A Comparative Analysis of Plagiarism across Written and Technical Disciplines

Abstract

Background: Plagiarism is usually discussed as one phenomenon with one definition and one remedy. Scholarly disciplines, however, differ in what they treat as an original contribution, how they attribute knowledge, and which artefacts they produce. Those artefacts range from argumentative prose to source code, numerical data, mathematical derivations and images. This study asks whether plagiarism means the same thing in written disciplines (humanities, social sciences, law and literary studies) as in technical disciplines (engineering, computer science, mathematics and the natural sciences).

Methods: A PRISMA-guided structured narrative synthesis was carried out, drawing on the PRISMA 2020 statement and its scoping-review extension. Eligible sources were peer-reviewed reviews, meta-analyses, empirical surveys, detection-technology evaluations, linguistic analyses and policy documents published before 2026. Findings were charted against seven analytical dimensions: definition, unit of borrowing, citation culture, detection feasibility, prevalence evidence, sanction and remediation, and susceptibility to generative artificial intelligence.

Results: Written disciplines concentrate risk in textual and ideational appropriation, including patchwriting, uncredited paraphrase and contract authorship. Technical disciplines distribute risk across code, data, figures, mathematical content and reused methods text. Text-matching software serves the first group reasonably well and the second group poorly. Prevalence estimates are not comparable across fields because the underlying instruments differ. Evaluations of AI-text detectors consistently show limited accuracy and demonstrable bias against non-native writers.

Conclusions: Plagiarism is best understood as a family of discipline-conditioned integrity breaches rather than a single offence. Policy, detection and pedagogy that ignore this conditioning will over-penalise some communities and under-detect misconduct in others. A discipline-sensitive, education-first integrity framework is proposed, together with eight citable propositions to guide later empirical work.