Using Large Language Models to Provide Explanatory Feedback to Human Tutors
learningai-educationcognitive-science
Abstract
Research demonstrates learners engaging in the process of producing explanations to support their reasoning can have a positive impact on learning. This work-in-progress demonstrates considerable accuracy in binary classification for corrective feedback of effective, or effort-based praise responses.
Summary
Relevance to Cognitive Load Theory
This paper connects explanation-based learning to tutor training, with important CLT implications:
The Generation Effect
- Students who generate explanations learn more effectively
- This aligns with the CLT concept of germane cognitive load - productive mental effort that builds schemas
Tutor Training and Load
- Training tutors to provide appropriate feedback is itself a learning task with cognitive load considerations
- LLM-assisted feedback can scaffold tutor development without overwhelming them
Real-Time Feedback Design
- The challenge of providing accurate, domain-specific feedback in real-time relates to managing both tutor and student cognitive load
- Explanatory feedback (vs. simple corrective feedback) promotes deeper processing
Implications for Instructional Design
- Feedback should guide students toward self-correction rather than simply providing answers
- Explanation generation is a powerful learning strategy that builds robust schemas
- AI can assist human tutors in managing the cognitive demands of tutoring