Towards Explainable Reinforcement Learning

Reinforcement learning (RL) is a machine learning technique capable of handling difficult sequential decision problems as well as control problems. In recent years, the development of powerful RL techniques has assured that RL will become an indispensable component in the industry (e.g., manufacturing, electric power systems and grid management). Additionally, RL-based solutions for applications such as (semi-)autonomous cars, socially assistive robotics, household solar storage management, also means that RL will find its way into daily human activities. Recent successes
like AlphaGo, prove that RL's time has come.
However, any powerful RL technique is built around complex function approximation procedures that transforms this framework in a black box approach. This may prevent its application in future critical domains, considering the upcoming European General Data Protection Regulation (GDPR). I believe it is a crucial moment to shift research efforts into creating an explainable RL framework. I believe it is a crucial moment to shift research effort into creating an explainable RL framework. This will be realised by opening the learning process to the user, by explaining what has been
learned, how the learning progressed and how the knowledge was applied. The approach taken is to augment existing reinforcement learning techniques, such that no loss in learning capacity is inflicted.