Bounded Rationality in Las Vegas: Probabilistic Finite Automata PlayMulti-Armed Bandits

We can think of the number of states of the automaton as a proxy for how computationally bounded the While traditional economics assumes that humans agent is. Neyman (1985) showed that cooperation can are fully rational agents who always arise if PFAs play a finitely-repeated prisoner's dilemma; maximize their expected utility, in practice, we work on this topic has continued to attract attention (see constantly observe apparently irrational behavior. Papadimitriou and Yannakakis (1994) and the references One explanation is that people have limited therein). Wilson (2015) considered a decision problem computational power, so that they are, quite rationally, where an agent must decide whether nature is in state 0 making the best decisions they can, or state 1, after getting signals that are correlated with given their computational limitations.