Causal Models of Decision Making: Choice as Intervention

Brown University, Providence RI

Investigators

Abstract

Decision makers are actively trying to understand their environments. They are busy constructing and using causal models that (hopefully) accurately predict the effects of their choices on themselves, on others, and on the world around them. The purpose of this research is to develop a cognitive theory of this causal knowledge that will explain how people's causal models get translated into action. The theory will be developed using a framework known as causal Bayesian networks that is being developed in the fields of statistics and artificial intelligence. Many of the ideas were originally developed by philosophers who observed that the traditional view that optimal decision making involves maximizing the probability of obtaining the most resources (evidential expected utility theory) fails to explain how people take the causes and effects of their actions into account when they make decisions. The key idea is that choice is not merely a selection of an option from a set of alternatives, but an intervention that changes not only the state of the world, but also the model we should use to represent how the world works. The key psychological claims of the theory are that people represent the world by decomposing it into autonomous mechanisms that support interventions and that choice suspends some of those mechanisms in the causal model relevant to a decision. Experiments will (i) test people's sensitivity to causal structure as well as the hypothesis that choice is an intervention; (ii) examine the proposed model by framing it in terms of causal expected utility and contrasting it to a more standard model of evidential expected utility in the context of both simple scenarios and two-player games; (iii) examine what people can learn about the world and about choice strategies from observing one's own and others' choices; and (iv) examine various actual and apparent boundary conditions on the theory such as the role of self-deception in choice. The research has implications for how people should make all kinds of decisions (personal, medical, policy, strategy, etc.). It also has implications for how to improve decision-making, specifically, by first understanding the causal structure of a situation before attempting to make decisions in it.