Social Cognitive and Affective Neuroscience

Fairness decisions often integrate affective responses within a social context, yet emotion regulation in this literature has been largely studied as a self-directed process rather than an interpersonal one. We examined how individual differences in other-directed emotion regulation--measured with the Emotion Regulation of Others and Self (EROS) scale--relate to behavioral and neural responses during fairness decisions in 138 adults completing a variant of the Ultimatum Game with human and computer partners during fMRI. Behaviorally, participants who more strongly endorsed worsening others emotions rejected unfair offers more frequently, and this tendency interacted with offer fairness to amplify rejection of unfair offers. At the neural level, the left anterior insula tracked offer unfairness more strongly in social versus nonsocial contexts, consistent with sociality modulating the neural encoding of fairness. Right dlPFC activation during socially unfair offers was greater among individuals who preferred to improve others emotions. Connectivity analyses revealed that social fairness sensitivity predicted stronger amygdala-orbitofrontal and amygdala-dmPFC coupling; the latter was further amplified among individuals higher in other-directed emotion worsening. Together, these findings identify interpersonal emotion regulation as an understudied source of variation in the affective and prefrontal systems supporting fairness-based social decisions.

We investigated the neural substrates underlying the brain network shared by social exclusion and physiological needs, both viewed as instances of deprivation. Using activation likelihood estimation (ALE) meta-analyses, we examined brain activation patterns from studies where participants perceived food/water while hungry/thirsty and social interactions while experiencing exclusion. This analysis revealed overlapping activation in the mid-posterior insula, caudate head and ventral anterior cingulate cortex (ACC), as regions consistently engaged when perceiving relevant stimuli across both physiological and social deprivation. Furthermore, we found high spatial correlation between this shared network and the distribution of dopamine receptors, and we identified a significant positive correlation with the 5HT4 among serotonin receptors. Our findings suggest that perceiving deprivation-related stimuli activates brain regions and neurotransmitters involved in aversive affect and goal-directed behavior. These results highlight a neural bridge linking basic physiological drives with complex social needs, offering new insights into the neurobiological architecture of human affective states--while representing just one part of a much larger puzzle.

Recent studies suggest that individual differences in empathic concern may be mediated by continuous interactions between self-other resonance and cognitive control networks. To test this hypothesis, we used machine learning to examine whether resting fMRI connectivity (i.e. the degree of synchronous BOLD activity across multiple cortical areas in the absence of task demands) of resonance and control networks could predict trait empathy (n=58). Indeed, resonance and control networks interconnectivity predicted empathic concern. Empathic concern was also predicted by connectivity within the somatomotor network. In light of numerous reported sex differences in empathy, we controlled for biological sex and also studied separately what aspect of these features could predict participants sex. Sex was best predicted by the interconnectivity of the visual system with the resonance, somatomotor, and cingulo-opercular network, as well as the somatomotor-control network connectivity. These findings confirm that variation in empathic responses to others reflects characteristic network properties detectable regardless of task demands. Furthermore, network properties of the visual system may be a locus of sex differences previously unaccounted for in empathy research. Finally, these findings suggest that it may be possible to assess empathic predispositions in individuals without needing to perform conventional empathy assessments.

Trust and fairness sustain cooperation, but social neuroscience has largely studied violations of these social norms in isolation. We asked whether neural sensitivity to betrayal carries forward into the evaluation of unfairness in a separate social exchange. In 132 adults who completed fMRI during the Trust Game and Ultimatum Game, betrayal and unfairness both recruited overlapping voxels within left anterior insula. Yet, individual differences in regional activation were uncorrelated across tasks, arguing against a simple shared-response account. Instead, cross-task structure emerged in connectivity: stronger anterior insula responses to social versus nonsocial betrayal predicted weaker anterior insula-ventromedial prefrontal coupling during social versus nonsocial fairness evaluation. These findings suggest that betrayal and unfairness are linked less by identical regional responses than by how salience signals engage valuation circuitry across contexts. By bridging two canonical social exchange tasks within the same individuals, this study reveals a network-level architecture for generalizing sensitivity to interpersonal norm violations.

