Assessment integrity without a camera

The obvious answer is a camera and an algorithm. We think it is the wrong answer, and this is the argument.

What a browser can honestly report

A cooperating browser can report that a tab lost focus, a window was resized, a paste happened. Each has an innocent explanation — a notification, a screen reader, a phone call — and none is evidence of anything alone.

What we do with them

• The candidate is told, while it is happening, that the signals are being recorded.
• A human marker sees them, with the caveat printed on the same screen.
• Nothing is automatic. No score is adjusted, no attempt is voided, no verdict is produced.

Why not a camera

Camera proctoring asks a learner to install surveillance on a device they own, in a room they live in, and to accept a false-accusation rate from a model that has never met them. The literature on differential false-positive rates by skin tone and disability is not comforting. Meanwhile the cheating it catches is the naive kind, which item analysis catches anyway.

What actually raises integrity

• Question banks large enough that two learners see different papers.
• Item analysis, so a question everybody answered correctly gets retired.
• Anonymised marking, so the marker does not know whose paper it is.
• Rubrics, so two markers agree.
• An audit trail on every grade override, with a reason.