Simulated Sense-Making or Social Knowledge? Artificial Intelligence and the Boundaries of Representation

Abstract

This article examines whether AI-generated texts—such as stories produced by large language models (LLMs)—can be considered social representations as defined by social representation theory. This paper argues that AI-generated outputs simulate communicative behaviour without participating in social processes of meaning-making. Although these texts contain familiar symbols, metaphors or narrative structures, they lack dialogical co-construction, intentionality and embeddedness in cultural practices. This paper introduces the concept of quasi-agents to capture the distinctive role that AI systems occupy in social interactions: entities perceived as social interlocutors, despite lacking genuine intentionality or social consciousness. This conceptual innovation extends social representation theory's analytical vocabulary, facilitating clearer distinctions between socially constructed meanings and algorithmically generated simulations. Misidentifying machine-generated texts as genuine social knowledge risks eroding the dialogical foundations of public discourse, particularly in education, media and policy contexts. Ultimately, meaning-making remains fundamentally a human and collective endeavour—one that AI may mirror but not originate.