We now bring you a new feature called Ask Dr. Strangedata, in which our resident data-mining expert David Auerbach, who previously worked in the software bowels of Google and Microsoft, answers your questions about how Internet companies know so much about you. Why is your crazy ex suddenly popping up on your suggested friends? How did Facebook know that you like Antiques Roadshow? Why did Facebook ask if you had a Chinese name? Dr. Strangedata has the—or at least a—diagnosis. (Please note that David is an engineer, not a doctor, damn it.)
Dear Dr. Strangedata,
Facebook recently suggested I become friends with two men I went out with once many years ago (not at the same time!) and never want to see again. Someone told me that if I still have their numbers in my phone, and I have Facebook on my phone, Facebook may mine my phonebook. I am sure I never (intentionally) told Facebook to search my contacts.
Mobile apps share their data promiscuously, sometimes without you realizing it. If you ever chose to “Sync” your Facebook contacts and your iPhone or Android contacts, Facebook certainly took the opportunity to go through your entire phonebook and see who turned up. (This was a possibility a Facebook representative suggested to me when I asked about your case.) Facebook has partnered with data-mining organizations like Turn that have access to nearly 1 billion profiles, so it will have no trouble getting from a phone number and a first name to that person’s Facebook profile. Perhaps you never synced your contacts or shared your address book with Facebook. But maybe your ex did. Like a police dog, Facebook and other services can pick up on the smallest lingering scent of old contacts in order to reconnect you with past acquaintances—even if you’d rather it didn’t.
Dear Dr. Strangedata,
Facebook suggested that I join the group “the Zembla diaspora.” I am not from Zembla and have never known anyone from Zembla. But when it promoted the group to me, it showed me four members, one of whom looks a whole lot like my daughter. A whole lot. But my daughter doesn't have a Facebook account, and I have never tagged her in a photo (though I have posted photos of her). What is going on? I. Am. Unhappy. About. This.
This is a fascinating case! We’re used to seeing Facebook suggesting ideas and groups based on text and structured data like our friends graph, our likes, and brand names we type. But that doesn’t seem to have happened here. Facebook told me it seemed like a coincidence; in other words, it wasn’t so sure, either! There certainly can be cases of false positives. But given how insistent Facebook is on getting information on people’s faces, I want to propose a hypothetical explanation. Facebook has developed increasingly powerful facial recognition algorithms. These algorithms still aren’t perfect; I get mixed up with Robin Williams fairly often, so there are a few algorithms out there that believe I’m dead. So just as Facebook suggests names of people to tag in your photos by grouping similar faces together, it might have located a group with a member who looked like your daughter, and passed some threshold at which Facebook decided to offer the group to you. At least, that’s my guess. When I asked Facebook about this scenario, it told me, “This isn’t something we do.” And there is always a selection effect whenever Facebook does something bizarre. But learning algorithms, which are what are involved here, don’t behave in predictable and controlled ways, as Pedro Domingos’ excellent The Master Algorithm explains, and so I don’t take Facebook’s response as ruling out something like this scenario.
Don’t feel singled out, though. It’s all part of the Facebook magic.
Dear Dr. Strangedata,
Facebook keeps suggesting that my landlord and me should be friends, and I don’t know why. I’m not friends with him, and we don’t have any friends in common. Has he been creeping on me on Facebook? He doesn’t seem like the sort, to be honest.
While it’s possible that your landlord has looked you up on Facebook, Facebook could well have drawn the connection on its own. As we saw in Ex Offender’s case, if Facebook found your landlord’s name and/or phone number on your phone (or vice versa), that might have prodded Facebook to make that connection. Or perhaps there’s a more indirect explanation. The two of you are probably connecting to Facebook from the same IP subnet, or else using Facebook on your cellphones from near-overlapping geocodes. Since Facebook knows you live in extremely close proximity, it probably thinks there’s a good chance you’re friends.
Dear Dr. Strangedata,
Facebook recently asked me whether I had a Chinese name, and did I want to add it to my profile. I am not Chinese, and no one would guess from my name that I’m ethnically Asian. I believe it had to have used facial recognition software to determine whom to ask, because I am as culturally white as can be: I was adopted young and don’t even have many Asian, let alone specifically Chinese, friends or co-workers, and the only thing about Asian stuff at all in my data is that I was born in South Korea. I know I don’t need to tell you Korea is not China and has its own language and names. Is it because Facebook thinks I look Chinese? Is Facebook doing racial profiling now?
—None of Your Business if I’m Chinese
As we saw above, face recognition algorithms are pretty powerful tools, and there’s no doubt that algorithmic photo analysis can cluster by skin tone. (This is one of the key ways in which porn images are identified, in fact.) I don’t think Facebook actually classifies people explicitly by race—this would be a bridge too far—but it might have decided that your face “looks” like a lot of other people who happen to have Chinese names. So I suspect it is not racial profiling in the strictest sense of the term—but neither is it wholly independent of race. Facebook ends up grouping people together by ethnic group whenever those ethnic groups coincide with other data common to them—location, language, interest, social graph, etc. In other words, if people are being grouped ethnically, it’s probably an emergent property of Facebook’s clustering algorithms rather than an explicit goal. Facebook may well have clustered your face with a lot of other faces of people who do have Chinese names in their profiles. Or it might have grouped your account with a bunch of other accounts from South Korea who happen to have Chinese (not Korean) names in their profiles. Or it could have been any combination of these and other factors pushing the algorithm over some threshold of likelihood. Unfortunately, Facebook is private about its algorithmic recipes, so there will be cases like yours where there isn’t a single obvious explanation. By not adding a Chinese name, however, you provide feedback to Facebook that will tune its algorithm in making future calls like this one. And so the circle of data profiling life continues inexorably.
That’s it for this time. If you have a question about a creepy social media experience, email me at email@example.com. Until next time, remember: When you use a free Internet service, you aren’t the customer, you are the product!
This article is part of Future Tense, a collaboration among Arizona State University, New America, and Slate. Future Tense explores the ways emerging technologies affect society, policy, and culture. To read more, visit the Future Tense blog and the Future Tense home page. You can also follow us on Twitter.