Normative and descriptive models of decision making: time discounting and risk sensitivity.

Abstract

The task of evolutionary psychologists is to produce precise predictions about psychological mechanisms using adaptationist thinking. This can be done combining normative models derived from evolutionary hypotheses with descriptive regularities across species found by experimental psychologists and behavioural ecologists. I discuss two examples. In temporal discounting, a normative model (exponential) fails while a descriptive one (hyperbolic) fits both human and non-human data. In non-humans hyperbolic discounting coincides with rate of gain maximization in repetitive choices. Humans may discount hyperbolically in non-repetitive choices because they treat them as a repetitive rate-maximizing problem. In risk sensitivity, a theory derived from fitness considerations produces inconclusive results in non-humans, but succeeds in predicting human risk proneness and risk aversion for both the amount and delay of reward in a computer game. Strikingly, and in contrast with the existing literature, risk aversion for delay occurs as predicted. The predictions of risk aversion for delay may fail in many animal experiments because the manipulations of the utility function are not appropriate. In temporal discounting animal experiments help the interpretation of human results, while in risk sensitivity studies human results help the analysis of non-human data.