Choice Deferral in Risky Decision Making: Using Cognitive Models to Understand Information Sampling Strategies

Many of life’s important choices require that the decision maker collects evidence to better inform their actions. However, excessive testing can be costly, so the decision maker must carefully choose when to terminate sampling and commit to a final decision. Such deferred decision making problems – where an individual can choose between two or more risky alternatives or defer choice and collect additional information – appear in many domains, from consumer choice to military decision making. We conducted two deferred decision making studies in which participants played the part of a physician diagnosing patients during a disease outbreak. Participants could purchase independent, probabilistic blood tests, at cost. When they were sufficiently confident, participants diagnosed each patient with one of two possible diseases. A monetary reward or punishment was applied for each correct or incorrect diagnosis, respectively. A small deduction was made when participants purchased each test. Our results show sensitivity to sampling costs – participants purchased more tests when sampling costs were low – and risk – participants purchased more tests when rewards and punishments were large. We tested several cognitive models and found that a sequential sampling model with decreasing decision thresholds provided the best account of participants’ choices. However, an alternative model focused on minimizing errors best fit the response data of a substantial proportion of individuals. These findings challenge the predictions of prominent normative and heuristic theories, and lay the groundwork for further investigation into the psychological processes underlying deferred decision making.