Case Study
Cognitive Tutors for Geometry
Submitted by:
Ken Koedinger, Carnegie Mellon University
Intervention Types
Process
Software
Applying AI and machine learning statistical algorithms to student learning data (collected by tutoring software technology) yield better cognitive models that predict student learning progress more accurately. Koedinger and colleagues asked whether such improved models actually lead to improved student learning outcomes. Working with a high school geometry tutoring program, they demonstrated that a redesigned tutor based on data-driven cognitive model improvements did help students reach mastery more efficiently. In particular, it produced better learning on the problem-decomposition planning skills that was the focus of the model. They conclude: “This is a great opportunity for AI and Education not only in mining educational technology data to discover better cognitive models, but in closing the loop by redesigning systems based on the resulting insights and testing them toward achieving better student learning.” (Koedinger et. al. 2013)
- Better cognitive models of learning, refined by insights from AI algorithms applied to learning data, can in turn, improve tutoring software to yield better student learning outcomes.
- Koedinger, et al., Using Data-Driven Discovery of Better Student Models to Improve Student Learning. In Artificial Intelligence in Education (16th International Conference, AIED 2013, Memphis, TN, USA, July 9-13, 2013. Proceedings.) Lane, Yacef, Mostow, and Pavlik. Springer-Verlag, 2013.
