Discursively negotiating AI: A social representation theory approach to LLM-based chatbots

To date, users have not merely interacted with large language model (LLM)-based chatbots. Notably, they collectively discussed about them, flooding the online information ecosystem with a sheer volume of social media posts about LLM-based chatbots. Despite research on users' reception of this equivocal technology is on the rise, it is mainly rooted in positivist and functionalist paradigms, leaving a finer-grained understanding of how early adopters collectively make sense of such novel and unfamiliar technology in dedicated online environments elusive. Drawing upon Social Representation Theory, this study employs a computationally grounded analysis of user-generated content to investigate how the social representations of LLM-based chatbots formed in online communities. Findings reveal that users, through different discursive and emotional anchoring and objectification mechanisms, represent the LLM-based chatbot as a “creative partner”, a “multistable artifact”, a “connective hackaton”, and a “technology of power”. This work contributes to the emerging literature about LLM-based chatbots acceptance by unveiling how users discursively make sense of such unfamiliar social objects, and how they renegotiate the agentic roles of both actants involved in human-chatbot interactions. It showcases an original text-mining protocol to study social representations based on social media data; and it offers managerial implications to AI service providers and policy makers.