Over the past few months, while building Datespots and reading The Mom Test, Thinking, Fast and Slow, and Justin Garcia’s The Intimate Animal, I’ve found myself reading more and more about game theory. Once you start looking for it, you see it everywhere: business, negotiation, friendships, relationships, even something as stupid as deciding when to text someone back.
Dating seemed like an obvious place to apply it. There are multiple people making decisions, everyone has preferences, nobody has complete information, and every decision changes what happens next. If two people like each other and broadly want the same thing, there should, in theory, be a fairly straightforward path toward an outcome both of them prefer.
But I think there is a problem with applying game theory to dating, or really to everyday life in general: you’re assuming you’re dealing with rational agents.
People aren’t particularly good at being rational agents.
Someone can know what they want and still behave in a way that makes getting it less likely. Fear does that. So do pride, jealousy, insecurity, anger and plain old confusion. Someone might walk away from something they want because they don’t want to be rejected first. Someone else might deliberately make themselves less available because they don’t want to appear too interested. We change our minds, misunderstand other people’s intentions and sometimes do things that we ourselves can’t properly explain afterward.
This isn’t some undiscovered flaw in game theory. Once you get past the introductory stuff, there are entire areas dealing with bounded rationality, incomplete information, signaling and the fact that real people aren’t perfectly rational utility-maximizing machines. What bothers me about dating is something slightly different.
You don’t know what game the other person thinks they’re playing.
You can be sitting across from someone on a third date thinking that this is going well, you clearly like each other and you’re both seeing whether it could turn into a relationship. They might be thinking the same thing. Or they might be six months out of a horrible breakup and determined not to get attached to anyone. They might be seeing three other people. They might like you but assume you’re the one who isn’t looking for anything serious. They might be dating because they’re lonely, because they’re bored, because they want sex, because their friends convinced them to download Hinge, or because they genuinely want a relationship.
Worse, they might not know which of those things is true yet.
It’s difficult to model another person’s preferences when the other person is still figuring them out.
Dating apps make the information worse
Even a match, which seems like the most basic useful signal a dating app can give you, is ambiguous.
Suppose I swipe right on very few people. Matching with someone means I looked at her profile and was genuinely interested. She might use the app completely differently: swipe fairly liberally, see who matches, and then decide who she actually wants to talk to. Neither approach is strange, but the exact same “like” means something different depending on who sent it.
Profiles aren’t much better. I don’t mean that people are necessarily lying on them. A profile is just an extremely compressed version of a person. You choose a handful of photographs from probably thousands that exist of you, answer a few prompts and add whatever other information the app allows. Someone then takes those fragments and constructs a person in their head.
Sometimes you meet and their approximation was pretty good. Other times five minutes into the date you realize the person in front of you is nothing like the person you thought you were meeting.
Attraction makes the whole thing even less predictable. You can meet someone who satisfies nearly every criterion you thought mattered to you and feel absolutely nothing. Then someone who doesn’t fit the specification particularly well comes along and suddenly the specification doesn’t seem so important.
So before anybody has started playing games, we’re already dealing with incomplete information and preferences that aren’t nearly as fixed as we’d like to believe.
Then the apps introduce another problem: there is always somewhere else to look.
You can have a genuinely good date with someone and open Hinge on the train ride home. It doesn’t have to mean the date went badly. The app is simply there, and looking costs almost nothing.
Before dating apps, meeting another plausible person required some friction. You had to go somewhere, get introduced through friends, meet through work or school, talk to a stranger at a bar, whatever. Now you can meet someone you like and, twenty minutes later, be shown another hundred people who might also be interesting.
I think that inevitably changes how we judge the person we already have in front of us. Someone can be attractive, funny, intelligent and easy to spend time with, and instead of asking whether we want to see them again, we can start wondering whether somebody slightly better is still out there.
