How Machine Learning Can Transform Online Dating
First time accepted submitter hrb1979 writes "Thought I'd share an interview with Kang Zhao — the professor behind the machine learning algorithm which could transform online dating. His algorithm takes into account both a user's tastes (in an approach similar to the Netflix recommendation engine) and their attractiveness (by analyzing how many responses they get) — enabling the machine to 'learn' and hence propose higher potential matches. His research was recently covered in both a Forbes' article and the MIT Technology Review, though this interview provides more depth and color."
All the algorithms in the world aren't going to help when the intersection of "people you'd care to date" and "people who'd care to date you" is empty. What we need is an algorithm to convince people to lower their expectations when they're unattractive, boring, unmannerly, old, poor and/or cheap, have baggage, etc.
His research was recently covered in both a Forbes' article and the MIT Technology Review
It was also recently covered on Slasdot.
This doesn't take into account chemistry, which is where feeling love comes from. If you want the logical mate, then sure you can probably create something like this. If you want a real match, you'll also need to send in hormones and blood samples for testing. It would probably also help if people didn't lie so much online.
Somehow I get the feeling that I won't have any success with this. I'm a typical Slashdot user.
This is where evolution backfired. It used to be the case that physical characteristics highly correlated with reproductive success, so we are hard-wired to look for them. Sometimes this takes strange detours into fetishes-obsessions, but most of the time we just looking for "good looking".
The perfect formula for matchmaking:
Males: Enter income.
Females: Enter attractiveness.
Match up most attractive to highest earners.
If telephones are outlawed, then only outlaws will have telephones.
Frink: Well, theoretically, yes. But the computer matches would be so perfect as to eliminate the thrill of romantic conquest. Mw-hurgn-whey.
SJW n. One who posts facts.
Kind of like job interviews. It's not that hard, really: if they don't like you after you're in the door, you've still gotten out of your mom's basement for a day and had a chance to steal some pens. It also gets you a chance to talk to the people who actually *work* at the company, who know what department actually needs your real job skills. Same with "si8ngles" dating. Even if the "single" is a real brown bagger, she may have friends who are interesting to meet and will "rescue" their friend by distracting you.
Nothing gets them wanting to go out like seeing that their friend is out with you. Think I'm kidding? Just watch them at a good party.
This is true even though a homely spouse makes for a much more attentive husband/wife.
Do they? Ugly people are just as capable of obnoxious behaviour as attractive ones.
SJW n. One who posts facts.
For people downvoting above - the reason you are still single is because you have unrealistically high opinion of your own charm, attractiveness, generosity, and achievements. Relationships are about getting along and building trust, not about putting yourself on a pedestal with a puffed-up profile. Unlike job interviews, it hurts relationships to over-sell yourself by introducing inevitable disappointment and distrust at the very early stages of it.
Being recently divorced (last year), I started poking around at the various dating sites. Let's see, I don't smoke, I effectively don't drink alcohol (one beer a year doesn't make me a "drinker"), I'm a gamer (you'd be amazed at the number of women who think gamers are "childish"), and I'm not into sports (lots of women who go to football/baseball/hockey games). I do like going hiking in the mountains, snowshoes, skiing, I like bicycling. But I'm not a fitness fanatic which also eliminated quite a few women. I'm not religious which eliminated a few more.
After eliminating the mis-matches, I started paring down the other issues. Based on profiles, I got down to about 60 women in the area who might be an match based on shared interests. I received no replies to my e-mails but I did receive three unrelated emails. One from a women in Australia. One from a woman in Texas who had pictures of her daughter leaning on a car (which was a bit creepy). And one from a woman who plays guitars who appeared to be looking for a man in every city.
Amusingly on my birthday (hit 56), my match list dropped to zero. Every one of the women were looking for guys 55 or younger. So I expanded my search until I got to a couple of women in a 250 mile radius.
For some of us (a small percentage I suppose), the dating sites really aren't helping. And since women receive all the emails, they have the choice of who to go out with.
Humorously I was chosen to moderate pictures on okcupid for a bit. The guys do send some very suggestive emails (and some not so suggestive!).
