UCT Pioneers New Era in Academic Integrity: Drops AI Detection for Ethical Engagement

UCT's Decision to Abandon AI Detection

The University of Cape Town has officially discontinued use of AI detection tools, including Turnitin's AI Score. Sukaina Walji, Director of UCT's Centre for Innovation in Learning and Teaching, explained the decision stems from fundamental flaws in detection technology.

"These tools are simply not reliable enough," Walji stated. "They're a form of AI that seeks to detect other forms of AI, but cannot consistently distinguish between human and AI writing."

The university identified two key problems:

1. False positives: Students being wrongly accused of using AI
2. False negatives: AI-generated content going undetected

"It creates mistrust between staff and students," Walji added, calling detection tools "fundamentally the wrong way to approach teaching and learning."

New Approaches to Academic Integrity

Instead of detection, UCT is implementing alternative strategies:

1. Observable Assessments

• Invigilated exams
• Oral examinations (vivas)
• Practical demonstrations
• Process-based evaluations

2. Critical AI Literacy

• Teaching students when AI use is appropriate
• Ethical discussions about AI tools
• Discipline-specific guidelines

3. Assessment Redesign

• Incorporating AI use where relevant
• Requiring draft submissions
• Reflection exercises on AI-generated content

"We have to really think and change our assessments," Walji said. "AI is here. Students are using artificial intelligence tools, so assessment strategies need to change."

Supporting Lecturers Through Transition

UCT has launched an AI Teaching Innovations Project to help faculty adapt:

• Developing use cases for AI in teaching
• Exploring multilingual education applications
• Curriculum redesign support

"We're providing spaces and strategies to work with lecturers to really look at how they're teaching in their courses," Walji explained.

Global Context and Open Sharing

While UCT is among the first African universities to take this step, Walji emphasized the global nature of the challenge:
"We've drawn very much on what our colleagues are doing in other universities... We release all our guidelines under open access licenses."

The university plans to continue sharing its framework as it develops more comprehensive policies for AI integration in academia.