Despite all our individual differences, there exist numerous regularities in human behaviour. Even in situations as complex as for example social interaction, we can still observe common patterns in how people position themselves and react to each other.
I am fascinated by these patterns and enjoy using the tools of science to try and figure them out – preferably to the point where we can reproduce them, for example in a robot. Two strongly related questions are particularly relevant and interesting to me in this respect;
- What are relevant cues for (social) behaviour and how can we (automatically) detect these cues?
- What responses are appropriate in which situation and why?
My background is in Cognitive Artificial Intelligence, which I studied at the Radboud University Nijmegen. For my bachelor’s thesis I investigated a computer model of human/animal reinforcement learning in situations with non-deterministic rewards. During my master I investigated different paradigms (or architectures) that could be used to create (robotic) behaviour. In my master thesis, I introduced a formal computational method for comparing such algorithms and used that to proof their relative efficiency in different kinds of situations.
Currently, I am a PhD student at the University of Twente, working on the Teresa project (2014-2017). Within this project, we are looking at situations where elderly are represented by a robot in a social interaction with their peers. The robot is then to semi-autonomously display social behaviours that will help it integrate in the interaction.
My focus is on the detection and understanding of the behaviour of groups interacting with the robot. On the one end of the spectrum, this involves using techniques from computer vision / social signal processing to investigate and detect relevant social cues. On the other end, this is about finding out what kinds of behaviours people would prefer the robot to show in response to those cues.