Playing Games with Ideas: when epistemology pays off
In standard game theory analysis, gameplay proceeds using rational strategies. While there is no explicit link between rationality and strategy selection, playing to equilibrium solutions often depends on choosing the optimal strategy. Yet what if strategy optimality is not obvious to the players? Often, people make seemingly irrational decisions based on beliefs or situational interpretations of the empirical world. These beliefs have high social import, but with little advantage compared to empirical reality.
We use a decision kernel to model how beliefs (an epistemological network) influence a rational strategy suite (top box in attachment). The feedforward network output is a set of beliefs that influence gameplay as a set of rational behavior (empirical strategy suite). This results in an epistemological-influenced social capital payoff function, and an empirical payoff function (bottom box in attachment). The joint distribution (or area of intersection between the payoff functions) is the epistemological payoff. This epistemological payoff can be small (low conditional social payoff) or large (high conditional social payoff). When small, players play their strategies close to the optimal expectation. Larger epistemological payoffs result plays that have with relatively higher social utility, but also little immediate utility when compared to the empirical world.
Attachment: whole-figure.png (280 KB)