Structured AI prompts produced more relevant facts for climate misinformation corrections

A study comparing three AI systems found that breaking the task into steps improved the relevance of facts used in debunkings.

Abstract editorial illustration accompanying the article: Structured AI prompts produced more relevant facts for climate misinformation corrections

Editorial illustration — not from the study.

The study combines large language models with climate misinformation categories and methods for detecting reasoning errors. It compares a basic system that uses one prompt with a structured system that divides debunking into subtasks, and a third system that adds information from authoritative external sources.

All three systems performed similarly overall. The structured system produced more relevant facts than the basic system, while the version using outside sources improved at identifying and explaining the reasoning errors behind misinformation.

What the AI systems produced

The researchers evaluated three AI-based debunking systems: a single-prompt “Vanilla” model, a “Structured” model that divided the task into subtasks, and a “Structured+RAG” model that added information from authoritative external sources. Overall, the models performed similarly. The Structured model was better than the Vanilla model at producing facts relevant to the misinformation being addressed. The Structured+RAG model improved at identifying and explaining misinformation fallacies, or flawed lines of reasoning.

Why structured corrections matter

Climate misinformation can spread quickly and widely, making entirely manual correction difficult to scale. The findings suggest that how an AI system organizes a debunking task—and whether it draws on curated, authoritative sources—can affect the relevance of its facts and its explanations of misleading reasoning. The study connects these design choices with psychological guidance on effective corrections, such as providing relevant facts and explaining fallacies.

Evidence and open questions

This is a comparative study of three large language model-based systems, evaluated using expert-annotated data designed to cover diverse climate misinformation. The abstract does not report the dataset size, the specific sources used, or whether the generated corrections changed people's beliefs or behavior. The results therefore show differences in the systems' outputs, not evidence that AI-generated debunkings work in real-world social media settings or at large scale.

Tuna and billfish fishing moved offshore mainly in the past 200 years

People have fished for tuna for about 44,000 years and billfish for about 10,000 years, according to a global review of archaeological evidence. But sustained, intensive fishing in offshore and oceanic waters appears to have developed mainly during the past 200 years.

10 hours ago·5 views

Abstract editorial illustration accompanying the article: Coastal residents weigh flood damage, costs and family ties when considering a move

Society & Economics

Coastal residents weigh flood damage, costs and family ties when considering a move

People considering whether to leave flood-prone coastal communities weigh disastrous flooding and rising costs alongside crime, healthcare and proximity to family and friends. In a survey of 182 respondents, most said they would not relocate, while others said they would move immediately or in the future.

18 hours ago·6 views

Society & Economics

Managers accepted reflective coaching from humans but wanted answers from AI

A study of eight managers found that they preferred human coaches for reflective questioning, even though they welcomed that approach from a person. When the same coaching protocol came from an AI system, they wanted it to provide answers instead.

1 day ago·19 views