Examples of delayed tracking behavior (top) and recalibration of perceived visuomotor simultaneity (bottom)

Sensorimotor delay adaptation    

Humans adapt both the behavior and time perception to compensate for modified feedback delays. What are the behavioral strategies to deal with delays? How exactly does this reshape our perceived sensorimotor timing? Does delay adaptation transfer between senses and body parts?





The temporal window of perceived agency (red)

Temporal processing, prediction and agency   

Temporal coherence between and temporal order of visual and motor/kinesthetic signals influences the sense of agency. How is agency perceptually inferred? On the other hand, we make predictions about the sensory consequences of our actions, which influence our sensory processing. How do our intentional actions influence the perception of stimulus timing and intensity?



The automated stroking setup used in my experiments

Body perception   

Visuotactile and visuomotor congruency influence our perception of body parts and the feeling of ownership over body parts. What is the role of intersensory and sensorimotor latencies in this? How do different levels of processing relate?



Example of robotic goal-babbling

Motor skill acquisition   

In robotics, the learning of inverse models of robots with redundant degrees of freedom in non-linear movement spaces is an engineering challenge. Researchers from the University of Bielefeld have developed a motor exploration strategy called 'goal-babbling' to sove this problem. This technique is inspired by work on motor skill-acquisition in human infants. Does the algorithm adequately describe the exploratory and learning dynamics of humans?

Collaboration with Jochen Steil, Kenichi Narioka, Marc Ernst and Lina Klein.



Example evolved behavior

The perception of other agency in interaction   

Minimal Turing test scenarios. What is it about a dynamical interaction with another person that makes me perceive the other as an intentional agent?

Embodied cognition and computational modeling   

Many apparent scientific problems or difficulties can be resolved by taking a step back and looking at the bigger picture to question the assumptions underlying one's research and consider non-obvious alternatives. This includes confusions between the mechanistic and functional levels of description, general lack of clarity about assumptions underlying a scientific investigation and a blind splot for dynamical and embodied/situated processes in perception, cognition and behavior.



Evolutionary Robotics   

Evolutionary Robotics is a method for the automatic generation of robots or robot controllers that optimize behavior according to a certain criteraion. Neural network control parameters are initialized randomly and then optimized in closed-loop interaction interaction with the environment. This underspecification of the cognitive architecture leads to surprising, counter-intuitive and dynamically non-linear solutions that might be a nightmare to the human engineer, but plausible as naturally evolved solutions.



Bayesian approaches   

In the early 2000s and 2010s, researchers in Cognitive and Perception Science realized that human performance in many tasks can be explained by statistically optimal models of optimal Bayesian inference. It has been shown time and again that we learn and use information about signal uncertainty and prior probabilities when performing inference for perception, action and cognition. Yet, the powerful Bayesian framework also holds a number of hazards if used incorrectly. For instance, when researchers take on board too many assumptions that cannot be easily justified, Bayesian modeling can turn into a curve-fitting exercise rather than a benchmark for human performance. Therefore, I promote the rigorous use of these methods.