Automatic Generation of Problems and Explanations for an Intelligent Algebra Tutor

Intelligent tutors that emulate one-on-one tutoring with a human have been shown to effectively support student learning, but these systems are often challenging to build. Most methods for implementing tutors focus on generating intelligent explanations, rather than generating practice problems and problem progressions. In this work, we explore the possibility of using a single model of a learning domain to support the generation of both practice problems and intelligent explanations. In the domain of algebra, we show how problem generation can be supported by modeling if-then production rules in the logic programming language answer set programming. We also show how this model can be authored such that explanations can be generated directly from the rules, facilitating both worked examples and real-time feedback during independent problem-solving. We evaluate this approach through a proof-of-concept implementation and two formative user studies, showing that our generated content is of appropriate quality. We believe this approach to modeling learning domains has many exciting advantages.