A new poster from the School of Emergent Priorities finds a student's first cover-sheet AI-use declaration predicts every later one, even as an independent reliance estimate moves considerably underneath it.
A term of 268 students' assignment declarations shows the first tick predicting every one after, while actual reliance keeps shifting underneath it
A cover sheet’s AI-use box gets filled in once, the University’s newest research poster finds, and then it simply repeats. Across a full teaching semester, a student’s first declared answer forecast every later one almost perfectly — even as an independent read on their actual reliance kept moving underneath it.
Once a box is signed, nobody comes back to ask if it still describes what’s happening. We wanted to know how long that first answer keeps being trusted anyway.
— Associate Professor Kwame Lindqvist, School of Emergent Priorities
Associate Professor Lindqvist and Dr Anouk Mensah tracked 268 students across six cover-sheet declarations each, pairing every tick against a separate, independently derived estimate of the student’s actual AI reliance at the time. The declared answer barely moved: 94 per cent of students were still ticking their Assignment One box by Assignment Six. The independent estimate moved a great deal more, drifting by a full band or further for well over half the cohort across the same stretch.
“A form doesn’t know it’s gone stale,” said Dr Mensah. “It just sits there being cited.”
The finding extends an agenda the Trajectory Analytics Group already carries — how a stated commitment, once recorded, comes to stand in for whatever it was originally meant to track — applied here to a form most people fill in without a second thought.
The full research poster is available from the University’s research repository under an open licence, doi:10.5555/slop.4vwn04.