A professor wants students to understand the limitations of generative AI (e.g., knowledge cut-offs). Which of the following assignment designs would be least effective at achieving this specific goal?
The least effective assignment design is having students brainstorm 20 potential essay topics and select one.
This specific task leverages the core strengths of large language models—rapid ideation, pattern matching, and text generation—without testing their boundaries. Brainstorming asks the system to be creative rather than factual, completely bypassing the intended lesson. Students finish the exercise seeing only what the tool does well.
To grasp system constraints, students need assignments that create deliberate friction. The alternative options successfully highlight specific weaknesses:
- Comparing an AI summary to current news reports directly tests knowledge cut-offs. Students see exactly where the training data ends and the real world continues.
- Pitting an AI’s generic community summary against real-world interviews demonstrates a distinct lack of hyper-local, on-the-ground awareness.
- Critiquing an AI’s solution to an ethical dilemma showcases its struggle with complex human nuance and deep moral reasoning.
Teaching true digital literacy requires tasks where the technology inevitably stumbles. Open-ended brainstorming fails to create those necessary teaching moments.