Send message to Author

Paper short abstract

By ethnographically interacting with a range of Large Language Models, this paper introduces the concept of “synthetic extremism” to critically analyse the cultures of contentious exchange afforded by artificial intelligence (AI).

Paper long abstract

This paper introduces the concept of “synthetic extremism” to critically analyse the cultures of contentious exchange afforded by artificial intelligence (AI). Famously, affordance theory describes the social outcomes a technological artefact allows, for better or worse (Gibson, 1986). In relation to digital media, the concept of affordances has been widely critiqued as relational to human subjectivities and cultural setting (Davis, 2018), and in co-dependence with adjacent actant technologies (Hopkins, 2017). These theoretical evolutions, however, do not consider the complexities of affordances in relation to AI.

By ethnographically interacting with a range of Large Language Model (LLM) chatbot assistants, including major services such as ChatGPT, but also fringe products (Gab.ai), the paper aims to make three critical interventions to broaden the conceptualisation of the term "extreme speech” (Pohjonen & Udupa, 2017; Udupa & Pohjonen, 2019). First, synthetic extremism acknowledges the post-human affordance of extreme content online thereby problematising assumptions about what is speech in the first place. Second, the LLM’s that produce synthetic extremism inhabit a broader assemblage of digital technologies, situating extremity afforded by AI in the wider ecology of platform capitalism. Third, synthetic extremism is manufactured in collusion with user data, demonstrating how extremity is a non-homogenous phenomena, but afforded preferentially reflecting user body politic including age, class, ethnicity and gender. Given the proliferation of LLM's across global digital cultures, the article suggests more critical research on the relationship between AI and extremism is a matter of urgency, and aims to provide a framework for further analysis.