Paper Overview
- Field: NLP (Learning Sciences / Educational Technology)
- Author: Conrad Borchers
- Published: 2026-03-25
- arXiv: 2603.24535
- Problem: Existing methods cannot reliably measure scaffolding in real, naturalistic tutoring conversations, and this limitation is increasingly important as remote human tutoring and LLM-based tutoring systems proliferate.
- Approach: The framework aligns the semantics of dialogue turns with problem statements and correct solutions using embeddings, allowing scaffolding dynamics to be quantified across the course of a tutoring session.
- Significance: It offers a scalable, data-driven measurement tool for both human tutoring and AI tutoring systems, bridging learning sciences research and NLP.
Abstract
Adaptive scaffolding enhances learning, yet the field lacks robust methods for measuring it within authentic tutoring dialogue. This gap has become more pressing with the rise of remote human tutoring and large language model-based systems. We introduce an embedding-based approach that analyzes scaffolding dynamics by aligning the semantics of dialogue turns, problem statements, and correct solutions.
Summary
The paper proposes an embedding-based alignment framework for studying instructional scaffolding—the adaptive support a tutor provides to a learner. Key aspects: