HCII PhD Thesis Proposal: Xu Wang, "Sourcing Student Open-Ended Solutions to Create Scalable Learning Opportunities"

Xu Wang

HCII PhD Student

Monday, May 13, 2019 - 1:30pm
Gates-Hillman Center 4405
Ken Koedinger (HCII), Co-chair
Carolyn Rosé (LTI and HCII), Co-chair
Jeff Bigham (HCII and LTI)
Chinmay Kulkarni (HCII)
Scott Klemmer (UCSD)


Getting deliberate practice is critical in developing skill mastery. However, providing quality support to learners requires a tremendous amount of time and effort from instructors. In college classrooms, for example, it is challenging for instructors to design sufficient high quality learning materials, including formative and summative assessments. In addition to the effort in designing learning materials, substantial grading effort is required when open-ended assignments are given, as they frequently are. From the students' perspective, such open-ended assignments do not always provide sufficient support for novices, and tend not to facilitate deliberate practice with immediate feedback. Techniques that could both magnify the investment of instructor effort while also improving student learning are highly desirable for scaling quality education.


To achieve these goals, I am developing UpGrade, a system to intelligently support instructors in creating scalable and quality learning opportunities. UpGrade takes past student open-ended solutions and past instructor feedback as input, and supports instructors in quickly creating multiple-choice questions at scale. On the one hand, UpGrade provides deliberate practice for students towards mastery learning with appropriate scaffolding, real-time feedback and repeated practice opportunities. On the other hand, UpGrade aims at reducing repetitive effort from instructors and complementing instructor expertise and effort with machine and crowd intelligence.


In my work thus far, I have focused on developing UpGrade to support student learning. I first conducted studies to understand the relative benefits of open-ended questions and multiple-choice questions in order to identify domains where multiple-choice questions can be beneficial for learning. A second study evaluated UpGrade's learning benefit through a 2-week classroom experiment. The study shows that using UpGrade reduces learning time by 30% without sacrificing learning outcomes compared to using traditional open-ended assignments. Furthermore, no manual grading is required.


To complete this thesis, I will continue developing UpGrade to support instructor usage and adoption and scale UpGrade to more classrooms. First, I plan to explore human-in-the-loop question creation in UpGrade through co-design with instructors. Second, I plan to conduct larger-scale classroom experiments at CMU to further investigate the learning benefit of UpGrade.


Queenie Kravitz

Upcoming Events