Cover Story: Finding Love through algorithms

This article first appeared in Digital Edge, The Edge Malaysia Weekly, on February 14, 2022 - February 20, 2022.
Cover Story: Finding Love through algorithms
-A +A

Can the process of finding a romantic partner be taken over by an algorithm? That’s the question many have been asking since online dating platforms burst onto the scene. Opinions on this vary widely, as do the algorithms used by online dating platforms.

On the one end are the more traditional players who draw insights from an expansive questionnaire that users have to complete; on the other end are dating apps that rely heavily on visuals, a catchy self-description and user behaviour to determine what one likes or does not like.

Many of these online dating platforms rely on proprietary algorithms that match users based on different factors. Some have ranked users’ “desirability” based on how many people “swipe right” (choose) them. People of the same rank are then matched with each other. The platforms also use machine learning and historical data to observe and predict the users’ preferences and then serve them with matches that it thinks they will like.

Others focus more on the compatibility of matches, based on a survey or personality test that users have to complete beforehand. In this case, the algorithm matches people based on their test results.

MatchMde pairs users based on their personality and love language test results. Users can also opt for a genetic compatibility test.

The differences in methodology underscore an important question about relationships that these platforms are wrestling with: What are the factors that make — or break — a relationship? And can they be determined by an algorithm?

The answer to the second question is mostly a no, say the experts that Digital Edge spoke to. Algorithms are, at most, helpful in determining the compatibility of individuals based on their preferences and personalities. Algorithms shorten the time needed for individuals to know whether the other person is a good match.

But otherwise, relationships are messy and influenced by many factors, including physical attraction, readiness to commit and body language, which cannot be captured by algorithms. 

“Relationships are all about hard work. Compatibility is one thing, but when the relationship evolves, so does your relationship. It all depends on the mindset of the person and also other factors, such as whether the person is financially stable or whether they are affected by some life-changing experiences,” says Fabian Foo, founder of Singapore-based dating service MatchMde.

“Even if we give you a match, there’s a high chance it won’t work because you said a wrong thing or you don’t speak the same lingo.”

Compatibility, whether it’s based on personality or preferences, is not enough. Chemistry is key in the initial stages of dating, says Dr Rashied Amini, creator of Nanaya, a personality test that aims to scientifically predict the future of one’s love life.

Another important factor is that people often do not know what they want or need. How would an algorithm learn to give them the right matches if that is the case?

“The process of finding a partner is very emotionally driven. I’m sure most people who have dated with a list of preferences didn’t always live by them,” says Rashied, who is a system engineer at NASA. He created Nanaya’s algorithms to understand the future of his own relationship several years ago.

A major reason why relationships end, he observes, is because partners were dishonest with themselves about what they were looking for.

“They thought they wanted something but they never really did, and after some time, the feelings of frustration could not be suppressed. If people cannot be honest with themselves, can they be honest with others? Can they be honest with a computer? How would a computer know? Maybe technology can address it once people learn to truly know themselves.”

This is why Nanaya focuses on helping people learn more about themselves through the test results, so they can date — online or in-person — with more self-awareness.

Nanaya is a personality test that aims to scientifically predict the future of a person’s love life

Rashied also points out that many popular online dating platforms measure attraction rather than compatibility, especially those that rely heavily on photos.

“Because they have so much data, they can calculate the odds of how often people swipe, which lets them design a swiping experience to keep you hooked to the app. Just like the algorithms of social media newsfeeds, they really practise psychological manipulation!”

Of course, some online dating platforms have introduced limits to the number of matches one is presented in a day to circumvent that issue.

Regardless, it seems that algorithms cannot create perfect matches owing to inherent human flaws. People do not have enough self-awareness, they are complex and also fickle, observes M Nazri Muhd, CEO of MyFinB, a tech company providing artificial intelligence (AI) services.

“To predict compatibility, it has to involve more than just assessing likes or dislikes. This is because people are fickle most of the time, so AI algorithms have to be very dynamic and sophisticated. AI has to go into the social-emotional learning aspects. A set of behavioural, attitudinal and personality analyses is needed, which is then linked to the user’s [behaviour when conducting] online activities. This has to be measured over time and needs to be validated,” says Nazri.