As avatars become more commonplace, understanding how the brain processes emotional expressions in virtual faces is critical. We compared behavioral and neural responses to real and virtual faces expressing seven emotions (anger, disgust, fear, joy, sadness, surprise, neutral). In Experiment 1 (n=61), participants rated the similarity between paired faces. Expressions conveying the same emotion were rated as highly similar across face types, whereas mismatched emotions yielded substantially lower similarity ratings, indicating perceived emotional meaning was preserved despite differences in face realism. In Experiment 2 (n=91), functional near-infrared spectroscopy was used to measure brain activity while participants viewed the same stimuli. General-linear-model analyses revealed greater activation limited to visual areas for 1) virtual faces and 2) surprise and neutral expressions. Functional connectivity analyses, however, revealed network level differences between face type and emotion across the brain. Real faces elicited stronger connectivity patterns across frontal, central-temporal, and parietal regions, whereas high-arousal emotions (fear, anger, and joy) were associated with broader network engagement than other expressions. Our results suggest face-type processing occur in early visual areas, and despite perceptual similarity, different emotions on real and virtual faces are associated with distinct patterns of network level connectivity across the brain.

Social decision-making relies on dynamic affect-cognition interactions across distributed brain networks, yet how incidental positive affect modulates these mechanisms at a millisecond timescale remains unclear. This study investigated the impact of music-induced positive emotion on the neural dynamics of decision-making in the Ultimatum Game. Fifty-six participants were assigned to either a happy music group or an active control (rain sound) group. Fifty-six participants were assigned to either a happy music group or an active control (rain sound) group, while electroencephalography was recorded to capture rapid neural dynamics. Behaviorally, happy music accelerated reaction times (RTs) and decoupled the ERP-RT correlations observed in the control condition. Neurally, positive affect amplified event-related potential amplitudes during early conflict detection (220-280 ms) and late valuation (520-560 ms) stages. Multivariate pattern analysis further revealed that happy music enhanced the neural separability and temporal stability of decision states (accept vs. reject). Moreover, using support vector regression based on functional network features, we found that decision acceptance rates were predicted with significantly higher accuracy in the happy music group (R = 0.60) compared to controls (R = 0.41). Crucially, feature weight analysis indicated a topological shift in decision strategy: while the control group relied on frontal-central edges (implicating executive control), the happy music group was characterized by central-temporal connections (suggesting integrative processing). Collectively, these findings provide novel evidence that incidental emotion intervenes at the millisecond timescale to bias social choices, offering a dynamic network-based account of the affect-cognition interaction.

This study investigated the neural mechanisms involved in feelings of interpersonal guilt and responsibility evoked by social decisions. In two studies (one during fMRI), participants repeatedly chose between safe and risky monetary outcomes in social contexts. Across conditions, each participant chose for both themselves and a partner (Social condition), or the partner chose for both themselves and the participant (Partner condition), or the participant chose just for themselves (Solo condition, control). If the risky option was chosen in the Social or Partner condition, participant and partner could each receive either the high or the low outcome of a lottery with 50% probability, independently of each other. Participants were shown the outcomes for themselves and for their partner on each trial, and reported their momentary happiness every few trials. As expected, participant happiness decreased following both low lottery outcomes for themselves and for the partner. Crucially, happiness decreases following low outcomes for the partner were larger when the participant rather than their partner had made the choice, which fits an operational definition of guilt. This guilt effect was associated with BOLD signal increase in the left anterior insula. Connectivity between this region and the right inferior frontal gyrus varied depending on choice and experimental condition, suggesting that this part of prefrontal cortex is sensitive to guilt-related information during social choices. Variations in happiness were well explained by computational models based on participants and partners rewards and reward prediction errors. A model-based analysis revealed a left superior temporal sulcus cluster that tracked partner reward prediction errors that followed participant choices. Our findings identify neural mechanisms of guilt and social responsibility during social decisions under risk.

