Artificial social agents: Beyond symbol-based interaction

Referent ist Dr. Ulysses Bernardet von der Aston University, School of Engineering and Applied Sciences.

Artificial social agents — human-looking robots and computer-generated virtual humans — will play an increasingly important role as stand-ins for biological humans in automated healthcare and education applications. I will argue for a bio-inspired approach to building artificial social agents that uses principles derived from biological systems to the construction of artefacts. Characteristic for this approach is that application and theory development are in a tight reciprocal connection: On the one hand, applications based on artificial social agents require the development of a specific control model to solve the given task. On the other hand, embodying a psychological or biological model in an agent is a heuristic for model testing.

While the term artificial social agent covers the whole reality spectrum, using virtual, computer-generated humans as the development platform has the advantage that they present a technically simplified case and are closer to humans in terms of surface realism. In my presentation will illustrate the topic and approach by elaborating on models we have built and studies we have conducted revolving around the topics of personality expression through nonverbal behaviour, social-spatial, and reflexive behaviour. By focusing on unconscious and autonomous processing, the presented work complements other approaches that focus mainly on the cognitive, symbol-manipulation level.