Singles hoping smart computer programs will identify their ideal romantic partner may be disappointed to learn they still must meet folks in person to see who clicks.
That's according to a new study in the journal Psychological Science that says computer-based algorithms can sometimes tell who has a general tendency to like others and who's apt in general to be liked, but when it comes to predicting the attraction between two specific people, technology seems to bomb.
"Compatibility elements of human mating are challenging to predict before two people meet," according to the study, titled, "Is Romantic Desire Predictable? Machine Learning Applied to Initial Romantic Attraction."
"We couldn't predict attraction for a specific person at all," said Samantha Joel, the study's lead author, who is an assistant professor of psychology at the University of Utah. "Zero percent. We could not find any predictive value; the models are useless."
That doesn't mean you have to cancel your membership with a dating service that claims to help people find a mate, though, she said. Dating services and apps provide lots of potential people that someone seeking a relationship can meet to test mutual attraction and compatibility.
"That's a really good service," she said.
For the study, the researchers used a computer, a questionnaire and two groups of speed daters, each made up of romantically unattached undergraduate students. Before a speed-dating event, the young adults filled out questionnaires about more than 100 traits and preferences that have been identified as "relevant" to choosing a mate. Then study participants met multiple people of the opposite sex in four-minute speed dates. After speed dating, they rated their interactions and attraction to each person they "dated."
Joel and her co-authors, Paul W. Eastwick of University of California Davis and Eli J. Finkel of Northwestern University, applied a machine learning algorithm to see if it could predict couple-specific interest in and sexual attraction between each pair before they met in person.
The technology randomly sampled different combinations or traits and preferences thousands of times to try to predict attraction as accurately as possible. The computer decided what worked best and created a model based on that, a cutting-edge strategy that's well-respected in research, Joel said.
Relevant traits and preferences cover a spectrum that have been vetted repeatedly in previous research, including neuroticism, attachment style, approach or avoidance goals, vitality and attractiveness and trustworthiness, among others. The researchers found their machine learning program could predict with about 18 percent accuracy study participants' general tendency to like other people, including how picky they are and their level of warmth. It did better predicting someone's tendency to be liked — "who's more attractive," as Joel said — at 27 percent accuracy. That wasn't surprising, she noted, as "there's some agreement on who's attractive."
But technology was a complete dud when it came to predicting relationship-level attraction, as in who is particularly well-matched. Because they used more than 100 different predictors, Joel said they expected it to find some predictive power. It found zero.
There are potential confounders in the findings, according to Joel. One is the fact that study participants were all of similar ages, late teens and early- to mid-twenties. The other was the short duration, by definition, of speed dates. Maybe if the couples had spent longer together to let attraction unfold, the technology's prediction might match up at least a little.
Then again, maybe not, she added. When people spend time together, they develop more inside jokes and shared history, but those are not necessarily the traits and preferences that two people bring to the relationship. They are, rather, the relationship itself.
"So I don't know if more time would make it easier to predict attraction or even harder," she said.
Joel also noted that online dating services might say the research looked at predicting attraction, while they try to pick long-term compatibility.
"While that might be true, attraction is a prerequisite," she said. "Most people don't date people they don't find attractive — or at least not for long."
The study suggests that in-person meetings are probably required to see if you are interested in someone. "Attraction might be comprised of idiosyncratic experiential things that we can't predict beforehand," is how Joel puts it.
"It may be that we never figure it out, that it is a property we can never get at because it is simply not predictable," said study co-author Eastwick in a written statement. "Romantic desire may well be more like an earthquake, involving a dynamic and chaos-like process, than a chemical reaction involving the right combination of traits and preferences."
Nathan DeWall, a psychology professor and director of the Social Psychology Lab at University of Kentucky, wrote this week about the research for Quartz. His story is of particular interest because he met his future bride through an online dating service. As Joel noted, services can provide a pool in which to find people you might not otherwise encounter.
"Her words jibe with my online dating experience," DeWall wrote. "Although I eventually married the woman a computer identified as my top match, I also went on dates with other women the computer thought I would like — and I didn't. But by taking action to join online dating sites, my dating pool expanded, increasing my chances of meeting the right person. All I had to do was practice patience and perseverance."
That sounds about right to Joel.
"I think you physically have to go out and meet people," she said. "There's not, from what we can see, a shortcut — no way to bypass the often lengthy and frustrating process of meeting lots of potential partners who you don't click with before finding the person you do click with."
The study was paid for by a postdoctoral fellowship from the Social Sciences and Humanities Research Council of Canada awarded to Joel, who hails from Toronto.