Oh computer scientists, is there nothing they won’t try to quantify? Amit Kaigen of Tel Aviv University and his team have developed a computer algorithm to recognize beauty:

In the first step of the study, 30 men and women were presented with 100 different faces of Caucasian women, roughly of the same age, and were asked to judge the beauty of each face. The subjects rated the images on a scale of 1 through 7 and did not explain why they chose certain scores. Kagian and his colleagues then went to the computer and processed and mapped the geometric shape of facial features mathematically.

Additional features such as face symmetry, smoothness of the skin and hair color were fed into the analysis as well. Based on human preferences, the machine “learned” the relation between facial features and attractiveness scores and was then put to the test on a fresh set of faces.

The article is well written, fair, and Kaigen is keenly aware of how early on this study is, as well as his own personal beauty failings. What struck me was that the same data wasn’t drawn on male faces, and furthermore that the numbers were so low, only 100 images and 30 participants. I appreciate the goal of Kaigen and his team, but would really like to see the same study done online, anonymously. Just put up 100,000 pictures of people from around the world and have each person sign in with vital statistics. I understand that the data wouldn’t be up to research standards, but it would give the algorithm much more data to work with and present a much broader understanding of beauty.

But let’s comeback to the original data set for a second: 30 Caucasian women. In short, Kaigen’s team preselected what were already “beautiful” people – white women – and then had people select from there. Having computers able to process a huge volume of data seems utterly wasted on pre-selecting the data-set with such an extreme bias towards race and then further restricting it to a single sex. Here are some ideas I’d love to see Kaigen and his team try in their research:

  1. Add other races.
  2. Add men.
  3. Take a data-set and alter the skin tone of the pictures, so that light-skinned people are dark-skinned, and visa versa.
  4. Mix and match facial features. Create deliberately androgynous faces and see what happens.
  5. Add false “aggregate scores” to see how much influence the opinion of others affects perception of beauty.
  6. Using blending software, create faces that are “beautiful” or “medium” or “ugly” but not unique, to see if minor flaws contribute to beauty, as well as if synthesizing levels out or undermines perceived non-beauty.
  7. Increase sample size dramatically, gather data from those who live in non-Western and/or non-Caucasian regions.
  8. Conduct the experiment with data/subject race correlation that isn’t Caucasian, then use the “beautiful” faces with non-correlating  data/subject groups.

Any other ideas? I don’t mean to argue here that this sort of research isn’t useful, but as aesthetics is an incredibly constructed and fickle form of judgment, there is a lot more work to be done here. Also, perhaps with a reasonably large and comprehensive data set, we can start to see where biological attraction and social attraction overlap and separate.

[h/t J. Hughes]

Tagged with: