Glossary · Assessment Science
Proctoring
Proctoring refers to the controls applied during an assessment to verify that the test-taker is who they claim to be, that they complete the test independently without prohibited assistance, and that detected behaviours are consistent with genuine individual performance. In online settings, it replaces the human invigilator present in a physical test room with a combination of technical and procedural controls.
Why it matters in hiring and assessment.
The validity of a pre-employment test depends on the assumption that scores reflect the candidate's own ability under standardised conditions. If a candidate can look up answers, use AI tools, have a third party assist them, or sit in a non-standardised environment, scores become unreliable signals — a high score may reflect resources, not ability. Proctoring is the control layer that defends that assumption.
Online proctoring takes several forms, with different trade-offs between deterrence, detection, candidate experience, and privacy implications:
- Browser lockdown: Restricts the candidate to the assessment interface — blocks tab-switching, new windows, copy-paste from external sources, and often keyboard shortcuts that could invoke AI tools. Effective against casual reference-checking; does not prevent second devices or pre-loaded content.
- Webcam monitoring: Records or streams video of the candidate during the test. Flags events such as the face leaving frame, multiple faces appearing, or gaze direction suggesting reference to another screen. Reviewed by a human reviewer, by automated analysis, or by a combination. Raises significant data-privacy considerations, particularly under GDPR and India's DPDP Act.
- Environment scan: Before the test starts, the candidate is prompted to show their workspace via webcam — desk, surrounding area, any visible screens or notes. Provides a point-in-time record of the test environment.
- AI anomaly flagging: Automated systems flag events for human review rather than making pass/fail decisions autonomously. The distinction matters — automated proctoring systems have documented false-positive rates, particularly for candidates with non-standard environments, disabilities affecting gaze patterns, or certain skin tones (bias in facial recognition systems is a documented issue).
- Timestamped event logs: Every detected anomaly (tab switch, focus loss, window resize, question revisit pattern) is logged with a timestamp, enabling post-hoc review of a complete attempt record alongside the score.
Effective proctoring is deterrence as much as detection. The knowledge that behaviour is being monitored and logged changes candidate behaviour, reducing the prevalence of casual cheating even if not all sophisticated attempts are caught. Over-reliance on automated pass/fail proctoring decisions without human review introduces legal and fairness risks that outweigh the convenience.
Example.
A company runs a 60-minute proctored coding assessment for 200 applicants. The platform enforces full-screen mode and logs 14 tab-switch events across the cohort, flags 3 attempts where a second face appeared in frame, and records 2 attempts where copy-paste was triggered from outside the browser. The admin reviews the flagged attempts alongside the submitted code and timestamps. One attempt is voided after review; the other flags are noted but the code quality does not suggest external assistance. The proctoring layer produced an evidence record — it did not make the decision; the reviewer did.
Related terms.
- Construct Validity
Proctoring failures (undetected assistance) threaten construct validity by introducing variance from sources other than the target ability.
- Reliability Coefficient
Systematic cheating inflates scores and reduces inter-candidate reliability — the score no longer reflects consistent individual performance.
- Adverse Impact
Biased automated proctoring systems that flag certain demographic groups at higher rates can introduce adverse impact through the monitoring layer rather than the test content itself.
- Cut Score
Proctoring flagged attempts should be reviewed before a candidate is screened out at the cut score — a flag alone is not grounds for disqualification.