Rationality & the Reflective Mind #5

More from Keith Stanovich's Rationality & the Reflective Mind follows.

Some people have argued that the research in the heuristics and biases tradition -- began by Kahneman and Tversky in the late 1970s -- has not shown human irrationality at all. They argue that the assumption of maximal human rationality is the proper default position, and have been dubbed the Panglossians. (Panglossian means naively optimistic and is based on the character Pangloss in Voltaire's Candide.) This position posits no difference between descriptive and normative models of performance "because human performance is normative." How do they explain the many observed errors of human thought? They argue that the normative model applied is not appropriate because the subject's interpretation of the task is different from what the researcher's is. They also argue that the modal response in the task makes sense from an evolutionary perspective (p. 8-9).

The Panglossians that Stanovich names are philosophers Nicholas Rescher and L. J. Cohen. Rescher argued that "to construe the data of these interesting experimental studies [of probabilistic reasoning] to mean that people are systematically programmed to fallacious processes of reasoning....is questionable. .... While all (normal) people are to be credited with the capacity to reason, they frequently do not exercise it well." Cohen attributes errors to "adventitious causes" and finds little interest in them. In his view human performance arises from an intrinsic human competence that is impeccably rational, but deviations may occur due to inattention, memory lapses, etc.

Johnson-Laird and Byrne articulate a view of rational thought that parses Cohen's distinction between  competence and performance. Stanovich says their view highlights the importance of the reflective mind. They hold that people are programmed to accept inferences as valid provided they have no mental model of the premises that contradict the inference. However, the search for contradictory models is "not governed by any systematic or comprehensive principles" (p. 165-6).