Social AI

The development of AI systems poses challenges in more than just in the technological domain. Take for example the social, ethical, philosophical, and societal challenges that need to be overcome in order to effectively collaborate with increasingly intelligent systems. In our research and development activities, we adopt a human-centered approach. A large body of our work is focused on studying and evaluating human-system interactions to facilitate human-AI collaborative experiences that are both efficient and pleasant. To this end, we develop and employ user models that can track the cognitive, social, and emotional state of a user.


For example, as part of a larger European project, our team has developed a Personal Assistance for a healthy Lifestyle (PAL). PAL facilitates self-management of children with a prolonged disease in order to support their autonomy, competence, and relatedness. Children can interact with a robot version of PAL in hospitals, and with a 3D-avatar of PAL at home. PAL uses a hybrid AI approach, with symbolic modeling (ontologies) to implement domain knowledge and to provide understandable explanations to a child, and sub-symbolic modeling (machine learning) to track the child’s knowledge.