The Effect of Games on Engagement and Performance in Intelligent Tutoring Systems

The purpose of this dissertation is to assess the relation between game-like elements, individual differences in gameplay, and engagement within an Intelligent Tutoring System (ITS). The current studies examined the incorporation of a game into an existing ITS, iSTART. The game, Self-explanation Showdown (Showdown) added game-like elements into the iSTART practice sessions. Incorporating games was expected to increase engagement while not affecting participants’ overall performance. However, the results of Experiment 1 indicated that game-based practice (Showdown) was more engaging than the non-game-based practice (Coached Practice), but produced lower quality self-explanation performance. The decrease in performance was attributed to the amount of pedagogical information available during the learning task. In Experiment 2, a second version of Showdown was created that added pedagogical feedback similar to the feedback provided in Coached Practice. The feedback-added version of Showdown (Showdown-FB) was expected to retain the benefits of engagement while mitigating the deficits in performance. Instead, Showdown-FB demonstrated a reduction in participants’ engagement to a level which was no longer significantly different from Coached Practice, and did not increase performance relative to the original version of Showdown. Finally, Experiment 3 investigated whether opponent difficulty would affect gameplay and how those effects may vary as a function of different types of game players (Achievers, Explorers, Socializers, Killers). The results of Experiment 3 indicated that opponent difficulty affected both performance and engagement. Participants were more engaged and produced higher quality self-explanations when playing against a highly skilled opponent. Follow-up analyses indicated that the differences in performance were likely a result of modeling responses from a highly skilled opponent. However, the effects of opponent difficulty were not affected by a participant’s gamer type