"Gaydar" Developed Using Social Networking Data: MIT

Gaydar is a term frequently used in sitcoms (and perhaps among male friends) to denote the ability to pick out homosexuals at a glance. Silly as that may be, two MIT students, while working on a term project for an ethics and law project, appear to have stumbled onto a working form, by leveraging data gleaned from social networking sites. This is actually a serious issue, as it just emphasizes further the need to be careful on social networking sites.

The students, Carter Jernigan and Behram Mistree, dubbed the 2007 project "Gaydar." They took information from Facebook involving friends a user might have on the site, and with software made used statistical analysis to make predictions based on the gender and sexuality of a person’s friends. The results, the pair said, was very accurate, at least for men.

Of course, none of this should be surprising using commons sense. People tend to associate with those who have similar beliefs, politics, etc. as they do. "Birds of a feather, flock together," as the old adage says.

Jernigan and Mistree “trained” their computer program by analyzing the friend links of 1,544 men who said they were straight, 21 who said they were bisexual, and 33 who said they were gay. Gay men had proportionally more gay friends than straight men, and that data was used by the software to make its predictions.

Once "trained," the students used their software to analyze 947 men who did not report their sexuality. Scientifically, here's where things get dicey. They did not have a way to verify their analsysis against all of those analyzed. All they could do was use their own knowledge of 10 people in the network who were gay but did not declare it on their Facebook page to check. All 10 people were predicted to be gay by the program.

However, while the analysis seemed to work on men, it did not appear to work as well on bisexual people or on lesbians.

While cautions have been raised about what you post on social networking sites such as Facebook, it hadn't been taken to this obvious conclusion. Once again, a little common sense thinking should have led to a theory about this sort of thing before.

Perhaps Google was right when they said complete privacy is a thing of the past. Facebook spokesman Simon Axten said:
“In general, it’s not too surprising that someone might make inferences about someone else without knowing that person based on who the person’s friends are. This isn’t specific to Facebook and is entirely possible in the real world as well. For example, if I know that someone has certain political views because that person makes them known in some way (say, by putting a bumper sticker on his car), and then I see the person walking out of a movie with friends I don’t know, I might assume those friends also have those political views.”
Jernigan and Mistree are still considering publishing their work in a journal so Facebook did not have an opportunity to review the study.