The annoying answer is almost certainly yes. There is someone hotter. Someone funnier. Someone wealthier. Someone who lives closer. Someone who shares more of your interests. Once you divide compatibility into enough dimensions, the imaginary alternative can always beat the real person on something.
But you never actually date the entire distribution. You date one person.
The temptation to keep searching doesn’t disappear just because you’ve found someone good, and whoever you’re dating is exposed to exactly the same temptation. This is one of the stranger things the apps have done: they’ve made finding people easier while making it harder to know when you’re finished looking.
We manufacture information from noise
The texting stuff is where this becomes almost comical.
Say I’ve been talking to someone who normally replies within an hour or two. One day she takes six hours. I have almost no useful information about why, but I have plenty of ways to interpret it. Maybe she’s losing interest. Maybe I said something wrong. Maybe she’s talking to someone else. Maybe I came across as too interested yesterday.
Or maybe she had six hours of meetings.
The trouble starts when I react to my interpretation rather than what I actually know. If I decide she’s pulling away, perhaps I wait before responding too. She notices that I’m suddenly replying differently and now has her own ambiguity to explain. Maybe she decides I’m losing interest and backs off slightly. I see her backing off and feel rather pleased with myself for having spotted the change early.
We can create the thing we thought we were detecting.
This is why I’m increasingly skeptical of advice like “match their energy.” It sounds sensible until you think about what happens when the initial change in energy had nothing to do with you.
If someone’s having a terrible week at work and becomes a little less talkative, I can respond by becoming less talkative myself. Now they notice that I’m being distant. Perhaps they withdraw a little because they think I’ve lost interest, which gives me even more reason to withdraw.
If I designed a system like that at work, I’d be worried about it. I’ve got poor observability, noisy inputs and a feedback loop that can amplify an error instead of correcting it. Yet in dating we’re quite comfortable telling people to run exactly this algorithm.
The same problem appears with the idea that you shouldn’t seem too available. Someone asks if you’re free Saturday. You are, but perhaps saying yes immediately feels too eager, so you tell them Saturday is difficult and suggest Sunday instead.
They make other plans for Saturday.
Nothing went wrong. The strategy worked perfectly. You wanted to communicate that your time was scarce, and the other person believed you.
That’s what I find funny about a lot of dating strategy. Sometimes the failure isn’t that the strategy didn’t work. It’s that it worked.
Two people can want the same thing and still miss each other
Suppose two people have been seeing each other for six weeks and both would prefer to be exclusive. Neither knows the other feels that way.
He doesn’t bring it up because he doesn’t want to put pressure on her. She doesn’t bring it up because she thinks that if he were serious, he would say something. That little bit of uncertainty makes both of them slightly cautious. He notices her caution and decides he was right not to push. She notices his and takes it as confirmation that he isn’t looking for anything serious.
Eventually one of them concludes that the relationship isn’t going anywhere.
Neither person necessarily behaved irrationally. Their decisions made sense given what they thought they knew; the problem was that each was making decisions based partly on a guess about the other person’s intentions. Those guesses then changed their behavior, and the changed behavior made the original guesses look increasingly correct.
Even very ordinary interactions have this problem. Say I offer to pay on a first date and she insists on splitting the bill. What have I actually learned? She might not be interested, certainly. She might also always split dates, dislike having someone she barely knows pay for her, earn plenty of money and see no reason I should pay, or simply be trying to be polite.
It’s still very easy to go home thinking, She insisted on splitting. Probably not a great sign.
Maybe. Or she wanted to split the fucking bill.
The behavior is easy to observe. The reason for it usually isn’t.
Even explicitly telling people what you want doesn’t completely solve this. “I’m not looking for anything serious” sounds wonderfully unambiguous until you’ve spent enough time around dating to realize how many different things people mean by it. For one person it means no relationship, full stop. Someone else means not right now. Another person means they’re happy to keep seeing you but don’t want expectations attached to it. Sometimes it means they don’t want a relationship with you specifically. And occasionally someone says it in complete good faith and is in a relationship a month later because they met somebody and changed their mind.
