Chapters
Shownotes Transcript
When you're running a small business, sometimes there are no words. But if you're looking to protect your small business, then there are definitely words that can help. Like a good neighbor, State Farm is there. And just like that, a State Farm agent can be there to help you choose the coverage that fits your needs. Whether your small business is growing or brand new, your State Farm agent is there to help. On the phone or in person. Like a good neighbor, State Farm is there.
Silicon Valley Bank is still the SVB you know and trust. The SVB that delivers human-focused, specialized lending and financial solutions to their clients. The SVB that can help take you from startup to scale-up. The SVB that can help your runways lead to liftoff. The only difference? Silicon Valley Bank is now backed by the strength and stability of First Citizens Bank. Yes, SVB. Learn more at svb.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.com.
So, Sangeeta, I did something for the podcast. Tell me about it. I started swiping again, but not on a real dating app. This was a simulated dating game, Monster Match. I thought it might help me understand how real dating apps work. Yeah, I don't think I've ever had a soundtrack to my swiping session. I guess that's because they're going to deliver some unpleasant news and they need me to have music. So, I'm going to do a little bit of swiping.
Monster Match works like a bizarro dating app. You create a profile. I get to choose my body,
Wouldn't that be an interesting option to have in life? I'm going to pick kind of like a snake-like body. It's kind of interesting to think about, you know, reptiles mating and dating. I ended up with an Iron Man robot mask, big hair with highlights, and a necklace. Very hot. Then I got to choose a background. I picked a city.
Then it was time to start swiping. First, I saw the profile of Count Daniel. For work, he has a startup that makes farm-to-table, 100% vegan blood for vampires. Good for you. I swiped right. Oh, Daniel has already started talking to me. I'm going to say you seem like bloody fun.
You seem like bloody fun. That's a great line. I mean, you have to work with what people give you. I'm glad I've still got some chat game. Then there was little Johnny Chestface. Larry, he was a dragon person. Cringe Chris, who is a self-identified nightmare date. Zoloth, the great demon.
He says, in the zombie apocalypse, I'd be the one dot, dot, dot, starting it all. Monster Match really lives up to its name, huh? Yep. And every so often, the game would reveal a little bit about how my swipes were affecting my recommendations. Basically, the profiles the app would show me next. And that is the point of the game, which is modeled on how real-life dating apps work.
Benjamin Berman is the guy behind Monster Match. The game was inspired by a friend of his who'd moved across the country to San Francisco and was single. Charming, handsome, friendly. Nine months of swiping led to one atrocious date after another. And I felt like either something is seriously wrong with him...
Something is seriously wrong with all the women in San Francisco, or something was wrong with the software. Software. That's something Berman, a game developer, could figure out. But what was it specifically? What element of the tech was making all of this feel so bad? It turns out to be the algorithm that recommends who you see. This is Land of the Giants.
Dating apps use algorithms to determine who users see and what kinds of matches they have access to. Some apps brag about their algorithms. Some won't talk about them at all. But none reveal exactly how they work. Which means users are at the mercy of technology that few people understand, but is all-powerful when it comes to matching them with potential partners.
So how good are these algorithms? And when users feel like they aren't working, can they hack the system? Asking a dating exec how their matchmaking algorithm works is like asking Coca-Cola for its top secret Coke formula. You never get a straight answer. But I asked who I could, what I could, when I could. Like Jonathan Badin, co-founder of Tinder.
There's like an urban legend around the Tinder algorithm. Did you guys ever rank people based on attractiveness or desirability, like hot or not? It was an algorithm that was essentially based upon
the number of right swipes that people would get. And that would eventually give you some sort of score, and you could then show people others that had more similar types of scores to themselves. Barin says this concept is called an ELO score, which is also how chess players are ranked in competition. If you beat a high-ranked player, your ranking goes up. A lot.
On Tinder, if a popular person swipes right on you, then you will see more attractive people on the app. That sounds like high school BS at scale. That sounds like a popularity contest. Doesn't it? Badin's no longer at Tinder, and the company says ELO scores are gone too.
