"researchers could predict personal information based on non-obvious cues. Not everyone successfully predicted to be gay, for example, had Liked “Gay Marriage.”
Supporting gay marriage is an obvious cue of someone being gay?
The most disturbing takeaway from this article is that it's not necessary to "like" marijuana legalisation to be identified as a stoner. It can be inferred from your taste of music, movies, and so forth. More disturbing: Even if you are careful what you "like", your social network is probably a sufficient indicator.
It is written poorly, "Not everyone successfully predicted to be gay" defines the set of people who are both gay and predicted to be gay, but had not clicked 'like' on "Gay Marriage."
The point I took away was that you could be successfuul predicted to be gay without, what some might see as an overt signal, "Liking" the topic Gay Marriage. Basically the other likes, the innocuous ones like Liking the show Arrested Development or Modern Family could predict things about you that you might not expect them to predict.
I hope (and expect) that eventually there will be the huge unliking.
The paper [1] is particularly intriguing. I've been wondering if anyone had done any serious correlation analysis of FB Likes with other personal traits. This is a good starting point.
There's a sea of personal information available, and most people are willingly sharing with apps/pages/sites or publicly with the world.
I blogged a little about this at the end of February after playing with the new Graph Search bar. I think when it's finally rolled out a lot of people are going to be surprised at the searches they're turning up in - it's incredibly easy to find people (friends, friends of friends, randoms) by their religion, political leanings, potentially objectionable activities (hunting, fighting, etc) etc.
If I understand it correctly, Graph Search uses only directly disclosed data. If you're searching for, say, Christians you'll only get people who list "Christian" in their profile. Facebook isn't doing statistical analysis to find people who don't list themselves as Christians, but are "probably" Christian.
I see that as a lot less troubling, mostly because it's much easier to understand what you do and don't write in your profile v.s. the unconscious patterns in your behavior.
I thought I'd seen "Friends of Christians who like...", "Spouses of friends of Christians who like..." which does a lot of the legwork for you in terms of inferences.
The problem with Graph Search is, well, it's a graph; I suspect many people are slightly surprised by graph theoretic results - Six Degrees of Kevin Bacon et al.
I think it'll wake a lot of people up to the information they're giving out to friends, but I think it's a very different thing than what this article outlines.
Yeah, you're right. But I still think there are going to be a lot of people surprised at what they're directly disclosing - I doubt my mum, for instance, who will never have adjusted her privacy settings, understands how much people can glean from her profile.
"A number of startups are helping banks decide whether to extend a loan by mining social media to predict whether a person is creditworthy"
I thought the whole point of going to the "credit score" system was the remove any risk of bias (or accusations of bias) from lending. Isn't social media mining just going to bring back the exact same issues?
15 comments
[ 1.6 ms ] story [ 41.0 ms ] threadSupporting gay marriage is an obvious cue of someone being gay?
This shit scares the fucking hell out of me.
The point I took away was that you could be successfuul predicted to be gay without, what some might see as an overt signal, "Liking" the topic Gay Marriage. Basically the other likes, the innocuous ones like Liking the show Arrested Development or Modern Family could predict things about you that you might not expect them to predict.
I hope (and expect) that eventually there will be the huge unliking.
Liking ‘Walking With Your Friend & Randomly Pushing Them Into Someone/Something’ correlates with having few friends. Go figure.
[1] is the full article of about four pages.
[0] http://www.pnas.org/content/suppl/2013/03/07/1218772110.DCSu...
[1] http://www.pnas.org/content/early/2013/03/06/1218772110.full...
The paper [1] is particularly intriguing. I've been wondering if anyone had done any serious correlation analysis of FB Likes with other personal traits. This is a good starting point.
There's a sea of personal information available, and most people are willingly sharing with apps/pages/sites or publicly with the world.
I see that as a lot less troubling, mostly because it's much easier to understand what you do and don't write in your profile v.s. the unconscious patterns in your behavior.
The problem with Graph Search is, well, it's a graph; I suspect many people are slightly surprised by graph theoretic results - Six Degrees of Kevin Bacon et al.
I thought the whole point of going to the "credit score" system was the remove any risk of bias (or accusations of bias) from lending. Isn't social media mining just going to bring back the exact same issues?