Updating impressions of others is essential to navigating social life. As we get to know an individual, we update our impressions of them in accordance with new information. How the brain dynamically revises impressions in real time under naturalistic conditions remains unclear. Here, we address this question using functional magnetic resonance imaging (fMRI) and natural language analysis in a naturalistic social cognition paradigm. Across 10 runs, participants viewed a character-driven TV episode, reported moments of insight, and described their impressions of the characters. Results reveal that impressions progressively evolved over time. Individuals with more similar existing impressions exhibit greater neural synchrony during movie-watching, which predicts convergence in post-movie impressions. Neural synchrony in the right superior temporal sulcus (STS) mediates the influence of initial similarity on later alignment. Insight moments accompany neural pattern shifts widespread across cortical regions, including the temporoparietal junction (TPJ), dorsomedial prefrontal cortex (dmPFC) and STS, and the magnitude of the shifts tracks the degree of impression updating. Specifically, distinct forms of insight selectively update complementary components of impressions: character insight updates person-centered representations, whereas non-character insight shapes social-event structure. Together, these findings show that people update their impressions of others through dynamic shifts in distributed brain activity patterns at moments of insight, providing a novel and ecologically grounded neural account of social cognition. Significance statementPeople do not simply form first impressions and keep them. As we learn new things about others, our judgments change--sometimes gradually, sometimes in a flash of insight. By tracking brain activity while people watched a TV show and described how their impressions of the characters evolved, we found that people who saw a character in more similar ways also show more similar brain responses when watching them onscreen, which in turn leads to convergence in subsequent impressions. Such updating is closely tied to "aha" moments, when a person is suddenly understood in a new way. These moments were marked by rapid shifts in the brain representational patterns, and larger shifts predicted greater changes in impressions. These findings offer a more naturalistic account of how the brain revises our understanding of other people in everyday life.

Anticipation of trust from a partner with high social closeness generates the preconditions for cooperative and reciprocal interactions to occur. However, if there is uncertainty in the interaction because the partner is a stranger or because the person has mistrusted us on another occasion, an aversive experience is generated. Hence, low social closeness or mistrust makes people keep track of the others behavior. In case of experiencing deviations from social norms, the person will monitor the intentions of the partner and update their priors regarding their social preferences. The anterior insula (AIns) seems to be sensitive to these social norm violations in functional magnetic resonance imaging (fMRI) experiments, however, the monitoring of partners with different levels of social closeness has not been investigated. In our study we wanted to find the brain regions related to (dis)trust anticipation from partners who differ in their level of social closeness. For this, we designed an experiment in which participants played an economic decision game with three people (trustors): A computer, a stranger, and a real friend. We covertly manipulated their decisions in the game so that they unexpectedly distrusted our participants. Using task-fMRI, our whole-brain analysis found that the AIns was active during the anticipation of the decisions from human partners (humans vs computer), but not during anticipation between high and low social closeness (friend vs stranger). However, using a psychophysiological interaction analysis, we found increases in functional coupling between the AIns and regions in the "mentalizing" network (such as temporal regions and parieto-occipital cortices) during trust anticipation between a high versus low social closeness partner. These results suggest that there may be a modulation of the AIns activity, specifically for high social closeness trustors, by regions that encode the intentions underlying the truster behavior.

Emotional experiences are never static but continuously evolve in response to internal and external contexts. Little is known about how neural patterns change as a function of these experiences, particularly in response to complex, real-world stimuli. This study aimed to identify generalizable neural patterns as individuals collectively engage and disengage from emotions dynamically. To do so, we analyzed functional magnetic resonance imaging (fMRI) along with subjective emotional annotations from two independent studies as individuals watched negative and neutral movie clips. We used predictive modeling to test if a model trained to predict a group emotional signature response in one study generalizes to the other study and vice versa. Disengagement patterns generalized specifically across intense clips. They were supported by connections within and between the sensorimotor and salience networks, maybe reflecting the processing of feeling states as individuals regulate their emotions. Prediction success for the engagement signature was mixed, but primarily linked to connections within the visual and between the visual and dorsal attention networks, maybe supporting visual attention orienting as emotions intensify. This work offers potential pathways for identifying generalizable neural patterns contributing to future affective research and clinical applications aiming to better understand dynamic emotional responses to naturalistic stimuli.

The present functional MRI study addressed how the brain maps different aspects of social information. We focused on two key dimensions of social knowledge: affect and likableness. Thirty participants were presented with audio definitions, half referring to affective (e.g. empathetic) and half to non-affective concepts (e.g. intelligent). Orthogonally, half of the concepts were highly likable (e.g. sincere) and half were socially undesirable (e.g. liar). We used a support vector machine to delineate how both concept dimensions are represented in a set of 9 a priori brain regions defined from previous meta-analyses on semantic and social cognition. We show that average decoding in semantic regions (e.g. lateral temporal lobe, inferior frontal gyrus, and precuneus) outperformed social ROIs (e.g. insula and anterior cingulate), with the lateral temporal lobe containing the highest amount of information about the affect and likableness of social concepts. We also found that the insula had a bias towards affect while the likableness dimension was better represented in anterior cingulate cortex. Our results do not support a modular view of social knowledge representation. They rather indicate that the brain representation of social concepts implicates a distributed network of regions that involves domain-specific social cognitive systems, but with a greater dependence on language-semantic processing.

