232. DatingOS: Bumble's New AI Matchmaking Service May be a Mismatch

That data suggests that this is what comsumers (especially women) want. But as I've explained before, Data Driven Design is not a substitute for asking your customers what they want.

My articles on DatingOS seek to explain how people are programmed to treat each other romantically by their culture, media, governments, religion, and other institutions. None of these influences are monolithic and are rarely coordinated. The result is a mismatch of beliefs and approaches that are typically contradictory and can lead to failure and frustration when attempting to forge romantic attachments with other people.

I surveyed (both in real time and in chats) ~250,000 women (mostly in the West but also Eastern Europe) from 2010 to 2020 for my dating economics research. [I’ve written a number of papers on Dating Economics in addition to the Dating OS papers below.] This is the opposite of using Data Driven Design, and is far more effective. I discussed some of the problems with DataDD two papers prior. Direct surveys are much more time consuming and can require other resources, but generates much higher quality data. In my last paper I discussed the critical importance of listening to people PRIOR to entering the product design stage.

Previous DatingOS papers:

• Soulmates

• Sexual Arbitrage

• Intimacy In Relationships

• Dating Operating Systems (Dating OS)

• Dating OS: Multi-Dating and Serial Monogamy

• Dating OS: “What Happened to All the Good Men?”

• Dating OS: Why Are Men “Uglier” Than Women?

In my survey of women on Bumble, I found the app somewhat limited in efficacy. Thus only ~10,000 of the women I surveyed were from Bumble. The biggest problem with the app is that its selling point is that women “message first”. It sounds empowering, but the truth is that women don’t want to message first. They feel like they are the prize in dating scenarios and that men should be doing the heavy lifting. Here I’m of course talking through a heterosexual lens. Gay men don’t suffer this asymmetry and have relatively effortless dating experiences. Lesbians have it hardest here since both sides feel privileged. They also have the highest IPV rates, but that’s a bit off topic.

So while the messaging gimmick led to positive press, the company has lost 90% of its stock value since its 2021 IPO. Now they have a new gimmick: AI matchmaking! An app specific AI called “Bee” is being rolled out, and there will be no more swiping. I think this is a very positive step, and brings the app much more in line with OkCupid which was my favorite dating app before it was bought and gutted by Match Group in 2011.

Note that this isn’t new/fresh or revolutionary as OkCupid came out in 2004. But if you are going to copy someone, that’s a great place to start. Match Group wrecking it in order to make their other products more competitive just opened the door for such a move. In both OkCupid (OKC) and this new Bumble 2.0 (B2.0) you will have lists of profiles to look through. Contact is made by liking the profile as usual and if it’s mutual then messaging can take place.

In my paper How to Dominate Dating Apps Without Lying I explained how to “break” the swipe-based dating app algorithms, giving you an unfair advantage without spending. Ethically, of course.

Both OKC and B2.0 ask you to answer a bunch of questions about yourself to find out what kind of person you are, what kind of people might like you, what you are looking for, and how to match you with the right people. I’ve personally talked to OKC engineers in private real time chats, so I know how to break the OKC algorithms entirely, but that’s a much more involved process. Here the difference with B2.0 is that Bee will interview you directly, making the process of finding out all your personal information faster and more efficient, and perhaps making it much harder to break. I’m sure you won’t mind all that personal information being out on the internet. After all, privacy is an illusion.

But how will B2.0 perform after the initial novelty spike? People are very unsatisfied by the current quality of dating apps and their monetisation. Match Group’s attempts to monopolize the space and then reduced the quality of acquisitions like OKC were aimed at forcing consumers to adopt slop, not assisting them in their romantic ambitions.

If an AI can find the right people for you, and then tell you who to date, in theory this might be as helpful as the old school human matchmakers, but automated at scale. I’m not opposed to this, and I contemplated similar setups when I was designing advanced social products at Arrivant, a company I cofounded for that purpose. The thing is, getting two people to meet each other is the easy part, and this solution might make the hard part even harder.