[John]
Shit better not happen!
> the reason you are still single is because you have unrealistically high opinion of your own charm
As opposed to " the reason you are still single is because you don't try"? Arrogance, which is what you're describing, can actually help. It can provide the confidence to actually ask out potential dates, to believe that you're worth spending time with, and get you past failures. Never trying is guaranteed failure, and it's too common.
I think there is some truth to this. I seldom see drop-dead gorgeous female engineers or scientists. On the other hand, female cheerleaders make me cringe and despair. You don't think there is a strong correlation between attractiveness and personal qualities/careers/etc.?
I seldom see drop-dead gorgeous female engineers or scientists....You don't think there is a strong correlation between attractiveness and personal qualities/careers/etc.?
Let us say that intelligence and attractiveness are uncorrelated.
The probability of being both at once is very low, since you're multiplying the two small independent probabilities associated with attractiveness and intelligence.
If, of course you chose any segment of the population not selected specifically for attractiveness, then the probability of any given member being attractive is low.
Drop-dead gorgeous female scientists and engineers are particularly rare because (a) female scientists and enigneers are rare and (b) drop-dead gorgeousness is rare. Even uncorrelated if you multiply those tow probabilities, a small number results.
If you have something where members are chosen for attractiveness then yes, the average member will be more attractive than the average.
Would I say that my fellow scientists and engineers are on average less attractive than a random sampling of the population? That's a hard call. There are whole segments of the population that I rarely mix with which makes such things hard to judge. I never have cause to visit grim, deprived, crime ridden former mining towns of the North for example. But I haven't noticed any particular difference. There's one guy in my office you could easily tell from the outside (long hair, overweight and wearing anime t-shirts is kind of a give away). The rest not so much.
SJW n. One who posts facts.
"His algorithm takes into account both a user's tastes (in an approach similar to the Netflix recommendation engine) and their attractiveness (by analyzing how many responses they get) — enabling the machine to 'learn' and hence propose higher potential matches."
I'd send them a hiring offer for a Photoshop Job, but that's just me.
So, in other words, the machine will tell you, "Hey dude, you might as well talk to this homely girl, because we've analyzed your interests and your apparent attractiveness and you're not going to do any better than this".
Proverbs 21:19
Totally wrong. Arrogance and lying actually work really well for many men. Those men get laid a lot more than the honest and humble men. You can debate whether it's worth it in the long term or not (he might be stuck with multiple child-support payments in his 40s), but it is a winning reproductive and dating strategy.
I seldom see drop-dead gorgeous female engineers or scientists.
That's because smart and gorgeous women know that those professions are underpaid and undervalued in Western (particularly American) society, so they avoid them (they also want to avoid all the sexual harassment). Instead, these women go into the medical field. I've met several nearly drop-dead gorgeous female physicians. And they certainly get better pay and better job security than I do as an engineer, while not being surrounded by creepy men or brogrammers.
Ugly people are just as capable of obnoxious behaviour as attractive ones
True statement. But attractive people (especially women with large breasts) tend to be conditioned by society to think that they are more desirable and hence that they are entitled to be more demanding. Same with guys who are good athletes or have lots of money.
There's a couple of problems here. 1) OKC doesn't have any way of blocking or filtering out people with kids, which apparently you're not open to (which is fine, it's a valid choice as a single man, esp. if you're younger; it gets unrealistic as you get older). From what I've seen, they only let you block people's messages if they're too far away from you. It'd be nice if there were more controls this way to block out people with deal-breakers (smoking, etc.) 2) There's no way for OKC to tell if someone is financially/emotionally unstable that I can think of. I don't even know how you would tell this about someone before meeting them in person and getting to know them, unless they're brazenly hitting you up for money when they first message you.
+1 spot on
never bring a twinkie to a food fight.
That's a nice ${ANATOMICAL_PART} you've got there.
Have gnu, will travel.
This is true even though a homely spouse makes for a much more attentive husband/wife.
