Humans learn quickly which actions cause them harm. As social beings, we also need to learn to avoid actions that hurt others. It is currently unknown whether humans are as good at learning to avoid others' harm (prosocial learning) as they are at learning to avoid self-harm (self-relevant learning). Moreover, it remains unclear how the neural mechanisms of prosocial learning differ from those of self-relevant learning. In this fMRI study, 96 male human participants learned to avoid painful stimuli either for themselves or for another individual. We found that participants performed more optimally when learning for the other than for themselves. Computational modeling revealed that this could be explained by an increased sensitivity to subjective values of choice alternatives during prosocial learning. Increased value sensitivity was further associated with empathic traits. On the neural level, higher value sensitivity during prosocial learning was associated with stronger engagement of the ventromedial PFC during valuation. Moreover, the ventromedial PFC exhibited higher connectivity with the right temporoparietal junction during prosocial, compared with self-relevant, choices. Our results suggest that humans are particularly adept at learning to protect others from harm. This ability appears implemented by neural mechanisms overlapping with those supporting self-relevant learning, but with the additional recruitment of structures associated to the social brain. Our findings contrast with recent proposals that humans are egocentrically biased when learning to obtain monetary rewards for self or others. Prosocial tendencies may thus trump egocentric biases in learning when another person's physical integrity is at stake.

SIGNIFICANCE STATEMENT We quickly learn to avoid actions that cause us harm. As "social animals," we also need to learn and consider the harmful consequences our actions might have for others. Here, we investigated how learning to protect others from pain (prosocial learning) differs from learning to protect oneself (self-relevant learning). We found that human participants performed better during prosocial learning than during self-relevant learning, as they were more sensitive toward the information they collected when making choices for the other. Prosocial learning recruited similar brain areas as self-relevant learning, but additionally involved parts of the "social brain" that underpin perspective-taking and self-other distinction. Our findings suggest that people show an inherent tendency toward "intuitive" prosociality.

人类可以迅速了解哪些行为会对他们造成伤害。作为社会存在者，我们还需要学习避免伤害他人的行为。目前尚不清楚人类是否像学习避免他人伤害（自我相关的学习）一样善于学习避免他人伤害（亲社会的学习）。此外，尚不清楚亲社会学习的神经机制与自我相关学习的神经机制有何不同。在这项功能磁共振成像研究中，有96位男性受试者学会了避免自己或他人遭受痛苦的刺激。我们发现，参与者在为他人学习时的表现要比为他们自己更好。计算模型表明，这可以通过亲社会学习过程中对选择替代方案的主观价值的敏感性提高来解释。价值敏感性的提高还与移情特质相关。在神经水平上，亲社会学习过程中更高的价值敏感性与评估过程中腹侧PFC的更强参与有关。此外，与自我相关的选择相比，腹侧PFC在亲社会期间与右侧颞顶交界处的连接性更高。我们的结果表明，人类特别善于学习保护他人免受伤害。这种能力似乎是通过与支持自我相关学习的神经机制重叠的神经机制来实现的，但是又增加了与社交大脑相关的结构的募集。我们的发现与最近提出的建议相反，即人类在学习获得自我或他人的金钱报酬时会自我中心偏见。

重要性声明我们很快学会避免造成伤害的行为。作为“社交动物”，我们还需要学习并考虑我们的行为可能对他人造成的有害后果。在这里，我们调查了保护他人免受痛苦的学习（亲社会的学习）与保护自己的学习（自我相关的学习）有何不同。我们发现，人类参与者在亲社会学习中的表现要好于自我相关学习，因为他们在做出其他选择时对所收集的信息更加敏感。亲社会学习招募了与自我相关学习类似的大脑区域，但另外还涉及“社会大脑”的某些部分，这些部分支撑了观点采择和自我与他人的区别。我们的发现表明，人们表现出一种内在的“直觉性”亲社会倾向。