Mitigating Marginalization in AI Education: Design-Based Research Insights from the US, UAE, Malawi and Chile

• In Event: Roundtable Session 4
  In Refereed Roundtable Session: The impact and challenges of artificial intelligence and digital tools in diverse educational contexts

Proposal

As AI becomes integral to the digital technologies pervasive in daily life, equity and justice-oriented efforts to promote AI literacy for students (and, to a lesser extent, teachers) have grown, often accompanied by claims about learning outcomes—yet implementation research remains scarce in this emerging field. While AI education interventions frequently employ design-based research and co-design methods to include the experiences of learners and educators in developing curricula and instructional tools, few have extended this involvement into research on how these interventions are implemented. Achieving UN Sustainable Development Goal (SDG) 4—ensuring inclusive and equitable quality education and promoting lifelong learning opportunities for all—requires a parallel commitment to implementation research that answers not just why interventions work for whom and in what contexts, but also how stakeholders and processes can be iteratively involved to scale data and social justice in digital literacy.

The **** project, grounded in constructionist learning theory, collaborates with education stakeholders to build AI-for-social-good projects for two primary purposes: 1) to iteratively design action-oriented K-12 curricula addressing AI fluency at both the student and educator levels, and 2) to research its implementation across diverse cultural contexts and learning settings, laying the foundation for expanded, scalable access to effective pedagogical and mathetic (Papert, 1993) practices around the responsible design of AI tools.

This paper presents findings from design-based implementation research (Fishman et al., 2013) that beta-tested a computer vision module for middle schoolers with nine teachers across four countries, including six US schools and educational settings in the UAE, Malawi, and Chile. Using 347 ratings as well as coded qualitative comments from teachers, collected via an interactive Google Docs version of the teacher guide or during unstructured recurring interviews, we compare how the implementation of curriculum components—activities, assessments, and teacher supports for AI fluency, responsible design, and technological, pedagogical, and content knowledge—varied between US and international sites.

We found that while both groups modified the curriculum, non-US sites skipped fewer activities, implemented assessments more consistently, and provided the only negative ratings of teacher supports. In addition, while many teacher modifications did not affect the curriculum’s substance, modifications from sites outside the US provided critical opportunities for improving the materials. These findings suggest that, rather than allowing nascent but essential AI education efforts—especially in the Global South—to be “described as stories of trial and error” (Call for Submissions, CIES 2025), design-based implementation research is a crucial step toward mitigating marginalization in AI-related societal opportunities, advancing data justice, highlighting the importance of teacher adaptability in curriculum implementation, and ensuring that educators’ expertise remains central to the rapidly evolving educational landscape.

Authors

• Christina Anderson Bosch, Massachusetts Institute of Technology

• Javier Campos Martínez, Universidad Austral de Chile

• Byamasu Patrick Paul, ADAI Circle

• Amisi Jospin Hassan, ADAI Circle

• Francisco Javier Francisco Marín, Escuela Pública Domingo Ortiz de Rozas

• Loreto Lizardi inalef, Escuela Teniente Serrano Montaner