If one is able to improve how people are matched, it would bring about a huge amount of utility for the entire world.
People would be happier, they would be more productive, there would be less of the divorce-related waste. Being in a happy couple also means you are less distracted by conflict in the house, which leads to people better able to develop themselves and achieve their personal goals. You can keep adding to the direct benefits of being in a good pairing versus a bad pairing.
But it doesn’t stop there. If we accept that better matched parents raise their children better, then you are looking at a huge improvement in the psychological health of the next generation of humans. And well-raised humans are more likely to match better with each other...
Under this light, it strikes me as vastly suboptimal that people today will get married to the best option available in their immediate environment when they reach the right age.
The cutting-edge online dating sites base their suggestions on a very limited list of questions. But each of us outputs huge amounts of data, many of them available through APIs on the web. Favourite books, movies, sleep patterns, browsing history, work history, health data, and so much more. We should be using that data to form good hypotheses on how to better match people. I’m actually shocked at the underinvestment in this area as a legitimate altruistic cause.
If an altruistic group of numbers-inclined people was to start working together to improve the world in a non-existential risk reducing kind of way, it strikes me that a dating site may be a fantastic thing to try. On the off-chance it actually produces real results, Applied Rationality will also have a great story of how it improved the world. And, you know, it might even make money.
There seem to be perverse incentives in the dating industry. Most obviously: if you successfully create a forever-happy couple, you have lost your customers; but if you make people date many promissingly-looking-yet-disappointing partners, they will keep returning to your site.
Actualy, maybe your customers are completely hypocritical about their goals: maybe “finding a true love” is their official goal, but what they really want is plausible deniability for fucking dozens of attractive strangers while pretending to search for the perfect soulmate. You could create a website which displays the best one or two matches, instead of hundreds of recommendations, and despite having higher success rate for people who try it, most people will probably be unimpressed and give you some bullshit excuses if you ask them.
Also, if people are delusional about their “sexual market value”, you probably won’t make money by trying to fix their delusions. They will be offended by the types of “ordinary” people you offer them as their best matches, when the competing website offers them Prince Charming (whose real goal is to maximize his number of one night stands) or Princess Charming (who really is a prostitute using the website to find potential clients). They will look at the photos and profiles from your website, and from the competing website, and then decide your website isn’t even worth trying. They may also post an offended blog review, and you bet it will be popular on social networks.
So you probably would need to do this as a non-profit philantropic activity.
EDIT: I have an idea about how to remove the perverse incentives, but it requires a lot of trust in users. Make them pay if they have a happy relationship. For example if the website finds you a date, set a regular payment of $5 each month for the next 10 years; if the relationship breaks, cancel the payment. The value of a good relationship is higher than $5 a month, but the total payment of $600 could be enough for the website.
That sounds a lot like really wanting a soulmate and an open relationship.
That’s a nice thing to have; I am not judging anyone. Just thinking how that would influence the dating website algorithm, marketing, and the utility this whole project would create.
If some people say they want X but they actually want Y… however other people say they want X and they mean it… and the algorithm matches them together because the other characteristics match, at the end they may be still unsatisfied (if one of these groups is a small minority, they will be disappointed repeatedly). This could possibly be fixed by an algorithm smart enough that it could somehow detect which option it is, and only match people who want the same thing (whichever of X or Y it is).
If there are many people who say they want X but really want Y, how will you advertise the website? Probably by playing along and describing your website mostly as a site for X, but providing obvious hints that Y is also possible and frequent there. Alternatively, by describing your website as a site for X, but writing “independent” blog articles and comments describing how well it actually works for Y. (What is the chance that this actually is what dating sites are already doing, and the only complaining people are the nerds who don’t understand the real rules?)
Maybe there is a market in explicitly supporting open relationships. (Especially if you start in the Bay Area.) By removing some hypocrisy, the matching could be made more efficient—you could ask questions which you otherwise couldn’t, e.g. “how many % of your time would you prefer to spend with this partner?”.
I wouldn’t jump to malice so fast when incompetence suffices as an explanation. Nobody has actually done the proper research. The current sites have found a local maxima and are happy to extract value there. Google got huge by getting people off the site fast when everyone else was building portals.
You will of course get lots of delusionals, and lots of people damaged enough that they are unmatchable anyway. You can’t help everybody. But also the point is to improve the result they would otherwise have had. Delusional people do end up finding a match in general, so you just have to improve that to have a win. Perhaps you can fix the incentive by getting paid for the duration of the resulting relationship. (and that has issues by itself, but that’s a long conversation)
I don’t think the philanthropic angle will help, though having altruistic investors who aren’t looking for immediate maximisation of investment is probably a must, as a lot of this is pure research.
You could create a website which displays the best one or two matches, instead of hundreds of recommendations, and despite having higher success rate for people who try it, most people will probably be unimpressed and give you some bullshit excuses if you ask them.
I think that’s the business model of eharmony and they seem to be doing well.
I wonder to what extent the problems you describe (divorces, conflict, etc) are caused mainly by poor matching of the people having the problems, and to what extent they are caused by the people having poor relationship (or other) skills, relatively regardless of how well matched they are with their partner? For example, it could be that someone is only a little bit less likely to have dramatic arguments with their “ideal match” than with a random partner—they just happen to be an argumentative person or haven’t figured out better ways of resolving disagreements.
