Normative theories may be superfluous in certain specific cases discussed in Elqayam & Evans' (E&E's) stimulating and informative target article. But the fact that some people may be tempted by a fallacious is-ought inference is not sufficient reason for abandoning normative theories in all cases.

A widely held position in philosophy of science is that all scientific observations are theory laden (e.g., Kuhn Reference Kuhn1962), partly because investigators' theories influence what data they consider it worthwhile to collect. Further, when interpreting behavior, we tend to apply Davidson's (Reference Davidson1973) “principle of charity,” originally intended for interpreting sentences but more widely applicable. This involves assuming that people are generally rational and interpreting their behavior in that light. Thus, behavioral researchers implicitly draw on a normative theory.

Normative theories have also been useful in generating powerful descriptive theories, using a style of theorizing that does not fall foul of the is-ought fallacy. For example, subjective expected utility (SEU) theory is evidently normative, specifying what choices rational agents ought to make in order to satisfy their own desires. By appending to SEU a hypothesis of weak rationality, according to which people try to do the best for themselves in any circumstances that arise, we derive the descriptive principle of methodological individualism (Weber Reference Weber, Fischoff, Rothe and Wittich1922/1978, Ch. 1), a mainstay of the contemporary social sciences, reflected in rational choice theory and behavioral game theory, both direct descendants of SEU theory (Elster Reference Elster1989).

It was Savage (Reference Savage1972), not von Neumann and Morgenstern (Reference von Neumann and Morgenstern1947), who introduced a normative interpretation into SEU theory: “One idea now held by me that I think von Neumann and Morgenstern do not explicitly support, and that so far as I know they might not wish to have attributed to them, is the normative interpretation of the theory” (Savage Reference Savage1972, p. 97). Normative considerations seem quite natural and useful in judgment and decision making research. If you invite people to make a snap choice between 96×69 cents and 87×78 cents, most will choose 96×69 cents; but if you point out that 96×69=6,624, whereas 87×78=6,786, they will swiftly change their minds, if allowed to (Binmore Reference Binmore2009, pp. 22–23). This illustrates two important facts: first, people generally try to act rationally in the sense of maximizing their expected utilities; but second, they are limited by bounded rationality and are prone to error.

A domain of higher mental processing within which normative considerations seem quite unavoidable is moral reasoning. Evaluative “ought” questions are the very essence of moral reasoning. We are currently engaged in a research project investigating judgments as to whether it is morally acceptable to sacrifice one life to save five in the following famous Trolley problem (Foot Reference Foot1967):

Most people (90%, according to Hauser Reference Hauser2007) say yes; but Thomson's (Reference Thomson1976) closely related Footbridge problem elicits very different responses:

Most people (90%, according to Hauser Reference Hauser2007) say no. Why do most people consider it morally acceptable to sacrifice one life to save five in one problem but not the other?

From the perspective of cognitive psychology, the differences in responses to the two problems are reminiscent of the classic demonstration of a framing effect, in which two different descriptions of a problem involving a certain number of lives at risk elicit difference responses (Tversky & Kahneman Reference Tversky and Kahneman1981). Greene (Reference Greene2007) has argued that the Footbridge problem tends to engage our emotions to a greater extent than the Trolley problem, and that our emotions deflect us from the utilitarian judgment in the Footbridge problem.

Some philosophers have argued that there are morally relevant distinctions between the two problems. Foot (Reference Foot1967) drew attention to the doctrine of double effect, first suggested by the medieval scholastic philosopher Thomas Aquinas, according to which harm is acceptable if it occurs as a foreseen but unintended consequence of an action serving a greater good, as in the Trolley problem, but not as a means to an end, as in the Footbridge problem. Quinn (Reference Quinn1989) argued that the difference in responses is justified by the doctrine of doing and allowing, according to which pushing the man off the bridge is unacceptable because the harm results from intentional action, rather than from an omission, or failure to act. However, some psychologists have argued that the distinction between omission and commission is the result of a psychological bias (e.g. Ritov & Baron Reference Ritov and Baron1992; but see DeScioli et al., in press).

Others have proposed a universal moral grammar or UMG (Hauser Reference Hauser2007; Mikhail Reference Mikhail2007), according to which normative moral principles, such as a prohibition of killing, are arrived at by an unconscious computational model, analogous to Chomsky's (Reference Chomsky1995) universal grammar for human languages, this grammar being in accord with the doctrine of double effect.

How could moral problems possibly be freed from normative considerations? Perhaps some theories of higher mental processing can manage without such considerations, but it is hard to see how this could (or why it should) be generalized to all domains of research.