Your students use GenAI. Let’s redesign learning so thinking still happens
Created in partnership with
Created in partnership with
You may also like
For years, higher education has rewarded students for producing answers efficiently. They learned how to absorb information, reproduce knowledge and satisfy evaluation systems built around outputs. Artificial intelligence did not create this model, but upon its launch, it exposed its weaknesses almost immediately.
Today, students can generate polished answers faster than ever before. Yet many still struggle to explain how ideas connect, why certain decisions make sense or how their reasoning evolved in the first place.
In my own classrooms, particularly while teaching design thinking, critical thinking, systems thinking and visible reasoning, I began noticing a pattern very quickly. Many students could produce work that looked sophisticated on the surface, yet the reasoning underneath often remained surprisingly fragile. They had become skilled at what I started thinking of as pseudo-polished reasoning: answers that appear intellectually convincing while masking shallow engagement with the thinking process itself.
- The best way to teach students to think critically about AI? Make them argue with it
- Banning AI won’t make students care about learning
- Campus Talks: What does ‘AI across the curriculum’ look like in practice?
The issue wasn’t simply that students were using AI. The issue was that many had already become dependent on externally generated structures long before AI arrived. AI simply accelerated the process.
This became particularly visible whenever students were pushed outside traditional academic formats. The moment activities moved beyond note-taking, memorisation or standard written responses, many students froze.
One of the clearest examples emerged when students were asked to externalise their reasoning rather than explain it through text alone. Instead of writing paragraphs, they had to build cognitive maps showing how ideas connected, evolved and interacted. They were immediately uncomfortable.
Students repeatedly asked whether there was a “correct format”. Others became visibly anxious once they realised they could no longer rely on polished wording alone. What became obvious very quickly was not simply uncertainty about drawing or mapping – it was uncertainty about their own thinking structures.
Higher education often trains students to internalise learning, but rarely teaches them how to externalise thinking – and suddenly that distinction has become far more important.
When students are asked to externalise reasoning, structure relationships spatially or explain how one idea leads to another, shallow understanding becomes much harder to hide. In many cases, students are no longer able to rely on memorised formulations, surface-level understanding or AI-generated phrasing. Either the reasoning exists, or it does not.
Yet externalising reasoning often reveals far more about a student’s actual understanding than polished essays ever could.
The challenge became even more visible once AI entered the equation.
Initially, many students approached AI in predictable ways. They uploaded assignment prompts directly into generative AI systems and waited for answers. In many cases, the goal was not exploration or learning. It was avoiding uncertainty itself.
This revealed another deeper issue inside higher education: many students now perceive not knowing as failure rather than as a stage of learning. Instead of struggling through ambiguity, testing possibilities or constructing understanding gradually, students often feel pressure to immediately produce correct-looking outputs. AI becomes attractive because it removes discomfort from the learning process.
But meaningful learning has always involved discomfort.
The classroom gets messier, and thinking gets better
Rather than banning AI or treating it solely as a threat, I began experimenting with redesigning classroom activities so students would still need to construct, test and defend their reasoning even while using AI tools.
The biggest shift came when AI stopped functioning as an answer machine and instead became part of a broader reasoning workflow. Instead of: Prompt → Answer; the classroom increasingly moved toward: Explore → Structure → Prototype → Test → Revise → Reflect.
One recurring activity I use while teaching design thinking is something called the Birthday Cake Project. Inspired partly by rapid design sprint models such as Stanford school’s Wallet Project, the activity initially appears deceptively simple: students are asked to design the “perfect” birthday cake for a highly specific client.
But the complexity escalates very quickly.
One group may need to design for a diabetic client, another for a client with coeliac disease, another for a professional athlete needing sustained energy without feeling physically slowed down. Students must simultaneously navigate constraints involving ingredients, texture, accessibility, portability, aesthetics, nutrition, structure and user experience.
Very quickly, students realised there was no perfect answer.
Every design decision created trade-offs. Improving accessibility increased cost. Prioritising health requirements affected taste. Adjusting one variable destabilised another.
The activity stopped being about finding the correct answer and became an exercise in negotiating complexity.
Most importantly, AI was intentionally delayed from the process until students first built their own reasoning structures, sketches and conceptual blueprints. The objective was never for AI to think in place of students. The objective was for students to build enough cognitive structure that AI became an extension of reasoning, rather than a replacement for it.
Once students finally reached the AI stage, the interaction changed dramatically. Students began using AI to visualise prototypes, test assumptions, simulate stakeholder perspectives and identify weaknesses in their own ideas. In many cases, the technology exposed flaws students had not initially noticed.
The most important shift, however, was not technological. It was cognitive.
Students slowly began moving away from passive answer consumption and toward more active forms of reasoning. They started testing assumptions instead of accepting first responses. They revised ideas more willingly. They became more comfortable questioning outputs, identifying contradictions and navigating ambiguity.
Some also began discovering that AI could help them build systems for their own learning rather than simply generate answers for them.
Tools such as NotebookLM became particularly interesting in this regard. Students realised they could feed their own notes, reflections, class materials and research into AI-supported systems and build personalised learning environments around their own knowledge structures. Instead of passively consuming information, they became active organisers of it.
AI exposed what passive learning was hiding
This is where I believe higher education now faces its real challenge. The issue is no longer whether students use AI – they already do.
The issue is whether universities continue designing classrooms around passive knowledge reproduction while the world increasingly rewards adaptability, systems thinking, autonomy and reasoning quality.
AI is forcing higher education to confront a difficult reality: learning environments built primarily around answer production no longer hold the same value they once did.
Knowledge is now everywhere. The real educational challenge is helping students learn how to navigate uncertainty, construct thinking systems, connect ideas, externalise reasoning and find multiple routes toward understanding.
For educators wondering where to begin, the shift may not require abandoning everything. But it does require redesigning learning so students remain cognitively active throughout the process.
A few shifts can make a significant difference across disciplines:
- Assess reasoning, not just outputs. Ask students to explain how conclusions were constructed, challenged and revised rather than simply presenting polished answers.
- Make thinking visible. Use cognitive mapping and visible reasoning practices that require students to externalise how ideas connect, evolve and interact.
- Sequence learning intentionally. Give students space to construct initial understanding before AI acceleration enters the process.
- Use AI to explore possibilities. Encourage students to prototype ideas, test perspectives, expose overlooked gaps and strengthen reasoning rather than outsource thinking.
- Design backwards from thinking. Build activities around the type of reasoning students should demonstrate, not simply the content they are expected to reproduce.
- Reward productive ambiguity. Treat iteration, reflection and uncertainty as legitimate parts of learning rather than valuing only polished immediacy.
- Keep classrooms cognitively active. Create environments where students build, test and defend understanding instead of passively receiving information.
AI may already be embedded into how students learn, research and solve problems. The current and future challenge facing universities is redesigning learning environments that keep cognition active throughout the process and across disciplines. And honestly, that challenge may prove far more complex than simply integrating AI itself.
Soukaïna Aijou is a lecturer at Al Akhawayn University.
If you would like advice and insight from academics and university staff delivered direct to your inbox each week, sign up for the Campus newsletter.
