Album review
Popularity and Attractiveness Bias
An analysis from Carnegie Mellon University and the University of Washington involving data from over 240,000 users on a major Asian dating platform over three months indicates a user’s chances of being recommended by the platform’s algorithm increase with their average attractiveness score. This reveals a bias toward recommending more popular or attractive users.
Popularity bias may be lower during a dating platform’s early stages of growth because a higher match rate can help build the platform’s reputation and attract new users. As the platform matures, the focus often shifts to revenue maximization, resulting in increased popularity bias.
Unintentional biases also arise from platform design choices. An example includes an experiment by OKCupid, where users were falsely informed they were highly compatible with certain people the algorithm deemed poor matches. The findings showed that users were more likely to have successful interactions when presented with a compatibility suggestion.
Racial and Ethnic Bias
Racial biases in dating app algorithms are common. A 2019 study highlighted that men frequently using dating platforms viewed multiculturalism less favorably and considered sexual racism more acceptable. Another paper presented research showing that black men and women are 10 times more likely to message white people than vice versa.
In 2016, a BuzzFeed reporter discovered that Coffee Meets Bagel showed users only potential partners of the same race, even when users indicated no preference. Allowing users to filter potential matches by race permits actions based on discriminatory preferences, preventing connections with partners they might otherwise start to appreciate.
A 2018 paper by Cornell researchers suggested that dating apps could combat discrimination by providing categories other than race and ethnicity for users to describe themselves. Additionally, inclusive community messages and algorithms designed to avoid discrimination were recommended. The paper also proposed policies against certain language and educating users about biases.
Effects of Dishonesty and Scams
Dishonesty in user profiles is another issue. A 2019 study reported that 31% of women and 36% of men admitted to lying on their online dating profiles “just for fun.” Men exaggerated their height by about two inches on average, and OKCupid estimated that its users report making about 20% more than their actual income.
The issue of romance scams also presents financial risks. In 2020, romance scams resulted in a cumulative loss of $304 million, a notable increase from $75 million in 2016. The median amount lost was $2,500, with users aged 40-69 most likely to report monetary losses. Scammers typically use fake profiles and plausible stories to extract information and money.
Also, your online search might point you toward smaller dating niches, such as someone looking to find a sugar baby rather than a person looking for a conventional relationship. It is fine if that’s what you’re looking for, but it will waste your time if it isn’t.
A 2023 survey found that 71% of online daters believed lying on profiles was very common, with an additional 25% considering it somewhat common. Approximately 10% of online daters quit within three months due to burnout from frequent swiping and unfruitful conversations.
Impact on User Engagement and Revenue
A study published in 2023 found that unbiased recommendations by dating platforms result in significantly lower revenue and fewer matches compared to popularity-based recommendations. Popular users help generate more revenue by boosting engagement through likes and messages, and they facilitate more successful matches, provided they do not become overly selective.
Another related piece of research is the work by Ben Berman on game design to expose biases in dating app algorithms. His “Monster Match” game illustrates how algorithms eliminate certain profiles based on early preferences, creating an echo chamber of tastes and discriminating against minorities.
The Match Group, which owns Tinder, Hinge, and other apps, faced a class-action lawsuit in 2024, alleging that the company prioritized profits over user interests. The lawsuit claimed that addictive features and opaque algorithms kept users continuously swiping without achieving meaningful matches. This frustration with dating apps has led some singles to opt for alternatives such as speed dating events.
As of 2024, it was estimated that there are over 8,000 dating sites and apps globally, with 366 million users. Approximately 35% of these users have experienced at least a six-month relationship through online dating, while nearly 14% of users have married someone they met on a site or app.
Recommendations for Mitigation and Improvements
Efforts are recommended to mitigate algorithm biases and improve user experiences. One suggestion is integrating algorithms that account for a variety of user-defined characteristics beyond attractiveness or popularity, providing a more inclusive platform. Another approach involves the transparent design of compatibility metrics that avoid reinforcing user biases.
Moreover, dating apps could benefit from increased user education on the realities of biases and the effects of dishonest behaviors. Implementing policies that discourage the presentation of false views and preferences can contribute to more authentic and fair interactions.
Thus, recognizing and addressing these biases is crucial for creating a more equitable user experience on dating platforms. Policies and design choices that promote diversity and authenticity hold substantial potential for improving both user satisfaction and engagement metrics while protecting users from deceitful practices and scams.
