AuthorsT. M. Gruschke and M. Jørgensen
EditorsC. Roland-Lévy
TitleTo Know Or Not to Know: When Does Feedback Lead to Better Assessment of Uncertainty of Own Beliefs?
AfilliationSoftware Engineering, Software Engineering
StatusPublished
Publication TypeProceedings, refereed
Year of Publication2006
Conference NameIAREP/SABE congress (Behavioural Economics and Economic Psychology)
PaginationO-226
PublisherElsevier
ISBN Number999-999-999
Abstract

People are frequently overconfident about the accuracy of their own beliefs. The goal of this experiment is to examine whether, and if so under what conditions, very large amounts of feedback lead to better assessments of the uncertainty of one's own beliefs. Fifteen participants answered the same 960 general knowledge questions over two days. Questions were selected from the board game “Who wants to be a millionaire?™”. Each question had four answer alternatives. When an answer alternative had been chosen, the participants were asked to assess the probability that it was the correct answer. They did this by choosing from a list of predefined confidence intervals. The questions belonged to one out of six difficulty categories, as decided by the board game developers. Participants answered piles of 80 questions of the same difficulty at a time. Feedback about the correct answer was received immediately, and feedback about correspondence between “hit rate” and confidence level was received immediately after a pile of questions had been answered. In addition, a summary of the first day's performance was provided at the start of Day 2. We found that thirteen out of the fifteen participants improved in the correspondence between hit rate and confidence level on the second day, which suggests that the feedback had an effect. The strongest improvement was achieved on questions with high “global” difficulty, i.e., high difficulty as assessed by the board game developers, and low “internal” difficulty, i.e., low difficulty as perceived by the participants.

Citation KeyGruschke.2006.5