What integrity scores actually measure

Research · 7 min · June 14, 2026 · by Dr. Renske Carter

Tab switches, gaze, faces, paste attempts — turning a stream of proctoring signals into a single number you can defend.

Proctoring produces a firehose of events: the candidate looked away, a second face appeared, a paste was blocked, the tab lost focus. On its own, each is noise. The integrity score is how we turn that stream into a signal a hiring team can stand behind.

Signals, grouped

We roll raw events into a handful of reviewer-facing categories — focus (tab switches, fullscreen exits), camera (face not visible, multiple faces, looking away), input (paste attempts), and device / network changes. Grouping is what makes a wall of events readable.

A score you can explain

The score starts at 100 and is reduced by flagged events, weighted by severity. Crucially, it's never a black box: every flag links to the exact moment — with the webcam snapshot and the analysis behind it — so a reviewer can confirm or dismiss it.

A number nobody can explain isn't evidence. Every point off the integrity score traces back to a moment you can see.

Dr. Renske Carter

Human in the loop

Integrity scores flag; they don't decide. A reviewer can dismiss a false alarm (a candidate glancing at a second monitor to read the question), keep a flag for review, or void a submission outright. The score focuses attention — the human makes the call.

← All articles