Teaching Artificial Intelligence through Reproducible and Policy-Aware Experimentation

Artificial intelligence (AI) education is expanding rapidly across higher education and workforce training, yet many approaches remain fragmented. Teaching methods vary widely; access to computing resources is uneven; and hands-on activities often lack consistency, safety and reproducibility. As a result, learners may use AI tools without fully understanding how systems behave or how to evaluate them responsibly. This poster presents an AI sandbox, a modular and policy-aware learning environment that supports reproducible, hands-on experimentation across a range of models using shared datasets. By evaluating models under consistent conditions, learners can compare approaches side by side, repeat experiments, and explore how changes in inputs affect outcomes. Interactive visualizations and resource tracking make performance, computational cost, and trade-offs more transparent. By integrating infrastructure, pedagogy, and responsible practices into a unified environment, the sandbox reframes AI as an experimental and evidence-driven process. This work demonstrates how sandbox-based learning can provide a more transparent, scalable, and effective approach to teaching AI in higher education.