Remote proctoring that stands up to a challenge

The point of a remote test is not that it is convenient. It is that the result is accepted by whoever has to act on it. A result nobody trusts has cost you the exam hall and bought nothing.

Acceptance rarely comes from more surveillance. It comes from being able to answer a specific question later: what happened during this attempt, and how do you know?

That means identity established at the start, and recorded. A photograph of the candidate and their identity document at session start, timestamped, is worth more than continuous monitoring with nothing retained.

It means a session record rather than a live judgement. Live invigilation is useful, but an invigilator's recollection is not evidence. What is evidence is the retained record: the video, the flag events, the timestamps.

It means anomaly flags that a human reviews, not a system that decides. Automated detection of a second face, a long absence from frame or a switch away from the test window is a reasonable signal. Treating that signal as a verdict is not, because the false positives are ordinary life — someone's family walks past, the power goes, a browser updates. A flag should open a review, not fail a candidate.

It means the candidate knowing the rules in advance, in writing, including what is recorded and how long it is kept. A challenge that begins "I was never told" is the hardest one to answer.

And it means proportionality. Browser lockdown and screen recording are appropriate for a certification that gates employment. They are heavy-handed for a formative quiz, and applying them there mostly teaches candidates that the process is adversarial.

The uncomfortable part is that most of this is administrative rather than technical. The technology to watch a candidate is easy to buy. The discipline to keep a defensible record, and to review flags rather than act on them, is what decides whether the result survives being questioned.