Teaching AI, teaching bias: addressing gender inequality in public health education

Artificial intelligence increasingly shapes healthcare, public health and health professions education. However, predictive and generative systems can reproduce or amplify longstanding inequalities embedded in data, institutions and established knowledge structures. Public health education therefore needs to connect technical approaches to identifying and mitigating algorithmic bias with their broader social, population health and institutional contexts. This conceptual article presents the sociotechnical Bias literacy for public health education framework (SBL-PHE), developed through a theory-informed conceptual synthesis drawing on conceptual framework analysis. A focused supplementary literature search in PubMed/MEDLINE, Scopus, ERIC and the ACM digital library was conducted to examine and refine its theoretical foundations. An evidence-informed composite case of AI-supported cardiovascular risk assessment traces algorithmic gender bias from knowledge production and data representation through modeling and implementation to monitoring. The framework comprises five interconnected competency dimensions concerned with recognizing, explaining, evaluating, addressing and reflecting on algorithmic bias. For master’s-level public health education, these dimensions are linked to learning objectives, learning activities, assessment methods and a prospective evaluation strategy. Sociotechnical Bias literacy is proposed as a foundational competency for public health professionals who use or interpret AI. Sociotechnical fluency describes an advanced, role-dependent capability, particularly for responsibilities in implementation, evaluation, translation and governance. SBL-PHE provides a theoretically grounded framework for integrating technical understanding, health equity, professional judgment and institutional responsibility into public health education. The framework and its proposed educational application have not yet been empirically validated.