Defending diversity.

Abstract

Why do some observations lead to broad generalizations whereas other obser-vations do not have as much influence on people's beliefs? One principle of evaluating evidence is the diversity principle, which states that more diverse evidence should lead to stronger inferences than a narrow sample of evidence. For example, if you see someone repeatedly acting aggressively while giving lectures, you might infer that this behavior is simply the person's lecturing style and not draw broad inferences about the person in general. On the other hand, ifyou observe someone acting aggressively in diverse contexts, such as in a lecture, at a restaurant, and at a party, you might infer that this is indeed an aggressive person overall. There are two views on the diversity principle. The first is the normative view. The emphasis in this view is on why people ought to reason in accord with the diversity principle-that is, why it will lead to successful reasoning. The normative view encompasses historical evidence of expert scientific prac-tice following the diversity principle, as well as various arguments and proofs intended to show that the diversity principle leads to useful inferences. The second view is descriptive. The descriptive view emphasizes showing the many situations in which ordinary reasoning does indeed follow the diversity princi-ple. Typically, the descriptive view has relied on experimental evidence inves-tigating various subject populations. This chapter is organized in terms of these views. The first two sections briefly summarize the positive cases for the normative view and the descriptive view. The following two sections describe some of the challenges to these views and offer some responses in defense of the diversity principle. Doug Medin's many contributions to this line of work and related research, inside and outside the lab, are gratefully and respectfully acknowledged.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 27 citations exist upstream.