Newswise — Researchers at the University of Massachusetts Amherst are looking for ways to program computers to improve the computers' own teaching skills based on observed student behavior. Under a $1.2 million grant from the National Science Foundation to the department of computer science, a team of UMass professors is collaborating to improve "computerized tutors" for students in high schools.
Professors Beverly Woolf, Andrew Barto, Sridhar Mahadevan, and Ivon Arroyo of the computer science department are working with Don Fisher from the department of mechanical and industrial engineering. "Our motivation is to improve computerized teaching so that students are excited about and engaged in learning," Woolf says. "Students will be more active and engaged if tutors can customize their responses to the student's learning needs. Using 'machine learning,' a computerized tutoring system can predict student performance and then adapt problems and help for each student, similar to one-on-one human tutoring."
Human teachers improve with experience as they handle different types of students and learn which methods work for different learning problems, the researchers say. How then, can a computerized teacher improve its skills over time? Among the issues being addressed is how to make a computerized tutor recognize that certain problems are too difficult for a student and to identify which hints it should offer to help the student reach a solution.
In the current research, machine learning techniques are being applied to Wayang Outpost, a tutor developed at UMass Amherst as part of an NSF-funded project conducted by Carole Beal of the psychology department. This tutor for Scholastic Aptitude Test geometry problem solving uses graphics and animation to motivate students. It offers multiple hints, animated explanations and characters pointing to salient portions of the problem. The tutor customizes the choice of hints for individual students based on consideration of a student's cognitive profile, gender, spatial ability, and math-fact retrieval speed.
Large-scale experiments conducted in local high schools will also help researchers determine the practical significance of each enhancement to the tutor, as well as the effect of machine learning techniques on students' attitudes towards science, mathematics and engineering.
RSS