I wonder to what extent the problems you describe (divorces, conflict, etc) are caused mainly by poor matching of the people having the problems, and to what extent they are caused by the people having poor relationship (or other) skills, relatively regardless of how well matched they are with their partner?
Well, the success of arranged marriages in cultures that practice them suggests the “right match” isn’t that important.
What makes you think these marriages are successful? Low divorce rates are not good evidence in places where divorce is often impractical.
Three main points in favor of arranged marriages that I’m aware of:
The marriages are generally arranged by older women, who are likely better at finding a long-term match than young people. (Consider this the equivalent of dating people based on okCupid match rating, say, instead of hotornot rating.)
The expectations people have from marriage are much more open and agreed upon; like Prismattic points out, they may have a marriage that a Westerner would want to get a divorce in, but be satisfied. It seems to me that this is because of increased realism in expectations (i.e. the Westerner thinks the divorce will be more helpful than it actually will, or is overrating divorce compared to other options), but this is hard to be quantitative about.
To elaborate on the expectations, in arranged marriages it is clear that a healthy relationship is something you have to build and actively maintain, whereas in love marriages sometimes people have the impression that the healthy relationship appears and sustains itself by magic- and so when they put no work into maintaining it, and it falls apart, they claim that the magic is gone rather than that they never changed the oil.
I also think most modern arranged marriages involve some choice on the part of the participants- “meet these four people, tell us if you can’t stand any of them” instead of “you will marry this one person.”
Forty-five individuals (22 couples and 1 widowed person) living in arranged marriages in India completed questionnaires measuring marital satisfaction and wellness. The data were compared with existing data on individuals in the United States living in marriages of choice. Differences were found in importance of marital characteristics, but no differences in satisfaction were found. Differences were also found in 9 of 19 wellness scales between the 2 groups. Implications for further research are considered.
Results from the analyses revealed that arranged marrieds were significantly higher in marital satisfaction than were the love marrieds or companionate marrieds.
Unexpectedly, no differences were found between participants in arranged and love-based marriages; high ratings of love, satisfaction, and commitment were observed in both marriage types. The overall affective experiences of partners in arranged and love marriages appear to be similar, at least among Indian adults living in contemporary U.S. society.
Multiple regression analyses indicate that wives in Chengdu love matches are more satisfied with their marital relationships than their counterparts in arranged marriages, regardless of the length of the marriage, and that this difference cannot be attributed to the influence of other background factors that differentiate these two types of women.
I’m not sure this is correct. That is to say, the empirical point that divorce is much less common in arranged marriage cultures is obviously true. But
a) I think there is some correlation between prevalence arranged marriage and stigma associated with divorce, meaning that not getting divorced does not necessarily equal happy marriage.
b) The bar for success in 20th-21st century western marriages is set really high. It’s not just an economic arrangement; people want a best friend and a passionate lover and maybe several other things rolled into one. When people in traditional cultures say that their marriages are “happy,” they may well mean something much less than what affluent westerners would consider satisfactory.
My instinct on this is driven by having been in bad and good relationships, and reflecting on myself in those situations. It ain’t much, but it’s what I’ve got to work with. Yes, some people are unmatchable, or shouldn’t be matched. But somewhere between “is in high demand and has good judgement, can easily find great matches” and “is unmatchable and should be kept away from others”, there’s a lot of people that can be matched better. Or that’s the hypothesis.
Seems reasonable, although I’d still wonder just how much difference improving the match would make even for the majority of middle-ground people. It sounded in the grandparent post (first and fourth paragraphs particularly) that you were treating the notion that it would be “a lot” as a premise rather than a hypothesis.
Well, it’s more than a hypothesis, it’s a goal. If it doesn’t work, then it doesn’t, but if it does, it’s pretty high impact. (though not existential-risk avoidance high, in and of itself).
Finding a good match has made a big subjective difference for me, and there’s a case it’s made a big objective difference (but then again, I’d say that) and I had to move countries to find that person.
Yeah, maybe the original phrasing is too strong (blame the entrepreneur in pitch mode) but the 6th paragraph does say that it’s an off-chance it can be made to work, though both a high improvement potential and a high difficulty in materialising it are not mutually exclusive.
The problem with dating sites (like social network sites or internet messengers) is that the utility you can gain from it is VERY related to how many other people are actually using it. This means that there is a natural drift towards a monopoly. Nobody wants to join a dating site that only has 1000 people. If you do not have a really good reason to think that your dating site idea will get off the ground, it probably wont.
One way you could possibly get past this is to match people up who do not sign up or even know about this service.
For example, you could create bots that browse okcupid, for answers to questions, ignore okcupid’s stupid algorithms in favor of our own much better ones, and then send two people a message that describes how our service works and introduces them to each other.
Is this legal? If so, I wonder if okcupid would take stop it anyway.
The chicken/egg issue is real with any dating site, yet dating sites do manage to start. Usually you work around this by focusing on a certain group/location, dominating that, and spreading out.
Off the cuff, the bay strikes me as a potentially great area to start for something like this.
and then send two people a message that describes how our service works and introduces them to each other.
