When to inspect? Recurrent inspection decisions in a simulated risky environment.

Participants scheduled inspections to detect costly events for which they were repeatedly at constant risk (probability of event onset) within a computerized environment. They were responsive to risk variations, conveyed either in advance or by experience with inspection outcomes, although experiencing unannounced increases in risk affected inspections more than experiencing unannounced decreases. Participants responded to variations in cost (time varying or fixed) when the effects were made perceptually salient. Compared with a normative model (R. L. Klatzky, D. M. Messick, & J. Loftus, 1992), some conditions showed near-optimal inspecting or had flat payoff functions that tolerated observed departures from optimal performance. Costly departures occurred particularly when combined cost and risk levels caused optimal responses to be extreme (always or never inspect). Results assess people's processing of relevant variables and indicate circumstances in which they may set substantially nonoptimal inspection schedules.