Sebastian Wallkötter

My host institution is the Social Robotics Lab Uppsala, Sweden. Here, I research how robots can learn the ethical consequences of what they do. The goal is to develop a system that can quickly adapt to different environments, realize if it acts unethical and avoid such behavior. To this end I want to create a robot that can assist teachers and promote such good values. I come from an international, cross-disciplinary background having studied mechanical engineering in Germany and computer-science in England. My work is driven by my strong interest in machine learning, AI and optimization problems of any kind.

Robot ethics is becoming a recognized field of research of great importance to social robotics20. However, little work has been done to integrate theories and computational modeling21, specifically in relation to teaching robots how to learn behaviours that are ethically appropriate via social learning. With the advent of educational robots, robot ethics is still an underexplored strand: what information should a robot tutor retain about individual learners and their learning? Who should this information be disclosed to (e.g. other peers, teachers)? What support strategies of a robotic tutor are ethically acceptable from a learning point of view? The objective of this ESR project is to develop computational social learning techniques grounded in ethical frameworks for robots to learn behaviours that are ethically appropriate in an educational scenario. In an educational scenario where the robot plays the role of an educational agent, a specific focus of this project will be on robot learning how to behave ethically, i.e., on identifying the appropriate values and the decision mechanism for promoting those values in human-robot interaction. The ethical values will be related to privacy et personal data protection. The robot will have to identify what information about its user to retain to respect their privacy and comply with human instructions. This may be achieved by using social learning mechanisms with different degrees of human intervention, based for example on environmental or teaching signals by human users (e.g., students, teachers) that could indicate to the robot what information it is appropriate to retain and what behaviours and teaching strategies should be used.