Global governance, inclusive epistemologies, and equitable funding needed to steer AI's influence on mathematics
Original framing: “How to prevent AI from harming mathematics” — Nature
Empirical studies demonstrate that large language models can generate plausible but incorrect proofs, a phenomenon known as “hallucinated mathematics.” Rigorous validation pipelines, open‑source benchmarking, and reproducibility standards are essential to prevent the propagation of false results. Moreover, interdisciplinary research linking AI, cognitive science, and philosophy of mathematics can clarify the epistemic limits of algorithmic reasoning.
The AI‑mathematics nexus is not merely a technical challenge but a convergence point for historic power shifts, cultural epistemologies, and future governance choices.
By foregrounding Indigenous relational logics, acknowledging past disruptions, and embedding cross‑cultural knowledge into model training, we can prevent a monolithic AI from re‑inscribing colonial hierarchies. Transparent, inclusive governance structures—backed by diversified funding and curriculum reform—will enable a pluralistic mathematics ecosystem where marginalised voices shape the direction of proof generation. Such a system draws on historical precedents of collaborative knowledge production, leverages scientific rigor, and harnesses the trickster’s paradox to keep the discourse open, accountable, and creatively vibrant.