Automating the Mentor in a Serious Game: A Discourse Analysis Using Finite State Machines

Serious games are increasingly becoming a popular, effective supplement to standard classroom instruction [1]. Similar to recreational games, multi-party chat is a standard method of communication in serious games. As players collaborate in a serious game, mentoring is often needed to facilitate progress and learning [2, 3, 4]. This role is almost exclusively provided by a human at the present time. However, the cost incurred with training a human mentor represents a critical barrier for widespread use of a collaborative epistemic game. Although great strides have been made in automating one-on-one tutorial dialogues [5, 6], multi-party chat presents a significant challenge for natural language processing. The goal of this research, then, is to provide a preliminary understanding of player-mentor conversations in the context of an epistemic game, Land Science [7].