Confidence Following Choice

Abstract

Two procedures for eliciting probability judgments are compared. In the choice method, subject answers a yes/no question and then states a probability that his/her answer is correct. In the no-choice method, subject states his/her probability that a given proposition is true. Previous experiments by Ronis and Yates (1987) and Fischhoff, Slovic, and Lichtenstein (1977) indicated that subjects are less confident (i.e. give less extreme probability judgments) in the choice than in the no-choice condition. The effect proves to be sensitive to methodolgical details. For questionniares embedded in a large packet which subjects complete rapidly, the reduction in confidence following choice found by earlier researchers is replicated and is extended to questions eliciting subjects ' willingness to bet. However, when the survey is completed at a more leisurely pace, in isolation from other studies and distractions, there is no difference in confidence between choice and no-choice subjects on average. For particular types of questions, however, confidence is enhanced following choice in an isolated setting, while for other questions confidence is reduced. A preliminary typology is suggested for predicting which effect will obtain, and initial tests tend to support it. Additional studies indicate that monetary valuation judgments and a Likert scale of attachment following choice do not exhibit the same pattern as confidence does, while evidential conflict does appear to affect confidence in a similar manner as the forced choice task. Results of Wallsten and Gonzalez-Vallejo (1994) and Tversky and Koehler (1994) are extended to show that the choice method biases yes/no answers toward "yes" but does not boost the judged probability of the focal proposition, i.e. binary complementarity holds for probability judgments using the choice method. The effects of choice on confidence appear to be orthogonal to these considerations. Results of all the studies point to the mutability and fragility of the effect of choice on subsequent confidence. The isolated-questionnaire studies suggest that the act of choice accentuates one's feeling of whether one knows enough to make a dichotomous choice in the available time, but other interpretations are also consistent with the data. This paper is a Ph.D. Dissertation, Department of Psychology, Stanford University, 1995.

Keywords: confidence, probability judgment, forced choice

JEL Classification: D81, D01, D03