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

PAOLO MONTI Università degli Studi di Milano Bicocca Dipartimento di Scienze Umane per la Formazione “Riccardo Massa” paolo.monti@unimib.it ABSTRACT Large Language Models (LLMs) are generative AI systems capable of producing original texts based on inputs about topic and style provided in the form of prompts or questions. The introduction of the outputs of these systems into human discursive practices poses unprecedented moral and political questions. The article articulates an analysis of the moral status of these systems and their interactions with human interlocutors based on the Habermasian theory of communicative action. The analysis explores, among other things, Habermas’s inquiries into the analogy between human minds and computers, and into the status of atypical participants in the linguistic community such as genetically modified subjects and animals. Major conclusions are the LLMs seem to qualify as authors that originally participate in discursive practices but do display only a structurally derivative form of communicative competence and fail to meet the status of communicative agents.

Introduction. Generative AI systems are becoming increasingly effective at producing original texts based on inputs about topic and style provided in the form of prompts or questions. Latest AI technologies, especially Large Language Models (LLMs) like GPT-4 by OpenAI or PaLM 2 by Google, develop their capabilities through a machine learning process that feeds on fragments of public discourse as found in internet webpages, books, and articles. These systems have become increasingly successful at engaging in areas of complex and specialized writing, like poetry or academia, with results sometimes indistinguishable from those of human writers. The diffusion of AI-generated discourse into the public sphere poses serious and unprecedented normative questions. Unlike other uses of AI technology, such as “deepfakes” – fabricated videos representing public figures in the act of saying words they never pronounced – AI writing cannot be reduced to a mere instance of forgery operated by human actors with the instrumental assistance of AI-based tools.

Discussion / Conclusion. The perspective of a pervasive presence of generative AI systems within the public sphere of liberal democracies inspires motivated concerns, especially at a moment in history when the advent of social media and the rise of populist movements haven’t yet exhausted their momentum and have abundantly shown how deeply technological transformations can affect the political realm (Sunstein 2017; Dijk and Hacker 2018; Urbinati 2019). It is to be noted, in this sense, that the interpretive framework sketched here, which sees AI systems are creative participants in highly sophisticated human practices without assigning them the full status of moral agents, can be applied also to other kinds of generative AI, beyond the case of LLMs. An obvious example are visual AIs like DALL-E by OpenAI, Midjourney by Midjourney Inc., and Stable Diffusion by Stability AI. The visual creative practices where these systems express their authorial capabilities are not as central in the Habermasian account as the medium of language and discourse are.