发表机构
Mahatma Gandhi International School（圣雄甘地国际学校）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究从人类神经行为数据校准愧疚信号作为亲社会奖励权重，嵌入多智能体强化学习环境后，其智能体的社会选择率最贴合人类数据，为亲社会奖励塑造提供了定量约束。
AI 中文摘要
合作型多智能体强化学习常向个体奖励中添加社会项，但这些项的规模通常由手动选择。我们探究是否可从人类神经与行为数据中校准愧疚信号，并将其迁移至人工智能体。使用公开的SoDec责任fMRI数据集（含40名参与者），我们拟合了瞬间快乐变化对结果类型计数的被试固定效应回归，恢复出愧疚权重为“伴侣负性”减去“社会负性”的对比值（$\boldsymbol{\tilde{w}}=1.118$，Cohen's $d=0.214$）。我们将该权重嵌入双智能体“社会彩票”环境，在四种塑造机制下训练独立的近端策略优化（Proximal Policy Optimization）演员-评论家：神经校准型、均匀常数型、零（自私型）及单位系数神谕型。每种条件下各进行1000次评估回合，结果显示校准型智能体最贴合人类的社会安全选择率（0.459，人类为0.484；KL散度=0.0012），其余三种条件的KL散度偏差为1至3个数量级。因此，人类神经行为先验可作为亲社会奖励塑造的定量约束。
英文摘要
Cooperative multi-agent reinforcement learning often adds social terms to individual rewards, yet the scale of those terms is usually chosen by hand. We ask whether a guilt signal can instead be calibrated from human neural and behavioural data and transferred to artificial agents. Using the public SoDec responsibility fMRI dataset (40 participants), we fit a subject-fixed-effects regression of momentary-happiness changes on outcome-type counts and recover a guilt weight as the Partner-negative minus Social-negative contrast ($\hat{w}=1.118$, Cohen's $d=0.214$). We embed this weight in a two-agent Social Lottery environment and train independent Proximal Policy Optimization actor-critics under four shaping regimes: neurally calibrated, uniform constant, zero (selfish), and a unit-coefficient oracle. Across 1{,}000 evaluation episodes per condition, the calibrated agents track the human Social safe-choice rate most closely ($0.459$ vs.\ human $0.484$; $\mathrm{KL}=0.0012$), while the other three conditions deviate by one to three orders of magnitude in KL. Human neurobehavioural priors can therefore act as quantitative constraints on prosocial reward shaping.
Comments12 pages, 6 figures, 3 tables
