Overconfidence in What-if Analysis: Active Involvement vs. Imaginability of Outcomes

摘要

Recent published studies suggest that the use of what-if analysis often leads decision makers to become overly confident in their resulting decisions. The purpose of the present research is to investigate the theoretical mechanisms underlying this inflated sense of confidence. The two candidate mechanisms examined relate to two prominent features of what-if analysis. First, the additional effort put forth physically entering decision scenarios into a what-if model compared to unaided decision making may be a form of "active involvement" which could lead to inflated expectations of success according to the illusion of control principle. Second, witnessing a broader range of simulated future outcomes via the what-if model might inflate the "imaginability" of successful outcomes, which theoretically could exaggerate decision makers' perceptions of the likelihood of those outcomes. The present research proposes an experiment to assess the relative roles of "active involvement" and "imaginability of outcomes" as underlying determinants of overconfidence during what-if analysis. The experimental design involves independently manipulating entry of model input and receipt of model results, and assessing their relative impact on decision confidence. Such research would have important implications for designing interventions aimed at reducing overconfidence using what-if analysis as well as for identifying possible sources of biased confidence in other types of decision support tools.