Multi-Level Barriers to Generative AI Adoption Across Disciplines and Professional Roles in Higher Education

URL SCAN: Multi-Level Barriers to Generative AI Adoption Across Disciplines and Professional Roles in Higher Education
FIRST LINE: arXiv cs.CY > Computers and Society

THE DISSECTION

This paper performs administrative theater dressed as empirical research. It surveys 272 staff at a Russell Group university about their perceived barriers to GenAI adoption, applies MLR, SEM, and semantic clustering to open-ended responses, and concludes that barriers are "structurally produced" by disciplinary context and institutional role.

What it's really doing: Cataloguing friction surfaces in a system that is already dead — higher education's slow-motion surrender to cognitive automation — while treating the symptoms as if they were the disease. The authors use the word "structurally produced" as though they've discovered something beyond individual-level factors, but they never interrogate why the structure itself is producing these barriers, which is to say, what economic logic is compressing institutional capacity to absorb AI in ways that produce exactly these friction patterns.

They find non-STEM academics report ethical/cultural barriers (academic integrity); STEM and PSs staff report institutional/governance/infrastructure constraints. This is a taxonomy of delay, not an analysis of displacement.

THE CORE FALLACY

The paper treats "adoption barriers" as the key variable — as if the problem is that universities are failing to adopt GenAI fast enough due to organizational dysfunction, and role-specific governance frameworks will solve this. This inverts the causality.

The real mechanism: Higher education is not failing to adopt AI quickly enough. It is being systematically hollowed out by AI in ways that make rapid adoption catastrophic for its existing labor architecture. The barriers aren't bugs — they are the last structural defenses of a workforce that knows, at some level, that adoption means replacement.

Non-STEM academics raise "academic integrity" concerns because their labor is most exposed: content generation, assessment, essay evaluation — the exact tasks AI performs at near-zero marginal cost. STEM academics and professional services staff focus on governance and infrastructure constraints because their labor mix includes more technical functions where AI is still partially contested. The barrier taxonomy is a map of displacement vulnerability.

The fatal assumption: The paper assumes universities can engineer their way through these barriers via "role-specific governance and support frameworks." This is like suggesting coal plants can be salvaged with better maintenance protocols. The structure isn't producing barriers — the structure is the barrier, because the institution's survival as a mass employer of human cognitive labor is structurally incompatible with the technology.

HIDDEN ASSUMPTIONS

1. GenAI adoption in education is desirable. The paper never questions whether adoption, as currently conceived, serves the institution's actual survival or merely accelerates its decomposition.
2. Universities are the relevant unit of analysis. By focusing on institutional roles and disciplinary contexts, the paper treats universities as closed systems capable of managing their own transition. They are not. The relevant forces are external to any individual institution — market pressures, credentialing dynamics, labor market signals.
3. Training and governance are meaningful interventions. The conclusion pivots to "role-specific governance" as the policy lever. This is the academic institution's reflexive reach for procedural solutions to structural displacement.
4. Survey data captures structural reality. Asking people about "perceived barriers" surfaces the domain of managed, articulable resistance — not the latent economic logic driving institutional paralysis.

SOCIAL FUNCTION

Transition Management / Lullaby. This paper performs the institutional version of organizing the deck chairs. It tells university administrators that the problem is knowable, divisible, and tractable — that the barriers can be mapped and addressed through governance reforms. It provides intellectual cover for continued investment in process-level responses while the underlying economic displacement accelerates. The 272 surveys are theater that signals seriousness without confronting the terminal diagnosis.

THE VERDICT

Higher education is not experiencing an adoption crisis. It is experiencing a structural contradiction: the technology that most threatens its core functions (content transmission, credentialing, assessment) is the technology it is being pressured to integrate into those same functions. The barriers this paper documents are not pathologies to be cured — they are the institution's last, structurally produced defense against its own obsolescence.

"Role-specific governance frameworks" will not resolve this. The universities that survive will be those that either (a) become AI deployment platforms for sovereign capital (STEM-adjacent research infrastructure), or (b) maintain credentialing monopolies enforced by regulatory capture — not through better training programs.

The paper's conclusion is empirically rigorous and structurally irrelevant. You cannot engineer around a mathematical displacement mechanism.