In this article, we will analyze the book by French journalist Judith Duportail “L'Amour sous algorithme” (Love Under Algorithm). In this book, the writer talks about her investigation in which she tried to find out how the application's algorithms work. Today, we publish abstracts and bullet points from the book on how Tinder matches users to others and analyzes our profiles.
About Elo rating system
Each Tinder user is assigned an internal rating – an Elo score. This is a term from the world of chess, where it is used to rank chess player skill levels. When someone comes across your profile, a chess-like mini-tournament takes place. If a “player” with a high coefficient is against you and he likes you, you get points. If he has a low rating and he swipes you to the left, the points are deducted.
The game continues
After every match, Tinder invites us to “keep playing.” The app uses vibrant colors and playful tools that trigger small bursts of serotonin into the brain with each match, forcing us to come back over and over again. Tinder sends us notifications with the number of new likes when we stop using the app. It informs us when our profile becomes displayed less often because we rarely use it. Tinder shows us candidates after candidates, giving us the feeling that there will always be someone new next.
Bypassing the algorithm
The technology and algorithms can be bypassed: Tinder allows users to escape from their own algorithm for money and thus feel like a leader in the Elo rating. This is a Boost feature that will take you to the top user for 30 minutes.
Success rate
Some photos have a success rate, which consists of ten digits, for example, 0.13131313131313133. This figure corresponds to the percentage of likes that a profile with this photo received. Tinder has neither denied nor confirmed this.
About keywords and fate
Roughly speaking, if you write on Tinder that you love Baby Yoda, you will see more men who mention Baby Yoda in their profiles. All of this data is brought together to assess the compatibility of your profile with others. Tinder realized that they did not analyze biographies texts enough, because many users simply did not write anything there. Then Tinder got Rekognition – an artificial intelligence created by Amazon for cataloging photos. If you're pictured with a guitar, you'll be classified as a creative person. It all is supposed to bring together people who have common ground.
The authors of the patent note that believing in fate is extremely useful when building a relationship or simply meeting a new person, because people care about symbols, especially when it comes to love. The server can be configured to match interests, place of birth, date of birth, university, first name, last name, nickname, and keywords to give the impression that users were destined to be together. When the algorithm finds such matches, there are two options: to either show the user these similarities or not. In the second case, the goal is for the user to find common ground himself and thus believe that this meeting happened because it was meant to happen.
About the conflict with Tinder and patriarchy
Tinder wrote to Duportail that many of the things that are mentioned in the patent are not currently being used. But the journalist was outraged because the patent should convey the company’s values, while she only found contradictory moments in it.
Duportail cites the opinion of Jessica Pidoux, professor of Digital Humanities at the École polytechnique fédérale de Lausanne, who says that the patent reflects a patriarchal model of heterosexual relationships.
It turns out that the algorithm can give preference to a match of an older man with a younger, less wealthy and less educated woman.
Here’s a quote from the patent:
As an example only, assume that Harry and Sally are registered users who have profiles in matching server. Harry has submitted a search request to matching server. While fulfilling this request, matching server evaluates Sally's profile since her profile is in pool. As part of the evaluation, matching server looks at the differences between Harry and Sally's stated age, income, education, ethnicity, and location. In this example, Harry is 10 years older than Sally, makes $10,000 more per year, and has a Master's degree while Sally has a bachelor's degree. Even with these disparities, matching server will give Sally's profile a high score which makes it more likely that Sally's profile will appear in Harry's result list. However, if it was Sally who submitted the search, and matching server was evaluating Harry's profile, a different score is possible. So, if it were Sally who was 10 years older, made $10,000 more per year, and had a Master's degree while Harry had a Bachelor's degree, matching server would give a low score to Harry's profile, making it less likely that his profile would appear in Sally's result list.
