Counterfactual Explanations for Continuous Action Reinforcement Learning
This addresses the adoption barrier in RL applications like healthcare and robotics by providing interpretable explanations, though it is incremental as it builds on existing counterfactual concepts for continuous actions.
The paper tackles the problem of interpretability in reinforcement learning for continuous action spaces by proposing a novel method to generate counterfactual explanations, showing effectiveness in domains like Diabetes Control and Lunar Lander.
Reinforcement Learning (RL) has shown great promise in domains like healthcare and robotics but often struggles with adoption due to its lack of interpretability. Counterfactual explanations, which address "what if" scenarios, provide a promising avenue for understanding RL decisions but remain underexplored for continuous action spaces. We propose a novel approach for generating counterfactual explanations in continuous action RL by computing alternative action sequences that improve outcomes while minimizing deviations from the original sequence. Our approach leverages a distance metric for continuous actions and accounts for constraints such as adhering to predefined policies in specific states. Evaluations in two RL domains, Diabetes Control and Lunar Lander, demonstrate the effectiveness, efficiency, and generalization of our approach, enabling more interpretable and trustworthy RL applications.