowak and I both agree that human language is a Very Special Way of transmitting information, but I say human language was not the outcome of a major evolutionary transition. Nowak says it was. We can’t both be right, so what’s going on? With respect, I think it’s Nowak who’s muddling things.

It was John Maynard Smith and Eörs Szathmáry who were actually responsible for popularizing the idea that human language was a major transition in evolution (see my essay and read between the lines; you’ll know whom I’m talking about even though I followed Brockman’s instructions to talk about ideas rather than the people who promote them). But as I wrote in my essay, Maynard Smith and Szathmáry made a category mistake when they did so: Here’s the first sentence from the description of their book, The Major Transitions in Evolution: “During evolution, there have been several major changes in the way that genetic information is organized and transmitted from one generation to the next.”

The critical word in the last sentence is “genetic.” Evolutionary transitions are about information stored in DNA, not about information in people’s minds. So, by Maynard Smith and Szathmáry’s own definition of major evolutionary transitions, human language categorically, absolutely cannot be one of them.

This has got to be incredibly obvious to anyone who takes a moment to think about it, so I’m not quite sure why influential people keep the misconception going. Equally puzzling to me, though, is why Maynard Smith and Szathmáry committed this error in the first place. Those gents are/were smart (Maynard Smith died in 2004; Szathmáry is still with us), and few people have ever had cause to doubt Maynard Smith’s judgment (though read Ullica Segerstrale’s biography of Bill Hamilton to learn about a striking exception to that rule). In any case, I can’t understand why Nowak continues to promulgate the notion that the evolution of human adaptations for nice things like human societies (as he did here) and human language (as in the Big Questions Online piece)—often using Maynard Smith and Szathmáry’s book for citation firepower—are comparable to actual Major Evolutionary Transitions that involved actual “major changes in the way that genetic information is organized and transmitted from one generation to the next.”

Human language is fascinating, puzzling, and a prime target for theory-building and research. Ditto for human cooperation and human societies. But these interesting features of human life are made neither grander, nor more comprehensible, by trying to get them into The Major Evolutionary Transitions club. They just don’t have the proper credentials.