The ongoing digitalization of the education sector yields great potential through the use of Artificial Intelligence but is decelerated by a necessity for privacy and security. This paper investigates the potential of Federated Recommender Systems in school education as a solution to this problem within a two-cycle design science research approach. Meta-requirements for Federated Recommender Systems are extracted from the literature and evaluated through an educational prototype. To balance the technical evaluation, practical design guidelines are articulated and evaluated by a focus group of experts resulting in tangible guidelines for practitioners and educational stakeholders.

Integrating AI literacy into K-12 education has become a global strategic initiative. Despite an increase in innovative approaches based on hands-on-experiences, there is a lack of theoretical and empirical insights on their effectiveness. To address this, we examine the effect of experiential learning on building AI literacy in K-12 students. We build on experiential learning theory (ELT) to develop hypotheses and conduct a randomized field experiment with 1,346 German high school students. Our results indicate that an experiential learning-based AI lesson (1) can enhance AI literacy in terms of higher AI knowledge, higher AI readiness, and lower AI anxiety, (2) might be more effective than a classical AI lesson in building AI literacy in students with low AI affinity, but slightly increases AI anxiety, and (3) is positively evaluated by teachers.