Rudimentary empathy in macaques' social decision-making

Abstract

Primates live in highly social environments, where prosocial behaviors promote social bonds and cohesion and contribute to group members' fitness. Despite a growing interest in the biological basis of nonhuman primates' social interactions, their underlying motivations remain a matter of debate. We report that macaque monkeys take into account the welfare of their peers when making behavioral choices bringing about pleasant or unpleasant outcomes to a monkey partner. Two macaques took turns in making decisions that could impact their own welfare or their partner's. Most monkeys were inclined to refrain from delivering a mildly aversive airpuff and to grant juice rewards to their partner. Choice consistency between these two types of outcome suggests that monkeys display coherent motivations in different social interactions. Furthermore, spontaneous affilitative group interactions in the home environment were mostly consistent with the measured social decisions, thus emphasizing the impact of preexisting social bonds on decision-making. Interestingly, unique behavioral markers predicted these decisions: benevolence was associated with enhanced mutual gaze and empathic eye blinking, whereas indifference or malevolence was associated with lower or suppressed such responses. Together our results suggest that prosocial decision-making is sustained by an intrinsic motivation for social affiliation and controlled through positive and negative vicarious reinforcements.

Keywords: emotions; eye blink; prosocial; social gaze; social neuroscience.

Fig. 1.

Task design. (A) Two monkeys (actor in black, partner in light brown) faced each other on either side of transparent touch panels on which visual stimuli were virtually projected. Both animals could observe the images and each other at all times. Tubes connected to solenoid devices allowed delivering the different outcomes. (B) Monkeys made social decisions regarding potential appetitive (offer J-S) and aversive (offer A-S) outcomes for the partner and nonsocial decisions regarding similar outcomes for self (offers J-nS and A-nS). For nonsocial decisions involving airpuffs and for social decisions, the actor was always rewarded with a drop of juice so as to maintain an adequate motivation level. (C) Typical trial sequence. A visual cue instructed the monkeys as to their role (actor or partner) in the current trial. The actor first touched this cue, triggering the appearance of two additional images. The monkey indicated its choice by touching one of these images. The unchosen image was then turned off and following a delay the partner’s and actor’s outcomes were delivered, preceded by unique 500-ms-long warning tones.

Fig. 2.

Social choice preferences. Positive values indicate prosocial decision-making, i.e., preference for granting juice (offer J-S) and avoiding airpuff (offer A-S) to the partner. Data are presented for M1-5 as actor and M1-4 and M6-7 as partner. Mean preference scores for each monkey pair across all experimental sessions were computed as [choice1/(choice1 + choice2)] − 0.5. *P < 0.05, **P < 0.01, ***P < 0.001. Significant preference for one of the two options (Wilcoxon signed-rank test); error bars represent SEM.

Fig. S1.

Choice preferences in nonsocial decisions and control experiments. (A) Nonsocial decisions. Positive values indicate rational decision-making, i.e., preference for granting juice (offer J-nS) and avoiding airpuff (offer A-nS) to self. (B) Control experiments in which monkey actors performed the choice task facing an empty nonhuman primate chair instead of an actual partner. Nonsocial trials were unaffected as monkeys chose to procure themselves juice and to avoid airpuff. On pseudosocial trials, monkeys chose indifferently between outcomes to nobody and to the empty-chair. Bar represented the average of 16 sessions performed by monkeys M1–M3. Mean preference scores for each monkey pair across all experimental sessions were computed as [choice1/(choice1 + choice2)] − 0.5. *P < 0.05, **P < 0.01, ***P < 0.001. Significant preference for one of the two options (Wilcoxon signed-rank test).

Fig. S2.

