AI evidence extraction

AI Writing Tools in Higher Education: A Narrative Review and Conceptual Framework for Ethical Use

Rowaidah AL Abdullah, Zakariya Al Shaaibi, Al Shaima Al Mahrami · International Journal of Drug Delivery Technology · 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/10.25258/ijddt.16.41s.114

Methodology & findings

Study design

Narrative review methodology.

Main result

The paper synthesizes literature on AI writing tool adoption across higher education, finding that "a majority of undergraduate and postgraduate students have used ChatGPT or comparable tools at least once for academic purposes, with usage rates ranging from 55% to over 80% depending on discipline, institutional context, and national setting." The review identifies genuine pedagogical benefits, noting that "students using AI tools generate first drafts faster and spend proportionally more time on revision, argumentation, and critical engagement with their sources than on initial generation tasks," and that "iterative engagement with AI-generated feedback can support metacognitive development." However, it also documents significant academic integrity challenges: "AI-generated content constitutes a categorically different problem: the text is original in the technical sense, it is not copied from any document, yet it is not the product of the student's intellectual labor."

Reports effect sizes.

Research paradigm

Interpretivist/Qualitative synthesis with theoretical grounding

Author conclusions

The authors conclude that "The integration of AI writing tools into higher education represents one of the most significant pedagogical challenges of the current decade, not because AI is inherently harmful to learning, but because its integration is proceeding faster than institutional capacity to govern it thoughtfully." They argue "the path forward lies neither in prohibition nor in uncritical endorsement, but in principled, differentiated governance grounded in clear conceptual distinctions about what different modes of AI use entail for student learning and academic integrity." They assert that "The three-level ethical framework, distinguishing Assistive, Augmented, and Delegated modes of AI writing tool use on the basis of student cognitive involvement and ethical risk, provides such a basis." The authors conclude that "the task before educators and policymakers is not to answer the question of whether students should use AI, but to develop the conceptual and institutional infrastructure needed to answer with precision: under what conditions, in what ways, to what extent, and with what transparency?"

Risk of bias

Selection bias: Non-systematic methodology with interpretive judgment in source selection and weighting; Publication bias: English-language literature only; limited to peer-reviewed sources, edited volumes, and recognized institutional bodies; Geographic bias: Predominantly US, UK, and Australian contexts; underrepresentation of Global South, Arab region, and non-Anglophone institutions; Temporal bias: Window limited to January 2022–March 2025; foundational scholarship selectively included; Social desirability bias: Self-reported usage rates for AI tools may be suppressed in contexts with prohibitive or unclear institutional policies; Confirmation bias: Framework proposed without empirical validation; theoretical grounding may reflect authors' existing assumptions; Narrative review methodology introduces interpretive judgment bias in source selection and weighting; English-language only inclusion criterion excludes non-Anglophone scholarship; Temporal constraint (January 2022-March 2025) may miss foundational works and recent developments; Geographic bias toward US, UK, and Australia contexts; Social desirability bias in self-reported student AI use acknowledged by authors; Selective inclusion of foundational scholarship from prior periods; Selection bias through inclusion criteria (peer-reviewed sources only; English language only); interpretive bias in thematic analysis (researcher judgment in source selection and weighting); geographic bias (predominantly Anglophone contexts); temporal bias (lower bound of January 2022 means earlier foundational scholarship included selectively); publication bias (opinion-based editorials excluded unless authored by recognized experts)

Limitations

The study acknowledges multiple limitations: "First, the methodology is narrative rather than systematic
While the thematic synthesis reflects careful, iterative analysis of included literature, it does not employ the quantitative aggregation procedures of a formal systematic review or meta-analysis." Second, "the literature base is temporally constrained in both directions
The lower bound, January 2022, means that relevant foundational scholarship on AI in education from prior periods is included only selectively." Third, "most significantly, the proposed framework has not been empirically validated
Its theoretical grounding is drawn from established learning theory and synthesized empirical literature, but it has not been tested through structured research examining how students, educators, and institutions respond to its application in practice." Fourth, "the reviewed literature is predominantly English-language and draws heavily on institutional contexts in the United States, United Kingdom, and Australia
Higher education systems in the Arab region, the Global South, non-Anglophone contexts, and vocational and technical education settings may face distinctive challenges...that are underrepresented in the present review."

Open questions raised

Empirical validation of the three-level framework in real institutional settings
Qualitative studies examining how students, educators, and policymakers interpret and apply the Assistive/Augmented/Delegated distinctions
Comparative studies examining academic integrity outcomes across institutions with graduated AI policies versus binary policies
Longitudinal research examining the effect of different AI use modes on student writing and reasoning development over time
Empirical testing of the skill atrophy hypothesis regarding delegated AI use and cognitive capability degradation
Research extending the framework's application to non-Western and non-Anglophone higher education contexts

Extracted from: pdfAgreement 70%

Explore related topics

AI Manuscript Writing Academic Writing Literature Review