Paper ID: 2210.13507
Causal Explanation for Reinforcement Learning: Quantifying State and Temporal Importance
Xiaoxiao Wang, Fanyu Meng, Xin Liu, Zhaodan Kong, Xin Chen
Explainability plays an increasingly important role in machine learning. Furthermore, humans view the world through a causal lens and thus prefer causal explanations over associational ones. Therefore, in this paper, we develop a causal explanation mechanism that quantifies the causal importance of states on actions and such importance over time. We also demonstrate the advantages of our mechanism over state-of-the-art associational methods in terms of RL policy explanation through a series of simulation studies, including crop irrigation, Blackjack, collision avoidance, and lunar lander.
Submitted: Oct 24, 2022