Graphical representation of monkeys’ decisions and behaviors predictions. All of these models considered the actors’ and partners’ identity as a random predictor. Individual data points are mean values on a given testing session. Dotted lines represent the SE of regression Plots in A–C show the best model account for different social and nonsocial decisions. (A) Prediction of the actor’s social decision in offer J-S. The model considered the choice preference in offer A-S and the anticipatory mutual gaze rate in offer J-S as predictor. (B) Prediction of the actor’s social decision in offer A-S. The model considered choice preference in offer J-S, anticipatory blink rate in offer A-nS and the interaction between actor’s and partner’s anticipatory blinks in offer A-S as predictors. (C) Prediction of the actor’s nonsocial decision in offer A-nS. The model considered the anticipatory blink rate in offer A-nS as predictor. (D) Correlation between actor’s blink in offer A-S and mutual gaze rate in offer J-S. (E) Correlation between actor’s and partner’s anticipatory blink rate in offer A-S.

Fig. 3.

Mutual looking rate for the different social decider profiles for juice to nobody and juice to partner choices. Thick lines below the plot indicate significant pairwise differences (permutation test, P < 0.05); shading overlays on the traces represent SEM. The number of sessions considered for benevolent, indifferent, and malevolent actors is, respectively, equal to 97, 30, and 37 sessions.

Fig. S3.

Gaze behavior of actor and partner during social decisions. (A) Pie charts of mean gaze distribution of all monkey pairs computed during the delay interval that followed decisions to grant juice to nobody or to partner (n = 164). Hatched and dotted arcs represent total social gaze of the actor and partner, respectively. (B) Results of single trial analysis of gaze coincidence. The permutation-based statistical procedure (see main text) allowed to classify mutual gaze rate in each trial as <, =, or > chance intersection of the two monkeys’ gaze within the face ROI.

Fig. 4.

Eye blink behavior when experiencing and when observing an airpuff. Net effect on blink rate of observing the partner receiving an airpuff, computed as the difference [airpuff to partner – airpuff to nobody] for benevolent, indifferent, and malevolent actors (n = 97, 30, and 37 sessions, respectively). Shading overlays on the traces represent SEM. Bar graphs show mean blink rate differences computed 300 ms before and after airpuff delivery. Thick lines below the plot indicate significant pairwise differences (permutation test, P < 0.05). *Significant pairwise difference (Wilcoxon rank sum test, P < 0.05); error bars represent SEM.

Fig. S4.

Examples of monkeys’ eyes blink behavior when experiencing and when observing an airpuff. (A) Time course of mean eye blink rate in the outcome phase for benevolent actor M1 in the M1–M3 dyad, when the actor, the partner, or nobody received an airpuff. (B) Time course of mean eye blink rate for the malevolent actor M3. Both plots show that the strongest eye blink response is to an airpuff to self. Both actors also reacted with an increase in blink rate to the delivery of an airpuff to nobody. However, relative to the latter blink response, M1 anticipated and overreacted to an airpuff to the partner, whereas M3 underreacted to this event. Thick lines below the plots indicate significant pairwise differences (permutation test, P < 0.05); shading overlays on the traces represent SEM.

Fig. 5.

Social affiliation structure of long-tailed macaques and relation to social decision-making tendencies. (A) Social grooming network. Circle diameter is proportional to the total percentage of time (in a 3-h cycle, 10 recording sessions) spent in allo-grooming activity by each monkey. Arrows show directionality of grooming and numbers, the percentage of time dedicated to grooming a given partner. (B) Schematic of the social decision network based on data from Fig. 2, allowing direct comparisons between spontaneous social behavior and decision tendencies. (C) Correlation between social decisions regarding aversive outcomes (offer A-S) and mean social grooming time in the home environment (R = 0.77, P < 0.05). (D) Correlation between mutual gaze rate of monkeys in the social decision task and mean social grooming time in the home environment (R = 0.83, P < 0.05). Mutual gaze (MG) was normalized using the ratio: MG_{[juice to partner]}/(MG_{[juice to partner]} + MG_{[juice to nobody]}). The dataset includes 10 recording sessions conducted during the same period that social decisions data were collected. Each dot in the correlations represents a monkey dyad.

Fig. S5.

Social affiliation structure of the minicolony of long-tailed macaques: proximity network. Circle diameter is proportional to the total percentage of time (in a 3-h cycle) spent in passive proximity to other monkeys [net proximity computed as (total proximity) – (total given grooming + total received grooming)]. Line segments and numbers represent the breakdown of this variable as a function of monkey dyad.