The true love and false promises of matching algorithms in online dating
Lena Orlova // Columnist
Jennifer Panata // Illustrator
Match, eHarmony, Plenty of Fish, OKCupid, Badoo, Grindr, Hinge, Tinder, Coffee meets Bagel, Her, Down, Happn, Bumble, Feeld, Clover, Pure, The League, Tastebuds, Hater, Sapio, Wingman, Zoe, Dig, with sprinkles on top. Take your pick a-la-carte off the menu of online dating platforms. There is something for everyone. The serious. The not-so-serious. Straight, gay, bi, tall, short, likes Friends, hates football. Dig matches you with fellow dog lovers. The League only invites you to use the app if you make the A-list of upscale people in your area. Hater connects haters, so they can hate the same things.
Hey, misery gets lonely too.
One thing is clear: people want to connect. These services engage our most basic, biological need for human relationships. The fulfilment of this need is “one of the strongest predictors of happiness and emotional well-being” concluded Ed Diener and Martin Seligman in their 2002 study on the “very happiest people.” By this token, any invention that gets people together will likely be popular.
Whereas the desire to be closer to another person hasn’t gone away, the way people go about connecting has radically changed. In the past, our gossiping relatives – those nosey match-making grandmas – set us up on blind dates with people they intuitively felt would be a “good match” for us. Internet changed all that because it allowed access to a vast number of potential partners without intervening mutual contacts.
The first computer-aided dating service sprang out of UK in 1964 under the bold entrepreneurship of Joan Ball; a woman weathering homelessness, joblessness and mental illness who developed a talent for making connections. Her marriage bureau used a system of punch cards to record likes and dislikes about a potential mate. The client then received a list of the good matches with their names and addresses. It saved time, energy and became a popular avenue for singles seeking marriage material.
In many ways, Ball’s service was the precursor to modern matching algorithms, which work on the premise that sharing common interests is a good predictor of hitting it off with a person. That’s why clubs, associations, and organizations exist. Common values draw people together. It’s a lot harder to bond with a person who disagrees with you. You don’t see a lot of vegans at Burger King chatting casually about food culture.
eHarmony asks users to complete a questionnaire of more than 100 questions as well as “29 Dimensions of Compatibility”. This section assesses areas like temperament, physicality, and style of interaction. Users also get to choose which questions hold the most value to them.
However, there are drawbacks to this method of matching.
“The matching algorithms that are used are based on personality adjectives, and on similarity. The algorithms are typically not published because they are proprietary, so we don’t really know how good they are.” pointed out Dr. Sally Walters, Capilano University Instructor of Psychology. “We don’t know whether or not they are capturing essential elements of compatibility and incompatibility, and these differ from person to person. The success rate is not published in peer-reviewed journals.”
Indeed, these algorithms are difficult to deconstruct, and every company has its own. Plenty of Fish (POF) refrains from sharing details about their custom software.
“The true purpose of great matching is not to find romantic suggestions that will definitely work, but rather to avoid those that have a large chance of not working” said POF Head of Research, Michael Dermott.
Kate MacLean, POF Public Relations Generalist, explained that it’s difficult to quantitatively assess the success rates of online dating. In general, users do not indicate whether they’ve found a relationship when they remove their accounts. While there are many testimonies of people having found their perfect match, there is much more to learn about user experience and what they want out of these dating services.
One surprising finding is that the modern dating culture is less promiscuous than one would think. “The vast majority of first year university student either report zero or one sexual partners in the previous year,” said Dr. Jocelyn Wentland, UBC Adjunct Professor “Most people are not out having indiscriminate sex with whomever they can find so that doesn’t support this idea of a rampant hook-up culture.”
Algorithms would invariably affect how people approach dating, because of the vast number of singles that rely upon them today. For example, Plenty of Fish – originally a Vancouver start-up – is now used by over “150 million registered members worldwide and available in 11 languages and more than 20 countries” shared Kate MacLean.
To begin with, the process of meeting a person online is very different than face-to-face. Anonymous viewers can see a user’s profile any time of day, anywhere, akin to online retail shopping.
“I can pick and choose; I can choose what size I want, it’s like buying a car, what options am I looking for. I can test drive it, eh it’s not really my fit, I’ll put it back and go try another car… You might say I only want to look for redheads today, so I’ll save the search where all my other criteria are the same, education, professional, but I only want to look for people who have red hair,” said a regular online dater quoted in Scripts for Online Dating by Bridget L. Long.
A profile has any sort of information: photos, bios, music preferences, zodiac signs, and emojis. It’s a two-dimensional snapshot of a person, and this snapshot isn’t always accurate. People strategize about the ways they can present themselves most attractively. A study of close to 22,000 online dating profiles by Hisch, Hortacsu, and Ariely in 2010 showed that women online were slightly thinner than the real-life average, while men online are slightly taller.
We want our best side to show and if not the best, then the embellished side.
In a 2014 article for the Guardian, James Bridle writes, “the development of online dating tells us more about our relationship with networked technology than with each other.” He refers to the way that altering our online identity can have a massive impact on the people we contact. He isn’t far from the truth.
Apps like Tinder track user behaviour as well as preferences. “Internal desirability” rating registers the number of right swipes a person gets. This means that the person in your queue shows up because your interests match and because they’ve been yes-ed by other users too.
This is the dark side of some algorithms: rewarding the most popular while marginalizing others. Online dating services are most useful to those with barriers to finding a mate offline. These obstacles vary from low accessibility – living in a far-away place – to just being shy about self-promotion.
“Some people have ‘low matchability’ which is unfortunate because these might be the people who have the hardest time finding partners in the real world,” reflects Dr. Walters.
The filtering software can also propagate bias.
“When it comes to real humans on real dating apps, the algorithmic bias is well-documented. OKCupid has found that, consistently, black women receive the fewest messages of any demographic on the platform. And a study from Cornell found that dating apps … [by their filtering options] reinforce racial inequalities in the real world,” writes Arielle Pardes in Wired. Biases influence not only our dating preferences, but whom we choose to interact with in day-to-day life.
As users become entangled in a world of their own preferences, they are less likely to be exposed to new information. We may become unknowingly entrapped in a narrow view of relationship culture as dictated by what’s popular on dating apps, and what could be wrong in reality.
The dating game is changing with evolving technology, but one thing stays the same: people want relationships. To be completely sure the online user is the person that fits your criteria, is best checked in person.