Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Towards Better Online Dating

blog.adamsmith.cc

Post reply on HN
Print view View original

1–10 of 88 posts

Re: Towards Better Online Dating

#2

Post by potatolicious »

I agree with everything the author said save this:

> "Then only allow people to access others they are likely to match against."

This only works if you are 100% confident that your matching algorithm is infallible. The odds of this being the case is about 0%.

I've thought about this problem some, and I think there are a few unacknowledged issues:

- Who you will get along with, and are a good match for, is often different from who you think you will get along with. Dorky Dillon, taking from the author's example, doesn't want page after page of plain janes similar to himself - like all humans he wants to date upwards... even though it is against his long-term interests and reduces his "efficiency" in the system.

- We have solid predictors of long term relationship success. Similar politics? Check. Similar religious views? Check. Suffice it to say, the odds of a hardline Republican and a extreme Democrat working out in the long run is pretty low, but would they enjoy hanging out for 5 minutes? 10? A month? Part of the dating site's job is to find dates, long term relationships build themselves out of that. IMHO there's a big difference between long-term compatibility (that the poor lovebirds will discover for themselves with or without your algorithm's help) and short-term compatibility (e.g. the ability to carry a conversation through dinner that won't put the other side to sleep). We have good predictors for the former, but IMHO not so much for the latter... and the latter is also much harder to infer.

So far dating sites - and even the vaunted OkCupid is guilty of this - conflating personality with opinion. Personality is what matters most in the initial stage, but none of the sites are collecting meaningful information on it.

I almost feel like matching people based on relationship compatibility (i.e., mid to long term) is the wrong way to go. IMHO it's a far better bet to predict whether or not two people can get through dinner without trying to kill each other... and just send them on a date. If they click, great. If they don't, spin the wheel again. The modern dating website, with its hyper-focus on long-term compatibility seems to be optimizing prematurely.

Re: Towards Better Online Dating

#3

Post by ookblah »

I've randomly thought about this and one thing i can say is that if you want to build a successful long-term dating site you HAVE to cater 90% towards the women haahha. that means making it easy for them to filter, having a safe place where they don't feel hit on by creepers, or provide an arena where they know that the men are "quality", etc.

You will always have enough male signups. So for the men it's about balancing the ratios somewhat so that it's not just completely ridiculous. Dating sites know this and probably don't care at all because they're getting revenue from any warm body.

While i agree that it just takes 5 minutes in person to do a quick BS/weirdo filter, that's not feasible online, especially for women. Women are spammed much more and have a tough time filtering, let alone meeting many people.

The only way this will work is if the matching algorithm he's proposing is extremely accurate, otherwise all it'd take is like 3-4 weeks of failed group dinner meetings to write the site off.

Re: Towards Better Online Dating

#7

Post by Detrus »

The best data is just hard to collect online. Photos of person in their social group to deduce personality type would probably fail since there are so many variables.

Lab scientists can measure biological or physical properties http://www.youtube.com/watch?v=XGfznq9OgdU&feature=playe... like sweat scent which reveals immune system compatibility. They're getting very good correlations for long term relationships from these simple physical metrics.

Some of these metrics can be collected online, high testosterone or estrogen is visible in the face. Has any dating site used them seriously? Because most approach the problem from a psychological perspective, with surveys, bios, chatting, etc.. But long term couples have a very high proportion of physical compatibility, which happens to be easier to measure.

Re: Towards Better Online Dating

#8

Post by patio11 »

It would be interesting to see an online dating site which treated users like mailing service providers do: "We really want you to succeed with this, but our priorities are to the ecosystem first and you a very distant second. You're guilty until proven innocent, we're going to invest a lot of effort into educating you in how to suck less, and if we catch a whiff of you being a nuisance you are banned for life."

(It would take some serious guts to say "Your presence in our community is a net negative. We wish you the best, as long as the best is found elsewhere.", particularly if you do that in a data-driven fashion. My intuition is that people's revealed preferences in dating are very, very, very far from socially acceptable.)

Re: Towards Better Online Dating

#9

Post by orangecat »

ookblah wrote: post #3

I've randomly thought about this and one thing i can say is that if you want to build a successful long-term dating site you HAVE to cater 90% towards the women haahha. that means making it easy for them to filter, having a safe place where they don't feel hit on by creepers, or provide an arena where they know that the men are "quality", etc. You will always have enough male signups. So for the men it's about balanc…

if you want to build a successful long-term dating site you HAVE to cater 90% towards the women

Of course. Seems like one obvious way would be to limit the number of emails that men can send per day/week/whatever, so women don't get spammed quite so badly.

Post reply on HN