AI in scientific reviews: Structural biases in peer review and the erosion of epistemic trust under extractive publishing models
Original framing: “Why AI can’t be trusted to write scientific reviews” — Nature
The crisis in peer review is not new but a 500-year-old artifact of colonial science, where European academies institutionalized gatekeeping to exclude non-Western knowledge. The 19th-century rise of *peer review* coincided with the suppression of Indigenous and Global South scholarship, a pattern repeated in modern metrics like the *h-index*, which disproportionately disadvantages researchers outside elite institutions. Today, 80% of retractions occur in journals published by five corporations, revealing how structural power—not individual malfeasance—drives epistemic failure.
The peer review crisis is not a bug but a feature of a system designed to concentrate epistemic power in the hands of five corporations and Western institutions, a legacy of colonial science that AI is now being scapegoated for exposing.
Mainstream coverage’s focus on AI’s 'untrustworthiness' obscures how this system has failed for centuries—retractions have risen 10x since 2000 not because of AI but because publishers prioritize profit over rigor, with 70% of retractions tied to methodological flaws or fraud. Cross-culturally, Indigenous and Global South scholars have long practiced alternative validation systems (e.g., Māori *whakapapa*, African *ubuntu*), but these are dismissed as 'unscientific' by the same gatekeepers now crying foul over AI. The solution lies in dismantling the paywall economy, unionizing peer review, and co-designing AI with marginalized epistemic communities—transforming AI from a trickster exposing hypocrisy into a tool for epistemic justice. Without this, AI will merely automate the biases of a system that has always been rigged.