Students at MIT have devised a way to identify people who are gay by analyzing their Facebook profiles.
The method works by examining who a person is friends with, and drawing conclusions based on the characteristics of those friends. Based on the principle that birds of a feather flock together, the technique assumes that if a person is friends with a high percentage of people who are themselves identified as gay, that person is guaranteed to be super-gay.
The linked article goes off into a discussion of online privacy in the face of ever improving technology and changing social norms, but once it started using complicated words like “homophily” and “principle” I kind of lost interest. Anyone trying to make a point without first breaking it down into a list of the 10 Most Hilarious Episodes of
The Gummi Bears
isn't going to get a lot of traction with me.
No, instead of caring about the actually-interesting privacy issues, my first though upon reading this was one of concern. You see, I haven’t been particularly discriminatory when adding friends on Facebook. As a ruggedly handsome Internet humor-smith, a significant number of people have realized their lives would be meaningless without adding me as a friend. And, perhaps due to childhood memories of loneliness and having to ride the teeter-toter with the school janitor, I haven’t rejected a single friend yet. As a consequence, my Friends list is now inflated with a significant percentage of people I’ve never met at all.
So my worry, given the probable-depraved nature of anyone who would find me amusing, is what would someone running a similar algorithm on me find? Based on the company I supposedly keep, would a researcher conclude that I’m a fan of mustache-rodeos? I had to find out.
So, using the computer knowledge that all white guys with glasses possess, I fired up my laptop, and during a techno laden montage of progress bars and really fast typing, hacked in to an MIT mainframe and stole a copy of the code used by the researchers. After reviewing it, I saw numerous opportunities to expand upon the student’s work, and over the next four minutes made several improvements to the original algorithm. Most notably, I implemented several heuristic generalizations into the code and gave it the ability to pivot on all possible classification criteria, not just homosexuality. Once activated the software would comb my list of friends, their lists of friends and so on, computing a list of attributes that all my accidental acquaintances possessed. If I had a significant number of friends with a given attribute, the algorithm's output would then indicate that I myself share the same attribute.
It was clear that activating such a wide-ranging piece of software would potentially be a violation of Facebook’s terms and conditions, set me up for felony charges and usher in a dystopic world where privacy was nonexistent, and government boots were constantly pressed to our necks. So I had a nice stiff drink first. The program’s output and my accompanying notes are recorded below.