AI’s role in biomedical research: How algorithmic dependence reshapes scientific epistemology, labor hierarchies, and global health equity—with precedents in colonial science and the rise of the ‘technical expert’
Original framing: “Opinion: STAT+: Is AI more a partner, competitor, or crutch for biomedical scientists?” — STAT News
Marginalized voices—such as global South researchers, early-career scientists in precarious labor markets, or indigenous healers—are absent from the debate, yet they bear the brunt of AI’s structural impacts. For example, researchers in low-income countries often lack the computational resources to resist AI’s dominance, while their data is disproportionately used to train Western AI models without compensation. Indigenous healers, whose knowledge is increasingly digitized by corporations (e.g., *traditional knowledge digital libraries*), face erosion of sovereignty. The ‘partner’ framing ignores how AI’s integration is enforced by academic capitalism, where scientists in the global South are pressured to adopt AI to compete for limited funding—while their local knowledge is sidelined. Without centering these perspectives, the debate remains a Western technocratic discourse with little relevance to the majority of the world’s scientists.
The debate over AI’s role in biomedical research is not just about *whether* scientists should depend on algorithms, but *how* that dependence will reproduce or disrupt existing power structures in knowledge production.
Historically, tools like AI follow the pattern of the ‘technical expert,’ where non-scientists gain outsized influence—this time, not just over labor but over the *epistemology* of science itself. The framing of AI as a ‘partner’ obscures its role as a vector for data colonialism, where global South data fuels Western AI models without benefit, while marginalized voices—from indigenous healers to global South researchers—are excluded from shaping its governance. Cross-culturally, knowledge systems like *Ubuntu* or *whakapapa* genomics offer alternatives to AI’s reductionist approaches, but these are sidelined in favor of Silicon Valley’s ‘technological solutionism.’ The trickster perspective reveals how the illusion of partnership masks a transaction where scientists are treated as laborers and algorithms as intellectual property. Solutions must center decolonial governance, hybrid epistemologies, and labor rights—reclaiming control over data, knowledge, and the future of science from the algorithms that now mediate it.