Tinder says that today, the app has a recommendation algorithm that factors in how active you are on the app, your profile details and photos, your swipe history, and how often your profile is liked, which still seems like a popularity contest.
When we asked Bumble about its algorithm, we got a somewhat opaque answer. A spokesperson told us that the algorithm learns from your history to serve up matches. That sounds so vague. It could describe a lot of things. Totally. There are apps that will share a tiny bit more about their algorithms because the algorithm is a big part of their marketing and their pitch to users. Oh, that sounds like Hinge. Exactly.
I asked former design and product VP Tim McGoogan about it. The most compatible feature on Hinge uses an algorithm that pairs people together and it uses something inspired by close to the Gale-Shapley algorithm. You'll actually see this come up in Hinge's marketing. Its special proprietary algorithm was modeled on the Gale-Shapley algorithm, a complex matching formula which won its creators, a couple of mathematicians, a Nobel Prize.
So that's cool, but what does that mean? How does it work? I'm not going to tell you how it works. The first rule of talking about your algorithm is that you don't talk about your algorithm. MacGugan left Hinge in 2022. We followed up with Hinge, and a spokesperson didn't have much else to add to MacGugan's explanation.
Matchmaking algorithms go back to the OGs. Match.com and eHarmony touted their recommendation systems to distinguish themselves from freewheeling online chat rooms.
Amarnath Tambrai, the CEO of Match Group Americas, told me that these recommendations were rudimentary at first. Users came in and they filled out, this is who I am and this is who I am looking for. And then the algorithms like, OK, you match each other on nine out of the 10 things you both say are looking for. Then that person became a recommendation for you.
That changed in 2008, when he joined Match as vice president of strategy and analytics and updated its algorithm. We said, you know what, we're just going to put that aside, what people say they want. And we're going to look at what people are doing and then build recommendation engines that entirely were based on user actions. It also helped that these sites had professionals to legitimize their systems and build trust in the algorithm.
eHarmony had Dr. Neal Clarke Warren. Match.com had Dr. Phil. So you both have these PhD psychologists, old white dudes as I call them, out on Oprah and the talk shows. Sam Yagen was CEO of Match Group when it went public in 2015. But he launched OkCupid in 2004. He'd already been thinking about the online matchmaking space for more than a decade.
They had this resounding message, which is usually talking to middle-aged women, you don't know what you're looking for. We know what the right match for you is.
And there's something very, to me, creepy about that. But there's also something very lucrative about that, which is if you can convince people that the reason you're single is very simple, you just don't know what you're looking for, and we have the answer. We know what a match looks like, and we know who your soulmate is going to be. That's a great business proposition. In other words, trust the algorithms. This same message has persisted through today.
Sure, an IRL meet cute sounds great, but it relies so much on chance. Plus, are you meeting enough people for that? And do you really want to talk to a stranger? Enter the apps. They filter thousands of people who are ready to mingle. They use quantitative analysis to create your dating pool. It's math. It's science. It seems kind of magic to us. We're not really sure why it works so well. We don't want anyone to see how dumb it is.
Cathy O'Neill is a mathematician, data scientist, and author of the book Weapons of Math Destruction. It's all about algorithms and their impact on society. She used to work in big tech. O'Neill says there's a reason why the apps are so cagey about their algorithms. She thinks that if we really knew how they worked, we might not want to put so much blind faith into them. You know, this is framed as really hard because they want you to sort of trust it and slash be intimidated by it.
but it's actually not hard. I asked O'Neill to break it down for me. She says that dating apps use predictive algorithms, and predictive algorithms are about digital pattern making. You basically are saying, if it worked in the past, it'll work in the future. Predictive algorithms just take historical data. They look for patterns of success or failure with respect to a specific definition of success.
and then they project the future from the past. They extrapolate. That's pretty simple. An app isn't smart. It doesn't really know what users want. It can't tell whether two people will have fun together, share passions, or be sexually compatible. But it knows if you've looked at someone's profile. It can count the number of words used in a chat, or tell when numbers were exchanged.
