AI evidence extraction

Artificial Intelligence and Academic Integrity in UK Higher Education: A Rapid Literature Review of Emerging Challenges, Institutional Policy Responses, and Future Research Directions

Aimesh Shaheen, Aimen Sadheer · Physical Education Health and Social Sciences · 2026

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

9/10

Relevance

1/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.63163/jpehss.v4i2.1400

Methodology & findings

Study design

Rapid literature review (RLR) design following PRISMA 2020 reporting principles adapted for rapid review methodology.

Sample

N = 38, 4 groups

Primary method

Thematic synthesis (not meta-analysis). Qualitative screening and coding of included sources based on inclusion/exclusion criteria. PRISMA 2020 reporting principles adapted for rapid review methodology.

Main result

The literature reveals a sector confronting a genuinely transformational challenge: "the scale of student AI use has already exceeded any reasonable expectation of universal prohibition. With 88% adoption among UK students, policies premised on AI-free assessment are increasingly detached from the reality of student behaviour and risk compelling widespread rule-breaking that delegitimises integrity frameworks as a whole." Additionally, "detection-centred integrity strategies are technically unreliable, ethically problematic, and strategically inadequate," with "false positive rates of up to 14% mean that substantial numbers of students who have not used AI may face misconduct proceedings on the basis of probabilistic algorithmic judgements."

Reports effect sizes.

Research paradigm

Interpretivist/Qualitative with policy analysis orientation

Author conclusions

The authors conclude that "the arrival of generative AI in higher education is not primarily a misconduct problem to be policed but an educational challenge to be addressed through fundamentally reimagined approaches to assessment, learning, and the cultivation of student capability." They further argue that "institutions that respond to AI primarily through detection, prohibition, and punishment risk expending substantial resources on an unwinnable technical arms race" and that "the central argument of the emerging literature, which this review endorses and extends, is compelling."

Risk of bias

Single-reviewer screening (not dual independent review); Rapid review design may introduce selection bias due to time constraints; Publication bias toward English-language sources; Potential for screening bias in 80% of records not double-checked; Exclusion of non-peer-reviewed sources and opinion pieces may limit perspective diversity; Single-reviewer screening (only 20% double-checked); Potential publication bias in selected literature; Time-bound search strategy may miss emerging evidence; Language restriction to English may exclude relevant international scholarship; Rapid review methodology trades comprehensiveness for timeliness; Single-reviewer screening (mitigated by 20% double-check verification); Rapid review design without full systematic review rigor; Publication bias toward published literature (policy documents and peer-reviewed articles may not represent all institutional responses); Geographic bias toward English-language sources; Temporal bias: rapidly evolving field means early 2026 publication captures incomplete evidence base

Limitations

The authors acknowledge that "The trade-offs of the RLR design, including the absence of meta-analysis and the use of single-reviewer screening verified through a random 20% double-check, are acknowledged as limitations." Additionally, "There is a near-complete absence of longitudinal evidence on the effectiveness of AI integrity policies
Almost all existing studies are cross-sectional, conducted at a single point in time in a rapidly changing environment." Furthermore, "UK-specific empirical research on student and staff experiences of AI integrity policy remains thin
The majority of empirical studies cited in this review draw on US, Australian, or broadly international samples."

Open questions raised

Near-complete absence of longitudinal evidence on the effectiveness of AI integrity policies - almost all studies are cross-sectional
UK-specific empirical research on student and staff experiences remains thin; majority of empirical studies draw on US, Australian, or broadly international samples
Equity implications of AI policy remain underexplored, particularly regarding socioeconomic background, disability status, and international student status
Limited research on perspectives and experiences of academic staff as frontline implementers of integrity policy
Lack of comparative research across different institutional policy models to identify which approaches produce better outcomes under what conditions
Near-complete absence of longitudinal evidence on the effectiveness of AI integrity policies

Extracted from: pdfAgreement 64%

Explore related topics

Institutional Policy Research Integrity Literature Review