Using State Transition Networks to Analyze Multi-party Conversations in a Serious Game

Brent Morgan, Fazel Keshtkar, Ying Duan, Padraig Nash, Arthur Graesser

As players interact in a serious game, mentoring is often needed to facilitate progress and learning. Although human mentors are the current standard, they present logistical difficulties. Automating the mentor’s role is a difficult task, however, especially for multi-party collaborative learning environments. In order to better understand the conversational demands of a mentor, this paper investigates the dynamics and linguistic features of multi-party chat in the context of an online epistemic game, Urban Science. We categorized thousands of player and mentor contributions into eight different speech acts and analyzed the sequence of dialogue moves using State Transition Networks. The results indicate that dialogue transitions are relatively stable with respect to gameplay goals; however, task-oriented stages emphasize mentor-player scaffolding, whereas discussion-oriented stages feature player-player collaboration.

The final publication is available at Springer via