Governing academic integrity: Ensuring the authenticity of higher thinking in the era of generative artificial intelligence

Synthesis: Coates, Croucher and Calderon argue that Academic Integrity in the era of Generative AI is a AI Governance problem before it is a detection problem, because contemporary academic governance is "not well positioned or poised" to handle threats to the authentication of student Assessment, which the authors place at the foundation of qualification integrity. The paper answers two questions: what information would help academic governors improve their work, and what reforms would ensure that information is used. Its answer to the first is an academic integrity indicator framework of 130 items under eight dimensions, developed from 2021 through a research review, five Australian university case studies, prototyping, expert confirmation and quantitative piloting. Its answer to the second is a reform programme targeting governance architectures, people, and technologies and resources, underwritten by external pressure from AI Regulation in Education, benchmarking and cross-institutional competition. The authors present the programme as formative and call for psychometric validation before the indicators are used for measurement or comparison.

Key Findings

1. Governance, not student behaviour, is the binding constraint. The authors describe assessment reform as "vexed, lethargic and devolved", typically decided for expertise-based reasons "with little input from institution-level governance", and conclude that digitalisation has carried the sector towards a tipping point. Academic governance is called remarkably resilient, meaning robust and flexible, yet not poised to meet GenAI-related threats to the authentication of assessment.
2. The framework holds 130 items under eight dimensions derived from a parameterisation of student assessment. The item pools run Designing (19), Developing (31), Training (19), Implementing (16), Analysing (7), Reporting (10), Evaluating (22) and Improving (6). The parameterisation is deliberately general so the indicators synthesise educational, institutional, technical and practical considerations across disciplines and levels.
3. The items are governance questions, not psychometric scales. Samples include whether the institution's top-most board or council receives updates on assessment processes and outcomes, whether key performance indicators cover assessment quality, what percentage of assessment resembles relevant and meaningful problems, whether induction and orientation include academic integrity, what percentage of students are known individually by the teachers who assess them, whether extreme low or high marks are cross-checked, and whether there is a simple process for referring contract cheating cases.
4. Expert confirmation reached 60 invited experts across six world regions, with 56 responses and nine in-depth interviews. Respondents were drawn from European, African, Middle Eastern, Eurasian, Asian, and North and South American systems, and included academics, university vice-chancellors and chief academic officers, evaluation and policy specialists, and governance experts. They affirmed the need and value of the indicators but questioned their prospects in a competitive information space, which the authors record as a dissemination and adoption problem rather than a design one.
5. The reforms target three internal levers plus one external one. Interventions are required in governance architectures such as policies and committees, in people including students and staff, and in the technologies and resources used for administration and learning, and these are unlikely to pay out without external affordance from regulation, benchmarking and cross-institutional competition.

How the framework was developed

The work began in 2021 and grew out of a larger programme of assessment reform research, so the design predates the late-2022 public release of GenAI. Five steps were run in sequence.

1. Multiyear review of existing research. The search was limited to the previous decade of work in globally exposed higher education systems, selected for their technology adoption, internationalisation and scale, with emphasis on the terms "university" or "higher education" and "generative AI" and "academic governance". The review was an instrument for insight into types and dimensions of misconduct, software platforms in common use, legal and educational framings, normative codes and standards, and regulatory activity, drawing on work such as Kier and Ives (2022), Dawson (2021), Davis (2023) and Harrad et al. (2024), alongside the regulator programme codified by TEQSA (2024a).
2. Multi-institutional case study analysis. Five institutions across three Australian states, all members of a national quality network, took part over several months during 2022. Each contributed management personnel, assessment materials, academic experts, students to complete assessment and feedback tasks, and executives to give formative feedback on reports. This step confirmed common practices and cultures, staffing and regulatory pressures, and blind spots, and revealed the lack of information on student assessment and integrity in particular.
3. Framework and data prototyping. The 130 items were detailed and organised under the eight dimensions, each with a named focus, an item pool from which a subset is sampled, and sample items.
4. Qualitative analysis with experts. The item pool was built into a website data collection and reporting tool and sent by email invitation during 2022 and 2023 to 60 experts, whose feedback covered integration with enterprise systems such as learning management, student information and similarity platforms, the coarseness of response scales, the diagnostic rather than judgemental character of the instrument, and its use as a benchmark for those in charge of assessment.
5. Quantitative analysis with multiple universities. Data came from the expert consultation and from small-scale application by the case study institutions and further sample universities. The authors state that this information was not designed to be representative or generalisable, educationally or statistically, and that it served to test the reporting metric. Results are scaled to a 100-point metric, with illustrative reports for two institutions each carrying sector comparisons, where sector values move with contextual covariates including institution type and the pilot's low sample size.

The eight dimensions

Designing covers academic design, governance and management activities (19 items); Developing covers academic and assessment development and production (31); Training covers development for people involved in assessment (19); Implementing covers implementation processes and conditions (16); Analysing covers data handling, marking and analysis protocols (7); Reporting covers academic and assessment reporting (10); Evaluating covers evaluation and detection activities (22); and Improving covers integrity improvement activities (6).

The instrument is broad by design because the authors treat assessment as the major touchpoint between students and the institution and the endpoint that yields the information aggregated to bestow credentials. They argue that collecting such data is relatively easy, and that the difficulty lies in creating a governance need for the intelligence and in having interpretative capacity to hand. Because the questions ask institutions to disclose their own practice, the framework's value depends on comparison across institutions rather than on audit alone.

