Abstract
In order to understand a reinforcement learning (RL) agent’s behavior within its environment, we propose an answer to ‘What is likely to happen?’ in the form of a predictive explanation. It is composed of three scenarios: best-case, worst-case and most-probable which we show are computationally difficult to find (W[1]-hard). We propose linear-time approximations by considering the environment as a favorable/hostile/neutral RL agent. Experiments validate this approach. Furthermore, we give a dynamic-programming algorithm to find an optimal summary of a long scenario.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Amir, D., Amir, O.: HIGHLIGHTS: summarizing agent behavior to people. In: André, E., Koenig, S., Dastani, M., Sukthankar, G. (eds.) Proceedings of the 17th International Conference on Autonomous Agents and MultiAgent Systems, AAMAS, pp. 1168–1176. International Foundation for Autonomous Agents and Multiagent Systems/ACM (2018). http://dl.acm.org/citation.cfm?id=3237869
Bastani, O., Pu, Y., Solar-Lezama, A.: Verifiable reinforcement learning via policy extraction. In: Bengio, S., Wallach, H.M., Larochelle, H., Grauman, K., Cesa-Bianchi, N., Garnett, R. (eds.) NeurIPS, pp. 2499–2509 (2018). https://proceedings.neurips.cc/paper/2018/hash/e6d8545daa42d5ced125a4bf747b3688-Abstract.html
Brockman, G., et al.: OpenAI gym. arXiv preprint arXiv:1606.01540 (2016)
Clouse, J.A.: On integrating apprentice learning and reinforcement learning. University of Massachusetts Amherst (1996)
Cruz, F., Dazeley, R., Vamplew, P.: Memory-based explainable reinforcement learning. In: Liu, J., Bailey, J. (eds.) AI 2019. LNCS (LNAI), vol. 11919, pp. 66–77. Springer, Cham (2019). https://doi.org/10.1007/978-3-030-35288-2_6
Danesh, M.H., Koul, A., Fern, A., Khorram, S.: Re-understanding finite-state representations of recurrent policy networks. In: Meila, M., Zhang, T. (eds.) Proceedings of the 38th International Conference on Machine Learning, ICML 2021, 18–24 July 2021, Virtual Event. Proceedings of Machine Learning Research, vol. 139, pp. 2388–2397. PMLR (2021). http://proceedings.mlr.press/v139/danesh21a.html
Darwiche, A.: Human-level intelligence or animal-like abilities? Commun. ACM 61(10), 56–67 (2018). https://doi.org/10.1145/3271625
European Commission: Artificial Intelligence Act (2021). https://eur-lex.europa.eu/legal-content/EN/TXT/?qid=1623335154975 &uri=CELEX%3A52021PC0206
Gajcin, J., Dusparic, I.: ReCCoVER: detecting causal confusion for explainable reinforcement learning. In: Calvaresi, D., Najjar, A., Winikoff, M., Främling, K. (eds.) EXTRAAMAS 2022. LNCS, vol. 13283, pp. 38–56. Springer, Cham (2022). https://doi.org/10.1007/978-3-031-15565-9_3
Greydanus, S., Koul, A., Dodge, J., Fern, A.: Visualizing and understanding Atari agents. In: Dy, J.G., Krause, A. (eds.) ICML. Proceedings of Machine Learning Research, vol. 80, pp. 1787–1796. PMLR (2018). http://proceedings.mlr.press/v80/greydanus18a.html
Guo, W., Wu, X., Khan, U., Xing, X.: EDGE: explaining deep reinforcement learning policies. In: Ranzato, M., Beygelzimer, A., Dauphin, Y.N., Liang, P., Vaughan, J.W. (eds.) NeurIPS, pp. 12222–12236 (2021). https://proceedings.neurips.cc/paper/2021/hash/65c89f5a9501a04c073b354f03791b1f-Abstract.html
Hasselt, H.: Double Q-learning. In: Advances in Neural Information Processing Systems, vol. 23 (2010)
Huber, T., Demmler, M., Mertes, S., Olson, M.L., André, E.: GANterfactual-RL: understanding reinforcement learning agents’ strategies through visual counterfactual explanations. CoRR abs/2302.12689 (2023). https://doi.org/10.48550/arXiv.2302.12689
Iyer, R., Li, Y., Li, H., Lewis, M., Sundar, R., Sycara, K.P.: Transparency and explanation in deep reinforcement learning neural networks. In: Furman, J., Marchant, G.E., Price, H., Rossi, F. (eds.) Proceedings of the 2018 AAAI/ACM Conference on AI, Ethics, and Society, AIES, pp. 144–150. ACM (2018). https://doi.org/10.1145/3278721.3278776
Juozapaitis, Z., Koul, A., Fern, A., Erwig, M., Doshi-Velez, F.: Explainable reinforcement learning via reward decomposition. In: IJCAI/ECAI Workshop on Explainable Artificial Intelligence, p. 7 (2019)
Lipton, Z.C.: The mythos of model interpretability. Commun. ACM 61(10), 36–43 (2018). https://doi.org/10.1145/3233231
Madumal, P., Miller, T., Sonenberg, L., Vetere, F.: Explainable reinforcement learning through a causal lens. In: The Thirty-Fourth AAAI Conference on Artificial Intelligence, AAAI 2020, pp. 2493–2500. AAAI Press (2020). https://ojs.aaai.org/index.php/AAAI/article/view/5631
