Thu, Oct 8, 2026
[Archive]
Volume 7, Issue 3 (SEPTEMBER ISSUE 2026)
johepal 2026, 7(3): 147-153 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Safaei A, Safaei M. (2026). Beyond Advisory Governance: A Three-Layer Model for Agentic AI in Universities. johepal. 7(3), 147-153. doi:10.66224/johepal.7.3.147
URL: http://johepal.com/article-1-1953-en.html
URL: http://johepal.com/article-1-1953-en.html
Abstract: (8 Views)
- This paper proposes the Three-Layer University Agentic AI Governance Model (TUAIGM), a proposed governance model for universities that are beginning to consider agentic AI in areas where decisions directly affect students.
- The paper argues that agentic AI exposes three governance weaknesses: diffused accountability, goal misalignment, and behavioural drift. These risks are especially difficult in universities because authority is already distributed across board, administrative, and faculty structures.
- The TUAIGM assigns oversight responsibilities to three institutional layers and connects them through escalation, authorisation, reporting, and vendor-accountability flows.
- The framework has not been empirically tested. It is a proposed governance model that tries to make responsibility, monitoring, and intervention more explicit before agentic systems become embedded in university operations.
- The main contribution is applying principal-agent theory to the multi-principal university setting, where board, administration, and faculty share authority but do not always share the same priorities.
Keywords: Agentic AI, Higher Education Governance, Principal–Agent Theory, Accountability, Shared Governance, Student Rights
Type of Study: Research |
Subject:
Special
Received: 2026/06/8 | Accepted: 2026/09/16 | Published: 2026/09/30
Received: 2026/06/8 | Accepted: 2026/09/16 | Published: 2026/09/30
Online Supplementary File [PDF 1635 KB] (2 Download)
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
