A Multiagent Framework for Understanding Addiction

In this paper, we provide a framework for examining the problem of addiction that considers both internal factors like self-control as well as external factors denoted as the environment. We do this by considering the Prisoner’s Dilemma, a game-theoretic concept examined by competitive multiagent systems researchers. In particular, we devise an iterated Prisoner’s Dilemma involving you and future you. We devote considerable effort in defining the notion of selfhood from previous literature in economics, as this is critical in examining addiction. The main contribution is a framework that enables calibration of alternate scenarios of behaviour for addicted individuals: in essence an application of artificial intelligence for the important social problem of modeling addiction, yielding some intuitive and explanatory results. We briefly comment on the main underlying assumptions and biases as well as mention future work that could be derived from this research, including commentary on how reasoning about both current and future reflections of self may be useful in general for multiagent decision making.