Istanbul Medipol University Contributes to Comparative AI Health Study
Researchers from the Social Pediatrics Department at Istanbul Medipol University have contributed to a comprehensive study evaluating the use of Artificial Intelligence in pediatric healthcare. The study, published in scientific journals, conducted a comparative analysis of major Large Language Models (LLMs)—including ChatGPT-4o, Gemini 2.5 Pro, Grok-4, and DeepSeek-V2—to test their reliability and readability when dispensing pediatric health advice.
The research assessed how these AI platforms responded to real-world, caregiver-style prompts. The findings revealed that while current LLMs can provide moderately useful health information and avoid overtly unsafe advice, they struggle to meet standard health-literacy targets. Responses often fluctuated between being highly reliable but too complex for average readers (e.g., college-level reading requirements) or highly readable but lacking structural rigor, highlighting the need for refinement before AI can be safely used for routine patient education.
Source: Nature / ResearchGate
No comments:
Post a Comment