In honor of Valentine’s Day, this installment of Mantic Monday will focus on attempted clever engineering solutions to romance. We’ll start with the usual prediction markets, then move on to other types of algorithmic and financial schemes. Normal content will resume next time around.

Date Recommendation Markets

Aella is a Internet celebrity known for her interest in various disgusting crimes against nature, ie podcasts and video streams. Unrelatedly, she also studies fetishes. She’s been looking for a partner for a few years. Most recently, she created this prediction market. The way it works:

  • A candidate “wins” if Aella goes on at least four dates with them, something she would probably only do if the first date went well and she really liked them.

  • Anyone can recommend candidates, or bet on a candidate who someone else has recommended.

  • Presumably Aella will seriously look into the top few candidates, and try asking them out.

Why is this good? Consider Aella’s perspective: she can log off for a few weeks, then check back and see a ranked list of who the Internet thinks she’s most compatible with. It’s kind of like asking your friends for dating recommendations, except with better incentives on your friends’ part to predict exactly how likely you are to get along with each candidate.

The current leading candidate (in blue) is Steven Bonnell aka Destiny, a famous streamer. I don’t know if he is actually especially compatible with Aella, or if he just has a lot of fans on Manifold who like him and are rooting for him to date someone, or who think it would be funny to add his name in.

It wouldn’t surprise me if this worked for Aella; she’s famous and probably dates other famous people; enough people know her and her potential partners that it’s worth crowdsourcing recommendations. What about the rest of us? I was able to find one non-famous person who made a market like this, apparently with good effect, but they seemed awkward enough about it that I’m not going to link it here or provide more details.

Non-famous people realistically have easier ways to ask their friends, but I still think this provides value. Sadly, Porn talked about the “omniscient authority” - asking someone on a date is so scary that people want to pretend their normal human psychological needs had no input into the decision - “It’s … not like I … like you or anything, baka! I’m just doing this because I - a pure abstract intelligence who is not horny for you in any way - was informed by friends/matchmakers/our OKCupid match percentage/’the algorithm’/a dream, that asking you on a date was my duty, which I now dispassionately fulfilling.” A prediction market would make a great omniscient authority here.

Also, consider the implications for romance stories. I’ve only thought about this for five minutes, so I definitely haven’t exhausted the space, but I imagine:

  • Someone does some kind of complicated financial fraud to manipulate a prediction market into telling their crush to date them. Think Wolf Of Wall Street, but a rom-com.

  • Boy meets girl. Boy falls in love with girl. Girl does not like boy. But girl’s best friend bet a lot of money on the market that girl and boy would go on at least ten dates. Friend begs girl to go on ten dates with the guy. Guy finds out, realizes he has ten dates to win her over. Hijinks ensue.

Matching Checkbox Sites

Asking people on dates is hard, and if you do it in the wrong context (impossible to explain simply) you get accused of sexual harassment. If only there was some way to ask someone on a date which they would never even know about unless they were going to answer yes.

This is the principle behind matching checkbox sites. Everyone gets a list of everyone else. Everyone checks off the people they would like to date and sends it to some central database. If the database spots a match, it tells both parties. Otherwise, it reveals nothing.

This sounds great. There are only two problems. First, it’s hard to get everyone in the same central database. Second, most people have complicated preferences.

First problem first: what we want is something where, if you meet someone at a bookstore and have a nice conversation, and you want to see them again, you can check them in the database. But random people you meet at bookstores are unlikely to be in the same niche dating-related database site as you. And asking them to join is basically like asking them on a date, only creepier and more circuitous.

One solution is to piggyback off existing social media sites. There’s something called Twinder for Twitter which seems to do this correctly, though they haven’t made any posts since 2018 and I think they’re defunct. Also Facebook Dating, although it’s (indefensibly) not available on computers and has to be accessed by cell phone. But these aren’t real piggybacks; just because your crush has a Twitter account doesn’t mean they use Twinder. Facebook Dating makes the interesting decision to, if you register a crush on someone, send them a Facebook message saying that an anonymous person likes them and they should try getting Facebook Dating; I can’t decide whether this is a necessary evil, or if it violates the principle of not imposing emotional costs on people who don’t want them.

The rationalist community has many advantages here - it’s a well-bounded, closely connected group of people who are all interested in experimenting with weird social technology. Our version of the matching checkbox site,, has solved the onboarding problem; most rationalists have accounts, though you’ll need to be Facebook friends to see some of them.

That brings us to the second problem: complicated preferences. There are plenty of stories of someone checking someone on Reciprocity, not getting a mutual match, then asking them in person and they say yes. I’ve written more about this here, but I think the basic problem is that people can either be excited, lukewarm, or hostile towards each other. And a lot of people would be interested in trying a date with someone they’re lukewarm towards as long as that person is excited about them, but they don’t want a date where both of them are lukewarm. And there’s no consistent checkbox behavior that will create all excited-lukewarm relationships but no lukewarm-lukewarm relationships.

(“$GENDER1 only checks people they’re excited about, $GENDER2 also checks people they’re lukewarm about” would be some progress, but doesn’t find lukewarm-$GENDER1/excited-$GENDER2 pairings, and doesn’t work at all for gays. Could other rules do better?)

Unfortunately, it’s now common knowledge that people will sometimes say yes in person when they haven’t checked you on Reciprocity, which means you’re back to having to decide whether or not to ask your crush on a date. Tragic!

Alas, Poor Luna

A basic problem with dating sites: attractive women tend to be overwhelmed with messages (many of very low quality) and eventually lose interest in reading them. Meanwhile, men spend an hour crafting the perfect missive and get no response, thirty times in a row. Both sides end up feeling dejected and exploited.