McDonald, R.: A study of global inference algorithms in multi-document summarization. In: Amati, G., Carpineto, C., Romano, G. (eds.) ECIR 2007. LNCS, vol. 4425, pp. 557–564. Springer, Heidelberg (2007). https://doi.org/10.1007/978-3-540-71496-5_51
Milani, S., Topin, N., Veloso, M., Fang, F.: A survey of explainable reinforcement learning. CoRR abs/2202.08434 (2022). https://arxiv.org/abs/2202.08434
Mishra, A., Soni, U., Huang, J., Bryan, C.: Why? Why not? When? Visual explanations of agent behavior in reinforcement learning. CoRR abs/2104.02818 (2021). https://arxiv.org/abs/2104.02818
Mnih, V., et al.: Human-level control through deep reinforcement learning. Nature 518(7540), 529–533 (2015). http://www.nature.com/articles/nature14236
Olson, M.L., Neal, L., Li, F., Wong, W.: Counterfactual states for Atari agents via generative deep learning. CoRR abs/1909.12969 (2019). http://arxiv.org/abs/1909.12969
Saulières, L., Cooper, M.C., Dupin de Saint Cyr, F.: Reinforcement learning explained via reinforcement learning: towards explainable policies through predictive explanation. In: 15th International Conference on Agents and Artificial Intelligence (ICAART 2023), pp. 35–44 (2023)
Sequeira, P., Gervasio, M.T.: Interestingness elements for explainable reinforcement learning: understanding agents’ capabilities and limitations. Artif. Intell. 288, 103367 (2020). https://doi.org/10.1016/j.artint.2020.103367
Shu, T., Xiong, C., Socher, R.: Hierarchical and interpretable skill acquisition in multi-task reinforcement learning. CoRR abs/1712.07294 (2017). http://arxiv.org/abs/1712.07294
Sutton, R.S., Barto, A.G.: Reinforcement Learning: An Introduction. MIT Press, Cambridge (2018)
Tsirtsis, S., De, A., Rodriguez, M.: Counterfactual explanations in sequential decision making under uncertainty. In: Ranzato, M., Beygelzimer, A., Dauphin, Y.N., Liang, P., Vaughan, J.W. (eds.) NeurIPS 2021, pp. 30127–30139 (2021). https://proceedings.neurips.cc/paper/2021/hash/fd0a5a5e367a0955d81278062ef37429-Abstract.html
Verma, A., Murali, V., Singh, R., Kohli, P., Chaudhuri, S.: Programmatically interpretable reinforcement learning. In: Dy, J.G., Krause, A. (eds.) Proceedings of the 35th International Conference on Machine Learning, ICML 2018. Proceedings of Machine Learning Research, vol. 80, pp. 5052–5061. PMLR (2018). http://proceedings.mlr.press/v80/verma18a.html
van der Waa, J., van Diggelen, J., van den Bosch, K., Neerincx, M.A.: Contrastive explanations for reinforcement learning in terms of expected consequences. CoRR abs/1807.08706 (2018). http://arxiv.org/abs/1807.08706
Watkins, C.J., Dayan, P.: Q-learning. Mach. Learn. 8(3), 279–292 (1992)
Yau, H., Russell, C., Hadfield, S.: What did you think would happen? Explaining agent behaviour through intended outcomes. In: Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M., Lin, H. (eds.) NeurIPS (2020). https://proceedings.neurips.cc/paper/2020/hash/d5ab8dc7ef67ca92e41d730982c5c602-Abstract.html
Yu, Z., Ruan, J., Xing, D.: Explainable reinforcement learning via a causal world model. CoRR abs/2305.02749 (2023). https://doi.org/10.48550/arXiv.2305.02749
Zahavy, T., Ben-Zrihem, N., Mannor, S.: Graying the black box: understanding DQNs. In: Balcan, M., Weinberger, K.Q. (eds.) Proceedings of the 33nd International Conference on Machine Learning, ICML 2016. JMLR Workshop and Conference Proceedings, vol. 48, pp. 1899–1908. JMLR.org (2016). http://proceedings.mlr.press/v48/zahavy16.html
Acknowledgements
The authors would like to thank Arnaud Lequen for his valuable suggestions that have led to the improvement of this paper. This work was supported by the AI Interdisciplinary Institute ANITI, funded by the French program “Investing for the Future - PIA3” under grant agreement no. ANR-19-PI3A-0004.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2024 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Saulières, L., Cooper, M.C., de Saint-Cyr, F.D. (2024). Predictive Explanations for and by Reinforcement Learning. In: Rocha, A.P., Steels, L., van den Herik, J. (eds) Agents and Artificial Intelligence. ICAART 2023. Lecture Notes in Computer Science(), vol 14546. Springer, Cham. https://doi.org/10.1007/978-3-031-55326-4_6
Download citation
DOI: https://doi.org/10.1007/978-3-031-55326-4_6
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-55325-7
Online ISBN: 978-3-031-55326-4
eBook Packages: Computer ScienceComputer Science (R0)