The modern world is fueled by matchmaking. Going out on the town? Uber pairs you with a driver you can choose based on ratings, proximity, and even car model. Craving a vacation? Simply filter getaways based on locale and price through Airbnb.
These matches are purely transactional, based on objective criteria such as price or venue. But what happens when emotion fuels the process? Could it increase efficiency and engagement?
Edward McFowland III, assistant professor at Harvard Business School, and his coauthors examine this matchmaking behavior through a high-stakes lens: online dating. Their research suggests that transparency on dating sites—as simple as allowing users to see the identity and photos of those who “swiped right” to show their interest—improves engagement for both men and women.
Since Match.com launched in 1995, online dating has ballooned into a $7 billion industry, by some estimates, with players such as Bumble, Tinder, and OKCupid vying to help people find love. While McFowland is not a dating expert, his work in machine learning and social sciences examines the efficacy of how people interact in online settings. McFowland notes that approximately one-third of marriages began with online dates.
“[Online dating platforms] are an extremely interesting application of the general question of influence and digital interactions,” he says, with lessons that can apply to other online engagement platforms that have nothing to do with love, although online dating has enormous influence on the American social fabric.
Knowing who likes you changes the game
McFowland and his fellow researchers partnered with a large North American dating platform, dubbed monCherie.com, for a 2021 working paper titled Strong-Signaling and Identity-Revelation in Online Dating: Evidence from a Randomized Field Experiment. McFowland coauthored the paper with Jui Ramaprasad of the University of Maryland, Ravi Bapna of the University of Minnesota, Probal Mojumdar of Indian Institute of Management Udaipur, and Akhmed Umyarov.
McFowland notes that online dating traditionally has been hindered by two trouble spots: Women don’t want to make the first move online. On the other hand, men come on stronger, sending an abundance of often fruitless messages. This leads to mixed signals, frustration, and user atrophy.
Consider: When McFowland met his wife (offline, the old-fashioned way), their connection deepened organically. She didn’t notice him at first but eventually discovered his sense of humor. Online interactions often don’t allow such give, take, or time.
One way to speed the process is by reducing information asymmetry: letting someone know the identity of their admirers right away, leveling the playing field and giving users a confidence boost. Picture walking around knowing exactly who likes you without needing to interpret signs like smiles, texts, and awkward banter.
The research team followed 100,000 newly registered heterosexual daters and gifted half of them with a premium “who likes you” feature. The control group couldn’t see their potential paramours, only their number of swipes. However, the test group would enjoy a reveal after one month, unblurring the photo of each user who “liked” their profile.
When women saw who liked them, they were more proactive, sending 21 percent more messages to potential dates compared with the control group. Women with the feature also increased their matches by 29 percent, while men increased their matches by 15 percent. The researchers defined a match as a series of four online exchanges. (They couldn’t tell if these online matches came to fruition in real life.)
Defining desirability
To sort matches, the researchers measured desirability, though not in the “supermodel” sense. Beauty is in the eye of the beholder: The team determined desirability solely based on “demand” or interest, as measured by the number of unique messages a user received during the beginning of the experiment, and divided the group into thirds. Physical attributes didn’t come into play, though there could be overlap between the number of messages and attractiveness.
“We basically let the market and natural interactions tell us what desirability is. If you’re more desirable, the theory is, you should be getting more attention via these messages,” McFowland says.
Extremely attractive users might receive 100 messages, whereas middling users might land 50, and lowly users might only snag a handful. As in high school, birds of a feather flocked together: Matches increased when both parties had comparable desirability.
Importantly, there were also heterogenous pairings. When a targeting male was highly desirable but a focal female was average, engagement increased. This provides “evidence of an encouragement effect,” McFowland says. In other words, if someone who a woman perceived as out of her league deemed her desirable, she would get a confidence boost and reach out to more people for dates. It’s the equivalent of George Clooney smiling at a woman across a crowded cocktail party.
“You might think: This person is higher than me on the desirability scale. Maybe I’m more desirable than I thought I was. Confidence leads to better outcomes,” he says.
Some love matches fizzled. When a targeting man was very desirable but a focal woman had low desirability, matches didn’t happen—maybe because the woman perceived these men as too far out of her league, he says. The reverse was also true: He observed that desirable women initiated fewer matches when voted on by a less desirable man, perhaps (understandably) deflated by the whole process.
Why user transparency matters, for better or worse
McFowland says this research is particularly useful for image-driven platforms such as online dating.
“Understanding this market is valuable,” he says. If dating platforms could increase user engagement—and therefore successful matches—by offering a “who likes you” feature to all users, they could drum up business. Sites tend to monetize this as a premium feature.
However, openness has a downside, he cautions. People can make snap judgments based on photos or other subjective bits of information. McFowland points to sites such as eBay as a prime example.
“If a buyer knows [the seller], their gender, their age, or other demographics, they may start making inferences about the product beyond just the product characteristics themselves—and it can lead to some negative consequences sometimes. [With dating], it could help some by reducing information asymmetry, it could degrade the experience of others, and we had no real idea what would prevail. So, this is where the scientific mystery lies,” he says.
Applying dating algorithms to other industries, cautiously
Platonic platforms could follow similar, industry-appropriate revelation models. For instance, LinkedIn could offer all job-seekers the ability to see how many companies are looking at their profile at any given time and offer organizations the same information about users to fuel efficient job-pairings. But it’s a slippery slope: Issues of bias come into play, especially if the feature isn’t distributed equitably to all users or creates consequences disproportionately felt by one group of users.
McFowland urges developers to be mindful when rolling out features with interpersonal ramifications.
“Platforms have tremendous power when to change behavior and to help lead to better outcomes for users, and so it’s really important to first understand these impacts and values. You set the governance, you set the standards, and you want to be very thoughtful about which ones you set,” he says. “The power you have could actually have enormous impacts on people’s lives.”
[iStockphoto/Martin Dimitrov]