Subjective sensitivity analysis

Previous studies have concluded that when making inferences, people tend to ignore various kinds of normatively important information (e.g., sample size, base rates). The basis for such conclusions has typically been the similarity of responses across groups presented with widely discrepant values of that information. The experiments reported here convert earlier, between-subject designs to within-subject formats. Each subject made several judgments as one kind of information varied across a range of possible values. When information regarding base rates or predictive validity was varied, roughly two-thirds of the subjects changed their judgments in directions dictated by normative considerations, although the magnitudes of these changes were too small. There were no appreciable shifts in response to changes in sample size information. Apparently, people know (or can figure out) somewhat more than what they have been given credit for in the past. Of course, modest sensitivity in these somewhat artificial conditions is not inconsistent with completely ignoring such information when, as is usually the case, the world presents but one set of values.