By Daniel Whitmore, MSc in Industrial and Organizational Psychology
In short
Proportional integrity means matching anti-cheating controls to the stakes of the assessment, disclosing exactly what is monitored, and treating every signal as context for human review rather than an automatic verdict. A tab switch on a low-stakes screening task is a note on a timeline; a certification exam justifies lockdown and recording. No flag should ever reject a candidate on its own.
Signals are context, not verdicts
A tab switch during a support work sample has many innocent explanations: a notification stole focus, the candidate re-read the instructions email, a laptop hiccuped, or they simply took a breath before answering a hard scenario. It also has one guilty explanation. A signal, by itself, cannot tell you which one you are looking at — and any system that pretends otherwise is guessing with a candidate's application.
That is why integrity events belong on a timeline, not in a gate. When paste events, focus changes, and timing anomalies are recorded in sequence alongside the candidate's actual work, a reviewer can read the story: a single early tab switch followed by forty minutes of steady, original drafting reads very differently from a paste of three polished paragraphs seconds after the question loaded.
The work itself is usually the strongest integrity evidence you have. A pasted answer that does not address the specific angry customer in the prompt exposes itself; a response that engages every detail of the scenario is hard to fake regardless of what the event log says. Signals earn their place by directing a reviewer's attention — never by replacing their judgment.
Match the controls to the stakes
Integrity controls carry a cost paid by every candidate, including the honest majority. Webcam capture, lockdown modes, and environment scans add friction, anxiety, and a tone of suspicion to what should feel like a fair audition. That cost is only worth paying when the stakes justify it — and a thirty-minute screening exercise for an inside-sales role does not carry the stakes of a proctored certification exam.
A useful habit is to write down, per assessment, what a successful cheat would actually gain and what the next stage would catch. A candidate who outsources a top-of-funnel support work sample still has to perform live in a role-play and then do the job every day. That containment is itself a control, and it means the early stage can stay light.
- Low stakes (screening work samples): passive signals only — focus changes, paste events, timing — recorded silently on a timeline for review if the work raises questions.
- Medium stakes (final-round exercises, timed simulations): the same signals plus identity confirmation and a stated review policy, disclosed up front.
- High stakes (certifications, regulated qualifications): active proctoring may be justified — and still feeds human review, never an automatic verdict.
- At every level: no control whose cost to honest candidates exceeds the realistic risk it mitigates.
| Stakes | Typical assessment | Proportional controls |
|---|---|---|
| Low | Top-of-funnel screening task | Attempt limits, question shuffling |
| Medium | Shortlist work sample | Tab-switch and copy/paste signals on a reviewable timeline |
| High | Certification or final-round exam | Lockdown with violation limits, webcam/screen recording, ID check |
Tell candidates exactly what is monitored
Transparency is an integrity control in its own right. Before the assessment starts, tell candidates plainly what is recorded — focus changes, paste events, time per task — what is not recorded, and how a flag is handled: a human reviews it in context, and no one is rejected by a sensor. Candidates who know the rules can follow them; candidates left guessing behave strangely, and strange behavior pollutes your signals.
Disclosure also deters the only group you actually want to deter. A candidate planning to outsource their answers reads "paste events are recorded and reviewed alongside your work" as a real obstacle. An honest candidate reads the same sentence as reassurance that the process is fair and that they will not be ambushed by an invisible tripwire. One sentence, two audiences, both moved in the direction you want.
The real cost of over-policing
Surveillance-heavy assessments lose candidates before a single question is answered. Strong applicants — the ones with options — see a webcam requirement and a lockdown browser attached to a screening task and quietly withdraw, not because they intend to cheat but because the process signals distrust before the relationship has begun. The candidates most sensitive to that signal are often exactly the experienced support and sales professionals you are trying to attract.
Over-policing also degrades the evidence you collect. Anxiety changes performance on exactly the dimensions support and sales work samples measure: tone, composure, judgment under mild pressure. A candidate typing under the eye of an aggressive monitoring stack is not showing you how they would handle a customer on a normal Tuesday. You end up measuring their tolerance for surveillance instead of their skill.
And there is a quieter cost: every false accusation is a story. A candidate auto-flagged and auto-rejected for a dropped connection tells that story to colleagues and review sites — and in tight markets for experienced support and sales talent, those stories travel fast. Employers using assessments carry the brand risk; proportionality is how you protect it.
Human review: how a flag should actually be handled
When a flag fires, the reviewer's first stop is the work, and the second is the timeline. Does the response engage the specifics of the scenario, in a voice consistent across tasks? Does the event sit at a moment that makes sense — a pause between tasks — or at the exact moment a suspiciously complete answer appeared? Most flags dissolve in under a minute of this kind of reading; the timeline exists to make that minute easy.
For the flags that do not dissolve, escalate like an investigator, not a judge. Weigh the pattern rather than the incident, get a second reviewer on anything that might affect the outcome, and when real doubt remains, prefer a follow-up — a short live exercise or a conversation — over a silent rejection. A genuinely skilled candidate will happily demonstrate the skill again; a fraudulent one usually will not.
Whatever the outcome, record the reasoning in the decision file: what was flagged, what the reviewer examined, and why they concluded as they did. That record protects the candidate and the reviewer alike if the call is ever challenged. And the judgment stays inside this hiring decision — a flag is not a reputation, and it never follows a candidate to another employer.
Key takeaways
- Integrity signals are context for a human reviewer, never an automatic verdict — a tab switch has too many innocent explanations to be a gate.
- Match controls to stakes: passive signals for screening work samples, heavier measures only where certification-level stakes justify the candidate cost.
- Tell candidates exactly what is monitored and how flags are reviewed; transparency deters cheaters and reassures everyone else.
- Over-policing costs completion, distorts the very skills you are measuring, and puts your employer brand at risk.
- Review flags against the work on a timeline, record the reasoning, and never let a flag follow a candidate across employers.