Awesome—that will fit right in between “I’m a Nigerian customs official with a suitcase of cash” emails and “Enlarge your manhood with our all-natural pills” ones.
P.S. Actually it’s even better! Imagine that you’re a girl and you receive an email which basically says “We stalked you for a while and we think you should go shack up with that guy”. Genius!
How can there be a monopoly if people can use more than one dating site?
Unless OkCupid bans you from putting your profile up on other sites, you can just as easily put a profile on another site with less people, if the site seems promising.
I don’t mean to be nitpicking, but a monopoly is a very specific thing. It’s quite different than it just being inconvenient to switch to a competitor. In very many cases in normal market competition, it’s inconvenient to switch to competitors (buying a new car or house, changing your insurance, and so on), but that doesn’t effect the quality of the product. Similarly, for a monopoly to effect the quality of OKCupid’s service, it would have to be a very specific situation, and different than what currently exists, which seems to be quite normal market functioning.
Unless OKCupid is hiring the government or people with guns to threaten other websites out of existence, there won’t be a drift towards a monopoly.
A monopoly isn’t created by one company getting the overwhelming majority of customers. A monopoly is only created when competitors cannot enter the market. It’s a subtle distinction but it’s very important, because what’s implied is that the company with the monopoly can jack up their prices and abuse customers. They can’t do this without feeding a garden of small competitors that can and will outgrow them (see Myspace, America Online, etc), unless those competitors are disallowed from ever existing.
You can keep downvoting this, but it’s a very important concept in economics and it will still be true.
Here is one improvement to OKcupid, which we might even be able to implement as a third party:
OKcupid has bad match algorithms, but it can still be useful as searchable classified adds. However, when you find a legitimate match, you need to have a way to signal to the other person that you believe the match could work.
Most messages on OKcupid are from men to women, so women already have a way to do this: send a message, however men do not.
Men spam messages, by glancing over profiles, and sending cookie cutter messages that mention something in the profile. Women are used to this spam, and may reject legitimate interest, because they do not have a good enough spam filter.
Our service would be to provide an I am not spamming commitment. A flag that can be put in a message which signals “This is the only flagged message I have sent this week”
It would be a link, you put in your message, which sends you to a site that basically says. Yes, Bob(profile link) has only sent this flag to Alice(profile link) in the week of 2/20/14-2/26/14, with an explanation of how this works.
Do you think that would be a useful service to implement? Do you think people would actually use it, and receive it well?
I wonder if a per-message fee for a certain kind of message would be a good business model for this. My suspicion is that it would work very well if all your users had that reluctance to ever spend anything online (people are much more willing to buy utilions that involve getting a physical product than to pay for things like apps)), but it breaks down as soon as someone with some unused disposable income realizes that spamming $1 notes isn’t that expensive.
Only being able to send a certain number of messages per week of a special type might be enough for indicating non-spam, as long as you could solve the problem of people making multiple profiles to get around it. Having a small fee attached to the service might help with tracking that down, since it would keep people from abusing it too extremely, and cover the cost of having someone investigate suspicious accounts (if more than one is paid for by the same credit card at around the same time, for example).
OKcupid solves the multiple account problem for us. It is probably better to not send a virtual rose than to make an account that you then have to answer all the questions to.
Our service would be to provide an I am not spamming commitment. A flag that can be put in a message which signals “This is the only flagged message I have sent this week”
Where will your credibility come from?
Alice receives a message from Bob. It says “You’re amazing, we’re nothing but mammals, let’s do it like they do on the Discovery Channel”, and it also says “I, Mallory, hereby certify that Bob only talked about mammals once this week—to you”.
Why should Alice believe you?
Things like that are technically possible (e.g. cryptographic proofs-of-work) but Alice is unlikely to verify your proofs herself and why should she trust Mallory, anyway?
I think if we had a nice professional website, with a link to a long description of how it all works that people won’t read anyway, they will tend to trust us.
Seconded—once you get as far as people trusting you enough to post their personal information and possibly pay you for the service, they’re not still suspecting you of letting people spam you with “certified” non-spam.
OK Cupid has a horrible match percent algorithm. Basically someone who has a check list of things that their match cannot be will answer lots of questions as “this matters a lot to me” and “any of these options are acceptable except for this one extreme one that nobody will click anyway.” The stupid algorithm will inflate this person’s match percent with everyone.
So, if you look at people with high compatibility with you, that says more about their question answering style, than how much you have in common.
This is why the algorithm is horrible in theory. In practice my one example is that I am getting married in a month to someone I met on OKcupid with 99% compatibility.
A good website design could change the answering style. Imagine a site where you don’t fill out all the answers at once. Instead it just displays one question at a time, and you can either answer it or click “not now”. The algorithm would prioritize the questions it asks you dynamically, using the already existing data about you and your potential matches—it would ask you the question which it expects to provide most bits of information.
Also, it would use the math properly. The compatiblity would not be calculated as number of questions answered, but number of bits these answers provide. A match for “likes cats” provides more bits than “is not a serial killer”.
Very consistently people that I know and like, when I see them on okcupid, have a high match percentage. When I meet okcupid people with a good match percentage, I usually like them. This seems to imply the algorithm is a lot better than your theoretical worst example of it. I think your situation is much more of a problem if you don’t answer enough questions.
