BI article: A Stanford professor thinks AI will be able to detect your politics, IQ, and sexuality — but not everyone agrees

http://uk.businessinsider.com/stanford-professor-thinks-ai-will-soon-be-able-to-detect-your-politics-iq-sexuality-2017-9
http://www.dailymail.co.uk/sciencetech/article-4876230/Face-reading-AI-detect-political-beliefs-IQ.html
from the same author of
Finally the perfect dating algorithm? Of course NOT, Big Five test is OBSOLETE.
http://onlinedatingsoundbarrier.blogspot.com.ar/2017/02/finally-perfect-dating-algorithm-of.html

Research at
https://globaldatinginsights.com/2017/09/08/this-algorithm-can-accurately-predict-your-sexual-orientation-using-just-a-picture-of-your-face/
https://psyarxiv.com/hv28a/
https://osf.io/zn79k/

Images/photos used in the samples:
Persons with facial surgery? or even worse Photoshopped images?
People with facial surgery / Photoshopped (nose, lips, eyebrows) can introduce bias in the algorithm.

208 Facial images. We obtained facial images from public profiles posted on a U.S. dating
209 website. We recorded 130,741 images of 36,630 men and 170,360 images of 38,593 women
210 between the ages of 18 and 40, who reported their location as the U.S. Gay and heterosexual
211 people were represented in equal numbers. Their sexual orientation was established based on the
212 gender of the partners that they were looking for (according to their profiles).

"Gay and heterosexual people were represented in equal numbers. "

Perhaps it is a flaw. Gay people are OVERrepresented in the sample.
Perhaps Gay people should be less than 10% in the sample.

If in real world heterosexual people are 90% of any REAL / REPRESENTATIVE sample of any real population, and 10% are only Gay and Bisexual, it will be easy to any algorithm to achieve 90% precision detecting heterosexual people!

It only needs to say all of them are heterosexual people and it will fail for a 10%,  i.e. 90% precision !

"Their sexual orientation was established based on the gender of the partners that they were looking for (according to their profiles)."
Another flaw? How many heterosexual people can be bisexuals and they are not represented in the sample.
605 to the quality of the ground truth: it is possible that some of the users categorized as heterosexual
606 were, in fact, gay or bisexual (or vice versa). However, we believe that people voluntarily
607 seeking partners on the dating website have little incentive to misrepresent their sexual
608 orientation.

Please see:

How facial features drive our first impressions
http://onlinedatingsoundbarrier.blogspot.com.ar/2014/07/how-facial-features-drive-our-first.html

and
we asked 252,900 straight people have you ever had a sexual encounter with someone of the same sex?
Almost a quarter answered ‘yes’.
https://theblog.okcupid.com/gay-sex-vs-straight-sex-49227971d33
http://www.huffingtonpost.com/frankie-thomas/the-paradoxical-world-of-bisexual-okcupid_b_7640726.html
The science behind a more meaningful understanding of sexual orientation
http://theconversation.com/the-science-behind-a-more-meaningful-understanding-of-sexual-orientation-42641


Ovulatory cycle and changes in face width: How women can tell when other females are ovulating using clues in their face - and how they may then try to hide their partners from the 'threat' of these fertile ladies
http://www.dailymail.co.uk/sciencetech/article-3417751/University-Bern-study-finds-women-know-females-ovulating-thanks-clues-face.html

Women's attractiveness changes with estradiol and progesterone across the ovulatory cycle
http://sites.uci.edu/dhbailey/files/2014/07/Puts-et-al-2012-Horm-Behav.pdf


Please see also:
Interesting and Controversial Chinese AI software judges
http://onlinedatingsoundbarrier.blogspot.com.ar/2016/11/interesting-and-controversial-chinese.html
Automated Inference on Criminality using Face Images
https://arxiv.org/pdf/1611.04135v1.pdf