People who innocently connect over social media may be unwittingly broadcasting their sexual preference, according to a study that correctly predicted whether men were gay by analysing their list of friends on Facebook.

The study, which was a project for a class at the Massachusetts Institute of Technology on internet law and ethics, revealed ¡°how little control one has over one¡¯s own privacy in online communities¡±, wrote the authors. The project was called Gaydar.

Two students, Carter Jernigan and Behram F.T. Mistree, used a software program to analyse 4,080 Facebook profiles of fellow MIT students and alumni. They discovered that the percentage of a male Facebook user¡¯s friends who identified themselves as gay could help predict his sexual orientation, even if the user did not reveal it himself. Mr Jernigan and Mr Mistree relied on private knowledge about some users who were gay to confirm their results: all were predicted by the program to be gay.

¡°Birds of a feather flock together,¡± the two men wrote. ¡°The lesson is that people are self-segregating, such that the composition of your friends reflects on you.¡±

ÁñÁ«ÊÓƵ

Although it was less successful at predicting whether women were lesbians or whether men or women were bisexual, the study found that gay men had between 3 and 18.3 times higher a percentage of gay male friends than straight men.

Mr Jernigan, who has graduated and now works as a software engineer, and Mr Mistree, now a doctoral student at Stanford, have worked to keep their own privacy intact. Neither responded to repeated requests for comment.

ÁñÁ«ÊÓƵ

Their computer-science instructor at MIT, Harold Abelson, also declined several times to speak about the project. ¡°I¡¯m afraid you¡¯ll have to ask the students,¡± he said. ¡°It¡¯s their study.¡± Nor did MIT have anything to say about it.

But privacy advocates said the research, published this week in a peer-reviewed online journal First Monday, raises issues about online confidentiality beyond controversial matters such as whether employers should be able to check Facebook and MySpace photographs to see if any showed prospective employees drinking excessively or doing anything illegal.

Researchers at the University of Texas separately analysed friends¡¯ and other information on 167,000 Facebook profiles to predict users¡¯ political views. And at the University of Maryland, researchers studying social networks found they were able to predict such things as where a user lived, even when the user didn¡¯t disclose it.

¡°I don¡¯t think anyone has ever had so much information about exactly who we¡¯re all connected to,¡± said Kevin Bankston, senior staff attorney at the Electronic Frontiers Foundation, which advocates for digital privacy. ¡°Even if you¡¯re not affirmatively posting something such as what your sexual orientation is, other information you do post can be used to help determine it.¡±

ÁñÁ«ÊÓƵ

He said people can set their social networks to prevent outsiders from viewing their list of friends. But Mr Jernigan and Mr Mistree bypassed the privacy setting by analysing Facebook profiles that were part of an MIT network of which they were also members, and to which they therefore had unfettered access.

¡°Here is a good reason why people should hide the list of who they¡¯re friends with,¡± Mr Bankston said. ¡°I¡¯m not sure it¡¯s a question of control. It¡¯s a question that people aren¡¯t exerting their control. They don¡¯t recognise how much information could be derived from their friend lists. It¡¯s also a matter of how many people don¡¯t know that they even have that option.

¡°The upshot is simply that in an age with so many tools for publishing information about ourselves, we have yet to sort out what is safe to publish or not,¡± he said.