Artificial intelligence in public health: a foundational shift or a technocratic distraction?

Artificial intelligence (AI) is increasingly presented as a transformative force in public health. From real time disease surveillance and outbreak forecasting to health system optimization and policy decision support, AI driven tools promise speed, scale, and analytical power beyond conventional approaches. Advances in computational capacity, coupled with the rapid expansion of digital health data including electronic health records, mobile health platforms, environmental sensors, and social media, have accelerated interest in applying machine learning and large multimodal models (LMMs) to population health challenges. Yet public health is not merely a technical or computational enterprise; it is fundamentally a normative, value driven discipline grounded in prevention, equity, and social accountability. Decisions in public health affect entire populations and often involve trade-offs that require ethical judgment, political legitimacy, and community trust. As emphasized by Wang et al. (2024), the integration of AI into health systems therefore raises questions that extend beyond performance metrics: Whose data are represented? Whose values are encoded? And who remains accountable when algorithmic recommendations shape policy? This Opinion article argues that AI can represent a foundational shift for public health only if it is explicitly aligned with core public health principles. When adopted uncritically, AI risks functioning as a technocratic distraction, privileging data centric efficiency over social context and reinforcing existing structural inequities. By examining methodological tensions, normative conflicts, and governance challenges, this article seeks to contribute constructively to ongoing debates on how AI should be positioned within contemporary public health practice.AI applications in public health typically rely on machine learning techniques designed to detect patterns within complex, high-dimensional datasets. Compared with traditional statistical models, these approaches can capture non-linear relationships, integrate diverse data streams, and adapt dynamically as new information becomes available (Topol, 2019). Such capabilities are particularly attractive in settings characterized by uncertainty and rapid change, such as infectious disease outbreaks or climate-sensitive health threats. However, methodological sophistication does not automatically translate into public health relevance. A growing body of scholarship highlights the tension between predictive accuracy and interpretability. Many high-performing models operate as "black boxes," producing outputs that are difficult to explain to policymakers, practitioners, or affected communities (Mittelstadt et al., 2016). In public health, where decisions must be justified transparently and democratically, opacity undermines accountability. Moreover, AI tools are often developed within siloed technical environments that are disconnected from the institutional realities of public health systems. (El-Sayed et al., 2025) note that the absence of standardized protocols for validation, deployment, and oversight remains a major barrier to meaningful integration. Without harmonized governance structures, AI risks becoming an add-on technology rather than a coherent component of population health strategy.A defining feature of public health is its focus on populations rather than individuals, with prevention as a central objective. In contrast, many contemporary AI applications emphasize individual-level risk prediction, aligning more closely with paradigms of personalized or precision medicine. While individual risk stratification can inform targeted interventions, an excessive focus on personal vulnerability may divert attention and resources from upstream determinants of health, such as housing, education, labor conditions, and environmental exposures (Frieden, 2010). AI aligns more closely with public health prevention when applied to collective risk assessment. Examples include environmental surveillance systems, early warning models for vector-borne diseases, and climate, health monitoring platforms. Machine learning-based dengue forecasting systems demonstrate how AI can support anticipatory action at the population level when embedded within preventive frameworks (Sophia et al., 2025). The public health value of AI therefore depends less on technical novelty and more on whether applications illuminate shared risks and inform structural interventions.Equity is the ethical core of public health practice. AI systems trained on historically biased or incomplete data risk reproducing, and amplifying, existing inequities. Empirical evidence shows that algorithms used in health management can disadvantage marginalized populations when socioeconomic context is inadequately represented. The landmark study by Obermeyer et al. (2019) revealed that cost-based proxies in population health algorithms led to systematic racial bias in care allocation. More recent analyses from low-and middle-income settings suggest that "algorithmic prejudice" is not a theoretical concern but a material harm. Models trained predominantly on urban or highresource data often fail to capture rural disease dynamics, informal care pathways, or gendered patterns of access. (Nasir et al., 2025) describe this phenomenon as a "digital shadow," wherein populations already underserved by health systems remain underrepresented in digital infrastructures. Realizing AI's potential to advance equity requires deliberate governance choices: inclusive data practices, routine equity audits, and mechanisms for community participation in model design and evaluation. Equity cannot be retrofitted after deployment; it must be treated as a design principle rather than an external constraint.Public health challenges are embedded within complex adaptive systems that span biological, social, economic, and political domains. In principle, AI's capacity to model interactions across multiple levels makes it well suited to such complexity (Greenhalgh rather, it depends on how it is governed, interpreted, and aligned with the discipline's normative foundations. As this article has argued, AI can constitute a foundational shift only when it reinforces public health commitments to prevention, equity, transparency, and collective accountability. When adopted uncritically, however, AI risks narrowing public health practice into a technocratic enterprise-one that privileges predictive efficiency over social context, algorithmic authority over democratic deliberation, and data availability over moral judgment. Public health decisions shape population trajectories and distributive outcomes; delegating such decisions to opaque systems without robust oversight threatens institutional legitimacy and public trust. Recent scholarship underscores that responsible AI integration must be explicitly equity-oriented, participatory, and embedded within public governance structures, rather than operating as a parallel technical domain (Dankwa-Mullan, 2024;Gozum without them, it risks becoming a technocratic distraction that obscures the very values public health exists to protect.