PhD Dissertation Defense - Xiangmin Fan

Type PhD Dissertation Defense
Start Date January 18, 2017 09:00 AM
End Date January 18, 2017 11:30 AM
Location Board Room 6329
Speaker Name Xiangmin Fan
Speaker Affiliation Ph.D. Candidate, University of Pittsburgh
Abstract Scalable Teaching and Learning via Intelligent User Interfaces
The increasing demand for higher education and the educational budget cuts lead to large class sizes. Learning at scale is also the norm in Massive Open Online Courses (MOOCs). While it seems cost-effective, the massive scale of class challenges the adoption of proven pedagogical approaches and practices that work well in small classes, especially those that emphasize interactivity, active learning, and personalized learning. As a result, the standard teaching approach in today’s large classes is still lectured-based and teacher-centric, with limited active learning activities, and with relatively low teaching and learning effectiveness.

This dissertation explores the usage of intelligent user interfaces to facilitate the efficient and effective adoption of the tried-and-true pedagogies at scale. The first system is MindMiner, an instructor-side data exploration and visualization system for peer review understanding. MindMiner helps instructors externalize and quantify their subjective domain knowledge, interactively make sense of student peer review data, and improve data exploration efficiency via distance metric learning. MindMiner also helps instructors generate customized feedback to students at scale.

We then present BayesHeart, a probabilistic approach for implicit heart rate monitoring on unmodified smartphones. When integrated with MOOC mobile client applications, BayesHeart can capture and collect learners’ heart rates implicitly when they watch lecture videos. Such information is the foundation of learner attention/affect modeling, which enables a ‘sensorless’ and scalable feedback channel from students to instructors.

We also present CourseMIRROR, an intelligent mobile system integrated with Natural Language Processing (NLP) techniques that enables scalable reflection prompts in large classrooms. CourseMIRROR can 1) automatically remind and collect students’ in-situ written reflections after each lecture; 2) continuously monitor the quality of a student’s reflection at composition time and generate helpful feedback to scaffold reflection writing; 3) summarize the reflections and present the most significant ones to both instructors and students. Through a combination of a 60-participant lab study and eight semester-long deployments involving 317 students, we found that the reflection and feedback cycle enabled by CourseMIRROR is beneficial to both instructors and students.

Last, we present ToneWars, an educational game connecting Chinese as a Second Language (CSL) learners with native speakers via collaborative mobile gameplay. We present a scalable approach to enable authentic competition and skill comparison with native speakers by modeling both the interaction patterns and language skills of native speakers asynchronously. We also prove the effectiveness of such modeling in a longitudinal setting.
Committee Dr. Jingtao Wang, Department of Computer Science, University of Pittsburgh
Dr. Milos Hauscrecht, Department of Computer Science, University of Pittsburgh
Dr. Diane Litman, Department of Computer Science, University of Pittsburgh
Dr. Muhsin Menekse, School of Engineering Education, Purdue University