They say money can’t buy love. But that was the bad old days of fiat money. Now there are dozens of love-based cryptocurrencies – LoveCoin, CupidCoin, Erosium, Nubilo – with market caps in the mid nine-figures. The 17-year-old genius behind CupidCoin just bought the state of Tennessee. You think I’m joking, but can you be sure? How weird is “too weird to be true” these days, and how confident are you in your answer?
Case in point: Luna, which bills itself as blockchain-optimized dating. They caught my attention by hiring Aella, previously featured on this blog for her adventures taking LSD megadoses weekly for a year. They kept it with their cutesy story about how the name “Luna” comes from founder Andre Ornish’s first word – adorable, until you consider that any baby whose first word is in Latin is definitely possessed. And they maintained it because – well, goodness knows we need new dating sites now that OKCupid has devolved into an off-brand Tinder clone. So let’s look through the white paper and see what they’ve got.
Most dating sites suffer from attention imbalance: men scrounge around for anyone willing to acknowledge their existence; women get inundated with countless desperate messages they don’t want. Luna solves this by making attention a commodity tradeable on the free market. Users who want to catch someone else’s attention can bid the local cryptocurrency, Stars, to get their message to the top of another user’s queue; all Stars spent in this way go to the user receiving the message. Stars can be bought with dollars and vice versa, so popular users can actually earn money reading all the messages sent to them.
This system has some pretty powerful advantages. Market forces are the known solution to the problem of connecting resources to their highest-value use. So if you treat user attention as a resource you can trust the market to allocate it optimally – in this case, to the guy who’s just realized he’s your soulmate, rather than the guy who’s spamming everyone with five dick pics.
But everywhere this solution is tried, it runs up against its one great weakness – rich people with mild preferences can outbid poor people with strong ones. I can’t predict how this particular market will clear, but this doesn’t necessarily have to be a big problem here. If the market rate for a certain user’s messages were $1, then even the poorest person could afford to send a message to a potential soulmate. And even a well-off person might hesitate to send out a hundred messages a day, every day. And if he didn’t? Well, getting paid $100/day to read messages on a dating site doesn’t sound like the worst outcome. Also, really good information about preferences in exchange for a biased system that favors the wealthy has been the deal Capitalism has been offering since Adam Smith first put quill to paper; it seems kind of weird to back out now.
A more practical issue: how long before someone finds a photo of a supermodel, limits their profile to “I AM A NYMPHOMANIAC”, and watches everyone trip over themselves to send paid messages? Luna alludes to vague plans to “verify” profiles, which could mean anything from “you have to Photoshop a picture halfway convincingly” to “you have to get an actual pretty girl to help with your scam”. Neither of these seem like too high a bar. Better is their offer to provide data, including how often users respond to messages and how often users meet with other users:
When choosing to attach Stars to his message, Bob should receive information such as the number of unread messages in Alice’s queue, an internally calculated reply quality indicator, and confirmation on whether Alice’s account is verified.
I’m still skeptical. I have bots pretending to be pretty women try to friend me on Facebook something like once a week, even though I have no idea what their endgame is or how this results in them making money. If Luna gives a real incentive for the scam, they’re going to have to beat Facebook pretty handily if they want to succeed here.
More promising than any individual claim they make about how they’re going to fix things, is their claim that they’re incentivized to fix them. OKCupid famously wrote about Why You Should Never Pay For Online Dating, the answer being that it incentivizes dating sites to keep you single – after all, the longer you’re single, the longer you’ll keeping spending money on dating sites. Even if that sounds a little cartoon-villainish, at the very least it doesn’t incentivize sites to do a good job matching you up. Luna claims that their model gives them a profit only when it succeeds:
At Luna, we intend to structure the token economy in such a way that our system is rewarded when users achieve their goals, thus aligning our own incentives with those of our users and ensuring that all data, AI, and machine learning technology will be used to actually connect people…the approach consists of two parts:
1. Fees which comprise Luna’s revenue only occur in the case of successful communication. As described in 3.1, when a user receives and reads a message boosted with Stars, they also receive the Stars used to boost that message. Luna intends to take a small fee for this transaction, but only if the recipient responds to the message within a window of a number of days yet to be determined. If the recipient does not respond, or only responds after more than this number of days, this fee will be re-paid to the sender. The number of Stars transferred to the recipient, however, will remain the same, whether they respond to the message or not. In this way Luna’s financial incentives will be aligned with users’ goals at Stage IV in the exchanging of messages.
2. Possibility of tipping in case of successful offline dates. Another way to provide incentive for Luna to help achieve its users’ goals is to allow users to tip the platform after the achievement of Stage V in the completion of a successful date. As described in 3.2.4, we intend to make feedback polls available after dates. Once users have rated their experience, Luna will then allow them to choose whether to leave a tip of their choice in the form of Stars. As this is a voluntary option, it should have no effect on user feedback. Tipping a platform is an infeasible idea in the context of currently existing dating apps; however, the free and direct-to-user benefits of Luna may register to users as something more resembling the mechanisms of Wikipedia: a free, friendly, and user-contributed service, rather than a platform like Match.com, which can feel exploitative. A tipping option may thus encourage a feeling of alliance with Luna in the user.
