Dealing with Dynamic Decision Problems when Knowledge of the Environment Is Limited: An Approach Based on Goal Systems

摘要

We experimentally analyzed decision procedures for dealing with a dynamic decision-making problem in which only qualitative information about the deterministic dynamics of the environment was available to participants. A participant's task was to maximize long-term profit in a computer-simulated monopoly market featuring delays and inertia. The design enabled a goal-system-based procedure, whereby a participant could select one or several short-term variables to be controlled (goal variables) and chose target values (aspiration levels) for each of them over a total of 50 periods. We report results based on a sample of 63 participants on the formation of goal systems and the process of aspiration adaptation. Our main findings are, first, that more frequently selecting goal systems that adequately reflect the causal structure of the underlying model is positively correlated with long-term profit; second, that goal persistence, a measure of a participant's tendency to stick to the current goal system, is positively correlated with long-term profit; and third, that aspiration levels tend to be adapted in strong agreement with certain basic principles of a benchmark model of aspiration adaptation. Our study thus suggests and provides empirical foundation for an approach to dealing with complex dynamic decision problems based on neither optimization nor learning.