Cornell Professor Joseph Halpern and Xinming Liu presented “Bounded Rationality in Las Vegas: Probabilistic Finite Automata Play Multi-Armed Bandits” at the 2020 Conference on Uncertainty in Artificial Intelligence.

In the paper, they leverage the multi-armed bandit problem to test the hypothesis that humans’ limited computational power helps explain their predictably irrational behavior. Using a probabilistic finite automaton to play a multi-armed bandit, they find that as the number of states of the automaton decreases, it exhibits a number of standard human biases.