This is the data that is being used to determine your success or failure as a mate. So what do you do if you feel like you're failing more than you're succeeding? Daters don't really know how the algorithms work, but they do know what it feels like to go on a bunch of terrible dates. So in the dark, they come up with strategies in an attempt to mitigate all of those bad matches.
Sangeeta, I want to tell you about Sarah Sodoroff. She's 37 and lives in Toronto. ♪
Sodoroff had burnt out on the apps at some point. The thought of having to go back onto them was almost worse than the breakup. It was an almost worse feeling of failure and having to like walk back into this cohort of rejects as I felt it to be. Honestly, I think I was paralyzed with doing it for weeks.
She was over the apps, but not over the idea of finding a partner. So she settled on a new game plan. This time, she'd say yes to everyone. I was very mission-driven, and I was very active. I think you're told a lot in Dating is a Numbers game, you got to get out there, you got to get your reps in, you've got to be giving yourself the maximum ability to meet with and match with the most amount of people. Her last cycle through the apps, she thought maybe she'd been a bit too picky.
Only right swiping on select profiles and interacting with certain kinds of guys. But then she felt like she wasn't getting the matches she'd hoped for. I tried to open that up to say, you never know if you're going to have a connection with somebody. On Hinge, I would say yes to anybody who had messaged me or I would match with anybody who had messaged me because that's the option that they give you and then have those conversations.
But Sodoroff's new approach bombed. She could feel the algorithm adjusting to her new preferences, but not in a good way.
She was getting further away from her goal. Time for a new strategy.
Hinge says its recommendation algorithm suggests compatible users who'll like each other back. It was a glimmer of direction, so Sauteroft grabbed onto it and leaned in. When you first start matching, swiping, and liking, it will say...
We're getting a feel for the kinds of person you like. And so I think you think, oh, the app is reading my behavior. I think it might make the user experience more enjoyable because you feel as they're getting people who are actually better tailored to you. She reset her criteria, made her own filters, tried to give Hinge's algorithm enough information to go on, tried to be a model Hinge citizen in hopes of getting better matches.
You get higher quality matches if you tend to interact with a prompt. I think it leaves you with some level of comfort. You're like, oh, the app is working for me. The app is looking out for me, which is a fantastic behavioral trick. But this comfort didn't last long. I'm often offended by the people they think I'm most compatible with.
I'd love to learn from them what that compatibility is actually based on, because you'll see that and say, I'm not physically compatible with this person. We don't have the same interests. So where does this algorithmically match me as a compatible person there? So that wasn't working either.
I think there is a certain sense of like, let go and let God. The algorithm is going to present to me people who are good matches for me. Maybe I'm a bit scorned, but I don't think it's working for me. I don't think the matches that I have been given are of any interest. I would say to me, that is just a big gray hole. When Soterov didn't give the app enough to go on, when she matched with everyone, she got irrelevant matches.
When she tried to feed the algorithm, the experience was still trash. And now Satorov is wondering if there's any strategy at all that could get her closer to finding a partner. Sounds like probably not. Okay, Lakshmi, now I have a story for you about someone who tried to make the algorithm work for him. That's in a minute. ♪
Support for Land of the Giants comes from Quince. The summer is not quite over yet, but shifting your wardrobe to the colder months could start now, little by little. You can update your closet without breaking the bank with Quince. They offer a variety of timeless, high-quality items. Quince has cashmere sweaters from $50, pants for every occasion, and washable silk tops.
And it's not just clothes. They have premium luggage options and high-quality bedding, too. Quince's luxury essentials are all priced 50% to 80% less than similar brands. I've checked out Quince for myself, picking up a hand-woven Italian leather clutch for my mom. As soon as she saw it, she commented on how soft and pretty the leather was and told me it was the perfect size for everything she needs to carry out on a quick shopping trip.
Make switching seasons a breeze with Quince's high-quality closet essentials. Go to quince.com slash giants for free shipping on your order and 365-day returns. That's Q-U-I-N-C-E dot com slash giants to get free shipping and 365-day returns. quince.com slash giants
On September 28th, the Global Citizen Festival will gather thousands of people who took action to end extreme poverty. Watch Post Malone, Doja Cat, Lisa, Jelly Roll, and Raul Alejandro as they take the stage with world leaders and activists to defeat poverty, defend the planet, and demand equity. Download the Global Citizen app to watch live. Learn more at globalcitizen.org.
Jeremy is 30 and lives in Philadelphia. He's on dating apps and has a very specific understanding of them. As an app designer, I pay a lot of attention to the patterns and behaviors and I guess just methodologies behind the apps that I use on a daily basis. Jeremy asked us not to use his last name because he works in tech and wanted to speak frankly about his experience.
These days, he does alright on the apps. My friends call me generically attractive and I think that that's something that I exploit or at least lean into on dating apps where I'm like, "Okay, well I'm a 30-something white male, decently, not ugly." I end up with quite a few matches on dating apps. But there was a time when that wasn't the case, when despite his efforts, he didn't have an abundance of matches.
He used to make his profile as true to his personality as possible. He talked about his actual interests, like going to museums and concerts. He gushed about his love for experimental film. But it all backfired. I realized, like, I'm just not going to care about constantly scrutinizing every detail and updating my dating profile because of, you know, it was just...
never really seemed to get anywhere. So we went back to basics. Literally. When I started getting a lot more matches on my profile, it was because I had just like things that were meaningless. It's definitely discouraging to know that like what's generic and watered down kind of works best.
What does generic Jeremy's profile look like? Can you give me a sample of one of your prompts and what are the photos? I think my current about me is my interests include getting lost and listening, and that one seems to convert the best. Getting lost? What the fuck does that mean? I don't know. It's just something I thought was really stupid and then it kind of worked.
For Jeremy, volume was a success. The more matches, the better. And that meant engaging in a bit of Jeremy erasure. He walked me through his photos. One with his mom, check. Another with a dog, check.
If you look up tips to improve your dating app profile, moms and pets are huge. This one is a new addition, just a photo of me blowing bubble gum. I think it works fine. Maybe it makes me look funny or something. It's still very generic. And then there are the prompts. This is a good one that says, we'll get along if you can laugh. That's a killer, you know, that's a real charmer there.
So basically, the more personality you show, the worse for you. Yeah, that's what I found, for sure. Jeremy says his strategy to lean into his generically handsome white maleness has gotten him more matches. But have they been better matches? I'm not like, wow, this person is, you know, I've never like, wow, they're really getting like who I'm going for, you know? I would say it's like almost the opposite. It just feels like an endless abyss of
For Jeremy, more matches was the goal. He'll venture into the abyss and sort it out from there. For Sarah Soteroff, quality matches were the measure of success. So, two cases. Two people trying to game the apps to get the results they want. Each of them had a hypothesis. They ran experiments and adjusted parameters. They gathered data on results and came to conclusions about how the apps work.
But the truth is, they don't know. They're just trying anything to make all of this feel a little bit better. But there are people who do know. They don't need to experiment in the dark because they've got all the variables laid out right in front of them. The apps may not be willing to demystify their algorithm, but we found someone who would. Fan Yin Zhang.
A leading dating app for intentional daters wanted to improve its algorithm and partnered with Zhang and some of her fellow researchers to help them figure out how to do so. It wasn't Hinge, but it is fair to say that this app functions a lot like other leading apps. Zhang is an assistant professor at Columbia Business School. She studies how data can improve business operations. For this dating app, she was asked to figure out how to increase the number of matches on the platform.
Zheng and team got access to all the inner workings of this app. Its user data, the algorithm, everything. Beep bop boop, they pulled a few lovers, and... Our algorithm improved the match rate by around 30%.
That seems significant. They did overhaul some things. One example. If there were tons of great people who could be a great fit for you, the new algorithm wouldn't flood you with them. You're a catch, and you might be juggling a dozen conversations on the app already. They updated the app to keep users a little bit thirsty.
What this means from the platform's perspective is that when you have some great potential profiles to show to your users, you don't want to give it to them all at once. You want to space out these attractive potential matches to your users, and this increases the chance of a successful match. They found that when a user was overwhelmed with options, they might miss a lot of great potential matches.
And Zheng says a bit of withholding benefited the company too. It may not be the best thing for the platform if everyone finds their right person and never return again, right? So there's a little bit of keeping their users active on the platform, at least for some time. That sounds like a conflict.
Many of the daters we've spoken to who are looking for a long-term relationship, they would love to download an app, find a partner, and never have to open it again. Yeah, I know a lot of people who would love that. Okay, so the next thing that Zhang looked at had to do with something truly fundamental about how this app and many others work.
It's about how the algorithms pick matches in the first place, how they figure out who might like whom, the secret sauce, the recommendation engine that shapes romantic prospects across the apps. Here's how it works. The algorithm needs data, lots of it, to make an informed recommendation. So you're lumped together with others who seem similar to you.
Then the data that you and thousands of others give the app, who you swipe right and left on, who you message, is crunched. Those millions of interactions help the app make judgments and predict the future. That sounds romantic. Take height. Looking at the data in aggregate, the algorithm might find, ah, yes, people who are 6'4 prefer to match with people who are at least 5'10".
So then you have Louise, who is 6'4", and only sees people who are at least 5'10" on the app.
But for this particular person, the algorithm might realize that sometimes when we recommend someone who's 5'6", but with all these other features, the user could show particular interest, right? And then I can build that into the algorithm. But Luis also loves music, and he also matches with people who are really into 70s prog rock. Aaron fits this bill. He wants to take a date to see Jethro Tull,
He could be a match, but he's 5'6". Will the app show Aaron to Louise? Will Louise be freed from the tyranny of the tall people bubble? Oh my god, you're killing me. What happens? Well, their fate is in the hands of the algorithm. If it makes inferences based on certain kinds of group behaviors, like tall people are into other tall people, it would actually keep Louise and Aaron apart.
To avoid these kinds of missed connections, Zheng decided to build more granularity into the system. We made the algorithm more personalized, really have the algorithm learn more about the user's preferences over time, not treating people as groups or only look at their average behavior, but look at the individual preferences.
What all this tells us is that these algorithms can leave the daters who depend on them wanting.
If I'm Asian, do I only get to see Asian profiles? And maybe that's not what I'm interested in. But in the end, this is the impact of the algorithm. And then it could actually lead to outcomes of matches. And how do we quantify the loss of welfare for individuals like that, right? How do we say, you know, is it just the preferences or is it actually the result of the algorithm?
But the companies have sold us on the idea that math and data are the key to romantic success. Trust the process. Trust the algorithm. Now that obviously makes sense for the companies, but does it make sense for you? We trust the data, that math and science couldn't be wrong, that it's on us if we're not finding the right people on the apps.
But what if you don't fit into the world the algorithm thinks you should? That's why I sometimes just say they make lucky people luckier and unlucky people unluckier. Mathematician Cathy O'Neill again. So, if you find yourself one of the unlucky ones, do you question the tech or do you question yourself?
We become more and more insecure and self-conscious about ourselves and how we are reduced and flattened into these little data points. The overall effect is icky and it doesn't actually improve our love experiences. Another way to say that, our trust, it's likely been misplaced. There's another company that is trying to build trust in online dating, not through technology, but through culture. And it's working.
It has 22% of the market share in the U.S. and presents a serious challenge to match groups' dominance. The story of Bumble, next time on Land of the Giants. Land of the Giants is a production of The Cut, The Verge, and the Vox Media Podcast Network. Oluwakemi Oladisuyi is the show's producer. Cynthia Batubiza is our production assistant. Charlotte Silver fact-checked this episode.
Jolie Myers is our editor. Brandon McFarland is our engineer and also composed the show's theme. Nicole Hill is our showrunner. Additional support from Art Chung. Jake Kasternakis is deputy editor of The Verge. Nishat Korba is our executive producer. I'm Sangeeta Sinkerts. And I'm Lakshmi Rangarajan. If you liked this episode, please share it and follow the show by clicking the plus sign in your podcast app.
