Oral presentation at the conference "Social cognition in humans and robots", held at the University Medical Center Hamburg-Eppendorf, Hamburg, Germany (September 27-28, 2018). This event was organized by the FET Proactive H2020 project "Socializing Sensorimotor Contingencies - socSMCs" (socSMCs.eu) in partner alliance with the European Society for Cognitive Systems - EUCog (eucognition.org). -- PDF available, keynote recommended -- ---------- Abstract **Basic mechanisms of social cognition unravelled with the use of humanoid robots** Cesco Willemse & Agnieszka Wykowska Social Cognition in Human-Robot Interaction Istituto Italiano di Tecnologia (IIT), Genova, Italy It would be difficult to imagine a day without social interactions, during which the brain routinely engages various mechanisms of social cognition. However, under which conditions these mechanisms are elicited, remains relatively unclear. In our research, we use humanoid robots to tease apart factors that evoke specific mechanisms of social perception and cognition. We follow a novel approach that combines naturalistic interactions with an embodied humanoid robot, whilst utilising methods of experimental psychology and social cognitive neuroscience. This approach allows excellent experimental control to be maintained while introducing embodied presence of an interaction partner, thereby increasing ecological validity, relative to standard screen-based experiments. In addition, we use technologies such as EEG and (mobile) eyetracking to complement our performance measures. In this talk, we present example studies that illustrate our research agenda. One series of studies comprised a gaze-cueing task with the robot. We found that the gaze cueing effects (behavioural and EEG) were modulated by whether the robot made eye contact with the participants before each trial, or avoided mutual gaze. Another experimental series investigated the “complementary side” of gaze cueing, namely gaze leading. We succeeded in integrating a mobile eye tracker with stimulus presentation software and the software controlling behaviours of the robot. In result, we were able to generate gaze-contingent behaviours of the robot. Participants were free to choose one of two laterally presented objects with their gaze. We operationalised the latency of return-saccades from the objects back to the robot’s face as a measure of attentional engagement. Intriguingly, our results show that this engagement depended not only on whether the robot followed the participants’ gaze in a given instance, but also on whether the robot usually did so. Through these example studies, we show how social cognition is grounded in basic perceptual and cognitive mechanisms related to action effect expectations, as well as others’ reactiveness to our own behaviour. More broadly, we highlight how the embodied presence of an agent casts new light on fundamental mechanisms of social cognition. Taken together, we argue that transferring methods from social cognitive science to human-robot interaction paradigms, as well as using the resulting parameters to implement human-like behaviour in robot design, informs basic research on social behaviour “in the real world”.