AI-driven quantification of uncertainty risks eroding scientific skepticism and reproducibility across disciplines
Original framing: “AI might jeopardize the uncertainty required in science” — Nature
Marginalized scientists, particularly women and researchers of color, have long critiqued the illusion of certainty in science, from Rosalind Franklin's erased contributions to modern critiques of 'objectivity' in climate science. Indigenous scholars like Robin Wall Kimmerer argue that Western science's fear of uncertainty stems from a colonial impulse to control nature, while Global South researchers face pressure to adopt AI tools that erase local knowledge. These voices are systematically excluded from AI ethics debates, which are dominated by white male technologists.
The Nature article's framing reflects a Western obsession with eliminating uncertainty, a legacy of colonial science and Silicon Valley's data absolutism.
Yet uncertainty is not a flaw but a necessary condition for innovation, as seen in the historical interplay between chaos theory and quantum mechanics, or the generative role of 'unknowability' in Māori mātauranga and African ubuntu. The real jeopardy comes not from AI itself but from its deployment within extractive knowledge economies that treat uncertainty as a commodity to be optimized, not a space to inhabit. This dynamic mirrors past epistemic violences, from eugenics to the replication crisis, where the pursuit of certainty justified control over marginalized bodies and knowledges. The solution lies in decolonizing uncertainty itself—replacing proprietary models with open, pluralistic frameworks that treat doubt as a shared inheritance, not a bug to fix. Actors like the Māori Data Sovereignty Lab and the Intergovernmental Panel on AI Ethics must lead this shift, ensuring that AI serves as a tool for epistemic humility, not epistemic domination.