Frame Of Stack

Non-equilibrium game theory.

Without equilibria, how do you decide what kind of agents to talk about? Random agents?

I think you want some kind of rational agent constraint, still; recursive "common knowledge of rationality", even. Just refuse to accept arbitrary equilibria.

Maybe accept unique equilibria when they exist. To first approximation, the idea is that agents are uncertain about what equilibrium they're in if there are several. This has obvious problems -- they're not in any equilibrium then, so why do they think they may be?

What does it mean to have common knowledge that all players are "rational" without knowing what function they implement?

Maybe you know that their function maximizes an expected utility, and you know their utility function, and you even know that there's a common prior between you wrt the game being played, but all you know about their beliefs about you is that they know those same things about you

In an antisymmetric coordination game (you like vanilla, I like chocolate, but we only get ice cream if we both choose the same flavor), if we have no way to break symmetry then we just have to flip a coin.