DIP: Exploiting Longitudinal Electroencephalogram (EEG) Input in a Reading Tutor

PIs: David Mostow, Kai-Min Chang
Carnegie-Mellon University

Automated (and human) tutors are limited in their ability to infer what is going on in students’ heads based on their observable behavior. The proposed work addresses this limitation by investigating how EEG input from a commercially-available device can be used as evidence about students’ mental states. In particular, the project focuses on adding EEG-enhanced feedback to Project LISTEN’s Reading Tutor, an intelligent tutoring system that helps children learn to read. The project seeks to answer two questions: (1) How can we use EEG to detect mental states that predict, indicate, or reflect student learning? (2) How can we use such detection to improve student learning? Analysis to answer these questions and to enhance the capabilities of the Reading Tutor draws on existing tools to explore annotate, and mine EEG data logged by the Reading Tutor. The research aims to tell us more about how to use EEG to identify mental states that predict learning and to use machine learning to make an intelligent tutoring system better, and it may also add to what is known about sources of reading difficulties. Expected technological contributions of this work include advances in relating EEG data to children’s behavior, cognition, engagement, and learning and advances in elucidating how intelligent tutors can robustly exploit noisy EEG input to better assist learning.

The technological innovation in this project is particularly important for those children who need extra help with sounding out, word recognition, and/or making simple inferences needed for understanding.

Tags: , , , ,