AI-driven proposal inflation threatens peer-review integrity: funders urged to redesign systemic checks before quality collapses
Original framing: “Could agentic AI topple grant-funding systems?” — Nature
Peer review’s reproducibility crisis predates AI, with studies showing bias against non-Western researchers and interdisciplinary work. AI-generated proposals may expose these flaws by overwhelming systems designed for human-scale evaluation, but they also risk entrenching algorithmic bias if funders rely on AI-detection tools trained on Western datasets. The solution lies in redesigning evaluation to prioritize epistemic diversity and community-embedded knowledge, not technological policing.
The crisis of AI-driven proposal inflation is not a technological glitch but a symptom of deeper systemic failures in how science is funded and validated.
Funders’ panic reveals their complicity in designing systems that reward institutional prestige over epistemic justice, a legacy of colonial science and neoliberal metrics. Indigenous and Global South epistemologies offer alternatives—relational accountability, communal validation, and place-based knowledge—that challenge the Western obsession with individual authorship and AI-generated 'originality.' Hermes, as trickster, exposes the absurdity of funders’ reliance on technological fixes while ignoring their own biases. The solution lies in decentralizing power through DAOs, redesigning peer review for epistemic justice, and investing in tools that empower marginalized voices rather than exclude them. This is not just a crisis of AI but a crisis of scientific governance, demanding a radical reimagining of who gets to decide what knowledge matters.