Another lonely night. Just you, your couch, a half melted pint of cookie dough ice cream, and the crushing absence of companionship.
So, like any good 21st-century single, you pluck your iPhone from between the sofa cushions and open up the matchmaking app Tinder. But all that swiping and judging and messaging is a pain and, frankly, exhausting. Instead, you press the home button until Siri (who will never abandon you) pops up.
“Hey Siri, find me a date,” you say. Siri jumps into action and within seconds she’s found tickets to a concert for a band that both you and a cutie a couple blocks away are into. And don’t worry, that nearby cutie will probably be your type: Tinder will take your personal romantic preferences into account when finding you a date.
This is the vision that Tinder CEO Sean Rad shared recently at the Start-Up Grind tech conference in Silicon Valley.
“It’s a little scary,” he acknowledged in an onstage interview. But scary or not, Rad said that we could see artificial intelligence serving as a matchmaker and date organizer within the next five years.
It’s a development that poses quite a few questions, obviously. While we’re used to AI lending its fleshy creators a helping hand in many fields, it hasn’t moved into an area as, well, human as romance. After all, this is kind of supposed to be our thing: Google the words “love makes us human” and you will be awash in a deluge of quotes from Tumblr teens and renowned philosophers alike. The question of whether Rad’s proclamation is actually feasible, though, has a definite answer: Oh yeah.
The technology required for creating a digital date assistant exists already.
“There’s been quite a lot of work done on matching algorithms,” said Ariel Procaccia, a computer science professor at Pittsburgh’s Carnegie Mellon University known for his research in the field of artificial intelligence. “I would imagine that roughly figuring out romantic preferences [would] not be that different from, say, figuring out your preference in movies.”
There are a lot of intangibles to consider in romance, of course, but there are also a whole lot of things that would fit in an online bio that can help predict compatibility. That, after all, is the whole premise behind services like OKCupid: Type out what kind of music you like, what your beliefs are, who your favorite Beatle is, and a program will go to work finding a match.
The trick to unleashing the algorithmic Tinder wingman of Rad’s dreams, though, would be getting users to reveal as much as they can to the app.
“Tinder is pretty well confined right now,” said Jay Schuren, vice president of field operations for Nutonian, a Boston tech company focused on data science and AI. “If it could loop into your Google calendar, your work schedule, as much personal info as you feel comfortable giving out, [Tinder] could easily make recommendations to set up a date.”
But if you’ve watched any science fiction movie about robots, you know that any discussion of AI comes with ethical issues to sort out. One that feels particularly pertinent given the divided climate in this country is the possibility that an AI matchmaker would discourage pairing singles of different political standings.
“If an automated system observes, perhaps correctly, that people with different political backgrounds tend not to get along, it might never try to match such people together,” explained Michael Littman, a computer science professor at Brown University, in an e-mail. “This kind of behavior is apt to reinforce existing biases instead of resolving them or working around them.”
Another potential problem area could be discrimination. There have already been instances of AI exhibiting downright bigoted tendencies. Remember when Microsoft’s disastrous Twitter bot turned into a racist and misogynist?
Such problems arise because AI systems need “training,” and that often comes from white male engineers (or in the Twitter bot’s case, the general public). Although unintentional, bias seeps through when the AI carries out what it’s learned.
“One could worry that a certain phenomenon would happen in dating where the options that you’re shown would be based on attributes that shouldn’t necessarily be taken into account,” said Procaccia. “This is really a problem for machine learning and AI in general.”
Given those concerns, plus the weirdness of handing an algorithm the responsibility of ending your loneliness, it’s easy to imagine lots of people being uncomfortable with the type of product Rad is talking about. But his target audience with Tinder is millennials, so that demographic’s interest is what really matters to the CEO.
And for a few young Tinder users at least, an AI component would be, if not entirely welcome, then at least something worth checking out. “It’s absolutely something that I would try,” said Skylar Davis, a 20-year-old at Boston University.
“It feels a little creepy, but I would be curious to see who it matches me with,” added 22-year-old Morgan Krush, a health sciences major at Northeastern University.
Their concerns (again, aside from the whole creepiness thing) are more centered on branding.
“I don’t know if that’s what I think the app represents. It’s for casual hookups, just based on my experience,” Davis said, of Tinder. “Most people don’t go there looking for dates or romance.”
Krush agreed. “People use it now just for hookups or even just when they’re bored. If it changed to more of a serious dating thing, I don’t know if it would boost the usage overall.”
Rad’s vision might be appalling to some and revolutionary to others. It all comes back to how much of a role you want AI playing in your life. Littman, who spends much of his life dealing with AI and machine learning, believes there is some promise in the idea of a date bot. But a dark side does loom.
“As a computer scientist, I’m taught to develop metrics for assessing how good proposed solutions are,” he said. “Then, I develop algorithms that search through possible solutions looking for ones with good scores. But, life satisfaction and human connection don’t really lend themselves to being captured by one-dimensional utility functions. The more we think about them that way, the more is lost.”