Argues that bundling the authentication of student performance with its quantitative evaluation is no longer sustainable under generative AI, and proposes a New Deal Model built on student-owned, machine-readable portfolios.
Abstract
This paper analyzes the deep structural mismatch between generative AI (GenAI) and the organizational logic of higher education through the lens of institutional economics and organization theory. GenAI defies the traditional worker-or-tool binary assumed by legacy organizational processes, while universities’ vulnerability stems from their reliance on student-made artifacts as proxies for learning and their hesitancy to embrace decisive adaptation efforts that would upset their internal consensus. Endogenous institutional change thus pushes toward marginal patches, especially the addition of layers of new rules while the institutional foundations drift underneath. The paper argues that patching triggers a provenance-verification arms race associated with surveillance creep and pedagogical regression while still failing to restore confidence in legacy assessment practices. Exploring radical alternatives to the standing university model, the paper finds both complete GenAI rejection and its wholesale embrace lacking as mainstream replacements. Its main takeaway from this analysis is that bundling the authentication of student performance with its quantitative evaluation is no longer sustainable within the university. In response, the paper proposes a New Deal Model, which gives up in-house quantitative evaluation (i.e., grades) and focuses instead on documenting the student’s journey. The result is an extensive, machine-readable portfolio, sealed by the university, owned by the student, and unlocked with the student’s consent for downstream gatekeepers, who query it using their own GenAI systems in light of the specific opportunities they offer. Process documentation and a shift toward mentoring-heavy tutelage mitigate concerns about cheating and fabrication. The New Deal Model preserves the community aspect of higher education and the marriage of teaching, research, and public epistemic authority while increasing institutional GenAI compatibility and future-proofness: as GenAI advances, thin, grade-based datasets depreciate while thick, data-rich portfolios appreciate. The paper charts a gradual transition path in which portfolios temporarily run in parallel with legacy grading. Such a transition allows a side-by-side test of the paper’s central bet, namely, that GenAI evaluation of authenticated process records will predict real-world performance better than current grade-based assessment does.
Generative AIHigher educationInstitutional economicsAssessmentOrganization theoryInstitutional change
Citation
Špecián, Petr. 2026. “Built for Humans Only? Why AI Adoption in Higher Education Requires a New Deal.” SSRN working paper.