Getting the Description Right

If Jane prefers a over b, and b over c, we normally expect her to prefer a over c too. Such transitivity of behavior is central to our normative assessment of the rationality of choices. However, experimental evidence shows recurring and stable violations of transitivity in choices over time. Are we ultimately incapable of satisfying our own criteria of rationality? In the present article, I shall argue against this conclusion. In ultimately choosing c rather than a, Jane might have been guided in her decision-making process by rational considerations (yet) unknown to the observer. New information or changed circumstances could have influenced the initial option c (and hence its assessment) such that it can be meaningfully refined and re-described as c*. As c* is preferred over a the transitivity of the preference ordering is preserved. Such a strategy, however, has been criticized by Paul Anand (1990). According to him, to re-describe prima facie intransitive behavior in transitive terms is necessarily ad hoc, arbitrary, and self-defeating. In the present article this conclusion is contested as Anand bases his reasoning on two questionable assumptions: a division between descriptive and normative decision theory, and observable choice being the sole criterion for describing preferences.