Empathy, perspective taking and attentional shifts differentially predict altruistic giving

Introduction: Human altruism is a complex phenomenon thought to involve numerous psychological processes. The current study aimed to identify some of the key processes at work in a simple altruistic giving decision-making task and to characterize their neural basis. In particular, we focused on the following three processes: affective empathic responses, cognitive perspective taking, and domain-general reorientation of attention. Methods: In an fMRI session, 32 healthy participants were endowed with €50 and determined on each trial how much of it they would be willing to donate to 60 different charities (Figure 1A). To determine the contribution of empathy and perspective taking to donation decisions, subsequent to scanning, participants rated their perceived empathy for the beneficiaries of each charity and the degree to which they tried to take their perspective. A subset of 22 participants also performed a Posner location-cueing task in the scanner that captures reorientation of attention. Results: Altruistic behavior, as measured by the size of the donations, varied considerably within and across participants (Figure 1B,C). Using multivariate pattern analysis (MVPA), we identified several brain regions that encoded the size of subject-specific donations (high vs low), including the right anterior insula (AI), temporo-parietal junction (TPJ) and posterior superior temporal sulcus (pSTS) (Figure 2A, p < 0.05, FWE corrected). Importantly, our data suggest distinct roles for AI and TPJ in the neural encoding of empathy and perspective taking during altruistic choices: Firstly, predictive accuracies in AI and TPJ reflected the degree to which participants used empathy and perspective taking as input for donation decisions (as measured by the behavioral regression weights based on the posttest ratings). In particular, participants that heavily relied on empathy during donation decisions exhibited higher predictive accuracies in the AI (Figure 2B). Likewise, increased weights of perspective taking as input for generous donations were related to improved decoding accuracies in the TPJ (Figure 2B). Further support for functionally distinct computations in AI and TPJ during altruistic giving comes from complementary multivariate regression (SVRs) analyses for both posttest ratings: Neural signatures in the AI during donation decisions (but not the TPJ) encoded trials-wise variations in empathy for the beneficiaries (Figure 3A), while response patterns in the TPJ (but not the AI) predicted the degree of perspective taking during giving decisions (Figure 3B). Neural networks involved in reorientation of attention have been found to overlap with those recruited for mentalizing (and empathy). Taking advantage of data from the attention task, we explicitly examined the role of domain-general attention shifts in altruistic giving. Using MVPA, we found that response patterns in the right TPJ (but not the AI) decoded stimulus-driven reorientation of attention (p < 0.05, FWE corrected) and that this cluster overlapped with two areas encoding generosity in the separate donation task (Figure 4A). However, supplemental cross-task classification found no evidence for shared neural code in the TPJ (Figure 4B) (p = 0.68), suggesting co-occurring, differential processes in the attention and donation task in this functional heterogeneous region. Notably, evidence for common neural code across both tasks was found for the pSTS (Figure 4B) (p = 0.007), suggesting that predictive information on generosity in this area is – at least partly – due to reinstated neural codes recruited for domain-general shifts of attention. Conclusions: Our findings support the hypothesis that the degree of altruism deployed in basic donation tasks is influenced by three distinct processes which are supported by dissociable neural substrates: affective empathic processes supported by the AI, cognitive perspective taking processes supported by the TPJ, and domain-general attention switches supported by the pSTS