Detecting Diabetes Misinformation: Leveraging Medical Expertise and NLP Models for Effective Detection

The proliferation of health-related misinformation on social media, particularly regarding diabetes, poses a substantial threat to public health decision-making and awareness. This study addresses the challenge of identifying false information about diabetes on Arabic social media platforms. A comprehensive dataset was collected directly from these platforms, with the verification of data authenticity conducted by medical professionals who accurately labeled the information. We subsequently evaluated various machine learning models for their effectiveness in detecting misinformation, achieving notable accuracy. Furthermore, the study incorporated an advanced Large Language Model (LLM), specifically GPT-4, to assess its proficiency in identifying fake news relative to the evaluations by our medical team. The findings highlight the efficacy of integrating human expertise with artificial intelligence to enhance the detection and mitigation of misleading health information.