How does a person feel about a situation and how does it differ from previous times? What are the reasons a person reacts this way? What is the likely behaviour of a person if a new event occurs? What are the factors that could change how the person feels? These are questions that the AI will have to grapple with.

“Practitioners should not aim for a perfect match using AI but use it simply to understand humans better, be it for personal or work matters. We cannot depend on AI to tell us whether to love someone or not,” says Nazri.

Lunch Actually relies on consultants to match potential couples

Can algorithms work for good in dating?

Foo does not believe that there are good or bad algorithms. They are, after all, built based on the logic of the programmer. Instead, people should choose the dating platforms using algorithms that they agree with.

For instance, if you believe in the Myers-Briggs Type Indicator (MBTI), which is a personality test, and love languages, you can use MatchMde. But if you believe in star signs, MatchMde might not be the right platform, he says.

It’s interesting to consider his perspective, given that Foo suffered from “dating-app fatigue” before meeting his wife. “It was always very superficial. These apps don’t go deep enough to know a person and give good recommendations,” he says.

He eventually met his wife through a friend’s recommendation. Not everyone is lucky enough to have such a good connection, he notes. This inspired Foo to set up MatchMde to help others, albeit through algorithms and an online dating platform.

So, why does he still believe in technology? Foo acknowledges that relationships are a combination of art and science, and MatchMde is mainly providing the latter for users. It offers as much relevant information as possible for users to make an informed decision without wasting too much time.

“Can you imagine asking someone on their first date whether they want to have children? We try to ask the difficult questions first and give you a report card so you don’t have to worry about such questions. There’s a real example where the couple had different opinions about having children and only found out about it after dating for two years. The relationship went downhill from there,” says Foo.

Meeting people offline is tough and takes a long time, he notes. “It’s all by chance. Even if you do meet someone, it’s always superficial at the first meeting. You have no inkling of the other person’s traits and it’s always about the physical appearance initially. Once you get to know a person, you have to go through a long stage of dating as you find out about their traits, nuances and bad habits.”

He tries to make MatchMde more personal. Instead of matching people based on physical appearances and short descriptions, MatchMde requires users to complete an MBTI and love language test. The algorithm then matches people who have complementing MBTI personality types and similar love languages.

The algorithm incorporates advice that Foo obtained from real-life matchmakers,. That is the reason MatchMde does not focus on matching people based only on similar interests. “Having the same interests doesn’t necessarily mean you are compatible,” Foo points out.

The algorithm is also dynamic. It will pick up on post-date feedback by users as well as the user’s behaviour on the platform. “The more you date, the more accurately it will recommend matches for you,” says Foo.

Users can opt for a genetic matching service with consent from both sides to determine the degree of natural attraction and other traits, he adds.

Of course, other things come into play, such as chemistry, which cannot be measured. “We also can’t measure things like laziness or financial responsibility,” says Foo.

The best MatchMde can do is screen matches to verify their identities, eliminate those in the community who misbehave and provide as much information as it can. An AI-powered bot also assists users in their conversations and prompts them to meet their matches.

“Typing messages is only two dimensional and you might misconstrue what they are saying. But by having conversations, you can see their expression and when you meet someone, you can observe their body language,” Foo says.

He believes online dating will eventually go into the metaverse and utilise blockchain for identity verification purposes, as well as virtual reality. That is something he wants to work on for Matchmde as well.

 

What an algorithm can’t do

Despite the proliferation of online dating platforms, there are still matchmaking services that do not rely on algorithms and swiping on apps. Violet Lim, co-founder and CEO of Lunch Actually, runs a service where dating consultants make matches based on a detailed questionnaire and interview.

“We ask them what they are looking for, which could be physical or personality preferences. We also ask them about their interests, hobbies, background, previous relationships and their values,” says Lim.

The consultants will use software to match these answers and generate a list of potential matches for an individual. But ultimately, each match is still handpicked by the consultant.

This is not because Lim does not believe in technology. Algorithms in online dating platforms are certainly helpful to find compatibility between matches, she says.

“I think it has worked for some people. But the challenge is how many people are willing to go through that. After a while, some people get burnout.”

Having dating consultants match couples has its perks. They can interview the clients and read between the lines. In comparison, online dating apps rely on the user’s description of themselves, which could be inaccurate if the users have a lack of self-awareness or choose to portray themselves in a certain way.

