Matching Algorithms in Dating Apps

Modern dating apps have long been a familiar part of our daily lives. Ten years ago, Tinder, Mamba, Pure, and others turned traditional ideas about dating and relationships upside down.

Whether we like it or not, dating apps have started to directly affect our relationships. Dating for any purpose has become much easier and faster. But will those relationships be successful? To understand whether you should confide your destiny to apps, you need to understand how their matching algorithms work and whether they are actually able to find the right partner for you. My name is Konstantin Berezin. I am a backend developer with ten years of experience, and in this article, I will tell you how everything works.

From Newspaper Ads to Tinder

First, a bit of history. Thirty years ago, the idea of using algorithms to find a life partner sounded bizarre. But with the spread of affordable internet and the advancements in technology, finding a partner, lover, or just a friend through dating apps has become a common thing for many people. Today, 32% of people in Russia use dating sites and apps.

The revolution happened in 2012, and it was connected with the emergence of Tinder. Before that, partners were sought, for example, through newspaper ads and SMS chats on television, where you could send a playful message. I think many people still remember "Krovatka," the legendary chat room of the 2000s, where there were sections with different topics such as "Dating," "Flirting," "Over 30," and even "Sex Chat."

In 2000, the eHarmony site was launched with a focus on building serious relationships. OkCupid followed shortly after with their unique Q&A system for matching compatible couples.

The industry peaked in the 2010s when online dating was no longer perceived as something shameful. Previously, having a profile on a dating app meant accepting a loser status, but now this stigma has disappeared. Thanks to this, many new platforms have emerged. Badoo, LovePlanet, Bumble, OkCupid, and Hinge developed in parallel with Tinder and most of them were powered by David Gale and Lloyd Shepley's algorithm.

This algorithm was described in 1962 by mathematicians David Gale of Brown University and Lloyd Shapley of Princeton University. In their article "College Admissions and the Stability of Marriages" in American Mathematical Monthly, they proved that for an equal number of men and women, it is always possible to find a partner for a stable marriage.

Of course, we can't ignore the impact the COVID-19 pandemic had on the rising popularity of dating apps. In March 2020, Tinder recorded an all-time high of 3 billion swipes per day. OkCupid user activity increased by as much as 700%. Video calls via Bumble have also increased, with 70% more video calls initiated in the system. It is obvious that despite the quarantine, people not only did not stop getting to know each other and communicating online but actually began to do it much more actively.

Today, dating apps remain the most common way of dating, with divorce rates about ten times lower than for people who met otherwise. 

This sounds very optimistic, so let's look into why dating apps work so well.

Behind the Scenes: How Matching Algorithms Work

Most dating apps operate on this principle: a person gives a description of themselves and, depending on how detailed and frankly they approach it, gets recommendations from potential partners. What that means is that the more reliable information in your profile is, the more chances you have to find the "perfect" match.

It's worth mentioning that Tinder, for example, previously also used the "Elo rating," which splits people into groups according to their level of attractiveness. French journalist Judith Duportey described this algorithm in detail in her book "Algorithmic Love. How Tinder dictates who we sleep with." As the creators of the app explain, it is a very complex system that calculates the degree of demand for a profile.

So, what machine learning algorithms are behind the process of finding the perfect partner? These are primarily classification, clustering, prediction, pattern recognition, information retrieval, and text analysis.

All of these tasks come down to collaborative filtering, or, in other words, predicting a user's preferences based on their history of previous events, searches, swipes, and interests of similar people.

It's worth noting that Tinder, for example, gathers information about users and their preferences from other services as well: it analyzes their Facebook and Instagram profiles, what music they listen to, and even what kind of purchases they make in online stores.

This is very result-oriented, as analyzing user behavior helps to more accurately determine customer preferences and lifestyles. In addition, such analysis reduces the possibility of distorting reality. Such distortions can be made intentionally or unintentionally by the user in order to present themselves in the most favorable light.

Neural networks based on Deep Learning algorithms are also actively used in dating apps. For example, some applications allow you to search for partners with a certain appearance. So you can set a photo of a celebrity that you like, and the algorithm will identify their features and characteristics, compare them with the data in its database, and, as a result, show users who resemble your preferences as much as possible.

It is widely known that in Russia, girls who look like Irina Shayk and Natalia Vodianova, as well as men who look like Sergei Lazarev and Roman Abramovich, are especially popular.

Ethical Issues

While dating apps have made starting a relationship and dating itself easier, they have also led to the emergence of ethical issues. As the process has become largely automated and emotionless, some psychologists note that this has created a culture of relationship consumption.

The dating app, in this context, is perceived as a store, and the search for a partner is nothing but a choice of goods that can be rented or exchanged. Thus, there is a risk of partner objectification when a person is instead perceived as an object with a necessary set of functions.

On the other hand, dating apps allow you to separate the social and sexual parts of your life so that they may never overlap. This important advantage of using dating apps is that it helps you find partners for short-term relationships, which are usually condemned by society. At the same time, dating apps provide a sufficient level of anonymity. A person who has found a one-night stand does not have to worry about this part of their life becoming known to their coworkers, for example.

Security and Privacy

Nevertheless, dating apps still do not create a 100 percent safe space. Sometimes, they become a place where users can face harassment, grooming, abusive behavior, and personal data leakage.

Just remember the Netflix documentary "The Tinder Swindler." A certain Simon Leviev created an image of a billionaire's son on Tinder and scammed millions of dollars from young women who fell in love with him.

In addition, some dating apps like Grindr had and still have problems hiding people's locations. They used the trilateration method, which determined the exact coordinates of the user using the "distance from me" feature. Another example is the 3fun app, which is considered even less secure. It used a GET request to find out the user's exact latitude and longitude and locate them all the way back to their home.

It is also known that JCrush, a dating app popular in the Jewish community, exposed the data of 200,000 users after a security breach.

The platforms themselves are actively addressing these issues. They regularly update privacy policies, tighten security measures, and introduce new rules of conduct. Unfortunately, this is often not enough, so caution and common sense are the main things to remember when dating online.

Conclusions

In my opinion, the most important and needed improvement to dating apps is increased security and reliability.

Many users still encounter fake profiles. On the one hand, there is a need for greater security, and on the other hand, privacy. There is already a practice of using predictive analytics, which allows you to look through profiles and detect those with a high probability of unreliable data.

Advanced analytics can also help analyze text and audio content and can even detect emotional content. Predictive analytics can help weed out users who post, for example, content that fails to meet security requirements.

Let's not forget that the goal of combining machine learning and data analysis is to make our lives easier. And dating apps should work with this goal in mind, not exposing users to risks but opening up new opportunities to find soulmates.