A new article in Science News spotlights groundbreaking research led by CCPP Graduate Student Affiliate Mowafak Allaham, co-authored by CCPP Faculty Affiliate Dr. Ayşe Lokmanoğlu (Boston University) and CCPP Director Dr. Erik C. Nisbet. The study, titled “Enhancing LLMs for Governance with Human Oversight: Evaluating and Aligning LLMs on Expert Classification of Climate Misinformation,” was presented at the AAAI International Workshop on AI Governance (2025).
The research evaluates how large language models (LLMs)—including GPT-4 and fine-tuned versions of GPT-3.5—can be aligned with expert human oversight to identify false or misleading claims about climate change circulating online. Findings show that fine-tuning GPT-3.5 on expert-annotated data significantly improves its accuracy, achieving expert-level reliability in classifying climate misinformation on social media.
The Science News article, “Can AI help spot climate misinformation? These experts think so,” discusses how the study underscores the importance of combining artificial intelligence with domain expertise to ensure responsible AI governance in combating misinformation.