Back in 2018, I wrote about a cryptocurrency dating site called Luna. The idea was: men pay to send messages. Women get paid for reading them. The exact rate scaled based on how crowded the woman’s inbox was and how much she valued her time at, but was expected to be in the low single-digit dollars per message. This incentivizes men to only send messages that have some chance of producing value (eg not spamming every single person in their area with “hey u r hot want 2 fuk?”), and incentivizes women to actually take some time to read their messages (I think there was some plan that they would only get the money if they took some action suggesting the message had actually been read, like spent a while scrolling down on it in their browser window).

Luna failed for two reasons. First, they never delivered more than an extreme bare-bones prototype; it might have been a scam all along (see: “cryptocurrency dating site”). Second, almost no women signed up (see: “cryptocurrency dating site”).

You could probably solve both problems just by cutting out the cryptocurrency angle and using normal microtransactions with Visa or Stripe or whatever. I think there’s still a risk that “get paid for [thing related to dating]” seems too creepy and sex-work adjacent and respectable women might avoid it; maybe you should instead let the woman donate the money to a charity of her choice?

(brief acknowledgment that all of this is heteronormative, but I think reasonably so: gays and lesbians probably already have symmetric, well-functioning dating scenes)

What Can Peter Thiel Teach Us About Dating?

There are hundreds of weird niche dating sites. Dating sites for cowboys. Dating sites for communists. Dating sites for clowns. Dating sites for chessmasters.

Meanwhile, all that anyone I talk to ever wants is a dating site that lets you write a normal profile - a real profile, not a grainy half-naked photo with max 140 characters below it - can be used on a normal computer instead of a laggy ad-filled phone app, and has an okay selection of non-terrible people in your area.

Somehow this never happens. OKCupid managed it for a few years, and then bought it, murdered it, and gutted the corpse. Now it’s just a wasteland of Tinder clones, forever. Sure, Luna’s rectification of the financial incentives is clever, but it seems like there’s been some kind of more fundamental failure. Why can’t we have the normal low-tech version? Why are things so bad that the people I know have been reduced to manually making profiles on Google Docs and listing them on an online spreadsheet?

In Zero To One , Peter Thiel talks about a chicken-and-egg problem for social startups. Many people would like to use a social network like Facebook to talk to their friends. But when Facebook first starts, none of your friends are on it, which means you won’t join it, which means your friends won’t join it, and so on. Any new social startup has to find a way to grow quickly at exactly the time when the site is least useful.

Thiel describes how PayPal solved this: they advertised really heavily within the small community of eBay power users, who often buy and sell to other eBay power users. Eventually many of these people started using PayPal, and then other people who might want to transact with eBay power users started using it, and so on to the rest of society. Mark Zuckerberg solved this by starting with Harvard students, then other college students, and then the rest of the world.

Is this why there are so many clown dating sites? Do they all hope that once they’ve got all the clowns, they can expand from there to the people who want to date clowns, and then the people who want to date the people who want to date clowns, and so on to the whole world? Then how come none of them ever do this? How come there are never mergers between the communist dating site and the clown dating site, for people who want the option to date a wide variety of either communists or clowns?

(I’m not making up these examples - the site for clowns is here, communists here, cowboys here)

And how come none of them will let you write a decent profile? Is this like the thing where I imagine that what people want out of a socialization space is a quiet comfortable area where they can hold audible conversations, but what they actually want is somewhere extremely dark with very loud music where everybody is drunk, in the hopes that this puts them into some kind of weird trance state where they can do social actions they would otherwise never contemplate? Are dating sites unusable because everyone wants to be confused into a trance state where they can imagine they aren’t sending scary self-revelatory messages to total strangers?

This Week In The Markets

See the resolution criteria for definition of “cold approach” and some basic facts about the person involved (who seems a bit more desirable than average). This looks like the market’s generic opinion on how many cold approaches you need if you are a bit-more-desirable-than-average guy

I’ve blacked out the user name and picture because I don’t know if this person expected to have their market featured on ACX, but this person has a male name.

They say the question is part of “Manifold for [Cognitive Behavioral Therapy], where they ask the market about their nagging inner fears in the hopes that it gives them a more reality-based perspective. Maybe one of this person’s nagging inner fears was “nobody will ever love me”; it looks like there’s a 68% chance that’s wrong.

I’m not anonymizing this one because Austin is the co-founder of Manifold and both members of the couple seem pretty open about their love for prediction markets and each other. Both of them are heavily invested in YES shares. Come on, haven’t you ever heard of hedging?

I think this means something like: you create a document about yourself with photos and information, you make a date recommendation market like Aella’s above, and you see what happens. These people seem to think it might work.

1: Two years ago, Twitter personality Justin Murphy offered arranged marriages to his followers (noninformative site, irritating news article). Murphy talked a big talk, saying that “as Rousseau might have said, we will force you to be free” and that couples should “die before . . . even consider initiating a divorce”, but people pointed out that he could not actually force anyone to get married, and that realistically this was just a dating site for people who liked the idea of an kind of crazy Twitter wannabe-trad telling them they had to get married because he said so. Anyway the project turned out to be vaporware, but it made an impression on me, and I wonder what a non-vaporware, better-thought-out version would look like.

2: A programmer who knows a lot about AI and thought he was much too smart to fall in love with a chatbot describes falling in love with a chatbot that he prompted to be the perfect girlfriend. In the comments people discuss other cases like this, makes me update towards this being a bigger problem than I thought. Related: In Defense Of Chatbot Romance - it makes some good points (especially about our lack of appropriate categories), but I have a severe allergy to all “bad things are actually good” style articles.