AI Helps Climate Sceptics Accept Reliable Information
tilburg, dinsdag, 14 oktober 2025.
Researchers from Tilburg and Toulouse have demonstrated that AI-effective headline adjustments can reach climate sceptics with reliable information. By subtly adjusting the titles of climate articles, sceptics were more willing to read more news about climate change. Their views even shifted towards the scientific consensus, which is a promising step to bridge the gap between science and public perception.
Subtle Adjustments to Climate Headlines
The research, conducted by researchers from Tilburg University and the University of Toulouse, used a simulated social media environment and 2000 participants. The study showed that subtly adjusting article headlines led to greater receptivity to climate change. For example, the headline ‘The Foods That Can Reverse Climate Change’ was adjusted to ‘Precision Farming and Gene Editing: The Future of Food Production?’ [1]. These adjustments resulted in 11% more positive responses from sceptics and were saved for later reading 7% more often [2].
Effect on Climate Sceptics
The adjustments had the greatest effect on the most sceptical participants. After reading the adjusted articles, participants’ beliefs shifted towards the scientific consensus, ranging from 0.8 to 2.7 points on a scale of 0 to 100 [2]. This is a promising step to close the gap between scientific consensus and public perception, especially at a time when climate change is an urgent issue [GPT].
Applications of AI in Education
AI can also help improve public education in other ways. Chatbots for public services, such as in libraries, can provide personalisation and automate tasks. A lecture on AI in the Bart Ensink library on 12 October 2025 discussed the impact of AI on the library sector, focusing on personalisation, automation, and improved search functions [3]. Bart Ensink, an AI expert, emphasised that the challenge lies in finding a balance between technological progress and maintaining human interaction [3].
Benefits and Challenges
While AI offers promising opportunities, there are also challenges. Privacy, inclusivity, and reliability are crucial aspects that require attention. It is important that AI is implemented responsibly by trusted parties to maintain public trust [1][2][3]. Researchers such as Bence Bagó from Tilburg University stress that AI not only poses risks but also offers opportunities for better understanding and trust in science among the general public [1].
Concrete Example: Successful AI Applications
A concrete example of a successful AI application is the study where Llama3, a language model from Meta, was used to adjust newspaper headlines. The study included 58 articles, of which 28 remained after fact-checking and pilot testing [2]. The adjusted headlines did their job: on average, they led to 11 percentage points more positive votes from sceptics compared to the original versions [2]. This demonstrates that AI-effective headline adjustments can have a significant impact on the perception of climate sceptics.