Virtual tutors are a promising technology, providing a rich interactive environment for children to learn in. However, the question of how they should behave in order to enhance pupils’ motivation remains unanswered. Using an embodied conversational agent platform, we tested human-computer interactions with 22 children aged 9–11 years. Children performed several numeracy exercises set by two different virtual agents. One agent provided solely verbal feedback (unimodal), while the other one combined facial expressions based on real muscle contractions with its verbal feedback (bimodal). Children then completed a perceived social support questionnaire. Qualitative and quantitative data were subjected to inferential statistical tests. Results showed that the overall duration of agent-pupil interactions varied, children found the bimodal agent more empathic, and produced significantly more correct answers. Moreover, there was a positive correlation between accuracy and mean reaction times for correct answers with the bimodal agent. The lack of a correlation for the unimodal agent is discussed in the light of empathy and motivation in social cognition.
ASJC Scopus subject areas