The headlines read like a synopsis for Spike Jonze’s next romantic drama: In the approximate future — in a strangely pastel L.A. — romantic loss will simply cease to exist! That’s because Match.com is rolling out facial recognition technology that can find you a clone of your ex. An exact lookalike. A twin. For $5,000, in short, you never have to move on.
It’s a wildly catchy narrative, mostly because it plays into prevalent fears and confusions about how facial recognition — and big data — work. Facebook can now, we’re told, recognize faces as well as the human brain. All-seeing, inscrutable algorithms on Google know us better than we know ourselves.
This Match partnership is additionally uncanny, because it implies that technology can somehow erode individual identity — make us interchangeable, even on the most intimate level. It’s a concern that sites like Match and OkCupid kind of fan already.
It’s also a concern that, in this case, is probably a bit overblown.
“I mean, I understand where people are getting that from,” said Talia Goldstein, the CEO of the company partnering with Match on this facial recognition tool. But in practice she says, the tool is actually based on a very intuitive, old-school, and not particularly sinister logic: “‘ Attractive’ means different things to different people.”
Goldstein’s high-end matchmaking agency, Three Day Rule, is actually a pretty conventional operation, as far as these things go. A former producer on E’s “True Hollywood Stories,” Goldstein quit her day job four years ago to pursue a budding hobby: setting up her lovelorn co-workers and friends. For a fee of $5,000, clients bought a lengthy personal interview with Goldstein and six months of curated matches with men or women her company sought out. (Under the Match partnership, the same fee structure will apply: Match users can, for free, opt-in to Three Day Rule’s database, where they’re eligible to be paired with Three Day Rule clients; or, for $5,000, they can become clients themselves — a package that entails six months of date-finding and screening by a professional matchmaker.)
Starting out, Goldstein always asked the usual questions: What personality traits are you looking for? Do you care about height? Education? Career? But she also asked clients to supply Facebook photos of their exes, just so she could see herself the types of people they were attracted to. Sometimes there was a range or a mixed bag; but generally, the stack of photos had some vague commonality — a hair color, for instance, or a particular face shape. Goldstein would include those notes as she looked for potential matches.
“If you look at photos of your exes all together — and you should play this game — you’ll notice some traits that look the same,” she said. Which is, of course, just a fancy way of saying that different people find different things good-looking.
As Goldstein’s company grew, that kind of manual photo-browsing became more arduous. She launched an online component, with all the user photos in a database. She licensed a facial recognition program that allowed Three Day Rules’ matchmakers to tag user photos as blonde or brunette, round-faced or oval-faced, deep-set eyes or prominent ones. Eventually the program could input those basic tags on its own.
Now, Goldstein says, the software essentially does three things: It identifies the hair color, face shape, eye shape and eyebrow structure of the person in the photo; it searches those tags against a database of user photos that have been similarly tagged; and it spits out a collection of approximate matches — which professional matchmakers can then check for other prerequisites, like education level or salary. At no point, Goldstein stresses, are they ever looking for — or finding! — “clones” of an ex-flame.
To test that, I sent Goldstein photos of three male friends and asked her to run them through her firm’s facial recognition program. She couldn’t run my photos against her actual user database, out of privacy concerns, but she agreed to run them against a database of World Cup player photos. As a demonstration of the software, the results are pretty uncanny.
But in the context of online dating, they don’t strike me as all that weird. Goldstein is quick to point out that her algorithms are human-mediated: Yes, a facial-recognition program cuts the dating pool from, say, 10,000 to 100, but a real live matchmaker then goes through to check things like personality and career and salary range. Compare that to the impenetrable cocktail of algorithms at play on other dating sites — where matches are made based on clicks, quizzes and cold probabilities — and it almost seems less ominous. In fact, while sites like Match and OkCupid and their ilk are pretty hush-hush about what goes into their formulas, it’s pretty safe to assume that they’re inferring your “type,” physical and otherwise, based on the user profiles you show interest in.
That’s not exactly the same thing, of course: There’s something about facial-recognition specifically that gives us the heebie-jeebies. But in actual matchmaking practice, Three Day Rule isn’t using that algorithm any differently than other popular dating sites use matching algorithms already.
Yeah, it sounds intuitively “creepy” to look for dates based on your exes. But maybe this algorithm only strikes us as creepier than others because it’s the only one we can actually see.