Romance, in most people’s minds, involves bold declarations of love and hanging upside down dressed as spiderman. Usually in the rain. But the real world is rarely so cinematic. People are busy and the weather hardly ever does what you want it to. In their quest to find convenient love, many have turned to dating apps and websites.
The online dating industry has boomed since taking off in the late 90s, helping to match millions of couples across the world. The American National Academy of Sciences reported in 2013 that over a third of people who married in the US between 2005 and 2012 met their partner online, half of them on dating sites. According to eHarmony’s ‘Married Couples by the Numbers’ report, meanwhile, 71% of female users and 69% of male users meet their future spouse on the site within a year of creating a profile. This success is, they say, largely down to their ability to apply algorithms to complex datasets about users themselves, and more general information about what makes a successful relationship, thereby turning the complex art of matchmaking into a precise science.
Data has never exactly been considered the language of love, despite data scientist being named the ‘sexiest job in the world' by HBR. Many are particularly wary when it comes to applying data to human beings. People are bewildering animals and understanding one is a difficult enough task, let alone understanding two and pairing them up. However, in the US, somewhere between 40-50% of marriages end in divorce. eHarmony claim that the divorce rate for married couples who met through their site is just 3.86%. Couples are often just fundamentally incompatible, and dating websites like match.com and eHarmony’s algorithms aim to ensure that they only bring together couples that demonstrate the same characteristics of successful relationships, essentially getting rid of the trial and error method that precludes most relationships. Though it may sound cynical, it’s really in the best interest of dating sites to find their customers better, more long-lasting matches than their competitors, and money is a more rational motivator in relationships than love or chance. But does algorithmic dating actually work?
The history of using data for matchmaking goes back to 1965, when Harvard undergraduate Jeff Tarr and a friend developed a personality quiz for students about their ‘ideal date’. They asked questions such as: ‘Is extensive sexual activity [in] preparation for marriage, part of 'growing up?' and ‘Do you believe in a God who answers prayer?’ The response was overwhelming, leading Tarr to start ‘Operation Match’. Tarr transferred the answers to a punch-card and fed them into an IBM 1401 computer for processing, which then produced a list of six potential matches. If both matches fitted one another’s ‘ideal’, they were then posted back to the applicant along with the address, phone number, and date of graduation.
Since then, the use of data in dating has grown exponentially more complex. eHarmony was founded on the principle that you can identify what makes a relationship work by analyzing successful marriages and comparing them to those that aren’t in order to pinpoint the most important factors. With more than 40 million users registered since it was founded in 2000, eHarmony has a lot of training data. They originally asked users 500 questions about their personalities and relationship preferences - a number that the company has since got down to 145 and is still trying to decrease - giving them a wealth of information about their matches that worked and those that that didn’t.
To match people effectively, you need to solve several fundamental problems. Firstly, you need to match people for the long term by looking at compatibility. To match people for the long term, you need to ask deeper questions than normal on a first date. Rather than asking where they went to school, their favourite music, and so forth - socially acceptable questions that may satiate basic curiosity and suggest a similar background - they look at questions research shows indicate things important to a successful relationship, such as how someone handles stress and whether they are happy with themselves. Obviously, you could ask these on a date, but you’d be unlikely to get a second regardless of the answers matching up or not.
Before you find out if they are compatible though, they first need to want to talk to each other, and to do that there needs to be an attraction. There are a number of features that people tend to look for. Firstly, height. eHarmony has found that the probability of communication strongly correlates to heights - women tend to go for men taller than them while men for those shorter than them. Food preference is also important. eHarmony asks what people eat, and vegetarians in particular are more likely to talk to each other, with a communication rate 44% above average.
eHarmony is also increasingly looking at the data around images to match people. Jonathan Morra, Director, Data Scientist at eHarmony, told us: ‘We do ingest information from images when doing affinity matching. We attempt to extract information about users’ faces including hair color, eye color, and facial hair. Judging attractiveness based on images in general is very hard and very subjective. We have done it in the past and found limited success. Using extracted features, though, has proven successful. I think image analysis is currently making great strides with all the work on deep learning, and I think that definitely has a place at eHarmony.’
But does it actually work at all? According to a 2012 paper published by Northwestern University’s Eli Finkel and four co-authors in the journal ‘Psychological Science in the Public Interest,’ there is is no evidence to suggest that the concept of matching algorithms works at all. Indeed, their skepticism was such that they called on the Federal Trade Commission to regulate claims about their effectiveness. The paper argued that relationship success basically depends on three things: individual characteristics, such as intelligence and kindness; the quality of you interaction, i.e. whether you actually get along; and external circumstances like race, health, and financial status. The paper argued that matching algorithms tend to focus on the first one of these metrics alone, yet all major, large-scale studies of married couples have shown that sharing a similar personality accounts for just half a percent of how happy someone is in relationship. And then there’s the question of whether or not you lie about what you’re like, or what you find attractive in someone else. One of the major issues around using algorithms is that people seeking partners are liable to enhance their own characteristics through mistruth. People’s photos, for one, rarely show anyone on an average day, only at their best, and the lies often run far deeper.
The second criticism, at least could be resolved by the next major advancement for the use of algorithms in dating: smart devices and IoT. The IHS forecasts that the IoT market will grow from an installed base of 15.4 billion devices in 2015 to 30.7 billion devices in 2020 and 75.4 billion in 2025. The data from smart devices will give us far more insight into ourselves than ever before. Therefore, it will provide a clearer, more objective picture of who people really are in terms of their interests, behaviour and preferences. For example, smart showers will show give us a better idea of cleanliness, which supposedly correlates strongly with levels of conscientiousness and organisation. Smart TVs will let us know our program choices, while wearables can give a true indication of how much someone exercises. eHarmony's own research indicates clothing style is actually an extremely precise and detailed reflection of a person's personality.
According to research conducted by students at Imperial College, an estimated four million relationships will have been created in the UK by 2026 through matching via smart technology and this will grow to 12 million by 2036. They further estimate that by 2036 more than 12 million UK adults will be matched to a compatible partner using the data that smart tech will record. There are, however, obvious privacy issues around taking all of the data from IoT. If algorithmic dating really is just a marketing ploy as Finkel claims, this is data they don’t really need.