Safeguarding Science with Artificial Intelligence: The Gatekeeper of Research Credibility 8 views

With the world’s rapid integration of artificial intelligence into academic research, knowledge has fundamentally changed in the way it is produced and evaluated. While AI tools provide increased efficiency and innovation, they also raise critical concerns about the authenticity and novelty of research outputs. My work bridges this tension by combining a technical investigation into detecting questionable research practices (QRPs) in algorithm design with a sociotechnical analysis of how AI reshapes standards of originality and credibility in academic communities. My capstone project addresses the growing difficulty of identifying true novelty in optimization algorithms inspired by nature. As researchers increasingly rely on AI-assisted generation, many of these so-called “new” algorithms are rather derivatives of preexisting ones, repackaged with surface-level changes. To address this, me and my team built a multi-level similarity detection framework with the purpose of evaluating algorithmic pseudocode across four dimensions: surface, structural, primitive, and semantic similarity. This system compares newly proposed algorithms against what is in the established MEALPY library. This tool provides a systematic method for improving research integrity and filtering out duplicates and meaningless contributions. However, the validity and societal impact can’t be comprehended merely with a technical lens. The increasing use of AI in research introduces complex social dynamics, including shifting norms surrounding authenticity, ownership, ethical responsibility and implications. My STS research examines how AI influences novelty in academic work and how the consequences trickle down on stakeholders. Using a sociotechnical framework centered on ethical responsibility, I analyzed how key factors including researchers and institutions for example respond or assist AI-generated outputs. I applied the frameworks of actor network theory and technological mediation, as well, in order to further analyze in structured methodologies other elements not touched on. The compilation of my STS paper represents a broad range of this topic with specific cases to give real world meaning and instances of how prevalent this is in our current day. Ultimately, preserving novelty in the age of AI is not a mere technical challenge, but instead a governance and cultural one. While tools like my capstone’s similarity detection system can flag redundancy and enforce higher standards, their effectiveness is dependent on how they are adopted. The manner in which these tools are interpreted and institutionalized by academic communities and all that comprises them (researchers, journals, institutions) calibrates the field of AI in research. It will continue to amplify superficial innovation rather than meaningful contribution if quantity-driven publication models are allowed to continue to bolster. Tools of accountability, transparency, authorship, and authenticity would redefine the space into something of much more beneficial future impact. In this way, my work highlights that the collective willingness of stakeholders aligns the technical capability with ethical responsibility that exists.

BS (Bachelor of Science)

Algorithm; Artificial Intelligence; Detection; Questionable Research Practices; Academia; Publications; Citations

English

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

2026-05-02