AI evidence extraction

Harnessing Generative Artificial Intelligence for Academic Research: Capabilities, Challenges, and Ethical Implications

Emmanuel U. Opara, Modupe Ojumu, Gbolahan Solomon Osho, Onochie Jude Dieli · International Journal of Higher Education · 2026

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

9/10

Relevance

0/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.5430/ijhe.v15n3p1

Methodology & findings

Study design

Qualitative descriptive research design using secondary data analysis of sixty scholarly sources (peer-reviewed articles, conference proceedings, editorials, policy documents) published between 2015-2025.

Main result

The study found that "STEM researchers report average time savings of approximately 50%, with relatively narrow variation across studies" and that "GenAI significantly boosts research productivity, particularly within data-intensive science, technology, engineering, and mathematics (STEM) fields." Additionally, the analysis revealed that "plagiarism as the most frequently reported ethical concern, with more than two-thirds of respondents citing it as a primary issue," while adoption patterns "differ markedly across disciplines, with STEM fields reporting the highest acceptance levels, social sciences showing moderate engagement, and humanities displaying the most cautious stance."

Research paradigm

Interpretivist/qualitative synthesis

Author conclusions

The authors conclude that "Generative artificial intelligence should not be viewed as a threat to academic integrity but as a catalyst for reimagining research practices when deployed responsibly. By treating GenAI as a co-pilot rather than a replacement, the academy can harness its transformative potential while upholding the principles of originality, rigor, and transparency. If balanced integration is achieved, GenAI will support and enrich the pursuit of knowledge, ensuring that innovation advances with scholarly trust and ethical responsibility." They further emphasize that "the challenge, therefore, lies not in whether GenAI will permeate research but in how its integration can be managed responsibly" and call for "a multifaceted response: researchers must commit to human-centered, critically reflective use; institutions and publishers must establish robust disclosure and accountability frameworks; and scholarly communities must invest in AI literacy and interdisciplinary dialogue."

Risk of bias

Selection bias: Inclusion criteria limited to sources explicitly addressing GenAI applications or academic integrity implications; early discussions on algorithmic decision-making (2015-2022) may underrepresent foundational critiques; Publication bias: Search limited to peer-reviewed sources, editorials, and policy documents; gray literature and negative findings may be underrepresented; Source type bias: Heavy reliance on cited studies (primarily Kasneci et al., 2023; Jakesch et al., 2023; Birhane & Raji, 2022; Lund et al., 2023) suggests potential citation clustering and selective emphasis; Temporal bias: Sharp increase in scholarship post-2022 (ChatGPT release) may skew findings toward recent, enthusiastic adoption narratives; Disciplinary representation bias: Literature review synthesis reflects available scholarship, which may overrepresent STEM and underrepresent humanities perspectives on GenAI risks; Author perspective: Authors appear to advocate for 'responsible integration' rather than neutral stance, potentially biasing interpretation toward balanced rather than critical framings; Secondary data analysis may introduce publication bias (preference for published sources over grey literature); Disciplinary biases in publication patterns could skew representation across STEM, social sciences, and humanities; Reliance on sources from 2015-2025 may not fully capture pre-ChatGPT era discourse on generative AI; Selection criteria requiring explicit mention of GenAI applications could exclude relevant theoretical or critical perspectives; Google Scholar inclusion alongside Scopus and Web of Science may introduce inconsistent quality thresholds; Selection bias: Dataset limited to English-language publications indexed in Scopus, Web of Science, and Google Scholar; Publication bias: Selective coverage of peer-reviewed sources and editorials; exclusion of grey literature or non-indexed materials; Disciplinary bias: Overrepresentation of STEM perspectives in available literature on GenAI adoption; Temporal bias: Heavy concentration of sources from 2022 onward following ChatGPT release; limited representation of pre-2022 perspectives; Narrative bias: Reliance on thematic coding by researchers, subject to interpretive variation

Open questions raised

Need for consistent disclosure standards across disciplines and institutions regarding GenAI use in manuscripts
Requirement for refined plagiarism detection tools and AI-audit mechanisms to capture both direct text reproduction and subtle forms of intellectual appropriation
Gaps in AI literacy training programs for researchers at all career stages to address technical limitations (hallucinations, error propagation) and ethical risks
Lack of established collaborative frameworks clarifying authorship, accountability, and ownership in GenAI-assisted scholarship
Need for interdisciplinary dialogue to develop balanced policies that preserve research diversity, equity, and rigor across STEM, social sciences, and humanities
Further investigation into epistemological tensions between Western-centric training data and marginalized knowledge systems in Global South contexts

Extracted from: pdfAgreement 80%

Explore related topics

AI Manuscript Writing Academic Writing Literature Review