A Mobile Application Front-End for Presenting Explainable AI Results in Diabetes Risk Estimation

XAI makes AI predictions understandable to patients: The research demonstrates that explainable AI methods like SHAP can be translated into formats non-experts can comprehend, moving beyond "black box" AI systems that provide predictions without justification—a critical step for building trust in health AI applications.

Visual and narrative explanations work better together: The application combined bar/pie charts showing risk factor contributions with personalized text narratives from GPT-4o, achieving an average user comprehension score of 4.31/5, suggesting that multiple explanation formats reinforce understanding better than technical outputs alone.

User-centered design is essential for responsible health AI: The developers prioritized user preferences through surveys to select visualization types, highlighting that responsible AI deployment requires understanding what formats actually work for the intended audience, not just what's technically possible.

Explainable AI can support preventive health decisions: By making diabetes risk factors transparent and understandable, the application empowered users to take preventive action—demonstrating that responsible AI isn't just about transparency, but about enabling informed decision-making that improves health outcomes.

Accessibility barriers remain in health AI adoption: The research addresses a real gap where many AI health applications exist but lack transparency, suggesting that without deliberate effort to make AI explainable and user-friendly, beneficial health technologies may fail to reach or help the populations that need them most.