Designing a better dating market

For fun and no profit

I got nerd-sniped into thinking about dating apps and matching markets despite zero experience with them¹. For something that gets discussed so much, I was surprised at how few people apply an economic lens to this stuff. For fun, let’s see if we can use economics and mechanism design to improve dating².

Matching market solution

Fortunately, computer scientists solved love back in the 1960’s. As long as everyone in the dating pool is able to rank potential partners, we can run a stable marriage algorithm on a regular basis to pair them up.

There’s a couple things we need to get right to make this work for real-world dating. One problem is that people entering the match system might be looking for different kinds of relationships (e.g. casual sex vs long term partnership). One fix is to have separate matching pools for different types of matches; charging larger fees for higher-commitment pools should keep out uncommitted suitors.

Another problem is that people might not like who they’re matched with and never meet them in person. It’s a perfectly reasonable choice, but it messes with the mechanism, which operates on the assumption that people actually spend time with their match. To see why this is a problem, imagine that Bob participates in the mechanism many times but never goes on any dates, this is relevant information for the people ranking Bob; they ought to know that he’s often unavailable so they can put a higher priority on someone that will actually go out with them. A simple solution would be to verify whether or not a user goes out with the person they’re matched with and display a metric for how “flakey” they are. I’ll suggest some ways to do this later.

But the biggest problem with matching algorithms is that they require participants to rank lots of people. In order for you to rank accurately, everyone needs to post a detailed profile, since its infeasible to meet everyone in person. This would have to be a prerequisite for participation. Regardless, as the number of individuals gets large, players may only be able to rank a handful of people. For these reasons, we have to keep the number of participants in each pool small, so they may have to serve specific geographic areas or age ranges.

A auction-based solution

Stable marriage algorithms are all well and good, but we live in a late-stage capitalist society, surely we can make our dating market more like an actual market?

That’s easy: have everyone post details about themselves and the kind of relationship they’re looking for, and then let others place bids to go on a date with them. The person receiving bids decides which suitor wins the auction based on a combination of the size of the bid and how attractive of a partner they appear to be.

We can go even further and allow people to propose contracts with various conditions such as the length of the relationship, relationship details, and compensation. Participants in such markets will often need to hire a matchmaker to find partners with mutually-agreeable contract terms³.

In practice, it would be hard to get users for such an auction and the similarity to prostitution would likely make it illegal. Sigh. One non-monetary approach would be to “bid” a certain number of future matches (or a short ban from the market) in order to signal interest. These are interesting possibilities, but auctions and contracts are overly complicated solutions to the problem of finding love.

Modifying existing apps

So none of the previous solutions were realistic, but we can at least think about how to set up the payment structure of dating apps to align users and company incentives. There are a lot of different goals a user might have on the dating market, and each requires a different payment system.

For casual dating, users aren’t looking for a long-term relationship and instead want to go on as many “dates” as possible. Ideally, the service charges the user per date, which incentivizes the company to set up as many dates as possible and the user pays for the dates they actually go on.

If the user instead cares about how quickly they get a date, the service can charge the user up front and return some of their money the longer it takes to find a date. That way the service wants to get them matched up as soon as possible.

If users are looking for a life partner, they should pay the company a large sum up front to use the service indefinitely. At that point, the company is incentivized to pair them up with a permanent partner so that they don’t come back.

The incentive schemes get more complicated if users have more specific needs. For example, if a user wants a lot of matches in a short period of time, some combination of an shrinking deposit and a pay-per-match system will be necessary⁴.

Other useful things

There are a couple of tools that would help many of these schemes work better.

Commitment sorting: It’s important to put people with similar relationship goals into the same matching market. Participating in these markets also has to credibly signal that you want that kind of relationship. If someone looking for sex can easily join a market for long term partnership it can create a lot of confusion. Charging a high fee to enter long-term dating pools can effectively filter participants. In addition, explicitly advertising dating pools for particular relationships or communities can create Schelling points for people⁵.

Complicating matters is that people often aren’t sure what kind of relationship they want and will participate in multiple dating pools⁶. Additionally, for things like casual sex, people might be too embarassed to affiliate with such markets, which will have to brand themselves in such a way as to give users plausible deniability.