Governance reforms: architectures, people, technologies and resources

The reforms were derived from governance-level work spanning a year, covering the design of a large institution's governance operations, cross-institutional consultation through benchmarking and collaborative research, and analysis of institutional reforms through Australia-wide consultation.

• Architectures. Infrastructure needs updating: a holistic institutional vision or position that makes the philosophical stance explicit, and committee structures that may need augmenting with standing or working groups holding specific expertise and oversight of academic integrity in the GenAI era. Policies must be updated along with the procedures and guidance notes that bring policy into practice, and integrity and cyber security expertise is treated as a core element of any governance capability matrix.
• People. Architecture change is "merely performative" without investment in training, so dedicated training and resources are needed for everyone holding a governance role, and formal courses on AI governance are beginning to appear. Students, academics and teachers, other staff, leaders, and parties outside the institution including employers and students' family and social groups are named as stakeholders with distinct interests; teachers and markers are described as integral to progress.
• Technologies and resources. Assessment reform is framed as imperative and ongoing, aided by indicator information, with course and delivery risk evaluation and multifaceted integrity and technology literacy education as priorities. The paper proposes borrowing red teaming from cyber security, applying an adversarial lens in which people try to break a process or system such as assessment integrity, to expose vulnerabilities and close the gap between prevailing practice and policy and frontier threats.
• Operations. The authors argue data is never by itself sufficient to substantiate a position or precipitate reform, so change to governance operations is required alongside the information, including a curriculum and assessment response from academics and convenors.

External pressure and the research agenda

External affordance matters because, in the authors' account, very little institutional development pays out even in large universities without it. Benchmarking, collaboration and pressure from quality and regulatory agencies are the levers. The paper cites emerging practice: TEQSA's analysis of submissions from all of Australia's registered higher education providers highlights integrity frameworks at the University of New England, Victoria University and Griffith University, working groups and committees at Monash University and the University of Western Australia, staff and student training across many institutions, and infrastructure transformation at Edith Cowan University and Western Sydney University, while the UK Quality Assurance Agency notes cases at Loughborough University and Oxford Brookes University. Quality and integrity agencies have also formed the Global Academic Integrity Network, and the Australian regulator asked every institution to report what it was doing about GenAI, which the authors credit with spurring national dialogue. Their longer-run proposal is a reputational economy built on academic Ethics rather than research reputation.

The paper is explicit about its limits. The quantitative evidence is a pilot, not a representative sample, so the authors call for further validation through more quantitative data subjected to psychometric validation, with benchmarking as the purpose. Conceptually they see a need for continuing review as digital threats change; technically they name integration with existing enterprise information systems as critical to generating engagement; and practically they note the absence of any sustainable international forum convening universities, governance and GenAI together.

Connected Concepts

• Academic Integrity — the paper reframes integrity as a governance responsibility for qualification authenticity
• AI Governance — the object of reform: boards, committees, policies, procedures and capability matrices
• Educational AI Policy — the policy and procedure layer that must be updated for the GenAI era
• AI Regulation in Education — external pressure from quality and regulatory agencies as a condition for reform
• Assessment — student assessment as the major touchpoint and the endpoint that yields credentials
• Assessment Validity — authenticity of student assessment as core to the validity of higher education
• Generative AI — the "apex technology" driving the tipping point the paper responds to
• Technologies — the platform, enterprise systems and tools the indicators ask institutions to account for
• AI Use and Disclosure Statements — the transparency and reporting gap the indicator framework is designed to fill
• Trust — academic ethics, honesty and fairness as the basis of qualification credibility
• Educational Measurement — the reporting metric, sector comparisons and psychometric validation still required
• Student Experience — students as stakeholders and co-producers of assessment integrity
• AI Literacy — integrity and technology literacy education named as a reform priority
• Change Management — governance reform as an organisational change problem in resilient institutions

Connected Articles

• Governing generative AI in higher education: a global Delphi study on policy and practice — Expert-consensus route to governing GenAI in higher education, complementary to this indicator approach
• Governing generative AI in higher education: Emerging policy approaches and support ecosystems at innovative U.S. universities — Policy-level analysis of how universities govern GenAI
• Policy Fragmentation or Institutional Alignment? Institutional Governance of AI in Universities and Business Schools — Institutional governance structures and AI, the layer the reform proposals target
• A Framework for Institutional Change in the Age of AI — A change framework for embedding AI reform in universities
• Generative AI as a Design Variable: An Evidence-Centered Framework for Principled Governance in STEM Assessment — Governance of GenAI in assessment, the operational core of this paper's indicators
• Purpose Before Policy: Academic Integrity, Generative AI, and Rhetorical Stance — Argues academic integrity needs purpose clarified before policy instruments
• How strong is the evidence in generative AI-related academic misconduct allegations? A mixed-methods analysis — Case-file evidence on what institutions actually hold when they allege GenAI misconduct
• The Integrity of Psychology Assessments in the AI Age: A Critical Examination — Programme-level evidence of assessment vulnerability that governance indicators would surface

Citation

Coates, H., Croucher, G., & Calderon, A. (2025). Governing academic integrity: Ensuring the authenticity of higher thinking in the era of generative artificial intelligence. Journal of Academic Ethics, 23, 2015–2028.