The confidence–hindsight mirror effect in judgment: An accuracy-assessment model for the knew-it-all-along phenomenon.

Abstract

This article presents a model for the effect: the accuracy-assessment model. The model is based on the assumption that participants in hindsight studies use the strategy of trying to reproduce the distribution of correct and wrong responses that seem appropriate in view of their assessment of the expected accuracy. The model provides precise, quantitative, parameter-free predictions about the extent and direction of hindsight bias. In particular, the model predicts good calibration in hindsight and a systematic relation between over/underconfidence in foresight and hindsight bias, referred to as the confutence-hindsight mirror effect. A novel and unique prediction by the model is a reversal of the knew-it-all-along effect in judgment domains characterized by underconfidence in foresight. This reversed hindsight phenomenon and other predictions by the model were tested and confirmed in 4 experiments. After the tragedy at the siege of a religious sect in Waco, Texas, in 1993, Swedish researchers on psychology of religion stated that with knowledge of such sects at hand, the outcome could not have been anything but violent. It is always easy to be wise after an event. Fischhoff (1975) coined the term creeping determinism for people's tendency in hindsight to believe that the world is more determinate and predictable than it indeed is. When subjective probabilities of an event are estimated in retrospect, with outcome information available, and judges are asked to make their estimation as if they had not received the outcome information, they are prone to be biased in the direction of the outcome, a manifestation of hindsight bias or the knew-it-allalong effect (Wood, 1978). Hindsight bias is measured by comparing hindsight estimates by participants who were given outcome information, either with those of a control (foresight) group without this information or with estimates given by the same participants at a prior occasion without outcome knowledge. Hindsight bias has been demonsurated in a variety of tasks and contexts, such as general knowledge items, political events, economic decisions, and sports events (see Christensen-Szalanski & Willham, 1991; Fischhoff, 1982; Hawkins & Hastie, 1990; for reviews). Motivational as well as cognitive factors have been proposed as causes of hindsight bias, with most recent research favoring cognitive explanations (Hawkins & Hastie, 1990).