Unconscious or, implicit learning (IL) is often described as being instrumental to human social functioning. However, most of the available IL tasks have limited external validity; they use surface stimuli that are not socially relevant. Additionally, the way in which participants exchange information within most of the available tasks departs from the way in which information is being exchanged in real-life social situations. This is a novel task, inspired from Broadbent et al. (1984), assessing the implicit and explicit learning of socio-emotional information in a dynamic environment. Participants interact with an animated virtual avatar that displays different levels of emotional facial expressions. Their task is to regulate the avatar’s facial expression to a specified level. Unknown to them, the relationship between their inputs and the avatar’s state is mediated by an abstract rule. Results from the validation study (Costea, A., Jurchis, R., Visu-Petra, L., Cleeremans A., Norman, E., and Opre, A., 2020 – now in preparation) indicate that learning occurred in the task, as participants gradually increased their ability to bring the avatar in the target state. We found evidence for both explicit (consciously knowing the appropriate response) and implicit (knowing the correct responses even when they based their responses on subjectively defined unconscious mental states) knowledge acquisition. This is one of the first studies to propose a task for studying the role of IL in interactive social situations.