Perhaps the way people tend to answer questions does not change very much from person to person, so this problem does not show up in practice.
However, if you are willing to change your style for answering questions, it is probably possible to game OKcupid in such a way that you get 90+% with anyone you would care about.
Selfdefeating The entire point of OKcupid is to find someone you will actually click with. Inflating your own match percentages artificially just makes OKCupid worse for you. Of course, this doesnt help if the site just isnt very popular in your city.
Eh. Radical: Have the government do this. Literally, run a dating site, have sex-ed classes teach people how to use it, and why gaming it is bloody stupid. That should result in maximum uptake, and would cost a heck of a lot less than a lot of other initiatives governments already run trying to promote stable pairbonds.
Now, how to get this into a political platform…
Still pointless! There is no upside to having a bunch of people you are not actually compatible with think the mirage you constructed is a good match. If they are not a match with your honest profile, you do not want to waste theirs or your own time.
If your actual goal is to have a bunch of one night stands, then make a profile that out and out states that so that you will be matched with people of like mind. Dishonesty in this matter is both unetical and nigh certain to result in unpleasant drama.
Proper use of this kind of tool is an exercise in luminosity—the more accurately you identify what you are truely looking for, the better it works.
Also, see radical proposal: If a site of this type is run by the government, sockpuppets are obviously not going to be an option—one account per social security number or local equivalent, because that is a really simple way to shut down a whole host of abuses.
I’ve had ideas sort of like this at the back of my mind since seeing Paul Graham pointing out how broken online dating is in one of his essays. (Not so much analyzing all of someone’s existing data, but analyzing IM transcripts to match people with IM buddies they’d be likely to make good friends with is a thing I considered doing.) Haven’t gotten too far with any of them yet, but I’m glad you reminded me, since I was planning on playing with some of my own data soon just to see what I find.
Do you think that not having dated much would be much of a comparative disadvantage in working on this problem? That’s one of the reasons I hesitate to make it my main project.
A possibly-related problem—why does every site I see that says it is for matching strangers who might like to be friends get full of people looking for a date? (Small sample size, but I’ve never seen one that didn’t give me the sense that the vast majority of the members were looking for romance or a one night stand or something.)
A possibly-related problem—why does every site I see that says it is for matching strangers who might like to be friends get full of people looking for a date?
So that people can look for dates without breaking plausible deniability.
So that people can look for dates without breaking plausible deniability.
I think it’s the web site, rather than its clients, that needs the plausible deniability. It cannot seem to be in the business of selling sex, so it has to have a wider focus.
If an altruistic group of numbers-inclined people was to start working together to improve the world in a non-existential risk reducing kind of way, it strikes me that a dating site may be a fantastic thing to try.
Why altruistic? If it’s worth anything, it’s worth money. If it won’t even pay its creators for the time they’ll put in to create it, where’s the value?
I am not convinced it is the optimal route to startup success. If it was, I would be doing it in preference over my current startup. It is highly uncertain and requires what looks like basic research, hence the altruism angle. If it succeeds, yes, it shouldake a lot of money and nobody should deprive it’s creators of the fruits of their labour.
It strikes me that it is much more plausible to argue that the dating market suffers from market failure through information asymmetry, market power and high search costs than to argue the same about economic activity. Yet although people search high and low to find (often non-existent) market failures to justify economic interventions, interventions in the dating market are greeted with near-uniform hostlility. I predict that, outside of LessWrong, your proposal will generate a high “Ick” factor as a taboo violation. “Rationality-based online dating will set you up with scientifically-chosen dates...” this is likely to be an anti-selling point to most users.
Obviously you’d take a different angle with the marketing.
Off the cuff, I’d pitch it as a hands-off dating site. You just install a persistent app on your phone that pushes a notification when it finds a good match. No website to navigate, no profile to fill, no message queue to manage.
Perhaps market it to busy professionals. Finance professionals may be a good target to start marketing to. (busy, high-status, analytical)
There would need to be some way to deal with the privacy issues though.
This might be a reason to start it out as a nice thing. Though, the problem is finding a niche that likes this proposal and has a decent gender ratio (or enough people interested in dates of the same gender).
Now that I think about it, existing dating sites do try to advertise themselves as being better because of their algorithm. If that advertising works, maybe the ick factor isn’t that strong?
Viliam_Bur sort of said this, but it doesn’t seem possible to outcompete the existing websites due to perverse incentives.
If I build a site optimizing for long term success, and another dating site optimizes for an intense honeymoon phase (which encourages people to come back and spread the word about the site) then I will lose. And optimizing for long term success is really hard since feedback occurs on the order of decades.
Of course I’m assuming that intense short term happiness and long term stability aren’t very highly correlated and I could be wrong. I’m also assuming that stability is desirable—I’d be curious if anyone disagrees.
Companies are trying, unfortunately the incentives seem sort of messed up to me. Dating websites have an incentive to encourage people to use their service, not get into wonderful long term relationships. Hence I would expect them to optimize for relationships with an intense honeymoon phase, rather than relationships with a high chance of long term success and compatibility.
