Inferring the private from the public: Facial recognition software in a post-Cambridge Analytica world


Photo by Abyssus 

In 2017, researchers at Stanford University released a paper that said their facial recognition algorithm, given a single facial image, was able to correctly distinguish a gay man from a straight man 81 percent of the time. Michal Kosinski and Yilun Wang’s paper quickly attracted criticism from LGBTQ groups, driving the researchers to release an author’s note defending their methodology.

For Kosinski, this was not his first brush with controversy. The researcher was instrumental in developing some of the methodologies later used by Cambridge Analytica to infer psychological data about people from their Facebook profiles.

We spoke to Varoon Bashyakarla, data scientist with Tactical Tech about the societal implications of the study, especially in countries where homosexuality is illegal. Bashyakarla says: “Some people have looked at Kosinski’s work and say that it’s as if he’s building and playing with bombs and complaining about how destructive war can be.”