Bounded rationality in games of strategy

Abstract

[ACCESS RESTRICTED TO THE UNIVERSITY OF MISSOURI AT AUTHOR'S REQUEST.] Traditional game theory predicts behavior contrary to how real people actually behave. And what traditional game theory prescribes as the rational thing to do is normally unattainable in real-­‐life. The problem is that game theorists have traditionally assumed that agents have no cognitive limitations and know all logical and mathematical truths. Hence, traditional game theory does not account for people's cognitive limitations—their bounded rationality. I remove the strong assumptions about rational agents and adjust the principles of rationality for real people. I focus on the Centipede Game, a sequential game, with multiple stages, where ideal agents predict moves at the last stage, and then use these predictions to predict moves at preceding stages, settling on a strategy for moves throughout the interaction —a procedure called backward induction. Applying backward induction makes heavy demands on agents' cognitive capacities and is unrealistic reasoning for them. Thus, I develop an account of bounded rationality that applies a simpler procedure for agents to begin their interaction, by exploring and testing others' behavior until they reach a moment in the sequential game when they are able to apply limited backward induction. This analysis of behavior better predicts how real people actually behave, and prescribes a course of action attainable in real-­‐life.