Successfully making connections with others is crucial to navigating the social world and general well-being, yet little is known about connection formation and its neurocognitive underpinnings. Increasingly, neuroscientists use interpersonal neural synchrony within the default mode network (DMN) to measure when two or more people subjectively experience something in similar ways. DMN synchrony as seeing eye-to-eye is typically observed when multiple people are passively observing the same stimulus. In this study, we tested whether the same DMN synchrony as seeing eye-to-eye pattern holds during social interactions. We conducted a between-subject naturalistic experiment with 70 pairs of strangers engaged in either shallow or deep conversations while brain activity was measured with functional near infrared spectroscopy (fNIRS). Stranger dyads successfully formed connections, as indicated by composite connection scores. Replicating Kardas et al. (2021), those in the deep conversation condition felt more connected than those in the shallow conversation condition. DMN neural synchrony significantly predicted self-reported connection, with synchrony in the DMN subregions of medial prefrontal cortex (mPFC) and right temporoparietal junction (TPJ) each correlating significantly with connection. Using machine learning classification, we distinguished high-versus low-connection dyads based on DMN neural synchrony and the perceived depth of conversation with 64.5% accuracy across 1,000 iterations. This effect was primarily carried by right TPJ, which alone classified connection strength at 62.6% accuracy. We consider implications related to the growing loneliness crisis and the importance of understanding how social connections can be formed and fostered in an era of increased social isolation. Significance StatementThe current loneliness epidemic has serious consequences on health and well-being. Forming interpersonal connections is crucial for alleviating loneliness, yet little is known about its neural basis. Neural synchrony, a potential biological marker of people being on the same page, may be an indicator of social connection. We recorded brain activity as strangers engaged in a get-to-know-you conversation and found that neural synchrony--specifically within the default mode network (DMN) and subregions including medial prefrontal cortex (mPFC), and right temporoparietal junction (TPJ)--predicted self-reported connection. Machine learning accurately classified high- and low-connection dyads based on DMN synchrony and perceived conversation depth. These findings suggest that deeper conversations, though more effortful, may foster stronger social bonds with measurable neural correlates.

Humans form selective and enduring pair bonds with romantic partners, a principal feature of human sociality. Neuroimaging studies have shown that romantic partners are differentially represented from other individuals in the nucleus accumbens (NAcc) and anterior insula (aINS), and that the specificity of partner representations in the NAcc diminishes as relationships mature. However, it remains unclear whether such differentiation reflects partner-specific coding or mere differences in familiarity with others, and whether these regions play different roles in romantic bonding. To address these questions, we applied multiple regression representational similarity analysis to fMRI data from 51 heterosexual male participants in early romantic relationships. The data were acquired during a social incentive delay task, in which participants anticipated social approval from their female romantic partner, a female friend, or an unfamiliar female individual. This approach allowed us to dissociate partner-specific representations from familiarity-related effects in the NAcc and aINS. We found that both regions exhibited partner-specific representations that could not be explained by familiarity. Consistent with previous findings, partner specificity in the NAcc was negatively associated with relationship duration, indicating that partner-specific coding in this region is established early in romantic relationships and diminishes as relationships progress. Moreover, greater partner specificity in the aINS was associated with more frequent intrusive thoughts about the partner. Together, these findings demonstrate that romantic partners are represented in the NAcc and aINS in a qualitatively distinct manner from other individuals, and that these regions support dissociable aspects of romantic bonding. Key PointsO_LIMultiple regression representational similarity analysis revealed partner-specific representations in the nucleus accumbens and anterior insula that cannot be explained by familiarity. C_LIO_LIIndividuals in longer relationships showed reduced partner specificity in the nucleus accumbens, consistent with prior findings. C_LIO_LIIndividuals exhibiting greater partner specificity in the anterior insula reported more frequent intrusive thoughts about their partner, indicating dissociable psychological functions across regions. C_LI

