Why You Swipe Right (PART 3)
We’re in the home stretch! Click to read Part 1 and Part 2.
More on racial discrimination: the users or the apps?
The disparate racial proportions encountered during my two months of Tinder swiping indicate that Tinder’s matching algorithm isn’t as unbiased as they’d like to believe. That algorithm is ultimately what dictates anyone’s chances of a match occurring, and, at the time of this research, it’s showing me people who are mostly Caucasian.
I can’t swipe on someone of a different ethnicity if I’m not being shown any people of that ethnicity at all! Which made me wonder: how often is my profile getting seen? What’s my desirability ranking, according to the Tinder gods?
Mobile dating apps that allow users to filter their searches by race—or that rely on algorithms that pair up people of the same race—reinforce racial divisions and biases, according to a new paper by Cornell researchers. As self-thinking machine learning products become a habitual part of our lives, the average person might not even question their results and just unconsciously perpetuate this cycle.
Dating apps thus have a responsibility to their users to implement fairness policies. In this case, Tinder’s matching algorithm should’ve been trained on a balanced dataset of ethnicities. Though I can’t be entirely sure, I don’t believe Tinder worked to balance the ethnicities for their matching algorithm. Perhaps they should consider a less racially-biased redesign.
The lopsided racial proportions here are reminiscent of another world heavily influenced by racial biases: fashion modeling. Sadly, fashion is another case in which the majority makes the rules. Cameron Russell discusses this in her famous TED talk:
_“Well, for the past few centuries, we have defined beauty not just as the health and youth and symmetry that we're biologically programmed to admire, but also as tall, slender figures, and femininity and white skin... In 2007, when an inspired NYU Ph.D. student counted all the models on the runway, of the 677 models hired, only 27, or less than four percent, were non-white.”_
Yuval Noah Harari, a renowned historian, also echoes this in his bestseller Sapiens:
_“American aesthetic culture was built around a white standard of beauty. The physical attributes of the white race—for example, light skin, fair and straight hair, a small upturned nose—came to be identified as beauty.”_
And Hugh Hefner blatantly popularized these preferences in Playboy magazine. He openly admitted to favoring blondes: “I would have loved to [date Marilyn Monroe]. I’m a sucker for blondes and she is the ultimate blonde,” he told CBS Los Angeles in 2012.
In an interview with the LA Times, Heffner recalled that seeing Alice Faye in films as a child was the moment that first fixed in his mind the specific ideals of beauty—busty and blonde—that he’d later made ubiquitous. This construct became pervasive in large part because of the personal biases of someone who had broad media outreach.
So if Caucasian features are favored in these contexts, are dating algorithms haphazardly doing the same thing as modeling scouts?
Let’s look at the big picture: dating apps are constantly showing people from the same demographic, and the media has been parading these same people around for centuries in magazines, billboards, and art. Anyone who’s heard of the mere exposure effect knows that when this occurs, we end up liking the things we’re most exposed to.
It should be clear why fairness policies need to be enforced; what about a platform which promotes diversity? Now that would be something.
Unfortunately, most apps go the other way—fully segmented, like Dil Mil, JDate, Christian Mingle, Minder, etc. (At least they’re upfront about it.) Sure, there are niche groups who only want to date their own kind but there’s nothing similar to “United Colors of Benetton” in the dating app world.
I’m sure I’m not the only person who’s met up with someone I matched with online, only to find that I wasn’t physically attracted to them… even though we got along well prior to meeting face-to-face.
If dating apps had a way to extract physical attributes from photos and combine them with personality data, they’d have much more success in creating couples. For example, OKCupid could pair the info gleaned from all those questions it asks users along with data about their physical attributes from swiping activity.
Better yet, what if no one had to fill out any lengthy, tedious questionnaires and instead, computer vision was used to infer more about personality? If an app could infer the activities you’re interested in from an activity-based photo (such as a pic of you surfing or doing yoga), maybe they could link people together based on the overlap of their favorite activities. Since it’s common to have access to a user’s Instagram feed, this should be even easier for an app to do and require minimal effort on the user’s part.
Tinder is moving in this direction with Top Picks, where it’s extracting info from user photos and bios to make personalized recommendations. But I don’t believe they’re using Instagram or Facebook photos yet, which is important because with these systems, typically, the more data, the better the recommendations.
Individual user insights are another feature worth digging deeper into: as Eckhart Tolle tells us, “Awareness is the greatest agent for change.” Like I mentioned before, there are often conflicts that occur between our conscious and unconscious brain—for example, you answered that you’re serious about dating someone of the same religion or race as you on your questionnaire, but your swiping activity indicates otherwise. Which is accurate?
And then there’s the big question: who’s attracted to you? What do the people swiping right on you have in common, both physiologically and psychologically? Do they all do yoga? Have an Ivy League pedigree? Have a dog or a cat?
Could there be an algorithmic tool developed that can identify your specific demographic even if you’re unaware that you belong to one?
Without being aware of your demographic, you’ll always have friction dating other people, says self-help/dating guru Mark Manson. He covers the concept thoroughly in his book Models.
This might be the most enlightening bit of insight for a person looking for love. People looking for answers to their dating struggles often blame things which aren’t relevant. Using the user insights described above, they’ll be able to pinpoint the source of their woes much more accurately.
Lastly, if apps put more stock in swiping activity at a personal level and if users were aware of this, maybe users would swipe with more care (just like I did, as I didn’t want to pollute the results of this study). That way, apps wouldn’t have to do things like limit users on how many times they can swipe in a given day. I’m positive this would result in a higher quality of matches.
What have I learned? What should you take away from all this?
Over the course of my analysis, I learned which facial features I find most attractive. This was was validated by my dating history—I had no idea I had such a bias towards women with sharp features.
It seems there’s no need for me to feel glum if online dating isn’t working out: the real world doesn’t mirror the algorithmic one online. The racial bias present in the online dating world is likely not a realistic reflection of user discretion but rather the matching algorithm’s weakness.
Dating apps have a responsibility to be aware of fairness practices and to level out the playing field, especially given their ever-rising popularity. This, combined with the lack of user-level insights and questionable algorithmic fairness, indicates that a redesign is long overdue.
Most importantly, through rigorously applying myself during this analysis, I’ve learned a lot about the science of attraction. It’s part nature and part nurture, which means we can make changes if we don’t consciously agree with our inherent biases.
We often get caught up in arbitrary societal norms, herd mentality, and the celebration of the fortunate few whose looks are constantly paraded around us, and are subsequently brainwashed to believe those select features alone are attractive. Recognizing this tendency for the social conditioning gives me more to think about when I find myself judging a potential match solely based on looks—maybe there’s more to it than that.
And here’s where we wrap things up! I hope you enjoyed this series—I really poured my heart into it (in more ways than one _🙂). _
Let me know in the comments if your experience with online dating lines up with my findings, or if you’ve got some additional insight to share!