Matchmaking: Better matchmaking systems can help people search through the possibilities more effectively. Fortunately, matchmaking doesn’t seem that hard, people generally just want a partner that is similar to them along a lot of dimensions, perhaps with some complimentary personality traits. Using behavioral analytics, a handful of experiments, AI, prediction markets, and genetic screening I think the matchmaking problem can essentially be solved⁷.

Verifying in-person interaction: This is crucial for schemes like stable matching that rely on people actually meeting each other.

One way to do this is to simply ask each person whether they went out. If Bob flakes on someone, they can rat on him as a punishment⁸. However, some pairs will come to a mutual agreement not to go out and falsely report that they did so in order to save their reputation.

You could also try to use peoples phones to verify whether two people actually met in person. Using geolocation data might work, but this comes with privacy issues. Another way to do this is by using a dynamic QR code on your phone; your date has to scan it with their phone in real time to prove that you actually met. It’s a neat idea, but I expect people will find a way to get around it.

Identity and reputation systems: If people can easily make new, fake profiles it opens up the dating market to bots, influencers, and scammers. Having a way to check that someone is a real person and attach a reputation to them can limit this behavior.

Event organization: Going to events where you can meet people slightly outside your social circle is a good way to find partners. The community you two share limits any bad behavior and ensures that you have something in common. A system to organize large-ish gatherings amongst friends or hobbyist communities would have the additional benefit of forming more relationships.

Conclusion

From the outside, it doesn’t seem like dating apps are that bad. Lots of people enjoy using them and many find love. I assume most of the negativity online comes from a handful of complainers⁹.

However, it’s clear that there are a lot of opportunities for improvement, at least in theory. I assume better systems will gain traction eventually, but the real challenge comes from making a profit and getting a critical mass of users.

These ideas aren’t as important or actionable as other things I’ve written about, but dating markets offer an interesting environment to think about mechanism design and economics.

Further Reading

Blogs:

Noble Matching - Marginal REVOLUTION

A Dynamic Theory of Romantic Choice | by Yichuan Wang | Medium

Towards Better Online Dating

How game theory improves dating apps

Why dating apps suck. - by Tomas McIntee

Stable Marriages and Designing Markets – Math ∩ Programming

Papers:

Double Matching Under Complementary Preferences

Price of Anarchy in Algorithmic Matching of Romantic Partners

Cupid’s Invisible Hand: Social Surplus and Identification in Matching Models

Matching with Transfers: The Economics of Love and Marriage

When Online Dating Meets Nash Social Welfare: Achieving Efficiency and Fairness

Pricing Strategies for Online Dating Platforms

Designing Approximately Optimal Search on Matching Platforms

Platform Design in Matching Markets: A Two-Sided Assortment Optimization Approach

The Mathematics of Love: Patterns, Proofs, and the Search for the Ultimate Equation

Thickness and Information in Dynamic Matching Markets

1. Also interesting because similar solutions can be used to form friendships.

2. On the off chance you want to turn one of these ideas into a dating app, remember that clever mechanisms can only get you so far, the real problem is not having users and the sacrifices needed to scale.

3. Or use prediction markets to find matches.

4. A more fine-grained approach is to look at how many times two people hang out and charge based on that. Users can target different relationship goals while paying an increasing amount per hangout. It’s tricky to design a mechanism where users honestly report how much they hang out if they’re being charged for it.

5. Though some combinations of preferences may be too niche to form a large-enough market. Would there be enough people to participate in a matching pool for artists to find a muse (someone they find physically and intellectually stimulating)?

6. Instead of self-selecting into dating pools, people could be grouped based on the number of previous matches.

7. Your social network could also help by suggesting dates for you. A match is formed if enough people like the pairing and early supporters are rewarded for successful matches.

8. Though to prevent greifing and avoid the difficulty of figuring out who messed up, both parties would have to get a bad mark.

9. Along with some legitimate issues from adverse selection, moral hazard, and a skewed gender ratio. Limiting the number of times a person can participate in a matching market could fix some of these problems.