Gaze and Head Movement

Looking away from the screen is the archetypal proctoring flag. The inference from it to cheating is weaker than any other signal in the field. The limits are similar to those around employee screen monitoring software: what appears on a screen or camera is context to review, not a direct measure of intent or performance.

What is measured

Head pose: angle away from the camera.

Eye direction, where the resolution allows it, which in a bedroom at night frequently it does not.

Duration above a threshold, and frequency of occurrence.

Some systems combine these into a single attention score.

Why people look away

Thinking. Looking away while composing an answer is normal and widespread.

Reading a long question on a second part of the screen, or on paper where permitted.

Fatigue, stretching, rubbing eyes.

Noise in the room.

A child, a delivery, a housemate.

And conditions that make sustained gaze at a camera difficult, which the fairness section covers and which is the largest systematic source.

The inference problem

To get from "looked away" to "cheated" requires assuming there is something to look at.

The system has no view of what is outside the frame, so the inference is entirely from the deviation.

Which makes gaze the weakest evidential signal in the toolkit and the one that should carry the least weight in a review.

How reviewers should treat it

As a reason to look at the recording, not as evidence in itself.

Look for what accompanies it: does the answer quality change, is there sound, does the pattern correspond to question difficulty.

Gaze alone should not support a finding, and a policy that permits it will produce outcomes that do not survive appeal.

Thresholds

Vendors ship defaults that generate a lot of flags, because under-flagging looks like failure.

Institutions can usually adjust these and rarely do.

Raise the threshold until the flag rate is reviewable, and measure what that does to the proportion resolved as nothing.

A threshold that produces reviewable volumes is more defensible than one that produces a backlog.

What to tell students

That looking away is expected and not itself a problem.

Because students who believe they must stare at the screen for three hours perform worse and are more distressed, and several report avoiding thinking away from the screen.

This is a real harm caused by poor communication rather than by the technology.

The accessibility overlap

Conditions affecting eye contact, attention, motor control and posture all produce gaze anomalies.

These students are flagged more, for reasons that are protected characteristics in many jurisdictions.

Its own note covers this, and it is the strongest single argument for down-weighting gaze.

What to check

What threshold is your system set to, and did anybody choose it?

What proportion of your flags are gaze-only?

Can a gaze-only flag start a case at your institution?

And do students know that looking away is normal?

The point

Gaze is the weakest evidential signal in the toolkit, because getting from looked away to cheated requires assuming there was something to look at.

A gaze-only flag should never support a finding.

Worth stating

Vendors ship thresholds that generate a lot of flags, because under-flagging looks like failure.

Institutions can usually adjust these and rarely do; raise the threshold until the volume is reviewable and measure what that does to the closure rate.

Also worth knowing

Conditions affecting eye contact, attention, motor control and posture all produce gaze anomalies, and those are protected characteristics in many jurisdictions.

That is the strongest single argument for down-weighting the signal.

And finally

Treat gaze as a reason to look at the recording, not as evidence in itself, and look for what accompanies it: sound, answer quality, whether the pattern tracks question difficulty.

Alone it should carry the least weight of any signal.

Summary

Raise the threshold until the flag volume is reviewable, then measure what that does to the proportion resolved as nothing. A reviewable rate is more defensible than a thorough one nobody can process.

In summary

Gaze is the signal most systems lead with and the one that should carry the least weight in a review.

Its link to misconduct is assumed rather than demonstrated, and the students it flags most are the ones least likely to have cheated. For wider institutional context, consult IEEE.