Whose story wins? LLM-powered chatbots as sites and agents of memory-political contestation and corporate greenwashing

Abstract

Abstract In this article, we explore the memory-political dimensions of LLM-powered chatbots. We are initially interested in how LLMs (large language models) can be utilised to manipulate political discourse and historical interpretations. By expanding on the idea presented by, for example, Makhortykh and colleagues (2024), we examine how open, commercial LLMs generate information on current political topics with contested historical and political dimensions in the Nordic region by examining the outputs of three different models – ChatGPT, DeepSeek, and Mistral – in Finnish, Swedish, and English. Our focus is on three themes that are intertwined in identity and memory-political contestations in the Nordic region: climate change, security politics, and colonialism (e.g., regarding Indigenous communities). In this context, we study whether and how LLM-powered chatbots mediate various actors’ contesting narratives engaging in and functioning as sites and agents of memory-political contestations. We discuss the ideological and epistemic dimensions of LLM-powered chatbots at a time when generative artificial intelligence (GenAI) increasingly impacts memory discourses in democratic societies.

Ähnliche Arbeiten