Artificial Intelligence and the Potential of Next-Generation Dating Apps

April 8, 2016 • Blog Post • BY ATLANTIC RE:THINK


  • Can artificial intelligence, machine learning and algorithms be better matchmakers than humans?

Advanced data analytics in dating apps today improve the odds that real-world romantic relationships will last

In the recent movie “Ex Machina,” a young programmer falls in love with a robot. An artificial-intelligence mastermind had engineered his robot so that “she” appears to think, feel and love in a way that was uncannily human. He had also mined and analyzed the young man’s online history deeply enough to create the woman of his dreams.

Long before anybody actually meets the perfect robot mate, however, intelligent, “learning” machines capable of advanced data analytics will be ever more precise in matching one human to another—and in predicting the success of computer-assisted matchups, which will improve the odds that our real-world romantic relationships will last.

Leaps in digital technology have already made drastic changes to 21st century romance. According to the Pew Research Center, 15 percent of American adults now use online tools to meet one another. Among digital and smartphone natives, texting has far surpassed physical contact and phone calls as the chief method for staying connected with significant others. And among this age group, flirting now happens almost as often online as it does IRL, or “in real life.”

The rules of courtship are also being rewritten by technology. Online dating, once the province of older people in “thin” dating markets, has skyrocketed among the young with usage by 18- to 24-year-olds tripling in just the last three years. “We’ve seen a big jump in usage by college-aged young adults, which is driven in large part by mobile dating apps,” says Aaron Smith, an associate director of research at Pew. But younger kids are using them too. Seven percent of Tinder users are now under the age of 18, and apps that permit underage users, such as MyLOL, Badoo and Skout, continue to proliferate.

The latest mobile dating apps—the most popular being Tinder, Happn, Grindr and Bumble—hint at how technology could change courtship in the years to come. Instead of earlier models, which are dependent on elaborate user-submitted profiles, newer apps run predictive analytics on data readily available online such as a user’s social media profile, using algorithms similar to the ones that deliver user-specific recommendations on Netflix or Amazon.

The paradox is that two-thirds of online daters, some 66 percent, have yet to meet in person someone they found through a dating site, up from 43 percent in 2005. Using date apps to browse but not actually date could be a result of how little information is sometimes used to match people. Tinder, for example, launched using just age, location and a profile photo to pair off users.

To correct for this, developers and computer scientists are looking to advances in machine learning and artificial intelligence. R&D work has already been funded by major tech companies from Facebook and Amazon to Hewlett Packard Enterprise and Tesla’s Elon Musk, among others.

“It’s a technological hurdle,” explains Sean Mahoney, VP of Sparks & Honey, a branding consultancy behind the 2013 report, The Future of Relationships. “The actual elements of listening, analyzing text, facial recognition and even GPS or biological data are inevitably going to be put in robust algorithmic language, and that's going to be able to connect people in ways that weren’t possible before.”

By analyzing such online fingerprints as our Spotify playlists and recent purchases from Amazon, for example, next-generation dating apps will be able to determine our preferences and tastes and how they are evolving. “It's kind of like the question "What's your favorite kind of music?" says Mahoney, referring to a common profile point required by some dating apps. “Well, your favorite kind of music is what you're into in that moment… So, what we'll see will be more robust and nuanced algorithms that can get in tune with your emotional state.”


One of the challenges for digital matchmaking is that we are actually quite bad at knowing and being able to specify what we want in another person, but cognitive computing and Big Data analytics are teaming up to try and eliminate that ambiguity by extrapolating our moods and desires from speech patterns in our online videos and audio clips and the language and pictures in our Facebook posts. “These are the techniques that we use to better understand what people are looking for,” says Harm de Vries, a researcher at the Montreal Institute for Learning Algorithms who recently began applying this technology to dating apps. “It turns out that it’s just way better to let the computer learn this kind of stuff.”

De Vries runs experiments on Tinder that predict with 68.1 percent accuracy if he will be attracted to a certain profile, and he’s working on a similar model to analyze whether or not a user’s profile is going to attract a potential match. Similarly, technologist Justin Long launched a bot called Tinderbox to automate profile selections on Tinder using facial recognition software. In February, he launched an AI dating program to share his algorithm strategy with others.

To increase the power and spread of these AI and machine learning tools for developers, tech companies are increasingly offering them up on cloud-computing platforms such as HPE’s Haven OnDemand. “In seconds, you can create a freemium account and use your own data to try the 60+ APIs without ever writing any code,” says Sean Hughes, chief community officer for HPE Haven OnDemand. That functionality includes advanced analytics such as speech-to-text, face detection, sentiment analysis and prediction. A recent HPE-backed hackathon that gave developers access to Haven OnDemand APIs produced a live video speed-dating app called Hey Blink Me. It provides icebreaker questions, lets the two sides chat for exactly two minutes, then goes dark and asks for both to say whether or not they would like to continue the conversation. If both say yes, they’re put in touch with each other directly. This app uses the face detection API from HPE Haven OnDemand to help keep unwanted “junk” off the screen. Hey Blink Me launched on the Apple App Store in January 2016.

Maybe the most important way that smart technology could improve online matchmaking is by increasing its safety, security and reliability. According to Pew Research, 54 percent of online daters say that someone has seriously misrepresented themselves on their profile. According to a Sparks & Honey report titled Generation Z 2025, teens are moving away from social media sites like Facebook in favor of Snapchat and other sites that allow for anonymity. “We're already seeing a backlash against technology because of its heavy associations with surveillance,” says Mahoney.

How could technology be used to make us feel safer and less exposed online? The capacity already exists to improve profile validation. With predictive analytics, says Hughes, “You could analyze profiles and flag those that score high for possible fake data”

An advanced form of sentiment analysis can also analyze the tone of written or verbal content and even screen for users who are unstable or post overly negative content online across social media platforms. “You could automate that screening to find out what this person is really like based on what they have posted publicly online,” says Hughes, “and see whether they’re likelier to be a good person or a bad person.”

De Vries takes it a step further. “You could also do analysis across a generation," he says. "You can use the information from your parents to better predict what kind of relationship is going to work for you, or analyze information across many years. All of that data could be used to better match everybody in society to each other."

The ultimate ambition of this marriage of AI, machine-learning and data-mining, says Mahoney, is “to quantify everything we do to basically make our lives easier and more frictionless—and one of those things is definitely relationships with other people. This is an interesting world we could get into.”

Big Data changes everything. HPE Software gives you the power to transform it into actionable intelligence so you can capitalize on new opportunities and solve real problems in the moments that matter.


Hewlett Packard Enterprise Provides Update on Houston Area Operations

News Advisory

Hewlett Packard Enterprise Enriches HPE IDOL Machine Learning Engine with Natural Language Processing

News Advisory

PODCAST: The Average Patient, Prescribing for One, and Memory-Driven Computing

Blog Post