Many prize synchrony as the ingredient that turns a discordant conversation into a delightful duet. Although failure to align is sometimes associated with discord, exploring divergent opinions can also foster understanding and agreement, satisfy curiosity, and spur imagination. Using fMRI hyperscanning and natural language processing, we tested if dyadic alignment or exploration during decision-making conversations was a more effective route to agreement. Dyads (N = 60) discussed pressing societal problems while being instructed to either persuade or compromise with their partner. Our analysis uncovered four key insights: First, individuals instructed to compromise rather than persuade tended to agree more at the end of the conversation. Second, hyperscanning and linguistic analyses revealed that encouraging compromise resulted in increased exploration during conversations; dyads given compromise instructions traversed more diverse mental states and topics. Third, heightened exploration was linked to greater agreement at the end of the conversation. Fourth, the effect of the compromise instructions on agreement was entirely mediated by the degree of exploration. Together, these results suggest that finding agreement may be spurred by exploration, something that happens spontaneously when people are motivated to compromise. Significance StatementSuccessful conflict resolution is not just about syncing perfectly--it sometimes means leaning into differences. Our study employed fMRI hyperscanning to examine if exploratory conversation strategies promote consensus. We paired participants to discuss societal issues with either a persuasion or compromise mindset. We found that dyads aiming to compromise reached more agreement because they delved deeper into each others minds and covered more topics; in contrast, dyads aiming to persuade their partner covered a narrower scope of topics and mental states and ended up disagreeing more as a result. The finding that exploration rather than simple alignment facilitates agreement could sharpen communication strategies in fields from politics to education and economics.

Facial perception is a central feature of everyday social encounters and a rich source of emotional information. Classic functional magnetic resonance imaging (fMRI) studies of emotional facial processing used static photos emotional expression to identify regions of the brain showing univariate differences in response magnitudes between different emotional categories. However, there has been much less work identifying how the brain represents dynamic emotional facial expression and the factors that drive the similarity among these representations. In the current study, incorporated dynamic facial expression stimuli and representational similarity analysis to compare three competing hypothesized models of similarity of each of the stimuli presented: action being made, valence of the expression, and the identity of the person being perceived. Participants were shown short videos of fourteen volunteer actors making positive or negative facial expressions directed either toward or away from the camera. Activation patterns were compared against competing models on a trial-by-trial basis using a full multilevel modeling approach. Results showed that the identity of the person in the video was a greater predictor of brain responses similarity than the action or valence across widely distributed brain systems, particularly in the default mode network and lower-level visual processing regions. This suggests that the specific identity of the stimulus being perceived is a central driver of neural response similarity during perceptual encoding in dynamic facial processing.

Social interaction perception and theory of mind (ToM) frequently co-occur, but their commonalities and distinctions at behavioral and neural levels remain unclear. Participants (N = 231) provided moment-by-moment ratings of four text and four audio narratives on social interactions and ToM engagement, which were reliable (split-half r = .98 and .92, respectively) but only modestly correlated (r = .32). In a second sample (N = 90), we analyzed the co-variation between social interaction and ToM ratings and fMRI activity during text and audio narratives. Social interaction and ToM activity maps generalized across modalities (spatial r = .83 and .57, respectively), both with significant, overlapping clusters in canonical mentalizing regions (FDR q < .01). ToM uniquely engaged the lateral occipitotemporal cortex, left anterior intraparietal sulcus, and right premotor cortex. These results suggest that perceiving social interactions automatically involves mentalizing, and ToM additionally engages brain regions for action understanding and executive functions.

Guilt and shame are key moral emotions that influence mental health and regulate social behavior. Although prior research has examined the psychological and neural correlates of these emotions, the cognitive antecedents that trigger them, as well as their transformation into social behavior, remain insufficiently understood. In this study, we developed a novel task to investigate how two crucial cognitive antecedents, harm and responsibility, elicit guilt and shame, and how these emotions subsequently drive compensatory behavior, by combining functional magnetic resonance imaging (fMRI) with computational modeling. Behaviorally, we found that harm had a stronger impact on guilt than on shame, whereas responsibility had a stronger impact on shame than guilt, which supports the functionalist theory of emotion. Moreover, compared to shame, guilt exerted a greater effect on compensation. Computational modeling results indicated that the integration of harm and responsibility by individuals is consistent with the phenomenon of responsibility diffusion. The fMRI results revealed that brain regions associated with inequity representation (posterior insula) and value computation (striatum) encode this integrated measure. Furthermore, individual differences in responsibility-driven shame sensitivity were associated with activity in theory-of-mind regions (temporoparietal junction and superior temporal sulcus). Guilt- and shame-driven compensatory behavior recruited distinct neural substrates, with shame-driven compensatory sensitivity being more strongly linked to activity in the lateral prefrontal cortex, a region implicated in cognitive control. Our findings provide computational, algorithmic, and neural accounts of guilt and shame.

