Making sense of AI: how lay users negotiate the shifting human–machine boundaries in talking about AI

Abstract

This paper examines how lay users make sense of artificial intelligence (AI) in everyday encounters and through practical reasoning and situated categorisation practices. Drawing on data from digital diaries and in-person group discussions, the study explores the local and situated ways lay users determine what kind of “thing” AI is, what it does, and how it should be judged. Rather than approaching AI as a fixed defined entity, the paper centres in participants procedures for meaning making. Specifically, the analysis focuses on three categorisation practices: the relational exclusion of the human-machine device, the assemblage of branched taxonomies, and the troubling of categorial boundaries. These practices reflect an ongoing negotiation of intelligibility, in which AI is explained not through formal and technically accurate definitions but through occasioned and socially recognisable use of every day categories. The paper draws on Membership Categorisation Analysis and Occasioned Semantics to show that the intelligibility of AI is both locally and culturally accomplished. In doing so, the paper reframes the notion of AI explainability as a situated achievement of human reasoning, rather than a system’s feature. An insight, we hope would have implications, if not at least a provocation, to the field of explainable AI and, more broadly, to the way we understand AI literacy, trust, and the everyday grammar of intelligence.