Pedagogical agents as team members: Impact of proactive and pedagogical behavior on the user

Julien Saunier, Alexandre Pauchet, Mukesh Barange

Abstract: In a virtual environment for the learning of cooperative tasks, learners have to coordinate with one or more autonomous agents in order to perform a task. This coordination requires that the human and agent team members reason and dialogue about their resources, action plans, and shared actions. This article proposes a new agent architecture called PC2BDI designed to generate reactive, proactive, and pedagogical behaviors of the agents based on three agendas: the task, the dialogues, and the pedagogical scenario. An experimental study, in the context of learning a procedural activity in a virtual environment involving 3 team members (1 human and 2 agents), is presented to evaluate the effect of the agent behavior on a learner. The results show that using proactive pedagogical agents improves the learner’s engagement and task learning.