Jimmy Soul - If You Want To Be Happy (1963)
If you wanna be happy for the rest of your life
Never make a pretty woman your wife
So from my personal point of view
Get an ugly girl to marry you
I'm guessing that Search Engine Optimization (SEO) techniques would come into play whereby geeks could artificially inflate their relevance/attractiveness to game the system. Once again the nerds get the supermodels... right?
I'm in my 40's and married, but back when I tried seriously using a few of the dating sites (never was willing to pay for the commercial ones, but gave sites like PoF and OKCupid a shot) -- I ran across a general theme for the individual who'd find the most success there. Basically, the formula seemed to be:
1. Into sports
2. Posted at least one sexy/arousing type photo instead of only head-shots
3. Somewhere in the "under 35" age range
4. Liked to "go out for a few beers/drinks" (but not "drink heavily", of course)
5. Claimed to have some type of job people perceived as "successful"
6. Those under 25 who bragged about "420 friendliness" (marijuana) seemed to do well with others in that age range.
Older people definitely tend to be the "outliers" on these sites -- with relatively few peers using them for dating. Worse yet, I think some of the 50 and 60 somethings use them to troll for much younger dates, sometimes even lying about their age to get the first in-person meetup.
I tried to be pretty honest about exactly what my interests were (and weren't), but found that didn't get me many initial contacts. I think just as in "real life", there's still sort of an expectation that the man is supposed to make the first move. Women will put a profile out there and just let the emails rolls in. Men put one out there and it gets viewed a number of times, but winds up only serving the purpose of getting reviewed closely if he contacts a woman first and she's trying to decide if she wants to write him back or not.
I'll tell you another thing that might be worth doing, if you want a "reality check" about your level of physical attractiveness to the opposite sex. Find a couple of photos of yourself you feel represent you well/accurately and post them on one of those "Hot or Not?" sites. Come back a few weeks later and see what your average rating was. I did this once, and frankly, it was pretty brutal. I wound up ranked about a 3 out of 10 -- despite being a guy I'm somewhat regularly told is "fairly attractive". I tried to analyze it, and discovered a few things -- like much younger women rating me really low, simply because I was "too old for them". But at the end of the day, I think it's important to realize that a lot of people using these dating sites will be clicking through guys' photos and profiles with this same mindset. "Ick... he's wearing an ugly shirt! Next!" "Looks like he could be my dad or something. Ugly! Next!" So without letting it shred your ego, I think it does provide some perspective at least -- when you get mad that "There are 1,000 women on this site right now and not ONE wants to write me back!?!"
This reminds me of the old story about college roommates going to a party. The guy who drove immediately walks up to a babe and whispers in her ear, and gets slapped. Walks up to another, gets slapped again. The other guy shakes his head and goes to get a beer and find a conversation. Later, having gotten nowhere and wondering about his ride home, he looks for the first guy and can't find him anywhere. Finally the driver reappears with a smile on his face, and in no mood to leave. "Why not? Aren't you getting slapped a lot?" "Yes, but I'm getting laid a lot, too."
Being ... attractive ... [is] always subjective
Yes and no. When it comes to physical attractiveness, if you take someone, and then have N randomly selected other people rate their attractiveness, you'll get a mean and a standard deviation. Of course that's not the most important thing to you (or your possible partner) as only your opinion matters (though the mean will of course affect how much attention someone gets).
Ok, that part's obvious, but there is an interesting effect. I wish I could find a link to the study, but I can't right now. Basically, if someone is fairly attractive, the SD of opinion will be comparatively small. However, if you look at the people that a given person finds the most attractive, there will generally be a much larger SD for them. In other words, most people more or less agree on who is reasonably attractive, but who they find most attractive is likely to be someone of who that opinion is not widespread.
I'm always surprised when I run into the contrary examples of women dating men who are by most measures less attractive than they are.
The two examples that come to the top of my head are both good friends of my wife and I.
One woman is in top physical shape (runs like 25 miles a week), high end corporate job (director of a hospital) and her husband is an easy 75 pounds overweight and some kind of rank-and-file finance guy. In three major respects, looks, earnings, and "job status" he's at a lower level than she is, yet they get along great.
