Large language models for whole-learner support: opportunities and challenges.

Area of Science:

• Artificial Intelligence in Education
• Educational Technology
• Cognitive Science

Background:

• Large language models (LLMs) are rapidly advancing and being integrated into educational settings.
• Personalized learning environments that cater to the 'whole learner' remain an open challenge.
• Existing educational technologies often do not fully address both cognitive and non-cognitive student characteristics.

Purpose of the Study:

• To explore the potential of LLMs in creating personalized learning environments.
• To identify and address key challenges in leveraging LLMs for whole learner support.
• To outline a vision for AI tutors that adapt to individual student needs.

Main Methods:

• Discussing approaches for improving LLM interpretability regarding learner representations.
• Examining adaptive technologies for tailored pedagogical support using LLM insights.
• Highlighting methods and challenges in authoring and evaluating LLM-based educational agents.

Main Results:

• LLM interpretability can be enhanced by analyzing internal representations of learners.
• Adaptive LLM technologies can provide context-aware feedback and scaffold non-cognitive skills.
• Natural language instructions offer opportunities and challenges for specifying AI tutor behaviors.

Conclusions:

• Addressing interpretability, adaptation, and authoring challenges is crucial for effective LLM educational applications.
• Personalized AI tutors can significantly enhance learning by considering diverse student characteristics.
• Future work should focus on developing robust and interpretable LLM-based educational agents.