Can You Buy the “Right” Result on a Lie Detector Test

Where this fear comes from

A fidelity test is ordered by one partner and taken by the other. Only one of them pays, so the other has every right to ask: does the result depend on who paid? The one who pays has a worry of their own: has my partner made an arrangement with the examiner in advance? Both fears are understandable: the decision people make after the test is too serious to trust the result blindly.

This distrust comes from the classic polygraph. There the device records breathing, pulse and sweating, and a person draws the conclusion: the polygraph examiner reads the charts and decides what they mean. That is no reproach to polygraph examiners; it is how the method works. But the client cannot check the conclusion from the outside and can only trust the signature.

How EyeDetect is built: the algorithm calculates the score

EyeDetect is a lie detector that reads eye reactions, with no sensors attached to the body. The person reads statements on a screen and answers “true” or “false”, while a camera records the pupil, eye movements and response time. The test is the same for everyone: the same instructions, the same pace, the same two buttons. Each question is repeated many times in different wordings, so a single accidental answer decides nothing.

After the test, the encrypted data is automatically sent to a secure server of Converus, the US company that developed the system. There the algorithm calculates a Credibility Score from 1 to 99 and generates the report, the same way for every person. The examiner cannot manually change the score: the examiner sees a result that is already final.

You open the original report yourself

A PDF file can be forwarded, printed and, in theory, altered. So we show more than the PDF. If you are next to our examiner during the test, you open the original report in the Converus system dashboard on our computer, in our presence. It shows the same result the algorithm generated, and you can compare it with the PDF you received.

This is why an EyeDetect report cannot be bought or adjusted: the algorithm calculates the score, the examiner has no way to change it, and you see the original with your own eyes.

The VerifEye online test works the same way in the part that matters: the Converus algorithm calculates the score, and the PDF report is generated automatically within 5 minutes after the test.

What the examiner does influence, and why it matters

It would be untrue to say that nothing depends on the examiner. The algorithm scores only what was asked. A vague question such as “have you cheated on me?” produces a precise score for an imprecise question: everyone understands the word “cheating” differently. And a test taken after a sleepless night more often ends with an inconclusive result.

Questions agreed in advance protect both sides. The client knows exactly what was asked. The person being tested will not meet anything in the test that they have not read: there are no surprises or traps. And they take the test only voluntarily, with written consent.

So it is worth choosing an examiner carefully, but for a different reason than people usually think: the examiner will not “draw” the score, yet whether the result is usable depends on the examiner's work. The examiner is responsible for three things.

• The questions. We word them together with you and agree them in advance. Before the test, the person being tested reads the full text with explanations: what counts as contact, what does not, and for which period.
• The conditions. A quiet room, clear instructions, a practice test. A person under the influence is not admitted to the test.
• Explaining the report. What the score means, where the line between “credible” and “not credible” lies, and what to do next.

The limits: the test can be wrong

The algorithm protects the result from human interference, but not from its own margin of error. According to Converus, EyeDetect is 86–90% accurate and the VerifEye online test 84–89%. For EyeDetect this is the share of correct results among tests with a conclusive outcome: 10–14 out of 100 such results may be wrong. An error is possible in either direction: the algorithm may doubt a truthful person or miss a lie.

Sometimes the algorithm cannot score the test at all: the person answered at random, broke the instructions, or eye or lighting problems got in the way. Such a result is called inconclusive; it is neither “truthful” nor “deceptive”. In that case we run the retest at our expense.

So the report is a basis for a conversation and a decision, not a verdict. It is not evidence in a Ukrainian court.

What to ask any examiner before you pay

These questions work for EyeDetect and for a classic polygraph alike, in Lviv or in any other city. A calm, specific answer to each of them is a good sign.

• Who calculates the result: a program or a person? Can the examiner change the score after the test?
• Will you show me the original report in the system, not only a PDF or a printout?
• Who words the questions, and will the person read them before the test?
• How accurate is the method, and which study does the figure come from? A promise of a result with no margin of error is a reason to be wary.
• What happens if the result is inconclusive, and who pays for the retest?
• Who receives the report, and is written consent required from the person being tested?