The other one is less of a mismatch in terms of jobs (he's a realtor, she works as an assistant to him on real estate), but she is very, very attractive -- very pretty. Without being too crude, a really amazing figure in all respects, even more so amazing in terms of the fact she's 40+ with two kids. Her husband is paunchy and not nearly as good looking as she is.
This happens often enough that I figure that simple "looks" is far less important to women than it is to men, and that many women are either willing to overlook some aspects of appearance or simply don't care.
I think this is relevant: How Amy Webb hacked online dating:
http://www.ted.com/talks/amy_webb_how_i_hacked_online_dating.html
- For the complete works of Shakespeare: cat
If its going to be anything like the Netflix recommendation engine then I feel like online dating will probably take a step backwards.
Everyone will get recommended the same stupid people they don't want to see over and over again.
1. Can not be bat shit crazy.
2. Not a midget. (How the hell did I end up on a blind date with a midget? It was like going out with a child, I think my 7 year old niece is taller. Before anybody asks it's true, I've actually been on a blind date with a little person, I'm not being sarcastic about that.)
Apparently it's going to be difficult to find both of those in 1 girl.
Did you know 80 to 90% of the moderators on slashdot wouldn't recognize a troll even if one dragged them under a bridge.
As it stands, I only get replies from bots and eharmony has scientifically proven that nobody on the face of the planet is compatible with me. Yes, I am serious. Yes, this is the main reason I brew very strong, very high quality mead. Inside work, I'm a meaningless drone in a stagnant occupation. Outside of work, the only company I keep is a three gallon jar. It's not a good conversationalist but it has a greater capacity for thought than my co-workers, which is something.
The older I get, the more I realize that Marvin, the Paranoid Android, was an optimist.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
Actually there is likely to be a positive correlation between attractiveness & intelligence, because we have evolved to look for healthy mates who are like to reproduce well. For example lack of symmetry tends (so there are plenty of counter examples, but they do not undermine the argument) to be associated with an underlying health problem. Also around the world men tend to consider women with a particular hip to waist ratio as being the most attractive - research found that more that women deviated from that ratio then the less likely they would be able to conceive, carry a baby successfully to term, or have a safe vaginal birth.
In the second world war the British found that the most effective way of determining if a young woman would be good at the job of evaluating reconnaissance taken by RAF planes was by looking at the girls's ankles - the work involved intense concentration for long hours. At the time they did not know why. I suspect now, it was because a shapely ankle correlate well with health and stamina - but feminists would be predisposed to jump to the wrong conclusions.
Certainly, when I was working at at with lots of PhD students, the girls seemed to be generally of above average beauty! (No I didn't attempt to date any of them, as I'm married and old enough to be their grandfather!)
I pretty much came in here to say this. People already game the system in so many intricate ways, but those ways at least are social engineering techniques that mean you at least have to have a clue about how people tick (and thus in a social situation, like a date, you can reasonably hold your own/not suck at being a conversationalist). But if you're optimizing against a machine algorithm the sky's the limit. Make a few fake profiles and hit yourself up to increase your attractiveness score, or convince your friends to do the same in kind. Figure out what keywords get you more attractive girls/guys, and build your profile around that. Then you finally end up on a date and...*pfft* now your out of your element. Then the dating site loses credibility and gets washed back into the primordial ooze from which it crawled.
There is much truth here. When I escaped my first wife at 38, I hadn't really ever dated (long, embarrassing story). I started out being a decent guy, and got little interest in the dating realm -- being 6' was a common reason for rejection. A friend at the time told me I was "too nice", which I initially dismissed. The turning point was when eHarmony matched me with a woman I was really into -- wicked smart, attractive, (and yes a bit overweight, as was I). After a couple of weeks she dumped me basically because I was treating her with respect. That hardened my heart and I started being a bit arrogant / a jerk, and the response rate from women skyrocketed. At one point I was in a coffee shop and overheard a couple meeting for the first time. The woman (who was okay looking, with a few extra pounds) for some reason was going on about how the last guy she'd met had wasted her time: "He was *TINY*, like 5'7"!"