Beyond selecting information: Biases in spontaneous questions and resultant conclusions

Abstract

Biases in trait hypothesis testing were examined from both subjects′ and observers′ viewpoints. Subjects formulated two questions to test each of three hypotheses and drew inference from hypothetical yes and no answers; observers read these questions and drew inferences from the same answers. It was shown that subjects tended to ask about features that were (a) consistent with the hypothesis and (b) either very probable or very improbable. Furthermore, subjects′ data showed that questions were less diagnostic when subjects tested a single hypothesis (e.g., extraversion) as opposed to a double hypothesis (e.g., extraversion versus introversion); observers′ data showed this effect only for the first question asked. Subjects also were likely to confirm the hypothesis that they were testing (relative to an alternative) and were more likely to confirm the hypothesis following hypothesis-consistent questions (relative to hypothesis-inconsistent questions). Both subjects and observers were too influenced by the hypothetical answers; that is, they overconfirmed the hypothesis when the answers were confirmatory and overdisconfirmed the hypothesis when the answers were disconfirmatory. © 1993 by Academic Press, Inc.