MA Defence: Bahareh Izadi, Philosophy

Beyond the Imitation Game: The Failure of LLMs to Participate in the Social Practice of Giving and Asking for Reasons

Bahareh Izadi

Supervisor: Ulf Hlobil

ABSTRACT: My thesis argues that Large Language Models (LLMs), despite their linguistic capabilities, fail to meet the constitutive requirements for participation in the social practice of giving and asking for reasons. Drawing on Brandom's Inferentialism, I contend that this practice is irreducibly social and governed by norms that determine not merely the success of one's moves in this practice, but whether one is genuinely playing this game at all. A person may be extraordinary at throwing a ball, yet if they use their hands, what they do, however skillful, is not playing soccer. The norms constitutive of soccer categorically exclude such conduct. The same categorical logic applies to discursive practice: the game of giving and asking for reasons admits participants only insofar as its distinctive normative constraints bind them. The constraining force of these norms can bind only a handful of participants; those who are members of a normative community.

While this thesis may appear far-fetched given the trajectory from early chatbots like ELIZA to contemporary LLMs such as Claude and GPT-5; I address these challenges by examining what participation in discursive practice requires. My conclusion is that the implications of this topic extend well beyond questions about Artificial Intelligence (AI), concerning reason and rationality itself.

To attend, click the button below: