Insights · Not measured · epistemic

A single attractiveness score out of 10why we won't measure it from a photo.

No such measurement exists.

Class of problem
epistemic
Who reports it anyway
Umax, LooksMax AI, and nearly every consumer app
Canthion
Refused

The reason

No such measurement exists. Rater consensus is real and high (Langlois 2000: within-culture r=.90) but it is an aggregate-of-many-raters figure; roughly 40% of the variance in whether any one person finds a face attractive is idiosyncratic to that person (Leder 2016; Germine 2015). A single number reports the shared 60% and hides the rest. Also: there is no commercially-licensed facial attractiveness dataset in existence — every one (SCUT-FBP5500, MEBeauty, CelebA, FFHQ) is research-only — so a legally-trained consumer beauty score would have to be built on the vendor's own consented, rated data, and none has published one.

Sources
What we measure instead

Ratios of near-coplanar frontal distances, angles relative to the interpupillary line, and skin appearance — each with an error bar, each against a population you choose. Start an analysis or read the methodology.

Full refusal list · all insights