Towards Interpretable Deep Reinforcement Learning Models via Inverse Reinforcement Learning

Sean Xie, Soroush Vosoughi, Saeed Hassanpour

Abstract: Artificial Intelligence, particularly through recent advancements in deep learning (DL), has achieved exceptional performances in many tasks in fields such as natural language processing and computer vision. For certain high-stake domains, in addition to desirable performance metrics, a high level of interpretability is often required in order for AI to be reliably utilized. Unfortunately, the black box nature of DL models prevents researchers from providing explicative descriptions for a DL model’s reasoning process and decisions. In this work, we propose a novel framework utilizing Adversarial Inverse Reinforcement Learning that can provide global explanations for decisions made by a Reinforcement Learning model and capture intuitive tendencies that the model follows by summarizing the model’s decision-making process.