Can LLM Code Explanations Adapt to Diverse Problem-Solvers' Needs?

LLMs show inconsistent ability to adapt code explanations to different problem-solving styles. The study found that while some models adapted their explanations more frequently than others, no LLM consistently tailored explanations across all five problem-solving style categories, suggesting current systems may not adequately serve diverse learners.

Thirteen specific linguistic adaptations were identified that can make code explanations more accessible. The research discovered concrete ways LLMs can modify their language—such as adjusting complexity, structure, or focus—to match different problem-solving approaches, providing a foundation for more inclusive AI-generated educational content.

One-size-fits-all explanations may hinder learning and workplace productivity. The paper emphasizes that when LLM explanations don't match how individuals naturally solve problems, people may become less productive at work and miss opportunities to develop their skills, highlighting a real equity concern in AI-assisted learning.

Inclusive design methods can guide AI adaptation efforts. By applying established inclusive design principles that consider five distinct problem-solving styles, the researchers demonstrated a systematic approach to making AI systems more responsive to human diversity, offering a model for responsible AI development.

Developers now have practical tools to improve LLM explanations. The study provides both declarative statements describing each adaptation and ten specific prompts designed to elicit problem-solving style adaptations, giving practitioners actionable resources to build more equitable code explanation systems.