Explainable artificial intelligence as AI literacy scaffolding
It is argued that AI explanations can function as a scaffold that supports users in learning not only about specific outputs but also about model behavior, limitations, and potential bias.