Watch the Proctor
Turn the cameras on the AI proctor: which innocent behaviours get students flagged?
The job
Finds the innocent behaviours that make AI exam proctoring flag students as cheating, proves the pattern with evidence and appeal outcomes, gets the vendor to change the rule, and checks that false-flag complaints fall at the next exam.
The moment
A wall of screens turns the cameras on the AI exam proctor, one screen per innocent behaviour that gets students wrongly flagged for cheating.
Screen 3 shows "face obstructed": 37% of this term's false-flag complaints come from students adjusting their glasses or pushing back their hair, and 88% of those appeals were upheld.
The dean sends the evidence pack to the vendor before midterms.
After midterms, flags for that cause are down 90%, screen 3 goes dark, and the appeals committee's queue halves.
What it does
- Send evidence pack to the proctoring vendor
- Log a setting change with expected effect
- Notify affected students of the outcome
- Save a cause as a reusable search
- Open cause verbatims
What you see
An inverted CCTV wall (monitors watching the proctor, one per false-flag cause) with anonymised pictogram reconstruction storyboards and a rule register
What it moves
False-flag complaints per exam week for each cause, and the share of appeals upheld, with the appeals committee hours saved.
Built for
- Compliance, complaints & legal
- Partners & ecosystem