They weren’t necessarily lying. Humans are allowed to update their preferences.
Again, terrible for the model.
There are more than two players
There’s another reason I think it’s a mistake to treat dating apps as a simple game between two people: they aren’t.
The app is part of the system too.
My objective might be to meet someone I really like and eventually delete the app. Her objective might be exactly the same. The company running the app has a much more complicated set of objectives. It wants people to have enough success that the product works, but it also has to think about subscriptions, retention, engagement, safety, marketplace health and probably a mountain of internal metrics I’ll never see.
I don’t think this requires a conspiracy theory about dating apps secretly wanting everyone to remain single. A dating app that never produced dates wouldn’t survive very long. It just means the incentives of the platform and the incentives of an individual user aren’t identical.
There are also all the other people we’re talking to, the recommendation system deciding who gets shown to whom, geography, timing and luck. Someone fantastic can download the app the day after you delete it. Someone you would have liked can be buried behind hundreds of profiles you don’t. Someone can disappear because they met somebody else before your first conversation went anywhere.
Once you include all of that, calling dating a two-player game seems almost quaint.
Knowing the strategy changes the strategy
There is also a point where dating advice becomes common knowledge.
People know confidence is attractive. They know desperation isn’t. They know having your own life is generally a good thing. They know not to over-invest immediately. They know the clichés about chasing and scarcity and being mysterious.
So naturally people learn to display those traits.
Then everyone else learns that people are deliberately displaying them.
Now you have to work out whether someone is actually relaxed or carefully trying to look relaxed, actually busy or trying not to seem available, genuinely confident or just very good at performing confidence. As people get better at understanding the signals, people also get better at manufacturing the signals.
Eventually the signal itself tells you less than you thought it did.
This is where game theory has become more useful to me, but not in the way I initially expected. I don’t think it’s particularly good at telling you how to “win” at dating. It’s much better at explaining how two normal people with decent intentions can arrive at outcomes neither of them wanted.
Chess is easy in one important respect: I know what the other person wants. They want to checkmate me.
Poker is full of deception, but I still know what everyone ultimately wants. They want the chips.
Dating doesn’t give you even that.
I don’t know exactly what you want, you don’t know exactly what I want, and there’s a reasonable chance that one or both of us will change what we want after learning more about each other. We observe each other’s behavior, but the observations are noisy. We have outside options that neither person can see. Emotion affects decisions. Previous experiences affect decisions. Sometimes we deliberately conceal information because revealing it feels risky.
Trying to turn all of that into a strategy can give you a false sense that you understand what’s happening.
That’s probably the part that bothers me most. Once you’ve built a model of someone in your head, almost anything they do can be made to fit it. A slow reply means they’re losing interest. A rescheduled date means you’re not a priority. A double text means they’re desperate. Someone showing a lot of interest must not have many other options.
All of those explanations can be true. None of them has to be.
After reading all of this, I’ve ended up with a conclusion that’s much less clever than the theory that got me there.
I think individual moves matter less than patterns.
One six-hour reply tells you almost nothing. Someone who regularly disappears for days tells you something. One cancelled date could be anything; five cancelled dates are a different matter. You don’t need to know whether someone is avoidant, strategically withholding attention, overwhelmed at work, talking to six other people or simply bad at texting to notice that you’re doing all the work.
At some point the explanation stops mattering.
And maybe that’s the useful limit of game theory in dating. It can help explain incentives, signaling, uncertainty and why people keep accidentally making things worse for themselves. What it can’t do is give you access to another person’s head.
You don’t need that access anyway.
You can be clear about what you want, behave roughly the way you actually feel, leave room for the possibility that you’ve misunderstood something, and pay attention to what happens repeatedly rather than trying to decode every individual move.
For all the complicated theory involved, that may be as far as the model gets you.