Effect on academic misconduct arising from unrestricted generative AI

Determine whether deploying Beacon during a live assessed module measurably reduces academic misconduct associated with unrestricted generative AI, including misconduct referrals or flagged submissions attributable to out-of-scope AI-generated content.

Background

Reducing misconduct linked to students’ use of unrestricted generative AI was an original motivation for Beacon, but the study evaluated the system after the relevant module had been completed and did not observe assessed submissions or misconduct cases. The authors explicitly characterize this benefit as un demonstrated and propose a before-and-after comparison during live assessment.