Do Face Scan Scores Match What Real People Think?

Two ways to judge a face

A face scan and a person look at the same face and do completely different things. The scan measures the positions of features in one still image. A person forms an impression in a fraction of a second, from a moving face, in a real room, with context attached. Asking whether the two agree is a bit like asking whether a thermometer agrees with a weather forecaster: both are about temperature, but they do not measure the same thing.

What the scan sees that people cannot

The scan has one advantage: it can be exact about geometry. It can report that the left eye sits two millimetres higher than the right, or that the ratio of face width to face height is a particular value. A person cannot produce a number like that, and would not care if they could. Precision is real, and it is also nearly irrelevant to how anyone responds to a face.

What people see that the scan cannot

Everything that actually drives attraction lives in the category the scan ignores: expression, warmth, eye contact, symmetry of movement, skin, and the way a face changes when someone talks. None of those exist in a single photo, so no photo-based scan can measure them. A still image throws away most of what a face is, then measures what is left with great care.

Why agreement is not the same as accuracy

Even if a scan and a person happened to agree, that would not settle which one is "right". The person is not running a measurement — they are having a reaction, and reactions vary with mood, hunger, context, and who else is in the room. The scan is running a measurement, but of a stripped-down version of the face. Agreement between a rough impression and a narrow measurement is not a standard anyone needs.

The consistency illusion

Face scans feel objective because they give the same-looking answer in the same-looking way every time. People do not — two people can look at the same face and feel opposite things, and one person can feel differently on a different day. That inconsistency reads as unreliability next to a machine. But it is the machine that is doing something narrow: it is being consistent about the wrong thing, and consistency is not the same as truth.

What a scan is actually good for

A face scan is a fine camera coach. It responds immediately to lighting, distance, angle, and expression, which is genuinely useful if you want better photos. It is a poor judge of how you come across, because that judgment does not live in a still image. Use the number for the thing it can do, and leave the question of how people see you to the people. That division of labour is the whole trick: let the machine do what machines are good at, and let people do what people are good at.

Frequently asked questions

Do face scan scores agree with people's opinions?

Rarely, and there is no reason they should. A scan measures geometry in one photo; people respond to a living face with context. Two different tasks, two different answers.

Are people more "accurate" than a face scan?

Neither is accurate, because they are not measuring the same thing. A person is having a reaction, not taking a measurement. The scan is taking a measurement, but of a fraction of what a face is.

Should I trust a face scan over a friend's opinion?

For photo quality, trust the scan — it can see light and angle precisely. For how you actually come across, trust the people around you, who can see everything the still image leaves out.