mlsanityData to 13 Oct 2026

Calibrating Artificial Guilt: Neurally Grounded Reward Shaping for Prosocial Multi-Agent Reinforcement Learning

arXiv 2608.04663 ↗First posted 5 Aug 2026cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

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 (hatw=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; 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.

Key facts. No citing papers in the map yet (data through 13 Oct 2026). Builds on 27 of its 56 references that are in the map. First posted on arXiv 5 Aug 2026.

Builds on

27 of 56 references.

References · 27

  1. 2022Training language models to follow instructions with human feedbackcited by 15,856
  2. 2017Proximal Policy Optimization Algorithmscited by 14,592
  3. 2017Deep reinforcement learning from human preferencescited by 3,871
  4. 2022Constitutional AI: Harmlessness from AI Feedbackcited by 3,011
  5. 2017Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environmentscited by 1,525
  6. 2017Curiosity-driven Exploration by Self-supervised Predictioncited by 1,180
  7. 2018Exploration by Random Network Distillationcited by 971
  8. 2021The Surprising Effectiveness of PPO in Cooperative, Multi-Agent Gamescited by 910
  9. 2018Diversity is All You Need: Learning Skills without a Reward Functioncited by 739
  10. 2020Monotonic Value Function Factorisation for Deep Multi-Agent Reinforcement Learningcited by 478
  11. 2018Scalable agent alignment via reward modeling: a research directioncited by 443
  12. 2016Cooperative Inverse Reinforcement Learningcited by 299
15 more
  1. 2017A Unified Game-Theoretic Approach to Multiagent Reinforcement Learningcited by 265
  2. 2017Learning with Opponent-Learning Awarenesscited by 220
  3. 2020PettingZoo: Gym for Multi-Agent Reinforcement Learningcited by 211
  4. 2017Multi-agent Reinforcement Learning in Sequential Social Dilemmascited by 206
  5. 2018A Survey and Critique of Multiagent Deep Reinforcement Learningcited by 174
  6. 2020Open Problems in Cooperative AIcited by 168
  7. 2018Inequity aversion improves cooperation in intertemporal social dilemmascited by 106
  8. 2016Towards an integration of deep learning and neurosciencecited by 77
  9. 2017A multi-agent reinforcement learning model of common-pool resource appropriationcited by 68
  10. 2017Maintaining cooperation in complex social dilemmas using deep reinforcement learningcited by 48
  11. 2017Prosocial learning agents solve generalized Stag Hunts better than selfish onescited by 47
  12. 2024Gymnasium: A Standard Interface for Reinforcement Learning Environmentscited by 41
  13. 2020Deep Reinforcement Learning and its Neuroscientific Implicationscited by 33
  14. 2019Learning Reciprocity in Complex Sequential Social Dilemmascited by 24
  15. 2022Human-centered mechanism design with Democratic AIcited by 3

Built on by

Not cited yet.