I will use molecular chemistry as an analogy. There are many chemical reactions that can be quite intense once they start, but are very difficult to start. Burning a wood log might be a good example. People can be dropped into a snowy forest surrounded by wood and still die of hypothermia because they can’t get the wood to burn. So at Arrivant my goal was to solve that problem: how to lower the chemical activation energy requirements so that fire started nearly spontaneously.

By explaining that process here, perhaps dating apps can utilize AI in a way that will actually help people find romantic success by:

• Teaching people how to date

• Making failing fun and safe enough that they will keep trying

• Increasing trust between people so that they are willing to take more risks

This is all possible if you approach dating as a game. Which it is. I first publicly explained this in a lecture I gave in Saint Petersburg Russia in 2016 (no video for the first 10 minutes or so).

Consent Culture is a Double Edged Sword

It wasn’t until the 10th year (2019 to 2020) of my research that I started to understand just how essential and problematic Consent Culture (CC) is. CC understanding is essential to build trust and make sure that safety is maintained. Having your boundaries overrun by an overly enthusiastic suitor can be psychologically or even physically harmful. The resulting outrage can then cause additional reputational harm to all parties.

I think the original Bumble “swipe first” gimmick was an attempt to facilitate Consent, but had absolutely no effect on consent. By having both parties like each other, consent to talk is already established. Like all permissions, they can also be revoked at any time, which is why you can also block anyone you previously liked. Making the woman “go first” just put more labor expectations on women who really want the men to do the work of convincing the women to go on a date.

I’ve gone on dates where the other person’s stated reasons for dating were because they wanted to learn all sorts of secretive/powerful things about love or intimacy from me. Fair enough, I love teaching. But then over a series of dates, they just sat there and not once asked me a single question about dating and never gave me any permissions. It wasn’t until that experience that I realized that many people don’t even know how to give consent. This isn’t one of those things you are born knowing how to do.

I give a lot of information on the three stages of consent in my paper on Intimacy In Relationships. Note that that paper has racy content and is intended for adults. Here I want to focus on just the mechanics of giving consent. Let’s say a woman wants a man to make a move, but he doesn’t. What’s wrong? Is she not attractive enough? Is the man inexperienced or lacking in confidence?

The problem here is that “making a move” was the way things were done before CC became the standard. Men were expected to do this and it was considered fairly innocent. Now in 2026, that might qualify as assault! Men are increasingly educated about this and won’t make a move without consent. If you consider how stressful it is for women to just send a first message, you can imagine how getting up the nerve to give consent (for anything) might just be too much for some people.

Of course there are degrees of consent, and one level is not the same as all levels. Further, any level can be revoked at any time. This is complicated, even for experts. Also looks can be deceiving. The person you are with might have looked like they wanted you to do something, only for you to get slapped or worse when you tried to go for a kiss or such. The risk of an aggravated allegation, real or imagined, is enough to make dating too high risk for many men.

Likewise, women often assume that men “always want it”. That’s not true, even if it often is. Women still have to follow CC just like men do, it’s just easier for them. Of course body language can do a lot of heavy lifting here, but from my experience the “art” of body language is a dying art. Modern people are largely disconnected from their bodies, and their ability to “read” others goes down every year as they spend more time reading electronic devices instead of each other.

Women tend to be more sensitive to body language and mood than men are. This causes women to assume that men will just understand what they want without them needing to articulate it. As someone that grew up isolated from people when I was young, and perhaps some flavor of autistic, reading people was really hard for me when I was younger. Because I was bad at it, any sort of intimate interaction, even expressing a feeling, was high risk. Because I went into nursing then fitness and physical therapy, I ended up doing a lot of touching by profession. Learning this was stressful but I’m a fast learner. If I didn’t have to learn when and how this was okay, I probably never would have.

In modern times our ability to read each other is the lowest it has ever been for our species. This is a skill, and it takes a lot of practice. An app that tells you who to date doesn’t help here at all. By speeding up the “getting to know each other” process, it could actually bring two people together way before they are ready to do that. As B2.0 will ship without consideration for this, I think the result could be a lot of hurt feelings.

