The Importance of What Might Have Been

Expected utility theory is often presented as a normative theory, resting on a few simple axioms that no reasonable person would wish to violate. Arrow (1972, p.21), for example, describes one version of these axioms as ‘a set of assumptions designed to characterise reasonable behaviour under conditions of uncertainty’ he describes particular axioms in such terms as ‘a hallmark of rationality’ (p. 22) and ‘highly acceptable’ (p.24). The claim that these axioms have a special normative status is sometimes used as a second line of defence for expected utility theory in cases where that theory does not predict observed behaviour well. It can be suggested that when people act contrary to the theory, they do so in error. As Morgenstern (1979, p. 180) puts it, ‘if people deviate from the theory, an explanation of the theory and of their deviation will cause them to readjust their behaviour’. Even some of the proponents of alternative positive theories seem to have been prepared to accept some of the normative claims made on behalf of expected utility theory. Kahneman and Tversky (1979, p. 277), for example, regard the behaviour that their theory predicts as #x2018;normatively unacceptable’, and Machina (1982, p. 277) notes the ‘normative appeal’ of expected utility theory before presenting his own theory.