“Real proof instead of over hyped up bullshit”: Exploring the content of what would change people’s minds about climate change using OpenAI’s GPT

What do Australians believe would change their current opinions about climate change? To date, research has explored how political cues can influence the stability of climate opinions. However, minimal research has focused on the content that might alter the beliefs of different climate opinion holders. In this study, we used audience segmentation analysis to distinguish between distinct groups of climate opinion holders within a representative sample of Australians (N = 5110). We asked them what it would take to change their current opinions about climate change and leveraged OpenAI’s Generative Pre-Trained Transformer (GPT) to identify the presence or absence of five themes in their open-ended responses (Nothing, Evidence and Information, Trusted Sources, Action, and Unsure) plus a sixth Not Applicable category. GPT performed at near-human levels, proving to be a highly useful tool for thematic analysis. Our analyses revealed that strong climate denialists and believers tended to have more dogmatic responses to this question, with a greater likelihood of stating that nothing would change their mind. The results also highlighted the need for diverse forms of evidence and information and the importance of trusted sources of information across audience segments. These findings provide support for GPT’s utility as a tool for managing large datasets in the social sciences and offer participant-informed insights into climate opinion change.