In this way, rather than recreating disparities which exist between the goals of current dating platforms and their users, Luna’s financial incentives and user goals will coincide.
I can imagine all sorts of horrible misalignments between maximizing-number-of-responded-to-messages and maximizing user satisfaction, but for now I’ll just admit this seems nice and I appreciate the effort.
Luna’s last major promise is to use cutting-edge machine-learning techniques to come up with a good match algorithm:
Despite significant technological advances in information processing, storage, and retrieval, online dating has yet to optimally integrate machine learning for the user’s benefit. A typical ML task for online dating might be to predict the level of compatibility between two users from a given set of input data, thus predicting for example whether one user is likely to respond to another user’s message […]
Luna may adopt a collaborative filtering algorithm developed by Dr. Kang Zhao. In addition, Luna may use advanced NLP techniques in conjunction with IBM Watson to integrate additional information from the contents of messages sent in-app, as well as from social media sources such as Twitter, if users choose to provide that information.
They’re right that this seems like a perfect use case for machine learning techniques, and that this possibility has been woefully underexplored. With all due respect to OKCupid’s excellent match algorithm (my girlfriend played with it one day and found that her two highest matches in the entire world were me and her ex), there’s a lot of room for improvement here. I find the idea of letting users link their social media accounts to provide more data really fascinating, and this reassures me that the attempts at incentive-alignment above really do have them thinking about how they can do better.
One part of the white paper I still don’t understand: why is it on a blockchain? They write:
For us to ever find out [how to design a match algorithm that really increases human happiness], we are going to require an open data ecosystem around computer dating. Blockchain is an integral part of that – it’s what pays the bills to do the science and, in the case of Luna, it nicely and accurately solves one of the key problems in the computer dating arena: cut-and-paste messages spammed over huge numbers of people, resulting in an ever-lower number of good quality genuinely interested messages, hidden in an ever larger sea of dating spam. Just getting rid of that dynamic once and for all would be a great result, but I think that Luna offers far, far more.
By establishing the decentralized paradigm in dating, Luna helps to remake dating culture. Luna is not a service or a place, like Tinder or a bar. Luna is a method, and a method which can be continually improved using techniques like A/B testing, until it is genuinely producing better lives for people. Because blockchain techniques allow for sophisticated tools to be developed to align economic interests between (say) search algorithm designers and individual users, or between users and other users who don’t like spam (i.e. everybody), the possibility exists to not only solve the questionable agency of the current generation of dating app providers, but to create positive agency to do something really, really new. We could pay the best people in the world to design algorithms to match other people, and make them happy.
I don’t know what they mean by “open data ecosystem”. I assume they’re not going to let everybody see all the data – I don’t necessarily want my parents to be able to know who I just sent a message to. But if not, how does “open data” help other people design match engines? Why is their crypto token more efficient than paying for Second Life in Linden Dollars, or any of the other silly token currencies that have existed forever on the Internet?
So what is blockchain doing for them? The null hypothesis might be: the same thing it does for Long Island Blockchain Tea and half a million other scams. I have hopes that Luna isn’t a scam. It employs some people I know and trust. The incentives and economics of their product seem very well-thought-through – so much so that if it’s a scam somebody else ought to be doing the same thing for real. And they’re promoting GiveWell and other effective charities in a way that suggests they want a lot of the money to go to altruism anyway. These are…signs. But I’ve been told you can’t be too paranoid in this area these days.
But I really do hope Luna isn’t a scam. Because if it’s real, it represents everything good about Silicon Valley. Some people use Intellect to wrest a secret from Nature: an elegant reduction of the chaos of human interaction into comprehensible and exploitable principles. To test their prize they build a Sampo, a machine churning out a hundred varieties of human happiness – from loving marriages to ecstatic sex to just sitting on the couch cuddling on rainy days. They give it to the public gratis. In the process they all get super rich and donate the money to curing malaria, good compounding upon good. Also, the whole thing is done in a weird and pointlessly-complicated format that adds nothing except a giant middle finger aimed at government regulators. What could be more beautiful than this?
One last thought on the blockchain issue: whenever I study intentional communities, I’m struck by how little the community’s principles matter, compared to the brute fact of it being an intentional community. An anarchist commune may have some spectacularly brilliant collaborative dispute-solving mechanism, but none of that matters, because the people involved will be the sorts of people who would join an anarchist commune, ie ridiculous and completely ungovernable.
So the most interesting and distinguishing feature of Luna, at least to start with, might not be the tokens, or the incentives, or the machine learning. It might be that it’s a place you can go to meet the sort of people who want to date on the blockchain. I could make a lot of cheap jokes here, but whatever weird hyperplanes through categoryspace further the difficult and desperate project of human-seeking-human are good and worthwhile in my book. I hope that lots of libertarian women find lots of security-conscious men and make lots of beautiful, high-price-volatility babies.