Using Eye-Tracking to Analyze Collaboration in a Virtual Role Play Environment

The ColCoMa environment supports the training of workplace-oriented conflict management strategies through virtual role play. The role play relies on a web-based environment in which the participants interact through chat dialogues. Two of the participants (the parties in conflict) are human actors whereas the third role (“mediator”) is occupied by a chatbot. Our study aims at exploring the potential of eye-tracking analyses to assess the quality of cooperation in this situation. The standard assumption is that a certain “convergence” of the visual foci of attention between cooperation partners indicates better coordination and consideration of the other party. In our scenario, this assumption has to be refined by taking into account the different roles (including the role of the chatbot) and the distribution of utterances on the chat history. The eye-tracking parameters are compared to quality criteria such as successful completion of the game or richness/mutuality of the chat interactions. There are quite strong correlations on the aggregate level (taking overall eye-tracking convergence as a global parameter), yet not in terms of synchronicity between convergent eye-tracking and chat interaction. This is possibly due to the specific distribution of roles in our virtual training environment.