A number of studies have reported that there is a strong tendency to ignore base-rate information in favor of individuating information, except when the former can readily be incorporated into a causal schema. In the present study, students in eight undergraduate classes were given problems in which the base-rate information was (1) either causal of noncausal and (2) either incongruent or congruent with the individuating information. In addition, twelve subjects were interviewed as they attempted to solve several versions of the one of the problems. We found (1) strong individual differences in the perceived importance of base-rate information and even in how the probability estimation task itself was interpreted, (2) little if any effect of the causality manipulations employed by Ajzen (1977) and Tversky and Kahneman (1980, and (3) greater use of base-rate information congruent with the individuating information than of base-rate information which is incongruent. The interview data indicate that it is difficult to determine from the answer alone whether or not the subject thought that the base-rate information was relevant. These data also suggest that subjects have different strategies for dealing with probability estimation problems. One of these we characterize as not only nonBayesian, but also nonprobabilistic.
The CAUSE Research Group is supported in part by a member initiative grant from the American Statistical Association’s Section on Statistics and Data Science Education