Search

We were interested in seeing if GPT4.0 explanations were better than explanations created by humans (teachers). Therefore we picked 100 problems with explanations that seem to be appropriate for this task from an intelligent tutoring database. We had 10 raters judge the quality of the explanation on a 5-point scale.

This is an example project for Neil Heffernan's User Modeling Class. This will be deleted at the end of the semester.

The experiment aims to keep students in conditions over a longer period of time by representing each condition as an associated educator to provide stronger claims on which educators are the most effective at generating student-supports. Additionally, it would allow us to see whether receiving feedback from the same educator may provide other benefits compared to a student-support condition of randomly delivering an available student-support.