Applications of an example subtlety measure to concept teaching in intelligent tutoring systems.

摘要

One way an Intelligent Tutoring System (ITS) can teach a concept is by the presentation of positive and negative examples. The positive examples provide a sense of sufficiency and the negative ones circumscribe necessity. Learning a concept in this situation is essentially learning an efficient way to specify the boundary separating positive from negative examples. Generally, the subtler examples occur at the boundary where the most delicate distinctions between positive and negative examples are made. This dissertation proposes a measure of the subtlety of examples that can be computed using only the example descriptions and classifications. The subtlety measure has three significant ITS applications. The first is the Pedagogical Example Continuum, which refines the dichotomy of positive and negative examples and supports the selection and sequencing of examples. The second is the Pedagogical Goodness Function, which assesses how close a student's provisional concept is to the target concept. Finally, the Pedagogical Path connects the student's provisional concept to the target concept along a path of pedagogically appropriate hints that can be tailored to the student's ability. Together these applications support many teaching strategies of established efficacy and provide novel ways in which an ITS can tailor instruction to the individual student.