Since we’re after long term success, feedback will occur on the order of decades—making this a very hard optimization problem.
If one is able to improve how people are matched, it would bring about a huge amount of utility for the entire world.
People would be happier, they would be more productive, there would be less of the divorce-related waste. Being in a happy couple also means you are less distracted by conflict in the house, which leads to people better able to develop themselves and achieve their personal goals. You can keep adding to the direct benefits of being in a good pairing versus a bad pairing.
But it doesn’t stop there. If we accept that better matched parents raise their children better, then you are looking at a huge improvement in the psychological health of the next generation of humans. And well-raised humans are more likely to match better with each other...
Under this light, it strikes me as vastly suboptimal that people today will get married to the best option available in their immediate environment when they reach the right age.
The cutting-edge online dating sites base their suggestions on a very limited list of questions. But each of us outputs huge amounts of data, many of them available through APIs on the web. Favourite books, movies, sleep patterns, browsing history, work history, health data, and so much more. We should be using that data to form good hypotheses on how to better match people. I’m actually shocked at the underinvestment in this area as a legitimate altruistic cause.
If an altruistic group of numbers-inclined people was to start working together to improve the world in a non-existential risk reducing kind of way, it strikes me that a dating site may be a fantastic thing to try. On the off-chance it actually produces real results, Applied Rationality will also have a great story of how it improved the world. And, you know, it might even make money.
Thoughts? Any better options?
There seem to be perverse incentives in the dating industry. Most obviously: if you successfully create a forever-happy couple, you have lost your customers; but if you make people date many promissingly-looking-yet-disappointing partners, they will keep returning to your site.
Actualy, maybe your customers are completely hypocritical about their goals: maybe “finding a true love” is their official goal, but what they really want is plausible deniability for fucking dozens of attractive strangers while pretending to search for the perfect soulmate. You could create a website which displays the best one or two matches, instead of hundreds of recommendations, and despite having higher success rate for people who try it, most people will probably be unimpressed and give you some bullshit excuses if you ask them.
Also, if people are delusional about their “sexual market value”, you probably won’t make money by trying to fix their delusions. They will be offended by the types of “ordinary” people you offer them as their best matches, when the competing website offers them Prince Charming (whose real goal is to maximize his number of one night stands) or Princess Charming (who really is a prostitute using the website to find potential clients). They will look at the photos and profiles from your website, and from the competing website, and then decide your website isn’t even worth trying. They may also post an offended blog review, and you bet it will be popular on social networks.
So you probably would need to do this as a non-profit philantropic activity.
EDIT: I have an idea about how to remove the perverse incentives, but it requires a lot of trust in users. Make them pay if they have a happy relationship. For example if the website finds you a date, set a regular payment of $5 each month for the next 10 years; if the relationship breaks, cancel the payment. The value of a good relationship is higher than $5 a month, but the total payment of $600 could be enough for the website.
That sounds a lot like really wanting a soulmate and an open relationship.
That’s a nice thing to have; I am not judging anyone. Just thinking how that would influence the dating website algorithm, marketing, and the utility this whole project would create.
If some people say they want X but they actually want Y… however other people say they want X and they mean it… and the algorithm matches them together because the other characteristics match, at the end they may be still unsatisfied (if one of these groups is a small minority, they will be disappointed repeatedly). This could possibly be fixed by an algorithm smart enough that it could somehow detect which option it is, and only match people who want the same thing (whichever of X or Y it is).
If there are many people who say they want X but really want Y, how will you advertise the website? Probably by playing along and describing your website mostly as a site for X, but providing obvious hints that Y is also possible and frequent there. Alternatively, by describing your website as a site for X, but writing “independent” blog articles and comments describing how well it actually works for Y. (What is the chance that this actually is what dating sites are already doing, and the only complaining people are the nerds who don’t understand the real rules?)
Maybe there is a market in explicitly supporting open relationships. (Especially if you start in the Bay Area.) By removing some hypocrisy, the matching could be made more efficient—you could ask questions which you otherwise couldn’t, e.g. “how many % of your time would you prefer to spend with this partner?”.
I wouldn’t jump to malice so fast when incompetence suffices as an explanation. Nobody has actually done the proper research. The current sites have found a local maxima and are happy to extract value there. Google got huge by getting people off the site fast when everyone else was building portals.
You will of course get lots of delusionals, and lots of people damaged enough that they are unmatchable anyway. You can’t help everybody. But also the point is to improve the result they would otherwise have had. Delusional people do end up finding a match in general, so you just have to improve that to have a win. Perhaps you can fix the incentive by getting paid for the duration of the resulting relationship. (and that has issues by itself, but that’s a long conversation)
I don’t think the philanthropic angle will help, though having altruistic investors who aren’t looking for immediate maximisation of investment is probably a must, as a lot of this is pure research.
I don’t think he was jumping to malice, rather delusion or bias.
I meant malice/incompetence on the part of the dating sites.
I think that’s the business model of eharmony and they seem to be doing well.
I absolutely agree, but I am not sure that anyone was even considering this as a way to make money.
Unfortunately, for all the same reasons we cannot make money, we cannot get people to sign up for the site in the first place.
Two proposed solutions for this:
1) Something like I suggested before that matches people without them signing up somehow.
