Safeguarding Scientific Algorithms; Analyzing the Sociotechnical Implications of the 2008 Hindawi Retractions 8 views

Academic misconduct increasingly occurs through algorithmic duplication, in which existing, lightly modified methods are presented as novel. This problem is especially relevant in metaheuristic optimization, a field that has experienced rapid growth along with concerns about derivative algorithms. This paper presents an AI-driven framework for detecting algorithmic plagiarism in metaheuristic research by comparing the computational logic of algorithms rather than their surface-level similarity. The proposed pipeline extracts pseudocode from academic PDFs, canonicalizes algorithms, and compares them against a reference database using four levels of similarity analysis: surface, structural, primitive, and semantic. Scores from each level are combined into an overall similarity score. The system was tested through controlled verification cases and real non-novel examples identified by experts. Results showed clear separation between disguised derivations of the same algorithm and genuinely distinct algorithms, supporting the framework's ability to measure algorithmic novelty. These findings suggest that automated detection of algorithmic plagiarism is feasible. It could support research integrity efforts by helping journals and reviewers identify submissions in need of close expert review.

BS (Bachelor of Science)

Algorithm design and analysis; Detection algorithms; Metaheuristics; Plagiarism

School of Engineering and Applied Science Bachelor of Science in Systems Engineering Technical Advisor: Matthew Bolton STS Advisor: Kent Wayland Technical Team Members: Halbert Nguyen, Colin Heffern, Gunni Kamran

English

All rights reserved by the author (no additional license for public reuse)

2026-05-09