Skip to main navigation Skip to search Skip to main content

Responsibility gaps meet Wittgenstein

Activity: Lecture / PresentationAcademic

Description

In this paper, I apply a Wittgenstein-inspired approach to conceptual engineering for understanding responsibility gaps in relation to LLMs. More specifically, I draw on Wittgenstein’s notion of language-games to clarify the idea of artificial speech. I will distinguish between two types of language-games: one in which an LLM’s linguistic output counts as speech, and another in which it does not. On this view, neither concept is inherently better or more correct than the other; each fits its respective game equally well. Both games also resonate with familiar patterns of use, since in some contexts we are content to describe AI output as speech, while in others we are not. The key question, then, is which model of speech we should adopt for legislative purposes. I will argue that the normative force of conceptual engineering does not lie in the concepts or games themselves, but in the societal goals we set. In legal contexts, this means we should adopt the concept of speech that best serves those goals. On this basis, I will suggest that LLM output should not count as speech for the purposes of legislation.
Period16 Oct 2025
Event titleWorkshop on Conceptual Engineering and Technology
Event typeWorkshop
LocationTokyo, JapanShow on map
Degree of RecognitionInternational