An Interdisciplinary Closed-Loop Framework for Adaptive GenAI Support in Self-Study and Problem Solving

Résumé fourni par la source

This research project proposes an interdisciplinary closed-loop framework for adaptive Generative AI (GenAI) support in self-study and problem solving. The framework uses evidence from learner–GenAI interactions to estimate latent learner states, including knowledge mastery, progress trend, tendency toward AI dependence, and estimation uncertainty, and subsequently regulates the level, form, and timing of GenAI support. Three interdisciplinary regulatory mechanisms are explored within the common framework: control theory for difficulty and scaffolding adaptation; adsorption–desorption dynamics as a phenomenological model for the accumulation and decline of AI-dependency tendencies; and pharmacokinetic dynamics combined with the ALARA principle for regulating the level and timing of hints. These analogy-based models are not intended to directly represent physiological, psychological, chemical, or mechanical processes in learners. The simulations presented in this project are mechanistic illustrations intended to examine the qualitative behavior of the proposed models and identify quantities requiring future empirical calibration and validation. They should not be interpreted as experimental evidence from human learners. This research project was developed by Team 8 — Solar Orbit at the 13th Vietnam Summer School in Science (VSSS’13), Quy Nhon, Vietnam, August 2026.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.