Universities have largely treated generative AI as an integrity problem-ban it or permit it with a responsible use disclaimer. A lecturer at the University of Queensland Business School argues that neither approach is sufficient because both fail to develop the critical evaluation skills employers now demand. She has designed two assessments that force students to argue against AI, turning the technology from a potential shortcut into a tool for sharpening their own reasoning.
"Workplaces need graduates who can interrogate a model's reasoning, identify where it is confidently wrong and take responsibility for the decision that follows," the lecturer said. These capabilities do not develop by accident, she added, and require structured adversarial engagement with AI outputs.
Two assessments built to challenge blind trust in AI
The first task is a live debate. In a postgraduate business information systems course, students are assigned positions on how AI is reshaping work-whether it will eliminate technical data skills, whether it should drive business decisions, and how it redefines individual roles. They argue without notes while fielding questions from peers.
The second assessment follows immediately. Students take one position they defended, ask an AI tool to produce the strongest possible counterargument, and then evaluate that output in writing. They must identify where the AI reasons well, where it overreaches, and where it fails to engage with disciplinary evidence. The task does not ask whether the AI was right; it asks whether the student can judge the quality of its argument and explain that judgement using course concepts.
Three design principles for developing AI judgement
- Create a stake in the question: Students need a position they have publicly defended before engaging AI. Asking them to evaluate an AI output without a stake leaves them nothing to test the model against. Debates and oral defences create that intellectual commitment.
- Be adversarial: The encounter with AI should not be collaborative. Instead of using AI as an assistant to refine work, students must ask it to challenge their own reasoning. They then explain why they agree or disagree, connecting their analysis to course concepts and evidence.
- Emphasize reflection as metacognition: Reflection must focus on the quality of judgement, not just the fact that AI was used. Students who simply note disagreement but cannot reconstruct their reasoning do not meet the standard. The highest marks go to those who can articulate where the AI made a fair point and explain why it did or did not shift their thinking.
Teachers looking to adopt similar tasks can explore structured professional development resources. An AI Learning Path for Teachers offers practical guidance for building assessments that go beyond responsible use policies and directly teach evaluation skills.
Many educators engaged in AI for Education discussions recognize that institutional policies alone cannot produce graduates who know when to trust the machine and when to override it. The design of the assessment itself must make that distinction visible.
Capabilities that align with employer needs
Student reflections and performance in these tasks suggest three capabilities develop through adversarial engagement with AI:
Evidence-based argumentation: The ability to construct a position, anticipate its weaknesses, and defend it against challenge. Public accountability from the live debate pushed students beyond surface familiarity with AI concepts to stress-testing their claims.
Metacognitive self-assessment: Examining how you arrived at a conclusion, not just what you think. Students said this was the harder skill because it required recognizing gaps in reasoning rather than rationalizing them away.
Calibrated judgement about AI: Understanding what AI can do and where it falls short. Students who entered with uncritical assumptions left with a grounded sense of where human judgement remains indispensable.
The lecturer argued that these capabilities-argumentation, metacognition, and AI discernment-align with what employers need from graduates working with AI. "Universities that get this right will graduate students with a scarce capacity: the ability to know when to trust AI, when to override it and why their own reasoning remains the thing that cannot be delegated," she said.
Why this matters for Education
The real challenge is not policing AI use but teaching students to evaluate it. Designing assessments that force students to challenge AI, weigh its claims, and take responsibility for deciding when to trust it builds the critical thinking that employers value. Without such tasks, universities risk graduating fluent users of AI rather than accountable thinkers who can work alongside these systems without delegating their own judgement.
Your membership also unlocks:
