Aggregation of student answers in a classroom setting

In a typical class, an instructor does not have enough time to poll all students for answers to questions, although it would be the best method for discovering students’ misconceptions. The aggregator module of a system called Classroom Learning Partner provides a solution to this problem by collecting answers students wirelessly submit on tablet PCs and placing them in clusters, which then are displayed to the instructor in histogram form. The student answers are compared, via syntactic parsing and similarity measures, to each other and to instructor-provided example answers to form clusters, which represent student misconceptions. In tests, the aggregator module consistently created relevant clusters, very similar to those created by humans working with the same data. Classroom Learning Partner, including the aggregator module, has been deployed successfully in an MIT introductory computer science class. Thesis Supervisor: Dr. Kimberle Koile Title: Research Scientist, CSAIL