Teaching AI awareness in middle school classrooms: Design, implementation and evaluation of two education modules on algorithmic bias and filter bubbles

The nascent digital footprints of adolescents make them particularly susceptible to being misrepresented by AI systems. We design, implement, and evaluate two educational modules on the causes and consequences of two particularly salient types of AI misrepresentation: algorithmic biases and filter bubbles. We implement these modules in middle school math and computer science classrooms by integrating our materials with the existing curricula and curricular requirements of the partner school. This integration, as well as the many confounding factors of the classroom, are nontrivial problems that have heretofore not been considered in privacy education research and required us to adopt methodologies from learning sciences, particularly design-based research and backward design. Quantitative results of the pre-/post-test and qualitative assessments of open-ended questions find that students were engaged with the materials and exhibited substantial learning gains. The findings emphasize the need for tailored educational interventions and avenues for future research to address AI-related challenges.