2) A bait and switch, where a site gets popular using the same tactics as other dating sites, and then switches to something better for them.
Neither of these solutions seem plausible to work at all.
I wonder to what extent the problems you describe (divorces, conflict, etc) are caused mainly by poor matching of the people having the problems, and to what extent they are caused by the people having poor relationship (or other) skills, relatively regardless of how well matched they are with their partner? For example, it could be that someone is only a little bit less likely to have dramatic arguments with their “ideal match” than with a random partner—they just happen to be an argumentative person or haven’t figured out better ways of resolving disagreements.
Well, the success of arranged marriages in cultures that practice them suggests the “right match” isn’t that important.
What makes you think these marriages are successful? Low divorce rates are not good evidence in places where divorce is often impractical.
Three main points in favor of arranged marriages that I’m aware of:
The marriages are generally arranged by older women, who are likely better at finding a long-term match than young people. (Consider this the equivalent of dating people based on okCupid match rating, say, instead of hotornot rating.)
The expectations people have from marriage are much more open and agreed upon; like Prismattic points out, they may have a marriage that a Westerner would want to get a divorce in, but be satisfied. It seems to me that this is because of increased realism in expectations (i.e. the Westerner thinks the divorce will be more helpful than it actually will, or is overrating divorce compared to other options), but this is hard to be quantitative about.
To elaborate on the expectations, in arranged marriages it is clear that a healthy relationship is something you have to build and actively maintain, whereas in love marriages sometimes people have the impression that the healthy relationship appears and sustains itself by magic- and so when they put no work into maintaining it, and it falls apart, they claim that the magic is gone rather than that they never changed the oil.
I also think most modern arranged marriages involve some choice on the part of the participants- “meet these four people, tell us if you can’t stand any of them” instead of “you will marry this one person.”
I remember seeing studies that attempted to measure happiness.
Links? I am also quite suspicious of measuring happiness—by one measure Bhutan is the happiest country in the world and, um, I have my doubts.
Source.
Source.
Source.
A contrary finding:
Source.
Why are you even asking for links to studies if you admit you don’t care what studies say?
I have a prior that the studies are suspect. But that prior can be updated by evidence.
I’m not sure this is correct. That is to say, the empirical point that divorce is much less common in arranged marriage cultures is obviously true. But
a) I think there is some correlation between prevalence arranged marriage and stigma associated with divorce, meaning that not getting divorced does not necessarily equal happy marriage.
b) The bar for success in 20th-21st century western marriages is set really high. It’s not just an economic arrangement; people want a best friend and a passionate lover and maybe several other things rolled into one. When people in traditional cultures say that their marriages are “happy,” they may well mean something much less than what affluent westerners would consider satisfactory.
Why does it suggest that rather than that the arrangers are better at finding the “right match” than the persons to be married?
My instinct on this is driven by having been in bad and good relationships, and reflecting on myself in those situations. It ain’t much, but it’s what I’ve got to work with. Yes, some people are unmatchable, or shouldn’t be matched. But somewhere between “is in high demand and has good judgement, can easily find great matches” and “is unmatchable and should be kept away from others”, there’s a lot of people that can be matched better. Or that’s the hypothesis.
Seems reasonable, although I’d still wonder just how much difference improving the match would make even for the majority of middle-ground people. It sounded in the grandparent post (first and fourth paragraphs particularly) that you were treating the notion that it would be “a lot” as a premise rather than a hypothesis.
Well, it’s more than a hypothesis, it’s a goal. If it doesn’t work, then it doesn’t, but if it does, it’s pretty high impact. (though not existential-risk avoidance high, in and of itself).
Finding a good match has made a big subjective difference for me, and there’s a case it’s made a big objective difference (but then again, I’d say that) and I had to move countries to find that person.
Yeah, maybe the original phrasing is too strong (blame the entrepreneur in pitch mode) but the 6th paragraph does say that it’s an off-chance it can be made to work, though both a high improvement potential and a high difficulty in materialising it are not mutually exclusive.
The problem with dating sites (like social network sites or internet messengers) is that the utility you can gain from it is VERY related to how many other people are actually using it. This means that there is a natural drift towards a monopoly. Nobody wants to join a dating site that only has 1000 people. If you do not have a really good reason to think that your dating site idea will get off the ground, it probably wont.
One way you could possibly get past this is to match people up who do not sign up or even know about this service.
For example, you could create bots that browse okcupid, for answers to questions, ignore okcupid’s stupid algorithms in favor of our own much better ones, and then send two people a message that describes how our service works and introduces them to each other.
Is this legal? If so, I wonder if okcupid would take stop it anyway.
The chicken/egg issue is real with any dating site, yet dating sites do manage to start. Usually you work around this by focusing on a certain group/location, dominating that, and spreading out.
Off the cuff, the bay strikes me as a potentially great area to start for something like this.
It’s spam and very likely violates the TOS.
Awesome—that will fit right in between “I’m a Nigerian customs official with a suitcase of cash” emails and “Enlarge your manhood with our all-natural pills” ones.
P.S. Actually it’s even better! Imagine that you’re a girl and you receive an email which basically says “We stalked you for a while and we think you should go shack up with that guy”. Genius!
