Eliciting the Type Space in Extensive-Form Game Forms

Ohio State University, The, Columbus OH

Investigators

Abstract

Game theory provides predictions of how individuals behave in strategic settings (`games'). It has an enormous range of applicability, from competition between firms to the evolution of a population of microorganisms. But its predictions stem from very strong assumptions on peoples' beliefs, preferences, and rationality. Recent laboratory evidence has shown that the standard predictions are often inaccurate, and several new theories have been proposed as replacements for the standard theory. But almost all existing data comes only from looking at the strategies chosen. The principal investigator proposes to dig deeper by eliciting information about how people think when playing these games. By gaining a better understanding of the underlying thought processes, more accurate game-theoretic models can be built. In addition to this research, the principal investigator will invite high school students to campus (particularly from under-privileged areas) to learn about economics, experiments, and campus life in general. In the centipede game, there is a pot of money and two players alternatively choose to `pass' or `take'. If they pass, the pot grows and the other player gets to pass or take. If they take, then they take a large fraction of the pot, the other player gets a small fraction, and the game ends. If nobody takes after 't' periods, they split the large pot equally. Backwards induction arguments predict that rational, selfish players will take at every node. But what should a player believe about her opponent if she observes him pass? Should she continue to believe in rationality, as prescribed by the backwards induction argument, or should she now admit the possibility that her opponent is irrational? Different assumptions lead to vastly different predictions about how the game will unfold. Past experiments are not especially insightful in answering these questions because they only observe players' actions. The principal investigator will run experiments in which beliefs about preferences and rationality are elicited during the game as well. This additional data will help to understand how beliefs about rationality evolve, and to develop new, more accurate predictions of how people behave in general sequential-move strategic settings.