Experimental tasks performed on separate days. (A) Social Perception Task. Participants made rapid yes/no judgments regarding others’ perceived deservingness (merit blocks, top row), need (need blocks, middle row), or factual inferences (control blocks, bottom row) while their brain responses were measured using fMRI. (B) Altruism Task. On each trial, participants accepted or rejected a monetary offer that affected the payoffs for themselves (“You”) and one of three partners (choice screen; displayed offer vs. constant default of $20 for both). The three partners (identified via colored geometric shapes) differed in their associated merit (merit cue: high/low/unknown) based on partner behavior in a separate exchange game played before the altruism task. Partner’s need (need cue: high/low) was manipulated on a trial-by-trial basis, indicated by the likelihood of a painful cold pressor task (CPT) for the partner after the altruism task (high need: 80±4%; low need: 20±4%). Partners could buy out of the painful CPT using funds from one randomly selected trial at the end of the altruism task. Participants were informed that they could help their partners to avoid the painful CPT by making generous choices.

(A) Distribution of participant-specific model-based estimates of perceptual sensitivity and (B) perceptual bias for each inference condition in the social perception task (merit, need, and control). (C) Need and merit inferences activate the mentalizing network to a similar extent (red illustrates brain regions activated for [merit - control], green illustrates brain regions identified for [need - control], and yellow indicates overlap; both contrast maps thresholded at p < 0.001 at the voxel level, FWE corrected at the cluster level at p < 0.05; R = right hemisphere). (D) Activity in the right temporoparietal junction (TPJ) during merit perceptions (merit - control) reflects individual differences in merit sensitivity scores estimated in the computational model of social perception.

Conjunction of neural activations during social need and merit inferences in the social perception task.

(A) Partner’s merit and need altered generosity in the altruism task. High (vs. low) need contexts elicited more generosity (p < 0.001). Compared to a low merit partner (circle), generosity was enhanced towards a high merit (diamond, p < 0.001) and an unknown merit partner (triangle, p < 0.001). (B) Condition-specific attribute weights (wself, wother, wfairness) for low, unknown, and high merit partner contexts. (C) Condition-specific attribute weights for low and high need contexts. All p’s ≤ 0.01, FDR corrected, for the six pair-wise comparisons of changes in attribute weights (high vs. low merit/need). Dots represent participant-specific estimates from the computational model of altruistic choice (see Supplemental Note S3 for detail); black lines illustrate the estimated means and 95% confidence intervals.

(A) Variance in individuals’ general sensitivity to others’ merit (social perception task) is positively linked with merit-related changes in generosity, (B) other-regard (wother), and (C) self-regard (wself) during altruistic choice for high vs. low merit partners. Higher values on the x-axes indicate increased generosity/other-regard and decreased self-interest when interacting with a high vs. low merit partner in the altruism task.