“I always tell my consultants that they are not waiters. Instead, they are nutritionists. If you were a waiter, you would just write the client’s orders and deliver the service. But if your client looks like he’s going to die of a heart attack and he orders a steak, a nutritionist wouldn’t just take that order. He would tell the person, ‘I understand you like things like this, but won’t you consider something else that might be more suitable for you?’ That is our role as a matchmaker,” says Lim.

She gives an example of a client she once profiled. She asked him what the things he could not live without were, and he said “sunrise”. After probing more, she understood that he liked simplicity and the little things in life made him happy.

“He really wanted a girl who was simple as well. He didn’t like girls who wore a lot of make-up. Many guys say this and I question if they mean it. Is it just that this means the woman’s make-up skills are so good that they [the guys] can’t tell [how the person really looks like]?” says Lim.

“Because of our interactions, I knew that he really meant what he said. A nice girl came to my mind and I thought they would be a good match. They are married and have kids now.”

A dating coach

The dating consultants also listen to feedback from each client after a date. If certain things are worth pointing out, they do so. “It helps them see their blind spots. From there, we give suggestions or coach them,” says Lim.

Such tweaks, sometimes, are crucial for success. “Most single people say they are single because they haven’t met the right one. Once they have, everything will be perfect. But having been in this industry for a long time, I realise that the people who succeed are those with the right mindset and skill set,” Lim says.

“When we put someone suitable in front of these people, they can connect and are attracted. If we put someone like that in front of a person without a good mindset or skill set, it doesn’t go anywhere.”

It could also be useful to eliminate the bad habits that people pick up from online dating. A common criticism of the online dating model is that it has created an illusion of choice, whereby people cannot commit to a person because they think there are many other better options out there.

“I call it elevator dating, where dates have become a commodity. We have clients like that. We give them a match that they are happy about, and then they ask us to find them another match who is just like that, but they add another criterion,” Lim says. And after that, the client asks for yet another person, with another added criterion.

“It’s a reality. We’re trying to slow things down. When people come to us, we tell them we are not a dating app. We are not going to give them 100 matches or dates. We are screening matches for them, so when they meet this person, they have to realise that we have made a lot of effort to pick this person for them,” she says.

Ironically, online dating platforms could narrow one’s choices owing to preconceived notions. For instance, a user might not like how a potential match looks in the photo and does not give them a chance. If they had met in real life, things could have been different.

“If my husband and I were to have met on a dating app, we would probably never have met up. That’s because he doesn’t even remember the first time we met, since I was completely not what he would usually look for. If we had been using a dating app, he would have just swiped left,” Lim says.

That is why Lunch Actually does not show potential matches photos of each other initially. “A photo is just a photo. But when you hear a person’s voice, it creates a different feeling. That’s where we are coming from. You also need to trust what we are telling you (as to why you both would make a good match) rather than just basing your assessment on photos,” Lim explains.

The company does, however, use technology to complement the work of its consultants. For instance, it is developing an artificial intelligence solution to help consultants answer questions and apps that allow clients and consultants to liaise with each other.

 

How do popular online dating platforms work?

OkCupid: Users have to answer an extensive list of questions about themselves. The algorithm looks at the users’ preferences and answers to create compatible matches.

Tinder: Considers users’ current location, age, gender preference. Prioritises potential matches who are active. Users swipe right (yes) or left (no) on profiles of potential matches.

Hinge: Users have to answer three prompts to activate their profile. They can like specific pictures or prompts of potential matches. Only eight free likes are given a day. Its “Most Compatible” algorithm pairs users with members whom they are most likely to like, and who are also most likely to like them as well.

Coffee Meets Bagel: Users identify their preferences, which are considered deal breakers. They also have to add information about their interests, social circles and other factors. The algorithm will not recommend someone who falls outside of the deal breakers. But if there are too many specific criteria, there might be fewer quality matches owing to a smaller pool of options for the algorithm to work with. The app gives users only one match a day.

Bumble: Allows users to match with potential dates, friends and mentors. Women have to initiate a conversation first within 24 hours of a match. If they do not, the connection expires. Not much is known about how Bumble’s algorithm works.