For the past months, my co-founder and I have been working on a matchmaking algorithm that we’re pretty excited about. It targets the following issue:

Dating sucks for many reasons, but one particular reason stands out. Imagine if today’s dating apps didn’t include the person’s gender. On half of your dates, you’d discover immediately you weren’t compatible because the other person wasn’t your preferred gender, and go home dejected having wasted another night on a pointless date.

We believe something similar to this is happening on most dating apps today, but for things harder to deduce: personality and values. Current available solutions struggle with both, so most of your dates will be with people who you just aren’t compatible with.

Unlike gender, online dating apps struggle to feature this information, and people are reduced to picking up faint signals from a few lines of text and a picture.

Dating apps partially solve the separate issue of value-alignment by targeting a certain segment of users who are more likely to be value compatible with each other than the baseline. There’s apps for christians, lgbt+, and even racists.

OkCupid was special because instead of selection effects, it attacked the value-alignment problem through an abundance of information on each user. Hinge attempts to tackle this problem by employing a recommender system ("youtube algorithm").

While these approaches improve the status quo, they far from completely solve the issue of incompatible values and are still barely selecting for personality.

Personal matchmakers and dating docs

Selecting on personality requires high bandwidth, and the intangible information that multiple choice questions cannot get at. Luckily, there are some options that do this.

Personal matchmakers have lots of information about the people they’re setting up, often also about personality. Matchmakers who are good at their job are very effective at filtering away bad dates based on hard-to-quantify intangible deal breakers that dating apps fail to spot. They are also very expensive.

Dating docs are a high effort attack at these problems. It has high bandwidth through free-form text with the intention of communicating who you are, along with powerful selection effects. As a result, they are able to send a lot of signal to potential partners about the sender.

But dating docs have numerous issues. Most importantly they require you to publicly disclose a lot of information about yourself, often deeply personal things, that you might not want the whole world to know. And don’t even mention how much effort it is to figure out what to communicate in the first place.

Our solution

We have built an algorithm that we think finds the same intangible signals that you send in dating docs and that personal matchmakers read. We create an internal profile based on these signals, and to date, our backtesting suggests that it is incredibly effective at reducing the amount of dates with people of an “incompatible gender” (personality) resulting in high quality matches and dates.

What are the components that make up this profile? Well, that’s the entirety of our alpha. As a hint, when signing up you’ll be asked questions such as:

How orderly is your home right now? (spoiler, the signal we’re after with this question isn’t whether your room is clean or not)

We are now opening Eeva to our first set of users. Unlike most dating apps, you’ll need to submit much more information than typical, as this is what enables us to do high quality matching.

Unlike dating docs, your profile is completely private.

Unlike matchmakers, it won’t cost tens of thousands of dollars.

Unlike both, it will have an easier time reaching compatible people outside your community.

Based on the information, we will suggest high quality dates when our algorithm has found a match, and provide an introduction if both parties show interest.

Sign up today, and we’ll find your partner

With what we have now, we are confident in our ability to put together high quality dates, and we're now looking for our first users. For every 100 users, we believe we’ll be able to create 3-7 high quality matches (depending on the exact split of gender, location, etc.).

As we are currently most interested in proving to investors that our method works, it is completely free to sign up.

We think the LessWrong-adjacent crowd is a great first test market, as many of us are already used to and comfortable with both the concept of dating docs and the idea of putting in effort to achieve something important.

Sign up here: eeva.site

We are immensely grateful to anyone taking a chance on us, and we hope to make it worth your while. With your help we’ll make the best (and the first) personality matching algorithm of all time, hopefully solving things such as: Dating, finding friends and more.

Thanks to MathiasKB for helping a ton with the post