This project addresses a core challenge in educational AI: how do we reliably compare the computational models that estimate what students know? Intelligent Tutoring Systems and adaptive platforms rely on student models — such as Bayesian Knowledge Tracing, Item Response Theory, and Deep Knowledge Tracing — to personalize learning. Yet comparing these models is difficult because studies vary widely in datasets, preprocessing, and evaluation methods.
You will build a unified benchmarking framework that standardizes how student models are evaluated across datasets and methods. The goal is a reproducible pipeline with a common data format, integrated modeling approaches, and consistent evaluation metrics — making it easier to determine which models best capture real learning processes under different metrics.
- begleitende Lehrperson: Irene-Angelica Chounta
- verantwortliche Lehrperson: Mona Münstermann
- verantwortliche Lehrperson: Kaimao Sheng