Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
184 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
45 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Agent-Specific Effects: A Causal Effect Propagation Analysis in Multi-Agent MDPs (2310.11334v3)

Published 17 Oct 2023 in cs.AI

Abstract: Establishing causal relationships between actions and outcomes is fundamental for accountable multi-agent decision-making. However, interpreting and quantifying agents' contributions to such relationships pose significant challenges. These challenges are particularly prominent in the context of multi-agent sequential decision-making, where the causal effect of an agent's action on the outcome depends on how other agents respond to that action. In this paper, our objective is to present a systematic approach for attributing the causal effects of agents' actions to the influence they exert on other agents. Focusing on multi-agent Markov decision processes, we introduce agent-specific effects (ASE), a novel causal quantity that measures the effect of an agent's action on the outcome that propagates through other agents. We then turn to the counterfactual counterpart of ASE (cf-ASE), provide a sufficient set of conditions for identifying cf-ASE, and propose a practical sampling-based algorithm for estimating it. Finally, we experimentally evaluate the utility of cf-ASE through a simulation-based testbed, which includes a sepsis management environment.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (49)
  1. Identifiability of path-specific effects. In International Joint Conference on Artificial Intelligence, pp.  357–363, 2005.
  2. Ethical considerations about artificial intelligence for prognostication in intensive care. Intensive Care Medicine Experimental, 2019.
  3. Boutilier, C. Planning, learning and coordination in multiagent decision processes. In Conference on Theoretical Aspects of Rationality and Knowledge, pp.  195–210, 1996.
  4. Woulda, coulda, shoulda: Counterfactually-guided policy search. arXiv preprint arXiv:1811.06272, 2018.
  5. Characterizing manipulation from AI systems. arXiv preprint arXiv:2303.09387, 2023.
  6. Chiappa, S. Path-specific counterfactual fairness. In AAAI Conference on Artificial Intelligence, pp.  7801–7808, 2019.
  7. Responsibility and blame: A structural-model approach. Journal of Artificial Intelligence Research, 22:93–115, 2004.
  8. Nested counterfactual identification from arbitrary surrogate experiments. Advances in Neural Information Processing Systems, 34:6856–6867, 2021.
  9. Causal mediation analysis with multiple mediators. Biometrics, 71:1–14, 2015.
  10. Multi-view decision processes: the helper-ai problem. Advances in neural information processing systems, 30, 2017.
  11. European Commission. Ethics Guidelines for Trustworthy Artificial Intelligence. URL: https://ec.europa.eu/digital-single-market/en/news/ethics-guidelines-trustworthy-ai, 2019. [Online; accessed 30-January-2023].
  12. Agent incentives: A causal perspective. In AAAI Conference on Artificial Intelligence, pp.  11487–11495, 2021.
  13. Path-specific objectives for safer agent incentives. In AAAI Conference on Artificial Intelligence, volume 36, pp.  9529–9538, 2022.
  14. Towards deployment of robust cooperative ai agents: An algorithmic framework for learning adaptive policies. In International Conference on Autonomous Agents and MultiAgent Systems, pp.  447–455, 2020.
  15. Towards formal definitions of blameworthiness, intention, and moral responsibility. In AAAI Conference on Artificial Intelligence, pp.  1853–1860, 2018.
  16. Heckman, J. J. Randomization and social policy evaluation. National Bureau of Economic Research Cambridge, MA, 1991.
  17. Causal shapley values: Exploiting causal knowledge to explain individual predictions of complex models. Advances in Neural Information Processing Systems, 33:4778–4789, 2020.
  18. Social influence as intrinsic motivation for multi-agent deep reinforcement learning. In International Conference on Machine Learning, pp.  3040–3049, 2019.
  19. On measuring causal contributions via do-interventions. In International Conference on Machine Learning, pp.  10476–10501, 2022.
  20. Income modeling and balancing. Lecture Notes in Economics and Mathematical Systems, 2015.
  21. The artificial intelligence clinician learns optimal treatment strategies for sepsis in intensive care. Nature Medicine, 24:1716–1720, 2018.
  22. The medical algorithmic audit. The Lancet Digital Health, 2022.
  23. Sample-efficient reinforcement learning via counterfactual-based data augmentation. arXiv preprint arXiv:2012.09092, 2020.
  24. Lynn, L. A. Artificial intelligence systems for complex decision-making in acute care medicine: A review. Patient safety in Surgery, 2019.
  25. Explainable reinforcement learning through a causal lens. In AAAI Conference on Artificial Intelligence, pp.  2493–2500, 2020.
  26. Counterfactual identifiability of bijective causal models. arXiv preprint arXiv:2302.02228, 2023.
  27. Game-theoretic modeling of human adaptation in human-robot collaboration. In ACM/IEEE international conference on human-robot interaction, pp.  323–331, 2017.
  28. Counterfactual off-policy evaluation with gumbel-max structural causal models. In International Conference on Machine Learning, pp.  4881–4890, 2019.
  29. Pearl, J. Probabilities of causation: Three counterfactual interpretations and their identification. Synthese, 121:93–149, 1999.
  30. Pearl, J. Direct and indirect effects. In Conference on Uncertainty and Artificial Intelligence, pp.  411–420, 2001.
  31. Pearl, J. Causality. Cambridge University Press, 2009.
  32. Detecting influence structures in multi-agent reinforcement learning systems. AAAI Workshop on Reinforcement Learning in Games, 2022.
  33. Learning to collaborate in markov decision processes. In International Conference on Machine Learning, pp.  5261–5270, 2019.
  34. Counterfactual harm. Advances in Neural Information Processing Systems, 35:36350–36365, 2022.
  35. Trial without error: Towards safe reinforcement learning via human intervention. In International Conference on Autonomous Agents and MultiAgent Systems, pp.  2067–2069, 2018.
  36. What counterfactuals can be tested. In Conference on Uncertainty in Artificial Intelligence, pp.  352–359, 2007.
  37. Complete identification methods for the causal hierarchy. Journal of Machine Learning Research, 9:1941–1979, 2008.
  38. Effects of treatment on the treated: Identification and generalization. In Conference on Uncertainty in Artificial Intelligence, pp.  514–521, 2009.
  39. Identification of personalized effects associated with causal pathways. In Conference on Uncertainty in Artificial Intelligence, pp.  530–539, 2018.
  40. Flow-based attribution in graphical models: A recursive shapley approach. In International Conference on Machine Learning, pp.  9733–9743, 2021.
  41. Flexible mediation analysis with multiple mediators. American Journal of Epidemiology, 186:184–193, 2017.
  42. Reinforcement Learning: An Introduction. A Bradford Book, 2018.
  43. Towards computationally efficient responsibility attribution in decentralized partially observable mdps. In International Conference on Autonomous Agents and Multiagent Systems, pp.  131–139, 2023.
  44. On blame attribution for accountable multi-agent sequential decision making. Advances in Neural Information Processing Systems, 34:15774–15786, 2021.
  45. Actual causality and responsibility attribution in decentralized partially observable markov decision processes. In AAAI/ACM Conference on AI, Ethics, and Society, pp.  739–752, 2022.
  46. Finding counterfactually optimal action sequences in continuous state spaces. arXiv preprint arXiv:2306.03929, 2023.
  47. Look before you leap: Safe model-based reinforcement learning with human intervention. In Conference on Robot Learning, pp.  332–341, 2022.
  48. Fairness in decision-making—the causal explanation formula. In AAAI Conference on Artificial Intelligence, 2018a.
  49. Non-parametric path analysis in structural causal models. In Conference on Uncertainty in Artificial Intelligence, 2018b.

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now