The El Farol Bar Problem: A Comparative Analysis of Expectation Models Used in Decision Making

Abstract

Arthur (1994) introduces the El Farol Bar Problem (EFBP) in his article “Bounded Rationality and Inductive Reasoning". He creates an agent-based model of the EFBP and uses it to explain the importance of the concept "bounded rationality". According to Arthur, deductive reasoning will not create decisions that will produce a desired behavior for the EFBP. Hence, boundedly rational agents using inductive reasoning in decision making is a must for this and similar type of problems. We extend Arthur's work by creating different types of agents and compare them in terms of performance measures such as mean attendance and standard deviation of attendance. We introduce adaptive learning agents that use inductive reasoning in forming their decisions expecting an improvement in the overall performance of the agents. Throughout the analysis of the EFBP, we discover the role of the heterogeneity and the detrimental effect of using the weekly attendance information. Unexpectedly, as a result of our findings, the behavior of adaptive learning agents converges to the behavior that would be expected from agents using deductive reasoning.