Slashdot Mirror
MIT Project "Gaydar" Shakes Privacy Assumptions
theodp writes "At MIT, an experiment that identifies which students are gay is raising new questions about online privacy. Using data from Facebook, two students in an MIT class on ethics and law on the electronic frontier made a striking discovery: just by looking at a person's online friends, they could predict whether the person was gay. The project, given the name 'Gaydar' by the students, is part of the fast-moving field of social network analysis, which examines what the connections between people can tell us, from predicting who might be a terrorist to the likelihood a person is happy, fat, liberal, or conservative." MIT professor Hal Abelson, who co-taught the course, is quoted: "That pulls the rug out from a whole policy and technology perspective that the point is to give you control over your information — because you don't have control over your information."
I am really curious if it thinks I'm gay (does it consider bisexuality?). Also, this could be useful as a dating tool; if you don't know if the object of your affections is gay or not, run them through MIT Gaydar, and then possibly feel more secure about asking them out.
Slashdot: Playing Favorites Since 1997
This is old news (and really pretty obvious) and have been known in the gay community since FB started :) I have ~250 friends and being gay, quite a few of my friends are gay too. Whenever I click on some new person I can usually tell whether that person is gay (at least if it's a guy) or not, simply based on the number of gay friends we have in common (i.e. I don't even need to look at that person's friends individually to see whom of them are gay). So if we don't have any friends in common at all, it's usually a sign that the person isn't gay. Now, being from a small country (Denmark, 5.5 mio. citizens) implies a smaller gay community, but I would still think this observation would be valid in other countries at least within cities.
The reason this works is of course that within all communities there are certain people who have _a lot_ of friends on Facebook and sort of serve as "magnets", in the sense that someone in the same community is likely to sooner or later run into that person and be added as a friend on Facebook - or at least run into one out of the "magnet" persons you are friends with.
But your friends know you. And they may, in fact, be posting information about you. Everything from tagging pictures to leaving notes. You have no control over this.
That's true to a point. But on some level that's not 'your information' that's others information about you. You couldn't stop your friends from outing you as gay or communist or vegetarian in the 60s and you can't today. Facebook isn't really a factor.
However, in terms of data mining and automated profiling etc its worse if you have a facebook account than if you don't. If someone tags a non-FB member its just a name attached to a photo. It doesn't really go anywhere. Its true that someone could see it or read a note mentioning you and connect it to you, or do sophisticated data mining to link all those references together and assemble a profile... but if you tag someone who is a fb member (the way they want you to) it creates a link back to that account, making it utterly TRIVIAL to connect it back to you.
I'm not on facebook. So while there may be some pictures on it with my name tagged to them, its not really any worse than the web in general. My name/photo is together in a few places online, but they aren't all linked together back to a single 'account' somewhere. If there are tagged photos of me on fb its the same, they are their but all disconnected. If you have a facebook account they'd all link back to that.
My 'privacy' isn't absolute. I don't expect it to be impossible for people find stuff about me online. But I do object strongly to stuff like facebook where a single company is handed tons of data self-documented by its own users... its idiotic that anyone would participate.
A couple friends of mine once wrote software to predict market commodities price changes. They had a huge dataset, the last 10 years worth of every commodity price. They tweaked it, and tweaked it, and tweaked it, and in the end it made a consistent profit over their entire dataset. Then they both invested $2,500 each, and it steadily lost every cent of it over less than a year.
It's easy to come up with a model that matches your data without even realizing it, this sounds like the exact same thing.
<xml><I><am><so><damn>Web 2.0</damn></so></am></I></xml>
"Do you actually have the slightest idea what you just said? I understand that America-bashing is fashionable these days..."
It would help if we didn't make it so easy. Like it or not, America is the gold-standard for 'sex is bad' (and 'skin = sex', therefore 'skin = bad'). Of course we inherited a goodly part of the from our English cousins, which brings us to:
"Ask Alan Turing about how tolerant Europeans can be about sexual orientation."
Alan lived and died in England, where his sexuality was illegal. I'm not sure if England considers themselves 'European' yet, but certainly most countries (not all) in continental Europe were more tolerant about sexual preference 50 years ago than most Americans today.
A.
...bringing you cynical quips since 1998