Domain Specificity of Inductive Rules

DOMAIN SPECIFICITY OF INDUCTIVE RULES

One of the major implications of the present viewpoint is that there should be a substantial degree of domain specificity of statistical reasoning. Its use should be rare for domains in which (1) it is hard to discern the sample space and the sampling process, (2) the role of chance in producing events is unclear, and (3) no cultural prescription for statistical reasoning exists. We have noted that many of the studies showing people’s failures in statistical reasoning examined judgments about events characterized by one or more of these factors. We have also noted that some of people’s few demonstrated successes in statistical reasoning have been observed for people’s reasoning about sports and academic achievements that seem to be characterized by clearer distributions for events, a more obvious role of chance, and, probably, cultural prescriptions as well. These observations were made across studies, across tasks, and across subject populations, however.

To demonstrate that the same subjects dealing with the same tasks in the same experiment are more likely to reason statistically for some events characterized by uncertainty than for others, Jepson, Krantz, and Nisbett (1983) presented subjects with two broad classes of problems. The first class of problems dealt with events that are assessable by objective means, such as abilities, achievements, and physical illness. The second class dealt with events that are assessable only by subjective means, for example, personal preferences among objects, assessments of leadership potential, and judgments about the need for sexual fidelity in relationships. It was reasoned that, in general, it is relatively easy to apply statistical reasoning to objective events because one is likely to have some idea of their distributions (or to be able to guess what the distributions might look like because the units of measurement and the sample space are likely to be relatively clear). In general, also, the role of chance is likely to be relatively transparent for those objectively assessable events that in fact have been observed under repeated, relatively fixed conditions. Finally, cultural prescriptions to reason statistically probably exist for many such events. In contrast, none of these things is true for most events that can be assessed only by subjective means. The original objective and subjective sample problems, and the experiment’s results, are not included in this excerpt.