How can there be a monopoly if people can use more than one dating site?
Unless OkCupid bans you from putting your profile up on other sites, you can just as easily put a profile on another site with less people, if the site seems promising.
It’s still more work to put a profiles on multiple sites.
Hi Eugine,
I don’t mean to be nitpicking, but a monopoly is a very specific thing. It’s quite different than it just being inconvenient to switch to a competitor. In very many cases in normal market competition, it’s inconvenient to switch to competitors (buying a new car or house, changing your insurance, and so on), but that doesn’t effect the quality of the product. Similarly, for a monopoly to effect the quality of OKCupid’s service, it would have to be a very specific situation, and different than what currently exists, which seems to be quite normal market functioning.
Coscott was talking about a “a natural drift towards a monopoly”.
Unless OKCupid is hiring the government or people with guns to threaten other websites out of existence, there won’t be a drift towards a monopoly.
A monopoly isn’t created by one company getting the overwhelming majority of customers. A monopoly is only created when competitors cannot enter the market. It’s a subtle distinction but it’s very important, because what’s implied is that the company with the monopoly can jack up their prices and abuse customers. They can’t do this without feeding a garden of small competitors that can and will outgrow them (see Myspace, America Online, etc), unless those competitors are disallowed from ever existing.
You can keep downvoting this, but it’s a very important concept in economics and it will still be true.
Forbidding anyone who hasn’t read “The Logical Fallacy of Generalization from Fictional Evidence” from watching any Hollywood or Disney movies about romance would go a long way. ;-)
So how would it be different from OK Cupid, for example?
As an aside, wasn’t the original motivation for Facebook Zuckerberg’s desire to meet girls..? :-D
Here is one improvement to OKcupid, which we might even be able to implement as a third party:
OKcupid has bad match algorithms, but it can still be useful as searchable classified adds. However, when you find a legitimate match, you need to have a way to signal to the other person that you believe the match could work.
Most messages on OKcupid are from men to women, so women already have a way to do this: send a message, however men do not.
Men spam messages, by glancing over profiles, and sending cookie cutter messages that mention something in the profile. Women are used to this spam, and may reject legitimate interest, because they do not have a good enough spam filter.
Our service would be to provide an I am not spamming commitment. A flag that can be put in a message which signals “This is the only flagged message I have sent this week”
It would be a link, you put in your message, which sends you to a site that basically says. Yes, Bob(profile link) has only sent this flag to Alice(profile link) in the week of 2/20/14-2/26/14, with an explanation of how this works.
Do you think that would be a useful service to implement? Do you think people would actually use it, and receive it well?
Scarce signals do increase willingness to go on dates, based on a field experiment of online dating in South Korea.
I wonder if a per-message fee for a certain kind of message would be a good business model for this. My suspicion is that it would work very well if all your users had that reluctance to ever spend anything online (people are much more willing to buy utilions that involve getting a physical product than to pay for things like apps)), but it breaks down as soon as someone with some unused disposable income realizes that spamming $1 notes isn’t that expensive.
Only being able to send a certain number of messages per week of a special type might be enough for indicating non-spam, as long as you could solve the problem of people making multiple profiles to get around it. Having a small fee attached to the service might help with tracking that down, since it would keep people from abusing it too extremely, and cover the cost of having someone investigate suspicious accounts (if more than one is paid for by the same credit card at around the same time, for example).
OKcupid solves the multiple account problem for us. It is probably better to not send a virtual rose than to make an account that you then have to answer all the questions to.
Where will your credibility come from?
Alice receives a message from Bob. It says “You’re amazing, we’re nothing but mammals, let’s do it like they do on the Discovery Channel”, and it also says “I, Mallory, hereby certify that Bob only talked about mammals once this week—to you”.
Why should Alice believe you?
Things like that are technically possible (e.g. cryptographic proofs-of-work) but Alice is unlikely to verify your proofs herself and why should she trust Mallory, anyway?
I think if we had a nice professional website, with a link to a long description of how it all works that people won’t read anyway, they will tend to trust us.
Especially if we use math.
Seconded—once you get as far as people trusting you enough to post their personal information and possibly pay you for the service, they’re not still suspecting you of letting people spam you with “certified” non-spam.
OK Cupid has a horrible match percent algorithm. Basically someone who has a check list of things that their match cannot be will answer lots of questions as “this matters a lot to me” and “any of these options are acceptable except for this one extreme one that nobody will click anyway.” The stupid algorithm will inflate this person’s match percent with everyone.
So, if you look at people with high compatibility with you, that says more about their question answering style, than how much you have in common.
This is why the algorithm is horrible in theory. In practice my one example is that I am getting married in a month to someone I met on OKcupid with 99% compatibility.
A good website design could change the answering style. Imagine a site where you don’t fill out all the answers at once. Instead it just displays one question at a time, and you can either answer it or click “not now”. The algorithm would prioritize the questions it asks you dynamically, using the already existing data about you and your potential matches—it would ask you the question which it expects to provide most bits of information.
Also, it would use the math properly. The compatiblity would not be calculated as number of questions answered, but number of bits these answers provide. A match for “likes cats” provides more bits than “is not a serial killer”.