Of course we have a name for that also…

Dating Resilience

I’ve discussed the general concept of Resilience in previous papers where I investigate the misandry epidemic. All genders are being reprogrammed without consent by social media, often to hate each other. Hate is fantastic for driving engagement. For whatever reason women seem much more vulnerable to this sort of content, and the data shows they are generating anger/hate much more rapidly than men are. This seems especially true in the UK for reasons I suspect I know but will not speculate here.

There may be differences between the genders in regards to psychological resilience. Women are much more adept with their emotions, and it follows that they are likely more sensitive. In countries like the USA, UK, and Australia, women are nearly twice as likely to self report seeking psychological treatment. Perhaps for men, being thick as a tree means more things bounce off of them. I think a lot of non neurotypical people, contrary to the stereotypes, tend to also be very sensitive and thus have psych profiles more like women regardless of their gender. This could also be why NNTs are 3 to 6 times more likely to be trans or non-binary instead of “staying in their lane” and being cis gendered.

As a society I think we intuitively understand that women need more protection than men do and men used to be wired for this. In older times we used to know the people around us, but now people are often strangers. As diversity increases, people can end up being even stranger. I think that’s a good thing, but it can increase the need for communication skills at a time when those are at an all time low.

“No plan survives first contact with the enemy”

— Helmuth von Moltke the Elder, 19th Century

Dating has unfortunately become increasingly adversarial over time. Going on a date can feel like going to battle. It is not unusual to lose your early battles, and the other person may not care how new you are at this. Your first losses are the most painful, coming at a time when this whole process is the most confusing.

Your ability to bounce back from a bad date and go on another one is Dating Resilience. Yes, that’s actually a thing. No I didn’t invent that one. There is a growing body of research on this. As my fellow neuroscientists like to say, “Biology always finds a way”. When puberty kicks in, the vast majority of people want to be in some sort of intimate relationship with another person. In modern times that’s generally considered inappropriate until they reach the age of consent, which varies by country and condition.

So it is fair to say that almost all of us start out in life as involuntarily celibate (“incel”) until we end up voluntarily non-celibate. Sure there are all sorts of involuntary sexual situations that occur, and both autistic and non-cis people tend to be magnets for these situations. Thus I unfortunately have a lot of experience there.

About one third of all “incels” are women.

If a person has a bad intimate experience they may end up voluntarily celibate. Both kinds of celibate seem to be on the rise, and I would attribute that to people feeling hurt during the dating process and being reluctant to “get back on the horse”. As biology is pretty strong, there must be a lot of hurt and/or threat out there to keep people celibate. If our biology wasn’t so strong, we wouldn’t be the dominant mammalian species on the planet. Yet that’s the way we are heading as we aren’t making enough new people to replace our population. I talk about this problem in a 6 part series on Demographic Collapse, papers 92 to 97.

We know from the research that Dating Resilience is going down. That makes sense from all the parallel metrics showing that people just aren’t getting it on like they used to. Biology still tries to find a way, so my fear is that dropping Dating Resilience is a leading indicator for increases in aggravated intimacy, infidelity, and other generally derogatory forms of intimate contact.

AI is powerful stuff. Because people are increasingly desperate for success in dating, they will end up giving up very sensitive personal information to an AI matchmaker to increase their chances of finding someone “compatible”. That compatibility can be overrated though. Biology finds a way, and biology doesn’t always care how much money you make or any other “civilized” characteristics.

[Biology does seem to care a lot about height for men, so short men are a bit screwed. Sorry guys. Move to a country where people are shorter. I live in an old mansion that is over 100 years old. I have to bend down to navigate low passages. People 100+ years ago were shorter. A lot shorter. I would attribute this not only to better nutrition and medicine, but also to the genetic “height advantage”.]

Thus dating has always been a bit of a “numbers game” as you can never be sure if your plans will go out the window on first contact. Perhaps the strongest chemistry I ever had on OKC was with someone I had a 53% compatibility with. I absolutely was not expecting us to get along at all. She couldn’t get over her side of that incompatibility so we ended up friends, but I was really shocked at how my body was trying to override my better judgement. Don’t underestimate biology.

