Tuesday, March 12, 2013

Report: "Liking" things on Facebook reveals that you are stupid

There's a pretty amazing news story going around that researchers have developed an algorithm that can analyze a Facebook user's public "likes" to deduce with startling accuracy various personal attributes, including the user's race, religion, politics, sexual orientation, drug history, and, most amusingly to me, their intelligence. According to TechNewsDaily:
For example, users who liked "Sephora," "Harley Davidson" and "I Love Being a Mom" were less likely to be intelligent, whereas likes that predicted high intelligence included "Thunderstorms," "Science," "The Colbert Report" and "Curly Fries."

I'm extremely skeptical of the veracity of this study. What metric are they even using to gauge the accuracy of the intelligence assessment? I kind of wish I could have participated in it. What would the machine have made of me? Well, probably not a lot, since I most likely haven't liked enough things on Facebook to provide a usable set of data toward drawing any conclusions. I know many people fill up their likes lists as basically an extension of their "About Me" section (though perhaps they never intended to reveal as much about themselves as this algorithm would allegedly infer). I originally would only like things if I wanted to get updates about them in my news feed. Then I got sick of all the ads, so I took back most of my likes. But, supposing all of my past likes were still out there for analysis, what would the machine conclude about this guy who "likes" animal rights activist groups, conservative Christian relief organizations, existential philosophers, lesbian singer-songwriters, Charlize Theron, Sanrio characters, sandwiches, and the Street Fighter series of video games? Who knows, maybe it would totally have my number, whatever that might be. It's funny to consider that a computer might, in a way, have us better figured out than other human beings ever could. Better perhaps than we know ourselves—how many of those guys in the study sample were actually confused about their own sexuality and didn't know themselves that they were gay, which the computer was supposedly able to very easily predict based on their music and movie interests?

The researchers did open up the algorithm for the public to get summaries of their own results. Unfortunately, all you get are broad statements according to a binary system across five categories describing your personality (e.g. in the category "Extraversion," "shy and reserved" versus "outgoing and active"). I gave it a try, and some of it was (broadly) accurate, while some of it was not. I'd agree that I'm more "liberal and artistic" than "conservative and traditional" on the "Openness" scale, but I would not consider myself "calm and relaxed," "well organized," or "warm, trusting and cooperative." Again, it might be that I don't have enough likes to draw any conclusive results. Or this might simply be an accurate representation of the five guys I pretend to be on Facebook, none of whom I actually am.

Of course, the researchers are also convinced that Google and Facebook have already long had their own sneaky algorithms that collect this same kind of data, gathered to help them tailor their advertising to the user. Now someone explain to me how that translates to my getting fed banner ads on YouTube for Muslim singles sites.

No comments: