Adult age differences in the integration of values for self and other

Abstract

Previous research suggests that older adults may display more prosocial behavior than younger adults. However, recent meta-analyses indicate that effects are heterogeneous, may be small, and are influenced by how prosociality is measured. Further, the precise cognitive and computational factors contributing to age-related differences in prosocial behavior remain largely unknown. In this study, we utilized a modified dictator game to combine a value-based decision framework with Bayesian hierarchical drift-diffusion modeling to investigate prosocial decision-making in a sample of younger (n = 63) and older adults (n = 48). We observed differences in how older and younger individuals incorporate information corresponding to potential gains for themselves (self) and another person (other) to reach a (potentially prosocial) decision. Younger adults integrated values for benefits for themselves and others in the decision-making process and demonstrated increased decision-making efficiency by effectively integrating both sources of information. In contrast, older adults showed improved decision-making efficiency when solely considering values for self and others separately. Interestingly, individual differences in the capacity of inhibitory control in older adults moderated the observed age effects: older adults with stronger inhibitory control abilities made decisions based on the integrated information of benefits for themselves and others. Together, these findings offer new insights into the behavioral and computational mechanisms influencing age effects in prosocial decision-making.

Keywords: Adult lifespan; Cognitive functioning; Drift-diffusion modeling; Inhibitory control; Prosocial behavior.

Fig. 1

Modified dictator game to assess the integration of values for self and others and schematic illustration of the drift-diffusion model. Note. (A) Illustration of the social choice task (modified dictator game based on). In each trial, participants accepted or rejected a varying monetary offer that affected payoffs for themselves (self) and another player (other) versus a constant default distribution of money (5€ for both players) using a 4-point scale (1 = strong no, 2 = weak no, 3 = weak yes, 4 = strong yes - toward the varying monetary offer; scale direction counterbalanced across participants). Outcomes were implemented in a probabilistic manner such that choices were implemented in 80% of trials or reversed in 20% of trials. (B) Nine proposed offer types, representing different monetary distributions between the participant (€self) and its partner (€other) and post-hoc classification of the offer types into self-serving, other-serving, and rational condition. Every offer type was shown 20 times (using a random jitter of up to 0.40€ to reduce habituation), randomly intermixed across the task. (C) The drift-diffusion model is used to conceptualize latent variables underlying the decision process, which are defined by the noisy accumulation of information (squiggly line). The non-decision time (ndt) controls for both sensory and motor-related processes. The v-parameter (drift rate) represents the speed of the accumulation process. The a-parameter (boundary separation) describes the distance between the two boundaries. The z-parameter indicates the initial bias of the accumulation process with respect to the two boundaries. A decision is made the moment one of the two boundaries is crossed (i.e., enough evidence accumulated). In the example trial (Fig. 1A), choosing “1” or “2” would correspond to rejecting the offer (lower boundary), whereas choosing “3” or “4” would correspond to accepting the offer (upper boundary). Upper boundary = accepting the offer (i.e., rejecting the default distribution), lower boundary = rejecting the offer (i.e., accepting the default distribution).

Fig. 2

Adult age group differences in choice behavior and RTs with respect to the interaction of €self × €other. Note. We identified significant 3-way interactions of age group × €self × €other (both for choices and RTs). The shaded areas around the lines represent the 95% confidence intervals. Note that there was no condition where both values (i.e., €self and €other) fell below 5€ (A and B) Predicted frequency of accepting the alternative offer. Here, the y-axis corresponds to the four choice options reflecting the extent to which the alternative offer is preferred (1 = strong no, 2 = weak no, 3 = weak yes, 4 = strong yes). Younger adults’ choices were modulated by an interaction of €self × €other. Higher values increased the probability of choosing the proposed offer, i.e., choosing either 3 or 4. Older adults’ tendency to accept an offer as a function of own benefits was not significantly modified by payoffs for the other: The probability of choosing the proposed offer increased with higher values of €self (regardless of €other). (C) Younger adults’ RTs were modulated by an interaction of €self × €other. They showed faster RTs when both values (€self and €other) were either high or low, and slower RTs when values diverged (i.e., one was high and the other low). (D) Older adults’ RTs were only modulated by €self, irrespective of €other. Faster RTs were observed for higher values of €self (regardless of €other).

Fig. 3

Significant moderation effect of inhibitory control on older adults’ choices behavior (mean inhibitory control – 1 standard deviation vs. mean inhibitory control vs. mean inhibitory control + 1 standard deviation). Here, the y-axis corresponds to the four choice options reflecting the extent to which the alternative offer is preferred (1 = strong no, 2 = weak no, 3 = weak yes, 4 = strong yes). Note. Significant interaction of €self x €other x inhibition ability. Note that we split up the sample in low vs. medium vs. high inhibitory control for illustration purposes only. Inhibition ability entered the model as a continuous predictor. In older adults with high inhibition abilities (right panel, mean inhibitory control + 1 SD), choices to accept the proposed offer were impacted by a multiplicative effect of both €self and €other. In older adults with medium inhibition abilities (middle panel, mean inhibitory control), choices to accept the proposed offer were impacted by values for €self and €other, but not by the interaction of €self x €other. Older adults with low inhibition abilities (left panel) were only influenced by the values for self (€self). The shaded areas around the lines represent the 95% confidence intervals. SD = standard deviation, OA = older adults.

Fig. 4

Differential effects of €self, €other and their interaction on drift rate in younger and older adults. Note. Bayesian posterior densities of drift rates (v-parameter) estimated from hierarchical regression drift-diffusion models and how they varied as a function of payoff values €self and €other. For an effect to be meaningful, 95% of the distribution has to be on the left or right of zero (dotted vertical grey line). (A) Main effects of €self and €other on the drift rate, separately for younger (YA; purple) and older adults (OA; green). All main effects indicated a positive impact of €self and €other on drift rates in YA and OA. (B) €self × €other interaction effect on the drift rates in YA and OA. Only younger adults demonstrated a positive interaction effect of €self × €other on drift rates.