Design of Embodied Conversational Agent to Elicit Participants' Reasoning for Their Decisions Taken Inside a VR Environment

Eliciting participants' reasoning for their decisions taken inside a VR environment is challenging. This research explores the use of an embodied conversational agent with a wizard-of-Oz method to elicit participants' reasoning while driving a car in complex public scenarios on Indian roads, simulated using VR. Participants were made to drive a car in a VR city environment, with a complex public scenario unfolding in front of them. The co-passenger inside the car, acting as the embodied conversational agent, assumed a peer role, with conversations triggered by the researcher based on the response given or dialogue said by the participant. Each participant's actions, gaze, and dialogues were recorded along with timestamp data. The method's effectiveness was evaluated by analyzing the frequency of appropriate elicitations from four participants by coding their conversations based on the noticing framework and mapping the action/gaze data with conversation triggers. Results suggest that the embodied conversational agent contributed to elicitations as much the same as the usual stimulated recall method conducted after the intervention, with the added advantage of not hindering the virtual presence of the participants.