Duke Builds An AI Literacy Course. It Focuses On Judgment. Novel Concept, Apparently.
Duke University's Office of Information Technology launched AI Foundations, a program where participants work through realistic Duke-specific scenarios involving privacy, security, accuracy, bias, and misinformation. The curriculum also covers prompt writing, refinement, and evaluation of AI-generated results. Michael Greene, director of Digital Experience and Academic Technology Solutions, emphasizes the goal is thoughtful decision-making, not simply increased AI usage.
This demonstrates what I would call the Competence Before Consumption principle. Duke correctly identifies that teaching people to use AI without teaching them to evaluate AI output is irresponsible. The mechanism here is scenario-based learning, which forces learners to confront failure modes like bias and inaccuracy in context rather than abstraction. The lesson for everyone: literacy means understanding limitations, not just capabilities. Most users skip this step entirely. That is why most users produce garbage.
Duke University's Office of Information Technology, specifically Michael Greene as director of Digital Experience and Academic Technology Solutions. The program is called AI Foundations and uses realistic institutional scenarios as its teaching method.
- Take any AI output you recently accepted without questioning it. Paste it back into the chat and ask the AI to critique its own response for bias, inaccuracy, or missing perspectives. Expected outcome: the AI will often flag at least one issue you overlooked.
- Write down three questions you should have asked before trusting the original output, such as 'What sources back this claim?' or 'Whose perspective is missing here?' Expected outcome: a personal checklist for evaluating AI responses.
- Apply that checklist to your next three AI interactions before using any output. Expected outcome: you begin building the judgment Duke is teaching, without enrolling.