Face Analysis

Picture faceanalysis.png Yuki recently blogged her experiments with FaceAnalyzer.com. It sounded fun, I decided to go see for myself.

FaceAnalyzer is a website that claims to be able to give you, from a mere picture, an automated breakdown of your “race” (which, may I remind you people, in the case of humans, is mostly a social construct, with hardly any ground in biology), along with an estimate of your “Intelligence”, “Ambition”, “Risk” (whatever the hell that may be), “Gay Factor”… complete with, I kid you not: Income bracket

Now I know what you think: quite a bit of pseudoscience, mixed in with a whole lot of outright quackiness…

Only one way to know, right?

Upon submitting a fairly broad sample of my likeness, in various states of drug-induced haze, hair color and cheekiness, the results were rather mitigated:
While I can certainly accept the possibility that I may have a fair share of East-Indian or Chinese blood in me (please no “ever had some [insert ethnicity] in you?” joke here), I find it a bit harder to accept that I’d be a woman (as the system insisted on pointing out, in every single case)… Although that would explain a certain fascination for lace undergarments and a secret passion for knitting I always attributed to my deviant morals… But in that case, wouldn’t all the lesbian action give me a “Gay Factor” at least somewhat above average?

Obviously not very conclusive. But science makes mistakes… One should not draw hasty conclusions and bluntly discard such breakthrough technology as “money-making snake’s oil scheme” without at least a second chance…

So I bravely submitted a picture of my brother’s cat, and I must say the results are troubling.

  • Indeed, the cat is FEMALE! Most definitely has a good deal of middle-eastern ancestry, in fact, her lineage can probably be traced to every other urban slum of the Eurasian continent, with the odd pedigreed tomcat here and there…
  • Frankly not sure about that “Low Promiscuity” rate: I mean, before she got spayed, the girl could be quite a little trollop around the neighbourhood.
  • It is a bit disappointing that she only displays “Average” ambition, but I guess we will have to find solace in the fact that she is destined to make at least $30,000 a year… not bad for a cat who doesn’t even have her high-school degree.

I’ll be contacting my brother shortly and urge him to stop wasting her talent on stupid things like chasing rats and rummaging around the neighbourhood: that cat has potential (Beta Academic potential!), she needs to be put to work!

By the way, physiognomy is not a new idea. It hasn’t been given the slightest bit of scientific credibility over the past 50 years, though.

As for how the website gets its “uncanny” success rate regarding geographical phenotypes: any first-year computer science student can get the same results with a big-enough sample database and some very low-grade face-matching algorithm. Any rating merely suggest a match with existing pictures of a certain ethnicity, which is well below the crease of scientific procedure, hence the wild margin of error. In fact a properly trained neural network would probably give vastly superior results on some criteria (such as gender, for example).

Needless to say that detection of personality traits such as “Gay Factor”, “Promiscuity” or even “Intelligence”, through face analysis, is complete and utter bullcrap. The last people with an academic background to have given it some sort of scientific credence ended up cracking capsules of cyanide in their mouth, approximately sixty years ago.

4 comments

  1. It amused me, but wasn’t very accurate. It tried to tell me I was middle eastern (i’m southern european) and it told me I was much less promiscuous and gay than I actually am. Hmph.

  2. It would be fun to put all the moblogging photos via this filter prior to posting so that it saves on any explanation (and typing) from the poster..

    I foresee nobody wanting to have their photo posted in future unlike now where I have people saying “Blog me, Blog me!!!)

  3. I tried two photos taken on the same day. The one where I wasn’t wearing cosmetics got me a 100% Korean/Japanese race analysis (!) but in the one with makeup I was 77% Eastern European and 23% Anglo. I think their algorithms need some fine tuning.

Comments are closed.