AI Enters Public Discourse: A Habermasian Assessment Of The Moral Status Of Large Language Models

The limited scope of this article aims to highlighting which insights can be drawn from Habermasian theory and what status can be assigned to LLMs that participate in discursive practices with humans in terms of responsibility for what they generate in that context. In recent years, Jürgen Habermas has discussed some of the implications of the new technological infrastructure of communication based on the internet and social media for the public sphere and deliberative democracy (Calloni et al. 2021; Habermas 2023). He has not substantially engaged, on the other hand, with the possibility that digital technologies could soon also produce a new kind of non-human actors of public discourse and deliberation. His vast philosophical project, however, offers relevant conceptual resources to attempt this undertaking as well. This account begins by looking at two areas, mutually connected but articulated in Habermas’s works at different times: first, the tension between the communicative origin of the person and the naturalistic understanding of the mind as a computer (2); second, the moral status of atypical members of the community of communicants like genetically modified individuals and animals (3). We will explore these two areas, to then attempt a characterization of the hybrid status of LLMs within our discursive practices (4) and outline a preliminary normative account of the moral responsibilities at play when fragments of discourse produced by LLMs enter public discourse and deliberation (5). The conclusions will briefly discuss how this account may fit within a larger consideration of the future impact of generative AIs on the ethics of democratic citizenship (6).