Here’s a new instance in a series of silly-dialogue blog posts (find others here, here, and here). Today, the study of the emergence of cooperation is a peculiar cocktail of applied math, behavioral economics, and theoretical population biology. I have found myself on either side of discussions about if this is the correct direction (depending on if I’m talking to Ivan Romic or Fredrik Liljeros, so shout out to them).
This is an extension of a comment available (and citable 😉) here. Please follow the hyperlinks, as they contain several great and recommendable papers.
😶 … ah sorry, I can’t. I’m going to help a friend move then.
🫥 Say what? What emergent norm is your friend adhering to so much that it warrants a major cooperative effort from you?
😶 Haha, you’re referring to that recent PNAS paper.
🫥 Well, I’m also referring to you never admitting that cooperation can simply be an act of friendship.
😶 It could! And in this case, it will be. But friendships have reasons. They can be explained. Explaining cooperation doesn’t stop at friends helping friends.
🫥 Maybe it does! Imagine a case where there really isn’t any deeper reason for an unselfish act other than friendship. How could you correctly classify such a situation if you assume people all have a utility function they keep optimizing?
When people interact socially, they do a myriad of things for a myriad of reasons. They’re not just, or even mostly, optimizing. They follow routines, gather knowledge, sadistically exert power, have fun, vent their frustration to others, etc. It’s like the Georg Simmel passage that Erving Goffman quotes at the beginning of his legendary Ph.D. thesis.
😶 “[T]here exists an immeasurable number of less conspicuous forms of relationship and kinds of interaction.”
🫥 Yes, “[t]aken singly, they may appear negligible. But since in actuality they are inserted into the comprehensive and, as it were, official social formations, they alone produce society as we know it.”
😶 But those other interactions, why are they there? Explaining them means finding the reason or motive for the behavior. If a magazine reports a crime, it will report the perpetrator’s motive.
🫥 Magazines! You really never leave the shallow end of the pool . . . If those are your standards, what whodunnit is solved by the killer’s motive alone? There would always be some mapping of social relations.
😶 Well, a theory should also explain the observed social relations.
🫥 What good is a theory if it’s wrong?
One cannot keep adding reality by extending a utility function. People don’t evaluate a utility function in their daily decisions, about cooperation or otherwise. Logically, the most pressing scientific question should be to estimate the fraction of decisions based on utility-function optimization. Still, nobody investigates that because, of course, the result would devalue the entire field.
😶 A wrong theory is at least a place to start. What do you have? Regression tables? There are good arguments that the reproductive crisis would have been less of a problem with more systematic use of theory.
🫥 Theory, theory. An ecosystem of competing theories is a state of non-knowledge. We need to find consensus and actually solve our open problems. There must be a description of reality better than others, given how society works both as a study object and a recipient of scientific information.
😶 Well then, why wouldn’t evolutionary game theory be a good basis for a consensus solution?
It has what many social theories are criticized for missing: It is a collective effort, beyond just a PI and a couple of colleagues, where participants know where they’re going. It is replicable. It has a working cycle of theory development and experiments. It has a…
🫥 OK, stop there. That cycle isn’t working. You need observation studies to take you away from the very narrow confines you are staying within. Narrow and extreme. Everyone is shooting at the same little spot way off the target. When are you ever in a situation that just happens to be well-described by the elegant math of game theory?
😶 Don’t be so cranky. You can learn about the normal from the extreme. Look at physics.
🫥 Physics experiments are extreme because the normal is already known. You study the extreme because that’s where you can do fancy math. And physics experiments don’t have your publication bias—every discovery of something not promoting cooperation stays in the file drawer.
😶 Come on. You’re still cranky. Game theoretic models of cooperation are not what they used to be. Many people are interested in the emergence of norms (like you already mentioned). What exactly can’t you cover with a model including individual rationality and social norms?
🫥 Quoting Jon Elster, “many actions are performed out of habit, tradition, custom or duty—either as a deliberate act to meet the expectations of other people and to conform to one’s own self-image, or as the unthinking acting-out of what one is.” Modeling social norms cover the “act to meet the expectations of others” part, but not the others. Furthermore, there might be different, yet-to-be-discovered, interactions and mechanisms.
Finally, and most importantly, we have no clue how these people combine all the different factors to reach a decision. Of course, people don’t run through all these factors and weigh them together, not even subconsciously. This is the frame problem that was the last nail in the coffin of modeling humans by flowcharts … ya know, the idea to measure people’s behavior in every specific situation, BF Skinner stylee, and integrate these atomic responses into a decision tree that would be to a realistic model of a person. But that just didn’t work—it was impossible to model what factors to include in the decision-making (beliefs, hunger, education level, age, values, etc.)—since these factors also depend on the entangled causality web called personality.
😶 But we are seeking scientific explanations. Per definition, such cannot be arbitrarily complicated. Maybe game-theoretic models of cooperation, including social norms, social networks,
🫥 … and beliefs
😶 and beliefs, do not converge to a complete predictor of every individual’s cooperative behavior. But that is not what we want. Explanations must be generalizable and understandable and, thus, not too complicated—the territory being the map, you know. Continuously adding features to them is the same as slowly building a black box.
This argument also holds if you replace models with data—“the richer the data, the more important it becomes to have an organizing theory to make any progress.”
🫥 If what you say is true, there might be situations where your million-parameter model called brain can come up with an accurate prediction, but we can’t explain why. Now, are you claiming that we don’t understand the situation? That we must dumb down to something communicable in English?
😶 Yap. If you can’t explain your intuition, you don’t understand. Prediction accuracy doesn’t matter.
🫥 Your thinking doesn’t seem too far from culture evolution, dual inheritance theory, and similar theoretical frameworks, no? First, there hasn’t been a clear link between payoff and behavioral contagion since pre-historic days, when we were subject to the same selective forces as any other animal. The second and worse problem is that such thinking doesn’t have a clear boundary to social Darwinism, eugenics, and the darkest depths of human intellectual history.
In today’s publishing climate, papers push far too strong conclusions. There are definitely claims about genes and cultures with desirable properties around the corner from that line of research. Even if your results come with caveats, you won’t have a chance to say that. Shady people will pick what they want from your paper. You could fuel scientific racism even if that wasn’t your intention.
😶 First, plenty of research in that direction stands on moral high grounds. But, yeah, I agree, there are disturbing descendants of social Darwinism, and those aren’t the only murky waters. People also sneak in deranged ideas through more benign-looking arguments like Hardin’s “Tragedy of the commons.”
🫥 Yes, it’s impossible to revoke families’ right to choose their number of offspring without denouncing people’s equal value. But that is far from the emergence of cooperation though.
😶 The universal declaration of human rights and the United Nations are shining symbols of collaboration, no?
🫥 No cooperation is the collective action of individuals. When institutions make us do something, that’s a different story, even though they originally came out of cooperation.
😶 Yeah, thus “symbols.” Duh.
🫥[Etc.]