This seems basically correct to me. But I think the example of dancing doesn’t really work.
Consider a population of agents that meet in pairs and play a complementary coordination game, like ballroom dancers that need to decide who should lead and who should follow. It’s kind of a pain if every single pair has to separately negotiate roles every time they meet! But if the agents come in two equally numerous types (say, “women” and “men”), then the problem is easy: either of the conventions “men lead, women follow” or “women lead, men follow” solves the problem for everyone!
This solution has its own costs, and only works if those are less than the costs of negotiating each dance.
People have to arbitrarily exclude half the population as potential dance partners. Otherwise, half the dances will have two “men” or two “women”, and there’s no convention to deal with that.
People don’t get to pick their dance role. Note that this means most people probably don’t have a strong preference between the two, but that also means the cost of negotiating is probably small.
(Though it could also be that people have a preference for “a consistent dance role, but I don’t care which”.)
If the labels really are just to solve a coordination problem, I think there are some other strategies that make a pretty strong showing: “taller person leads, shorter person follows” (or vice versa); “just play rock paper scissors if you don’t have opposite preferences”; “choose a role and wear an indicator of it”.
In reality, I think this convention only works because “men” in general want to dance with “women” in general—and then the labels are no longer arbitrary, we just have men and women.
(And I observe that the dance events I used to go to would sell tickets according to gender, not role.)
That is, when it comes to dancing, this convention has to be downstream of gender roles. It doesn’t work to explain gender roles as “the kind of thing that happens when you try to solve the dancing coordination problem”.
Apologies—my blog distillation of “what I learned” is glossing over a lot of stuff that the actual book covers properly: the difference between models where agents only meet the other type vs. also their own type is discussed in §3.3.2–3, and “taller person leads, shorter person follows” is an example of what O’Connor calls “gradient markers” in §2.3.2.
As far as dancing goes, I think it’s kind of like how we give cute mnemonic names like “Hawk–Dove” to payoff matrices of a particular form that don’t quite make sense as a literal story about literal hawks and literal doves, but evolutionary game theory in general really is useful for understanding the behavior of animals (including birds).
This seems basically correct to me. But I think the example of dancing doesn’t really work.
This solution has its own costs, and only works if those are less than the costs of negotiating each dance.
People have to arbitrarily exclude half the population as potential dance partners. Otherwise, half the dances will have two “men” or two “women”, and there’s no convention to deal with that.
People don’t get to pick their dance role. Note that this means most people probably don’t have a strong preference between the two, but that also means the cost of negotiating is probably small.
(Though it could also be that people have a preference for “a consistent dance role, but I don’t care which”.)
If the labels really are just to solve a coordination problem, I think there are some other strategies that make a pretty strong showing: “taller person leads, shorter person follows” (or vice versa); “just play rock paper scissors if you don’t have opposite preferences”; “choose a role and wear an indicator of it”.
In reality, I think this convention only works because “men” in general want to dance with “women” in general—and then the labels are no longer arbitrary, we just have men and women.
(And I observe that the dance events I used to go to would sell tickets according to gender, not role.)
That is, when it comes to dancing, this convention has to be downstream of gender roles. It doesn’t work to explain gender roles as “the kind of thing that happens when you try to solve the dancing coordination problem”.
Apologies—my blog distillation of “what I learned” is glossing over a lot of stuff that the actual book covers properly: the difference between models where agents only meet the other type vs. also their own type is discussed in §3.3.2–3, and “taller person leads, shorter person follows” is an example of what O’Connor calls “gradient markers” in §2.3.2.
As far as dancing goes, I think it’s kind of like how we give cute mnemonic names like “Hawk–Dove” to payoff matrices of a particular form that don’t quite make sense as a literal story about literal hawks and literal doves, but evolutionary game theory in general really is useful for understanding the behavior of animals (including birds).