The University of Arizona

Events & News

Cohen, Beal, Chang Win $300,000 Award

November 24, 2008

Paul Cohen, Carole Beal and Yu-Han Chang won a one-year, $300,000 seedling project award from DARPA for work on educational technology.

One supposed advantage of intelligent tutoring systems is individualized instruction for each student. This is usually accomplished by a pedagogical strategy that considers the student's learning history and goals and then selects a problem, a hint, or some other educational material. As thousands or millions of students use these systems, and we gather data about their performance, statistical models can replace or augment conventional, knowledge-based
pedagogical strategies. Amazon does something like this when it recommends books. However, recommender systems like Amazon's do not try to construct an effective sequence of recommendations, as tutoring systems must. The problem of generating a sequence of events that maximize future rewards is called a sequential decision problem, and can be solved by a variety of algorithms. DARPA wishes to evaluate the feasibility of treating pedagogy as a sequential
decision problem.