The use of robots as peers is more and more studied in human-robot interaction with co-learning interactions being complex and rich involving cognitive, affective, verbal and non-verbal processes. We aim to study the co-learning interaction with robots in the light of perspective-taking; a cognitive dimension that is important for interaction, engagement, and learning of the child. This work-in-progress details one of the studies we are developing in understanding perspective-taking from the Piagetian point of view. The study tried to understand how changes in the robot’s cognitive-affective state affect children’s behavior, emotional state, and perception of the robot. The experiment details a scenario in which child and the robot take turn to play a game by instructing their counterpart to reach a goal. The interaction consists of a condition in which the robot expresses frustration when the child gives egocentric instructions. We manipulate the robot’s emotional responses to the child’s instructions as our independent variable. We hypothesize that children will try to change their perspective more when the robot expresses frustration and follow the instructions wrongly, e.g. does not understand their perspective. Moreover, in the frustration groups, we are interested to observe if children reciprocates the robot’s behavior by showing frustration to the robot if it is egocentric. Consequently, we expect our analyses to help us to integrate a perspective-taking model in our robotic platform that can adapt its perspective according to educational or social aspect of the interaction.