Ranking Algorithms in the Context of Dating Apps
Increasingly, people are relying on dating applications such as OKCupid, Zoosk, or (and by far the most common of the bunch,) Tinder to meet potential partners. While online dating was previously negatively stigmatized, its results speak for themselves – “Today, nearly half of the public knows someone who uses online dating or who has met a spouse or partner via online dating” (Pew 2016). But how have mere applications been able to emulate real-life attractions? How have they been able to pair like people together in such an efficient fashion? Perhaps the answers to these questions lie in the underlying algorithms of these apps.
Most dating apps nowadays rely on two primary methods to connect potential partners:
- Allowing users to create profiles. People with similar profiles (i.e. similar specified interests, hobbies, or backgrounds) are ranked higher and are displayed to one another in hopes of creating a match.
- Dynamically ranking “desirability” on the basis of “liking” or “disliking” other users. People with similar “desirability” scores are displayed to one another in hopes of creating a match.
Given how homophily (the concept detailing how a person is drawn to others similar to him/herself) plays a role in attraction, the former of these methods is straightforward. The latter is a little more complex, but also vaguely familiar – we’ve seen something like it before. The latter method, which is the core of Tinder’s matching algorithm, closely resembles Google’s page-rank algorithm (fastcompany 2016).
Consider person A with a Tinder account. Person A encounters person B on the app, but is not impressed. Person A “dislikes” person B. Next, Person A encounters person C. Person C has caught the fancy of person A, so person A “likes” person C. We can think of the above scenario as person A sending a link to person C, but not to person B. Objectively speaking, person C is more desirable than person B.
But what happens if person A “likes” all of Persons B through Z, and “dislikes” nobody? Clearly person A’s rating system is unreliable. While the people that person A “liked” will have increased rankings, they will not increase by much. This parallels how a site linking to many other sites does not contribute much to the rankings of those other sites.
Next, let’s consider persons A and B both “liking” person C. Person A is highly ranked in terms of desirability, while person B is not. Peron A’s “liking” of person C will impact person C’s desirability far more than person B’s “liking” of person C did. This parallels how sites with higher rankings have more impact on sites they link to than sites with lower rankings.
Thus, these dating websites have created a system in which people are dynamically ranked and easy to match with one another. Who could have guessed that an algorithm initially intended for the use of search engines could be used to match single millennials?
Works Cited:
http://www.pewresearch.org/fact-tank/2016/02/29/5-facts-about-online-dating/
https://www.fastcompany.com/3054871/whats-your-tinder-score-inside-the-apps-internal-ranking-system