Classifying Climate change misinformation using Large Language Models

A growing interdisciplinary literature demonstrates the role of digital communication in underpinning radicalization processes, mobilizing supporters of extremist ideas, and even promoting terrorist attacks. The ease of communication in online environments has fuelled the rapid dissemination of extremist material across a diverse range of platforms, where digital tools are increasingly exploited by an ideologically and organizationally diverse set of groups-from Salafi-jihadists to right-wing extremists. The growing scale and diversity of online extremist discussions poses serious challenges for security practitioners
tasked with understanding, monitoring, and responding to this communication.

My project will contribute to meeting this challenge by building on and extending existing computational approaches to understand extremist language. I focus on what is arguably the foremost theoretical concept in the study of extremist language: the tendency for sectarian discussion to be structured by sharply separate, immutable in-group and out-group labels. I begin by addressing the following ethodological question:

What are the most accurate approaches to identify specific in- and out-group actors in a large body of text across a diverse range of extremist groups?

While this may seem like a narrow methodological advancement, a computational approach to learn about the specific ways in which extremists mobilize particular group labels to structure their language is arguably a necessary first step to unlock a more comprehensive and sophisticated theory of extremist language. Though social psychologists and researchers of extremist communication rely heavily on the twin concepts of in- and out-groups, the way extremists actually refer to these groups in their language is surprisingly scarce. Moreover, the way extremists mobilize hate through the construction of particular sorts of groupings is even less theorized, with existing attempts falling short on several dimensions. As such, I will draw on existing theoretical approaches-particularly the work of Reicher and
colleagues - to address the following substantive questions on extremist communication:

Are there general linguistic markers for extremist language? To what extent is the language used by ideologically diverse groups similar or different?