Of course B2.0 doesn’t understand biology. Opposites often attract, and we might find out that B2.0 creates a lot of mediocre matches that people just find boring. More importantly, the more risk that is involved in a date, the higher the dopamine hit if it goes well. That’s the risk/reward pairing that is so critical to nail in game design. Again, that’s biology. If B2.0 can be programmed to sort dates into “safe” and “risky”, I think that would be a fantastic option. People may complain about risk in dating, but off camera they are going to prefer risky dates to safe ones. Of course some people are just super risk adverse, and those people will appreciate “safe” dates. That also might be the best place to start for people who are new to dating or who are low in Dating Resilience.

It never hurts to learn to ride a bicycle with training wheels on first. If your first ride finds you pinned under a car, you may never ride again. I always recommend starting with boring dates, with people who you think are rated lower than you in the dating market. Start with the training wheels on, then take them off, and eventually work your way up to the motorcycle. Once I became a dating expert, I tended to prefer people who had a lot of riding experience. I didn’t want to ruin a novice. Now that I’m 60, if I become single again I might relax that a bit since I’m not as athletic as I used to be and I take life a bit slower.

What Bumble 2.0 Will Be Missing

AI can be used for human augmentation, or human replacement. My “north star” paper, paper number 1 on my substack, is my Third Tier of Game Development from 2010. It explained how in the future we could use AI and strong authentication systems to vastly speed up human education. It also would eliminate the “catfish” problem. Both are huge problems in the labor pool, and also in the dating pool.

While the major platforms currently developing AI told me they read Third Tier, ultimately those organisations abandoned human augmentation and went for human replacement. Since it’s a 16 year old paper, the people who actually knew about Third Tier might have aged out. That’s why I ended up writing the much darker Fourth Tier series, starting with paper 197. That explains where we are heading in our current trajectory. I’ve completed 5 papers in that series.

Since B2.0 will be sold to humans for the purpose of augmenting their dating experience, the contents of Third Tier are useful here. If B2.0 was built around that model, it would be much more useful. It’s not enough to put you in front of a person you look compatible with on paper, and just hope biology works out.

Just like in any other game (and dating is absolutely a game), the higher the difficulty level the faster you level up. Or crash out. So naturally you start off at the highest difficulty level you can handle and raise the difficulty when you are ready. Leveling up is critical because the biggest mismatches in dating are between low level people and high level people. That power asymmetry is fraught with risk for the lower level person.

You can be 50 years old and still be low level. It all depends on your experience and how challenging that experience was. My partner is much less experienced than I am, despite her being 54 years old. She trusts me not to abuse the power asymmetry, but that’s not the usual situation. From my side of that, this is a situation I will always have so I lean on my compassion.

B2.0 could estimate player levels by turning dates into quests, helping to suggest when to give or raise consents, and notify the other person when a consent has been granted. This would make dating much safer for everyone. It could also then notify the other person if a consent is removed. This is then also recorded which could be useful legally. Pairing people by level would be much more effective than pairing people by political orientation or other such characteristics.

The risk here is that the more that “Bee” does the work for you, the lower the difficulty of dating. That translates to you leveling up more slowly. Your level is the most important characteristic in dating. Or working. Or any sort of leadership position. Those romantic movies you’ve watched where Fate just made everything align, that’s the universe lowering the difficulty or raising the motivation just enough to make something legendary happen. If you are high enough level to do the right thing, history could be made. But if it happens when you are still too low level, that ship could just keep sailing and you might regret it your whole life.

I could go into a lot more detail, but that’s how I make my living. Bumble, feel free to call me any time, especially if B2.0 runs into difficulties.

What sorts of features would my readers like to see in B2.0? There could be a surge in similar apps if B2.0 is successful. Of course I’m counting on that, that’s why my subtack is about games and dating apps. But I’m also counting on a lot of failures on the road to success. Maybe we could call that Game Development Resilience. Can your game blow up, but you still get back on the bike? Your experience could be a lot better if you had an expert coach there to guide you.