De-CoDe Risk. A Novel Tool for Identifying Climate Change Denial Processes on Social Media

Socio-psychological processes, such as denial, moral disengagement, and psychological distance, can impede the adoption of effective coping behaviours in response to risk events, such as climate change. To build effective risk communication, crucial for risk management, it is essential to recognize and address these processes in a timely manner. In this contribution, we employ the Risk Co-De model (Author, year) to identify psychosocial processes associated with climate change risk denial, as articulated in natural language within social media communication. This system combines various nuances of consciousness, disengagement, distance, and denial into four distinct macro categories and nine micro categories. We employed Mistral7b, a large language model based on artificial intelligence and instructed using the Risk Co-De model, to categorize 21,042 English tweets related to climate change. The tweets were collected between February 18 and October 10, 2019, a period during which the discourse on this topic was highly active on the social media platform Twitter (now X), thanks to the Fridays for Future movement. The classification result underwent manual verification to assess its coherence to the Co-De categories, revealing a satisfying accuracy in the classification task. Additionally, tweets exhibiting forms of denial were scrutinized to delve deeper into how these processes are expressed in natural language in English. The discussion of the results will center on the contribution that studying psychosocial processes in natural language through social media can provide to the understanding of risk perception.