ChatGPT, a colonialist agent of lifeworlds: A Habermassian analysis of conversations

Abstract

Conversational generative AI-based systems (CGAIS), like ChatGPT, seem capable of taking part in conversations with such fluidity that we may not distinguish them from a human. The global integration of CGAIS, however, bears various risks, including the one of colonizing the social through data. In this article, we are interested in the colonization of a new territory: that of conversations. We therefore investigate the nature of conversations individuals have with these technologies. Following Habermas, the realm of the communicative is what enables to construct a common social world, a lifeworld. Conflating the communicative and the instrumental threatens lifeworld construction and more broadly our democracy. We explore CGAIS technical and conceptual properties, and show that they do not support communicative action. Rather, CGAIS should remain confined to the realm of instrumental action. Yet, as they give the impression to belong to the realm of the communicative, we conceptualize them as colonialist agents. Notably, they are imperialist agents because they are increasingly used for all types of activities at work and in the private and public spheres. This reduces the space for communicative action. In addition, they are derealization agents as they distort conversation, by giving the illusion of communicative action. As a result, they threaten the co-construction of a common lifeworld, that forms the basis of our democratic societies.