"Academic integrity moves to a different stance. not saying everybody's cheating, but questioning whether our teaching approaches are still up to speed and whether we're doing the right thing."
– Dr Gad Christ, University of Greenwich
The rise of generative AI has forced universities to confront a fundamental question: how do we maintain academic integrity when students have access to powerful writing tools? Many institutions initially responded with detection software and outright bans. Today, the evidence is clear that this approach has failed.
At the University of Greenwich, Dr. Gad Christ and Emma Connor took a different path, embracing authentic assessment through simulations rather than fighting an unwinnable detection battle.
See simulations that are AI-proof and designed for real-world learning.
Independent testing reveals that premium AI detection tools achieve at best 70% accuracy, nowhere near reliable enough for academic decisions. The fundamental problem lies in how AI generates text: it uses training data to create new content, making detection inherently unreliable.
Detection software creates additional problems. Non-native English speakers are frequently flagged incorrectly because they use standard phrases and limited vocabulary learned in school, exactly what detectors identify as AI-generated. Students who translate work from their native language face similar false accusations.
The solution isn't better detection. It's assessment redesign.
Learn how simulations eliminate detection concerns:
Rather than restricting AI use, Greenwich developed clear policies for acceptable use while redesigning assessments around authentic learning experiences. Emma Connor's Marketing Principles module exemplifies this approach.
Connor embedded Edumundo's Trainer Startup simulation throughout her 30-credit module, creating dynamic scenarios where traditional cheating becomes impossible. Students manage virtual footwear companies, making pricing, marketing, and strategic decisions based on constantly changing data unique to their team's choices.
The key breakthrough: Students must reflect on their simulation decisions in every assignment answer, connecting theory to their specific gameplay experience. This creates an assessment that is naturally AI-resistant while developing critical thinking skills.
Simulations create multiple layers of protection against inappropriate AI use:
Closed Systems: End-to-end encrypted platforms prevent external AI tools from accessing simulation data directly.
Dynamic Data: Each team's decisions create unique datasets that change throughout the simulation, making it impossible for AI to predict outcomes or generate relevant responses.
Cost-Benefit Reality: The effort required to extract meaningful insights from simulation data and prompt AI effectively often exceeds the work needed for authentic engagement.
Process-Focused Assessment: Success depends on decision-making processes and reflection rather than just final outputs.
Connor's results speak volumes: zero plagiarism detection across all student submissions, high-quality source usage, and dramatically increased student engagement. Students became visibly competitive, worked beyond class hours, and produced sophisticated analysis connecting theory to practice.
The 5% assessment weighting for reflection on simulation experience proved crucial. Students could not fake engagement because their reflection had to connect to specific gameplay decisions and outcomes unique to their team's experience.
Ready to transform your assessment approach?
Universities initially banned AI or invested in detection software. Both approaches failed because they treat AI as a threat rather than a professional tool students must learn to use ethically.
Greenwich's approach recognizes that AI will be integral to students' careers. Instead of preventing AI use, authentic assessment teaches appropriate application while making superficial misuse pointless.
The future belongs to educators who embrace AI as a learning facilitator while designing assessments that develop genuine competencies. Simulations provide the perfect vehicle for this transformation.
Key Takeaways: AI detection software is not reliable enough for academic use, authentic assessment makes AI misuse impractical rather than impossible, and simulations create naturally AI-resistant learning environments that develop real-world skills.