When educational technology platform Turnitin adjusted its AI detection model last year, many educators noticed a sharp increase in assignments being flagged as “AI-written”. That raises a difficult question: What happens when a student who genuinely completed their own work is wrongly flagged?
Since generative AI entered classrooms and universities, much of the conversation has centred on one question: How do we ensure students are still genuinely engaging with learning?
It’s an understandable concern.
Educators worry that overreliance on AI tools may weaken critical thinking, compromise learning outcomes, and create unfair advantages in assessment. Yet the real issue may not be AI use itself, but how we assess learning in an AI-enabled environment.
AI detectors are far from perfect. False negatives allow AI-generated work to slip through, while false positives mistakenly flag authentic student work. The consequences of false positives can be serious.
Students may lose confidence in their abilities. Trust between students and educators may erode. Institutions may unintentionally create environments where suspicion overshadows learning.
Perhaps the better question is not detection – but design.
From Detection to Design
Instead of relying heavily on AI detection tools, educators can rethink how assignments are designed in the first place. At SUSS, as part of our critical thinking curriculum, we focus less on whether AI was used and more on whether the work demonstrates three things:
Veracity: Is the information accurate and supported by evidence?
Relevance: Does the response meaningfully address the question?
Understanding: Can the student explain the logic and reasoning behind their work?
These criteria shift the focus back toward the quality of students’ thinking.
Good assessment design can naturally discourage superficial AI use. This might include:
- asking students to engage with highly course-specific concepts
- requiring references with verifiable links
- incorporating recent developments or breaking news
- following up orally with students when needed
Ironically, if a student uses AI thoughtfully enough to produce work that demonstrates accuracy, relevance, and coherent reasoning, they have likely engaged more deeply with the material than we often assume.
That does not mean “anything goes” with AI. But it does mean the conversation should become more nuanced.
Professor Cayce Meyers of the School of Communications at Virginia Tech University recently posed an important reframing of this issue. Instead of asking, “How do we stop students from using AI?”, perhaps educators should ask, “What kinds of learning do we want to protect?”
That distinction matters enormously.
AI is not disappearing. Today’s students will graduate into workplaces where AI tools are embedded into everyday professional life. The goal, then, cannot simply be prohibition. It must be responsible use, critical engagement, and intellectual accountability.
Detection still has a role. But it should exist within a broader ecosystem that prioritises thoughtful assessment design and authentic learning experiences.
Ultimately, we should spend less energy trying to detect “AI writing” and more energy identifying poor academic thinking. When students lose marks, it should be because their arguments are weak, unsupported, or poorly reasoned – not because an algorithm believes their writing style resembles AI.
Education has always evolved alongside technology.
The challenge now is not whether AI exists in learning, but whether we can teach students to use it responsibly while still developing the habits of critical thought that matter most. And perhaps that is the real opportunity in front of us today.
Associate Professor Adrian Kwek is from the College of Interdisciplinary & Experiential Learning at the Singapore University of Social Sciences. His research examines critical thinking and issues in applied ethics.