Infant stimuli elicit widespread neural and behavioral response in human adults and such massive allocation of resources attests to the evolutionary significance of the primary attachment. Here, we examined whether attachment-related cues also trigger cross-brain concordance, generating greater neural uniformity among individuals. Post-partum mothers were imaged twice in oxytocin/placebo administration design and stimuli included four ecological videos; two of infant/mother alone (Alone) and two mother-infant dyadic contexts (Social). Theory-driven analysis measured cross-brain synchrony in preregistered nodes of the parental caregiving network (PCN), which integrates subcortical structures underpinning mammalian mothering with cortical areas implicated in simulation, mentalization, and emotion regulation, and data--driven analysis assessed brain-wide concordance using whole brain parcellation. Results demonstrated widespread cross-brain synchrony in both the PCN and across the neuroaxis, from primary sensory and somatosensory areas, through insular-cingulate regions, to temporal and prefrontal cortices. The Social context yielded significantly more cross-brain concordance, with PCNs striatum, parahipporcampal gyrus, superior temporal sulcus, anterior cingulate cortex (ACC), and prefrontal cortex displaying cross-brain synchrony only to mother-infant social cues. Moment-by-moment fluctuations in mother-infant social synchrony, ranging from episodes of gaze aversion to rhythmically-coordinated positive bouts, were tracked online by cross-brain concordance in the pre-registered ACC. Findings indicate that social attachment stimuli, representing evolutionary-salient universal cues that require no verbal narrative for immediate comprehension, trigger substantial inter-brain concordance and suggest that the mother-infant bond, an icon standing at the heart of human civilization, may function to glue brains into a unified experience and bind humans into social groups. Significance StatementInfant stimuli elicit widespread neural response in human adults, attesting to their evolutionary significance, but do they also trigger cross-brain concordance and induce neural uniformity among perceivers? We measured cross-brain synchrony to ecological mother-infant videos. We employed theory-driven analysis, measuring cross-brain concordance in the parenting network, and data-driven analysis, assessing brain-wide concordance using whole-brain parcellation. Attachment cues triggered widespread cross-brain concordance in both the parenting network and across the neuroaxis. Moment-by-moment fluctuations in behavioral synchrony were tracked online by cross-brain variability in ACC. Attachment reminders bind humans brains into a unitary experience and stimuli characterized by social synchrony enhance neural similarity among participants, describing one mechanism by which attachment bonds provide the neural template for the consolidation of social groups.

Withdrawal StatementThe authors have withdrawn this manuscript because the authors discovered a fundamental error in the article. Therefore, the authors do not wish this work to be cited as reference for the project. If you have any questions, please contact the corresponding author.

The evolution of human social cognitive capacities such as mentalizing was associated with the expansion of frontoparietal cortical networks, particularly the default network. Mentalizing supports prosocial behaviors, but recent evidence indicates it may also serve a darker side of human social behavior. Using a computational reinforcement learning model of decision-making on a social exchange task, we examined how individuals optimized their approach to social interactions based on a counterparts behavior and prior reputation. We found that learning signals encoded in the default network scaled with reciprocal cooperation and were stronger in individuals who were more exploitative and manipulative, but weaker in those who were more callous and less empathic. These learning signals, which help to update predictions about others behavior, accounted for associations between exploitativeness, callousness, and social reciprocity. Separately, we found that callousness, but not exploitativeness, was associated with a behavioral insensitivity to prior reputation effects. While the entire default network was involved in reciprocal cooperation, sensitivity to reputation was selectively related to the activity of the medial temporal subsystem. Overall, our findings suggest that the emergence of social cognitive capacities associated with the expansion of the default network likely enabled humans to not only cooperate effectively with others, but to exploit and manipulate others as well. SignificanceTo navigate complex social lives, humans must learn from their interactions with others and adjust their own behavior accordingly. Here, we show that humans learn to predict the behavior of social counterparts by integrating reputational information with both observed and counterfactual feedback acquired during social experience. We find that superior learning during social interactions is related to empathy and compassion and associated with activity of the brains default network. Paradoxically however, learning signals in the default network are also associated with manipulativeness and exploitativeness, suggesting that the ability to anticipate others behavior can serve both the light and dark sides of human social behavior.