Very consistently people that I know and like, when I see them on okcupid, have a high match percentage. When I meet okcupid people with a good match percentage, I usually like them. This seems to imply the algorithm is a lot better than your theoretical worst example of it. I think your situation is much more of a problem if you don’t answer enough questions.
Perhaps the way people tend to answer questions does not change very much from person to person, so this problem does not show up in practice.
However, if you are willing to change your style for answering questions, it is probably possible to game OKcupid in such a way that you get 90+% with anyone you would care about.
Selfdefeating The entire point of OKcupid is to find someone you will actually click with. Inflating your own match percentages artificially just makes OKCupid worse for you. Of course, this doesnt help if the site just isnt very popular in your city.
Eh. Radical: Have the government do this. Literally, run a dating site, have sex-ed classes teach people how to use it, and why gaming it is bloody stupid. That should result in maximum uptake, and would cost a heck of a lot less than a lot of other initiatives governments already run trying to promote stable pairbonds. Now, how to get this into a political platform…
Not if you have an honest account too so you can check compatibility while still broadcasting higher compatibility than you actually have.
Still pointless! There is no upside to having a bunch of people you are not actually compatible with think the mirage you constructed is a good match. If they are not a match with your honest profile, you do not want to waste theirs or your own time. If your actual goal is to have a bunch of one night stands, then make a profile that out and out states that so that you will be matched with people of like mind. Dishonesty in this matter is both unetical and nigh certain to result in unpleasant drama. Proper use of this kind of tool is an exercise in luminosity—the more accurately you identify what you are truely looking for, the better it works.
Also, see radical proposal: If a site of this type is run by the government, sockpuppets are obviously not going to be an option—one account per social security number or local equivalent, because that is a really simple way to shut down a whole host of abuses.
I’ve had ideas sort of like this at the back of my mind since seeing Paul Graham pointing out how broken online dating is in one of his essays. (Not so much analyzing all of someone’s existing data, but analyzing IM transcripts to match people with IM buddies they’d be likely to make good friends with is a thing I considered doing.) Haven’t gotten too far with any of them yet, but I’m glad you reminded me, since I was planning on playing with some of my own data soon just to see what I find.
Do you think that not having dated much would be much of a comparative disadvantage in working on this problem? That’s one of the reasons I hesitate to make it my main project.
A possibly-related problem—why does every site I see that says it is for matching strangers who might like to be friends get full of people looking for a date? (Small sample size, but I’ve never seen one that didn’t give me the sense that the vast majority of the members were looking for romance or a one night stand or something.)
So that people can look for dates without breaking plausible deniability.
I think it’s the web site, rather than its clients, that needs the plausible deniability. It cannot seem to be in the business of selling sex, so it has to have a wider focus.
Why altruistic? If it’s worth anything, it’s worth money. If it won’t even pay its creators for the time they’ll put in to create it, where’s the value?
I am not convinced it is the optimal route to startup success. If it was, I would be doing it in preference over my current startup. It is highly uncertain and requires what looks like basic research, hence the altruism angle. If it succeeds, yes, it shouldake a lot of money and nobody should deprive it’s creators of the fruits of their labour.
It strikes me that it is much more plausible to argue that the dating market suffers from market failure through information asymmetry, market power and high search costs than to argue the same about economic activity. Yet although people search high and low to find (often non-existent) market failures to justify economic interventions, interventions in the dating market are greeted with near-uniform hostlility. I predict that, outside of LessWrong, your proposal will generate a high “Ick” factor as a taboo violation. “Rationality-based online dating will set you up with scientifically-chosen dates...” this is likely to be an anti-selling point to most users.
Obviously you’d take a different angle with the marketing.
Off the cuff, I’d pitch it as a hands-off dating site. You just install a persistent app on your phone that pushes a notification when it finds a good match. No website to navigate, no profile to fill, no message queue to manage.
Perhaps market it to busy professionals. Finance professionals may be a good target to start marketing to. (busy, high-status, analytical)
There would need to be some way to deal with the privacy issues though.
This might be a reason to start it out as a nice thing. Though, the problem is finding a niche that likes this proposal and has a decent gender ratio (or enough people interested in dates of the same gender).
Now that I think about it, existing dating sites do try to advertise themselves as being better because of their algorithm. If that advertising works, maybe the ick factor isn’t that strong?
Have you seen this TED talk?
fantastic, thanks!
Viliam_Bur sort of said this, but it doesn’t seem possible to outcompete the existing websites due to perverse incentives.
If I build a site optimizing for long term success, and another dating site optimizes for an intense honeymoon phase (which encourages people to come back and spread the word about the site) then I will lose. And optimizing for long term success is really hard since feedback occurs on the order of decades.
Of course I’m assuming that intense short term happiness and long term stability aren’t very highly correlated and I could be wrong. I’m also assuming that stability is desirable—I’d be curious if anyone disagrees.
Companies are trying, unfortunately the incentives seem sort of messed up to me. Dating websites have an incentive to encourage people to use their service, not get into wonderful long term relationships. Hence I would expect them to optimize for relationships with an intense honeymoon phase, rather than relationships with a high chance of long term success and compatibility.
Since we’re after long term success, feedback will occur on the order of decades—making this a very hard optimization problem.