Professor Eli Finkel recently gave a great talk at TEDx called What Would a Successful Matching Algorithm Look Like?
Professor Finkel breaks down the data into two types: individualistic, and couple-level.
Sites like eHarmony use individualistic data–that is, data about each person, separately–to determine matches. Professor Finkel believes strongly that what we need instead is couple-level data, because his research has shown that individualistic data is not likely to be enough.
Couple-level data is about what happens when two people interact. Similar to the ideas in my previous blog post about what we can learn from analyzing speed daters’ language, it’s what happens when two people are together that gives us information about whether or not they will be a successful match.
Professor Finkel has an interesting proposal, which is to have online daters do virtual speed dates–4-minute video chat sessions. From these, a dating site/app could analyze language clues, such as whether people get in sync about how they structure their sentences. In a study of the chat data of undergads who had been in a relationship for 6+ months, 78% of the couples still together had high “language-style matching”.
Another idea includes examining heart rate when you’re interacting with someone–when two people have chemistry, their heartbeats get in sync.
If an online dater spends 4 hours in one week going on these virtual dates, they could interact with 60 people. Their speed dates could be analyzed, and for those people they say they’d be willing to go on a first date with, it could give them a predictive score on who is most likely to be compatible.
Professor Finkel was kind enough to answer some of my questions via email. I asked him how he felt about using text/chat data–would it be as useful as speech? He believes that the data will be useful, but perhaps not as rich:
My guess is that, as data roll in regarding this question over the next couple of decades, the story is going to be about a continuum that we might call “social richness.” At the high end are face-to-face communications, which involve many channels at one (linguistic, paralinguistic, sensory, etc.). At the low end are asynchronous text-based methods like email. In guessing that, to the degree that a communication method is high in social richness, people will be able to use that information to evaluate how compatible they are with their interaction partner.
I also asked him if he thought people would be up for such an intensive round of virtual speed dates. Would people balk at the idea of the time investment? His reply:
There’s no question that 4.5 hours is a substantial time commitment, but are people really wary of that level of investment in finding a compatible partner? What if it were proven that this method is especially effective? People spend years of their lives going on dates with people they meet online, going to bars, even going on singles cruises. My sense is that they’ll be willing to hang out in their dining room for 90 minutes each evening for three consecutive evenings.
I find his proposal intriguing. I have been searching for the answer to what can we do BEFORE people meet up, but perhaps that’s the wrong line of thinking. Using methods like these, we still save people from having to go on in person dates with every potential match. It may be harder for introverts, but if it does prove to be successful, the extra effort of “meeting” so many people might be worth it.
I can’t help but believe there is still room for the individual data, too. If I were online dating, I know I wouldn’t want to date a smoker, or someone extremely religious, or whose favorite movie is Meet the Fockers. And I keep going back to the pheromones.. if I don’t like the way someone smells, that seems like a big indicator as well.
In the meantime, I’d love to see a site like Match.com dive into couple-level data analysis. With video technology and chatting becoming so ubiquitous, it seems like a no-brainer.