AI evidence extraction

Generative AI and Academic Integrity in Higher Education: A Systematic Review and Research Agenda

Kyle Bittle, Omar El-Gayar · Information · 2025

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

107

Citations

44.52

FWCI

Top 10%

Impact

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.3390/info16040296

Methodology & findings

Study design

Systematic literature review following PRISMA 2020 guidelines.

Sample

N = 41, 3 groups

Primary method

Qualitative synthesis using Braun and Clarke's thematic framework. Inductive coding approach for identifying patterns across literature. Data extraction and screening by multiple reviewers to minimize bias. No quantitative statistical analysis or meta-analysis performed.

Main result

The review found that "the rapid integration of generative artificial intelligence into educational environments brings with it a spectrum of both potential benefits and inherent risks that educators and institutions must navigate." The thematic analysis revealed three reoccurring topical categories: Risks of Academic Dishonesty and Cheating, Pedagogical Implications and Ethical Use of GenAI, and Impacts on Student Learning and Educational Practices. Key concerns include that GenAI can generate "undetectable, ghostwritten assignments" and current plagiarism detection tools are "inadequate for the needs of educators," while simultaneously offering "unprecedented opportunities to enhance the learning experience through more personalized, accessible, and engaging educational content."

Reports effect sizes.

Research paradigm

interpretivism

Author conclusions

"Integrating Generative AI (GenAI) in higher education offers transformative possibilities but also introduces significant challenges, particularly in maintaining academic integrity. While GenAI tools like ChatGPT can enrich personalized learning experiences and make education more accessible, they pose risks, such as enabling academic dishonesty and diminishing critical thinking skills." The authors conclude that "a balanced approach is necessary to harness the benefits of GenAI while safeguarding foundational educational principles" and that future research must address detection technologies, educator training, ethical guidelines, and assessment redesign.

Risk of bias

Language bias: English-language only publications included; Publication bias: Only peer-reviewed journal articles and conference proceedings included; Database selection bias: Search limited to four specific databases; comprehensiveness depends on database indexing; Recency bias: Focus on 2021-2024 may miss important earlier work; Screening bias: Multiple reviewers conducted data extraction but potential for subjective interpretation in qualitative synthesis; Geographic bias: Likely underrepresentation of non-Western perspectives; Language bias (English-only publications); Publication bias (peer-reviewed journals only, excludes grey literature); Selection bias in database choice (may exclude relevant publications from other sources); Screening bias mitigated by multiple reviewers but specific inter-rater reliability metrics not reported; Subject bias (opinions on GenAI integration vary widely among included studies without quantitative consensus); Language bias: English-language papers only; Publication bias: peer-reviewed journals and conference proceedings only; Temporal bias: limited to 2021-2024 timeframe; Geographic bias: potential underrepresentation of non-Western perspectives; Selection bias in database choice and search string formulation

Limitations

The review was "limited to English-language papers" as explicitly stated: "Finally, the review focused on English-language-only papers
Future studies could extend their search to papers written in different languages and from non-Western literature sources." Additionally, the rapid pace of technological change may have rendered some findings less representative, and the literature itself exhibits divided opinions on integration approaches without definitive empirical resolution of key debates.

Open questions raised

Lack of comprehensive analysis balancing GenAI benefits against dishonesty risks
Limited synthesized overview integrating findings across isolated studies
Inadequate assessment of long-term implications on educational practices
Need for advanced detection tools capable of identifying AI-generated content
Development of pedagogical frameworks integrating GenAI responsibly
Redesigning assessments for GenAI-enhanced learning environments

Extracted from: pdfAgreement 66%

Explore related topics

Hallucination Control Research Integrity