Challenges of Transparency for Learning Robots

With the importance of transparency for autonomous intelligent systems (AISs) becoming clear to the robotics community, definitions of transparency and guidelines for its implementation have started being proposed. In this paper, we adopt from the literature a model characterizing different aspects of transparency in the light of human-robot interaction and apply it to the specific case of robots learning from humans. Requirements deriving by the specific nature of learning systems are remarked and the main challenges highlighted and discussed.