Situated AI ethics: a cultural-historical and ecological framework for education

Synthesis: Raffaghelli, Vartiainen, Bower, Ronci, Shelton, MacCallum, Lee, Webb, Chtouki & Smith (2026) propose a situated AI ethics framework for education that fuses Bronfenbrenner's ecological systems theory with Cultural-Historical Activity Theory (CHAT). Moving beyond universalist, principle-based ethics, they frame AI as a non-neutral socio-technical assemblage whose ethical implications are historically produced and locally negotiated. Applied as a critical-ecological activity model across five ecological levels (self, classroom/group, state/geopolitical, cultural norms, global), the framework is used to comparatively analyze seven national cases (Australia, Finland, England, France, Italy, New Zealand, South Korea). The analysis finds teachers are routinely positioned as moral gatekeepers of AI use while lacking structural, institutional, and epistemic support, and argues that ethical AI engagement requires context-sensitive, collective, and transformative agency that extends AI literacy beyond technical skills toward critical, political, and ecological forms of action.

Key Findings

• Situated AI ethics as a critical-ecological framework. The paper proposes a 'situated AI ethics' framework grounding ethical decision-making in both ecological systems theory (Bronfenbrenner's micro, meso, exo, macro, chrono systems) and Cultural-Historical Activity Theory (CHAT), which foregrounds contradictions, power relations, and transformative agency. AI is treated as a non-neutral socio-technical assemblage whose ethical implications are historically produced and locally negotiated, rather than captured by abstract or universal principles.

• Ethical decision-making across five ecological levels. The critical-ecological activity model maps AI ethics onto self, classroom/group, state/geopolitical, cultural norms, and global levels. Ethical action emerges as a negotiated, context-dependent process shaped by intersecting systems—not an isolated individual choice, and not a matter of compliance with checklists.

• Teachers as moral gatekeepers without structural support. Across all seven national contexts, teachers are frequently positioned as moral gatekeepers of AI use while lacking adequate structural, institutional, and epistemic support to exercise ethical agency. The analysis surfaces recurring tensions: uneven teacher autonomy, regulatory overload, fragmented policy guidance, cultural anxieties about automation, and global pressures around competitiveness, labor markets, and platform dependence.

• Transnational comparative analysis. The framework is applied to a comparative analysis of seven national cases (Australia, Finland, England, France, Italy, New Zealand, South Korea), derived from expert gray-literature analysis within the EDUsummIT Sessions (TWG5). This mapping shows how situated ethical practices diverge across cultures, regulatory environments, institutional conditions, and global pressures.

• From AI literacy to critical, political, ecological agency. The paper argues that ethical engagement with AI requires context-sensitive, collective, and transformative approaches that extend AI literacy beyond technical skills toward critical, political, and ecological forms of agency—including resistance, contestation, and the co-creation of more just digital futures.

• Critique of techno-solutionism and compliance-based ethics. The authors reject ethics as checklists, 'ethical washing,' or compliance, and critique techno-solutionist and neoliberal framings of AI in education. They foreground global dimensions such as environmental costs, labor exploitation, surveillance capitalism, and the marginalization of Global South knowledge.

Educational Significance

The situated AI ethics framework offers a way to move AI ethics in education beyond abstract principle lists toward contextually grounded practice. It directly engages debates on Teacher AI Competency and AI in Education, providing conceptual tools for teacher professional development and policy. By foregrounding Ethics as a negotiated, ecological and socio-historically produced phenomenon, the framework supports equitable and culturally responsive approaches to AI integration across Higher Education and K-12 settings, linking to Learning Theories and Equity.

What this means for practice

• Teacher educators. Build professional learning around the concrete ethical dilemmas teachers face at each ecological level — self, classroom, institution, state, culture, global — rather than abstract principle lists they cannot convert into action.
• Policymakers. Pair any expectation that teachers act as moral gatekeepers with the structural conditions the analysis shows they lack: protected time, autonomy over adoption decisions, and coherent cross-level policy guidance instead of compliance-driven mandates.
• Policymakers. Fund critical, political, and ecological forms of AI Literacy — including the capacity to contest, resist, or refuse adoption — rather than literacy confined to technical operation.
• Instructors. Treat AI adoption as a contestable pedagogical decision and design classroom activities that make a tool's embedded power relations and ethical trade-offs visible to students, instead of presenting adoption as inevitable.
• Researchers. Use the five-level model to design comparative studies of how the same tool is negotiated differently across settings, treating the levels as an analytical heuristic for discussion rather than as fixed, distinct categories.

Limitations

• Conceptual and comparative rather than empirical: the paper collects no primary data, so it can specify what an ethical framework for education should contain but cannot demonstrate that the framework changes teacher judgment or classroom practice.
• The seven national cases were assembled through expert gray-literature analysis within a single EDUsummIT working group (TWG5) rather than systematic sampling, so case selection reflects the expertise and reach of that group.
• All seven cases (Australia, Finland, England, France, Italy, New Zealand, South Korea) come from high- or upper-middle-income systems, with none from the Global South whose knowledge the paper itself argues is marginalized.
• The authors present the five ecological levels as a useful structure for discussion rather than a claim that they are distinct, so the model still awaits empirical operationalization.

Citation

Raffaghelli, J. E., Vartiainen, H., Bower, M., Ronci, M., Shelton, C., MacCallum, K., Lee, J., Webb, M., Chtouki, Y., & Smith, D. (2026). Situated AI ethics: a cultural-historical and ecological framework for education. Computers and Education Open, 11, 100368. doi:10.1016/j.caeo.2026.100368.