Are we still able to recognize pearls? Machine-driven peer review and the risk to creativity: An explainable RAG-XAI detection framework with markers extraction

Automation of peer review could inadvertently suppress scientific innovation: The paper warns that when machine learning systems assess research submissions, they may systematically favor conventional, pattern-matching studies while rejecting unconventional ideas that require human contextual judgment—potentially stifling the paradigm-shifting discoveries that advance science.

There's a risk of "epistemic homogenization" in research: As researchers learn that algorithmic systems evaluate their work, they may optimize submissions for machine approval rather than pursuing genuine discovery, creating a feedback loop that narrows the types of research considered acceptable and reduces intellectual diversity.

The paper proposes a detection framework to identify machine-generated reviews: The researchers developed an explainable AI system (RAG-XAI) that can identify when peer reviews are likely machine-generated, achieving 99.61% accuracy. This transparency tool aims to help maintain human oversight and accountability in the editorial process.

Fully automated editorial pipelines pose a governance challenge: The paper highlights a concerning trajectory where LLM-generated reviews could lead to algorithmic editorial decisions, creating a completely automated evaluation system with no human judgment—raising questions about who is responsible when flawed decisions are made.

Transparency and explainability are essential safeguards: By developing methods to detect and explain automated review patterns, the authors argue we can preserve human creativity and accountability in science, suggesting that responsible AI integration requires maintaining visibility into where machines are making decisions.