ChatGPT-HealthPrompt. Harnessing the Power of XAI in Prompt-Based Healthcare Decision Support using ChatGPT

Domain expertise improves AI reliability in healthcare: The study demonstrates that incorporating medical knowledge from interpretable machine learning models into ChatGPT prompts enhances decision-making quality, suggesting that AI systems perform better when they're designed with input from domain experts rather than used as standalone tools.

LLMs can work with limited data in clinical settings: The research shows ChatGPT can perform effective binary classification tasks even in data-scarce scenarios through strategic prompt design, which is particularly valuable for rare diseases or specialized medical conditions where large datasets are difficult to obtain.

Prompt design matters significantly for medical AI: The paper emphasizes that how questions are asked—including task descriptions, feature explanations, and expert knowledge—substantially affects AI performance in healthcare, suggesting that careful communication with AI systems is as important as the technology itself.

Explainability supports clinical decision-making: By extracting feature importance from interpretable models and integrating it into prompts, the approach creates more transparent AI recommendations that clinicians can understand and verify, rather than "black box" suggestions they must blindly trust.

Few-shot learning offers flexibility for healthcare applications: The study explores how ChatGPT can learn from minimal examples (few-shot learning), potentially allowing healthcare systems to adapt AI tools to specific patient populations or institutional needs without requiring extensive retraining.