Estimating the Use of Higher-Order Theory of Mind Using Computational Agents

When people make decisions in a social context, they often make use of theory of mind, by reasoning about unobservable mental content of others. For example, the behavior of a pedestrian who wants to cross the street depends on whether or not he believes that the driver of an oncoming car has seen him or not. People can also reason about the theory of mind abilities of others, leading to recursive thinking of the sort ‘I think that you think that I think…’. Previous research suggests that this ability may be especially effective in simple competitive settings. In this paper, we use a combination of computational agents and Bayesian model selection to determine to what extent people make use of higher-order theory of mind reasoning in a particular competitive game known as matching pennies. We find that while many children and adults appear to make use of theory of mind, participants are also often classified as using a simpler reactive strategy based only on the actions of the directly preceding round. This may indicate that human reasoners do not primarily use their theory of mind abilities to compete with others.