Researchers Build AI System to Match Social Media Health Claims to Actual Science
Health claims flood social media daily: fruits shouldn't pair with milk, protein shakes damage kidneys, a 30-day workout fixes PCOS. Each comes with charts, citation links, and enough scientific language to sound credible. Few people click through to verify the source.
The disconnect between what scientists publish and what circulates online is the problem Ritwik Banerjee set out to address. Banerjee, a research assistant professor of computer science at Stony Brook University, works on a question that matters for anyone reading research: "How do we get from what scientists actually know to what people are saying online?"
"Right now, that bridge is very noisy," Banerjee said. "Claims slip, get simplified or just made up."
Building a verification tool
Banerjee's team-including graduate student Parth Manish Thapliyal, undergraduates Ritesh Sunil Chavan and Samridh Samridh, and Chaoyuan Zuo at Nankai University-developed an automated system to solve this gap. The work was submitted to CheckThat! 2025, a competition organized under the CLEF conference in Madrid.
The system does one specific job: when a social media post cites "science," it determines whether an actual research article matches the claim being made.
Most people assume citations are honest, especially when claims align with what they already believe. Banerjee said the field needs "reliable tools to trace what's real."
For researchers and professionals who evaluate information for their work, this type of verification matters. Misinformation spreads fastest through posts that sound scientific without being scientifically sound.
Learn more about how Generative AI and LLM systems are being applied to research and verification tasks, or explore AI Research Courses that cover these applications.
Your membership also unlocks:
