Understanding human reactions to robot behaviors is essential for a better integration of robots into our daily lives. Social robotics, with increasingly humanoid appearances and programmed emotional expressions, presents opportunities to explore authentic human responses to these technologies [1]. Our hypothesis is that social cognition processes unfolding in natural interaction can manifest in human interaction with social robots. We propose an experimental protocol aimed at studying how individuals respond to Pepper, a humanoid robot, during a rock-paper-scissors (RPS) interaction [4]. In the experiment, a participant plays RPS 30 times with Pepper (Fig. 1). Evaluation metrics include expressed emotions, attention level and cognitive engagement, assessed through subjective feedback (pre- and post-experiment surveys) and physiological data such as EEG and heart rate [3]. A data glove is used to communicate the participant’s choice to Pepper. The experiment incorporates diverse robot behaviors (conditions: pleasant, neutral, unpleasant) including errors (genuine mistake and cheating act [6]). We use the continuous valence/arousal model to identify emotions based on EEG measures [5], like frontal asymmetry in alpha and beta bands. For attention and engagement, we analyze alpha, beta, and theta bands in both the frontal and parietal cortex [2]. Error-Related Potentials are also examined to evaluate perceptions of robot behavior in error scenarios [7]. In addition, we are exploring the potential for predicting users’ hand movements using kinesthetic motor imagery during a movement preparation phase, in an aim to control Pepper’s skill level and provoke varied emotional responses. Following a pilot study with a small sample of subjects, we obtained encouraging results, setting the groundwork for larger experiments to test our hypothesis. In the future, this work could help enhance the immersion of human-robot interaction by using the participants’ reactions as feedback.