Seung Y. Lee

Real-Time narrative-centered tutorial planning for story-based learning

Interactive story-based learning environments offer significant potential for crafting narrative tutorial guidance to create pedagogically effective learning experiences that are tailored to individual students. This paper reports on an empirical evaluation of machine-learned models of narrative-centered tutorial planning for story-based learning environments. We investigate differences in learning gains and in-game performance during student interactions in a rich virtual storyworld. One hundred and eighty-three middle school students participated in the study, which had three conditions: Minimal Guidance, Intermediate Guidance, and Full Guidance. Results reveal statistically significant differences in learning and in-game problem-solving effectiveness between students who received minimal guidance and students who received full guidance. Students in the full guidance condition tended to demonstrate higher learning outcomes and problem-solving efficiency. The findings suggest that machine-learned models of narrative-centered tutorial planning can improve learning outcomes and in-game efficiency.

Modeling Self-Efficacy Across Age Groups with Automatically Tracked Facial Expression

Affect plays a central role in learning. Students’ facial expressions are key indicators of affective states and recent work has increasingly used automated facial expression tracking technologies as a method of affect detection. However, there has not been an investigation of facial expressions compared across age groups. The present study collected facial expressions of college and middle school students in the Crystal Island game-based learning environment. Facial expressions were tracked using the Computer Expression Recognition Toolbox and models of self-efficacy for each age group highlighted differences in facial expressions. Age-specific findings such as these will inform the development of enriched affect models for broadening populations of learners using affect-sensitive learning environments.

Director agent intervention strategies for interactive narrative environments

Interactive narrative environments offer significant potential for creating engaging narrative experiences. Increasingly, applications in education, training, and entertainment are leveraging narrative to create rich interactive experiences in virtual storyworlds. A key challenge posed by these environments is building an effective model of the intervention strategies of director agents that craft customized story experiences for users. Identifying factors that contribute to determining when the next director agent decision should occur is critically important in optimizing narrative experiences. In this work, a dynamic Bayesian network framework was designed to model director agent intervention strategies. To create empirically informed models of director agent intervention decisions, we conducted a Wizard-of-Oz (WOZ) data collection with an interactive narrative-centered learning environment. Using the collected data, dynamic Bayesian network and naive Bayes models were learned and compared. The performance of the resulting models was evaluated with respect to classification accuracy and produced promising results.

Modeling narrative-centered tutorial decision making in guided discovery learning

Interactive narrative-centered learning environments offer significant potential for scaffolding guided discovery learning in rich virtual storyworlds while creating engaging and pedagogically effective experiences. Within these environments students actively participate in problem-solving activities. A significant challenge posed by narrative-centered learning environments is devising accurate models of narrative-centered tutorial decision making to craft customized story-based learning experiences for students. A promising approach is developing empirically driven models of narrative-centered tutorial decision-making. In this work, a dynamic Bayesian network has been designed to make narrative-centered tutorial decisions. The network parameters were learned from a corpus collected in a Wizard-of-Oz study in which narrative and tutorial planning activities were performed by humans. The performance of the resulting model was evaluated with respect to predictive accuracy and yields encouraging results.

Investigating director agents' decision making in interactive narrative: a Wizard-of-Oz study

Interactive narrative planning offers significant potential for creating engaging narrative experiences that are tailored to individual users. Orchestrating all of the events in a storyworld to create optimal user experiences calls for effective narrative decision-making. A key requirement of this endeavor is understanding the role that different knowledge sources play in narrative decision making. To investigate knowledge sources for interactive narrative, a corpus was collected in a Wizard-of-Oz (WOZ) study conducted with a narrative-centered learning environment. With narrative planning and natural language dialogue functionalities provided by wizards, the data from the WOZ study offers insight into the knowledge sources involved in narrative decision making and suggests how these knowledge sources can be effectively utilized by a narrative planner to create engaging interactive narratives.