Abstract
Universities govern AI adoption through artefacts. Policies, rubrics, assessment briefs, vendor evaluation frameworks, and design templates cross organisational boundaries and carry meaning differently in each context. This paper applies the concept of boundary objects within a cultural-historical activity theory (CHAT) framework to analyse which artefacts effectively mediate AI governance across academic, institutional, and commercial activity systems, and which fail. Drawing on the author’s framework for negotiated legitimacy in interacting activity systems and reflexive practitioner experience as an education design leader, the paper develops a typology of AI governance boundary objects classified by origin, interpretive flexibility, and failure mode. The analysis identifies three structural conditions that determine whether a boundary object supports or undermines the legitimacy of learning designers who broker AI adoption across systems. These are co-design across activity systems, support for transformation rather than coordination alone, and resistance to capture by a single system’s logic. It also identifies a failure dynamic specific to generative AI, in which an unstable referent causes artefacts to decay. The paper offers practical design principles for governance artefacts and connects them to a wider question, namely how institutions keep human and pedagogical judgement central as they adopt digital technologies.
Keywords
Boundary objects, AI governance, learning design, Activity theory, professional legitimacy.
