Business Case March 2026 UiT The Arctic University of Norway
AI Assisted Feedback: Findings from a small-scale pilot at UiT
What six dentistry students made of AI-generated feedback on their immunology answers.
This is relevant to you because:
- You want to offer formative feedback at scale, but marker workload makes individual comments impractical.
- You're evaluating AI feedback tools and want evidence from real students, not vendor claims.
- You need to know how much prompt design affects the quality of the output.
- You're weighing where human oversight remains necessary once AI is drafting feedback for students.
Abstract
A pilot at UiT tested whether the AI-assisted feedback module in WISEflow could produce useful, consistent feedback grounded in student answers and a marking guide. Six second-year dentistry students submitted written responses to five immunology tasks, and four evaluated the feedback they received.
Students found the feedback clear and relevant, and all four agreed it identified strengths and areas to improve. The dominant theme was timing: feedback arriving after the exam was judged far less useful than a mid-course checkpoint would have been. The assessor found output consistent when one standardised prompt was applied across all six students, and noted that default summaries were less reliable than explicitly prompted analysis.