Like cognitive function, decision making across the life span shows profound age-related changes

Abstract

It has long been known that human cognitive function improves through young adulthood and then declines across the later life span. Here we examined how decision-making function changes across the life span by measuring risk and ambiguity attitudes in the gain and loss domains, as well as choice consistency, in an urban cohort ranging in age from 12 to 90 y. We identified several important age-related patterns in decision making under uncertainty: First, we found that healthy elders between the ages of 65 and 90 were strikingly inconsistent in their choices compared with younger subjects. Just as elders show profound declines in cognitive function, they also show profound declines in choice rationality compared with their younger peers. Second, we found that the widely documented phenomenon of ambiguity aversion is specific to the gain domain and does not occur in the loss domain, except for a slight effect in older adults. Finally, extending an earlier report by our group, we found that risk attitudes across the life span show an inverted U-shaped function; both elders and adolescents are more risk-averse than their midlife counterparts. Taken together, these characterizations of decision-making function across the life span in this urban cohort strengthen the conclusions of previous reports suggesting a profound impact of aging on cognitive function in this domain.

Fig. 1.

Experimental design. (A) An example of a risky loss trial. The subject has a choice between losing $5 and equal chances of losing $8 or nothing. (B) Example of an ambiguous gain trial. The subject has a choice between a gain of $5 and a lottery that pays $20 with a probability that is not precisely known. (C) All ambiguous lottery bags used in the experiment.

Fig. 2.

(A) Expected earnings calculated separately for each age group as though every risk counted for payment. The horizontal green line indicates expected earnings for a risk- and ambiguity-neutral chooser. (B) Frequency of violations of first-order stochastic dominance. (C) Proportion of choice situations in which subjects changed their mind, that is, out of the same choice set they chose both a safe option and a risky option at least once. Graphs show means +/−1 SE.

Fig. 3.

Maximum-likelihood estimates of (A) risk and (B) ambiguity in the gain and loss domains. The green lines coincide with risk neutrality (α = 1) in A and ambiguity neutrality (β = 0) in B. Data behind hatched bars are presented for comparison and were originally reported in ref. . Graphs show means +/− 1 SE.

Fig. 4.

Relationship of (A) risk and (B) ambiguity attitudes between the gain and loss domains. (A) Scatterplot of individual risk attitudes (calculated as the proportion of risky choices) in the gain versus loss domains. (B) Scatterplot of individual ambiguity attitudes (calculated as the proportion of ambiguous lottery choices corrected for risk attitude) in the gain versus loss domains. Green lines indicate risk and ambiguity neutrality.