Analyzing big data to personalize education
Advancing the way instructors teach and students learn
Diane Litman, Professor
Asked to describe the optimal learning environment, both teachers and students would probably say one-on-one instruction. A custom-designed teaching process tailors the feedback to each individual student. In writing courses, peer and/or instructor reviews have historically played this role. Yet, how do we know if the peers consistently provide helpful insights? When it comes to tests, outcomes would improve if teachers could differentiate between a student who truly knows the material from one who made a lucky guess. Students of all levels — elementary school through college — can benefit from emerging educational technologies.
The research by Diane Litman uses artificial intelligence, specifically natural language processing and machine learning algorithms, to improve education.
With a joint appointment as a professor in the Computer Science Department and a senior scientist with the Learning Research and Development Center (LRDC), as well as holding a secondary appointment in the Intelligent Systems Program Litman applies her technical expertise to a range of educational issues. Her research has enabled students to improve their ability to write and review arguments, learn physics, and reflect on material at the end of a lecture.
She’s worked with colleagues from Pitt’s Department of Psychology, School of Education, and School of Law on a research version of the web-based reciprocal peer review system called SWoRD, which is used throughout the world to support writing instruction from high school through graduate school. SWoRD was developed by Professor Christian Schunn at the LRDC and is now licensed by Panther Learning Systems under the name Peerceptiv.
When using the research-version of SWoRD that was developed for a study sponsored by the Department of Education, students write an argumentative paper. The goal is to improve their argument construction and to help them to provide their supporting evidence in a coherent way. To increase the quality of the peer reviews, a natural language processing component evaluates the usefulness of these reviews.
This helps the student who wrote the paper and the student reviewer. If an auto-grade mechanism is built into the system, it could lead to tutoring the writer since teachers rarely have time to give detailed feedback on student drafts.
This same personalized interaction is built into the AI-based tutoring system she created to teach physics. The system provides a different question based on the student’s answer. Even if the student answered a question correctly, the technology can infer other things that might be more relevant such as if the student was guessing or if they were confident. “This isn’t meant to replace the teacher, but to supplement the instruction and to provide tutoring at any time,” explains Litman.
Her work has involved both voice-activated programs and those where the student interacts with the computer by typing. The beauty of voice-activated systems is the ability to monitor if the student is engaged by analyzing the person’s voice, the words they use, and the way they answer questions.
Before joining the university, she was a member of the Artificial Intelligence Principles Research Department, AT&T Labs – Research (formerly Bell Laboratories). “The technology is very similar, but in education, the assumptions of what the dialogue system should optimize are different so it pushes my research into a different direction from other people’s because the application has different assumptions.”
Diane Litman is a professor of Computer Science, senior scientist with the Learning Research and Development Center, and faculty director of The Intelligent Systems Program, all at the University of Pittsburgh.