Tuesday, January 28, 2014

Bogus Title, Good Story

The title is "How a Math Genius Hacked OK Cupid to Find True Love." There is no evidence in the story that the protagonist is a math genius, merely a doctoral candidate in math at UCLA. But it does describe, in reasonable detail, how he first data mined OK Cupid, a large dating site, and then used statistical techniques to analyze the data in order to figure out how to construct a profile that would attract women he was likely to find of interest—and do it without lying.

It took more than fifty dates to find one that worked. They are now engaged.


Unknown said...

There is a Ted Talk with Amy Webb called "How I hacked online dating".


I don't know how different their methods were, though iirc there were some similarities. What's interesting is that she did this over a year ago.

George Haley said...

I'm not sure what definition of 'math genius' you're using that wouldn't include 'enrolled in doctoral program at very selective university' and 'gamed web page in clever way using statiscial analysis,' as evidence of being one. 'Genius' isn't really a rigorous, clinical term.

David Friedman said...

George: I was enrolled in a doctoral program at a university slightly more selective than UCLA—Chicago, physics not mathematics—and I wouldn't describe most of my fellow students as geniuses. It isn't a precise term, but neither does it mean "moderately smart person."

Russ Nelson said...

Girls can't do math, so they can't be math geniuses. QUOD ERAT DEMONSTRUM.

Anonymous said...

For a good fraction of the American people, voluntarily taking any course -- not to mention degree -- that requires math beyond high school algebra makes one a "math genius".

David Friedman said...


I can't offer real world examples of female math geniuses, although I expect they exist, but I have certainly known women who can do math. I like to claim that I fell in love with the woman I am now married to when I heard her explaining calculus to a fellow student.

If you are willing to accept an (I hope) believable fictional example, I think the female protagonist of my Salamander (available as a Kindle on Amazon) qualifies.

Power Child said...

I won't comment on what qualifies someone as a genius, but I'm certainly unimpressed with this doctoral mathematics candidate. Fifty dates...pathetic. Do regular OK Cupid users even take that long?

As an adult, I probably went on around a dozen dates--and with only...let's see...three women--before I started exclusively dating the woman who would become my wife.

I've never been strong at math (I bribed the teacher for a passing grade in my senior year of high school), but a passing glance at the numbers shows that I have done a lot better than this doctoral mathematics candidate, and with much less use of math.

(Plus, I've never used online dating--only the much less efficient "meatspace" social networks that were available to me as a student: of the three women I dated, two I met in class and one I approached cold at a restaurant.)

In fact, I'd expect a guy going for his doctorate in math who hacked the OK Cupid database and, analyzing the data with his expert math skills, constructed an optimized online dating profile, to find his perfect match in one or two dates, not fifty!

The report says he held himself to the constraint of not lying, so maybe he's just a thoroughly unappealing person.

Power Child said...

Here's the next big news story for you:

How A Kenetic Genius Hacked His Body to Move Across Room

Reilly Smahrt, a physical therapy doctoral candidate at Rutgers, hacked his body by analyzing and arranging the messages his brain was sending to his legs. By mining data from an expensive electroencephalograph, a lactic acid detector, and the fancy machine that goes ping, Smahrt was able to arrange a series of matrices showing where commands were originating in the brain, how they were flowing down through nerve cells, and how signals were being interpreted by telling muscle fibers to fire, then how these signals shot back up through the nervous system to the brain, confirming they had been executed. The doctoral candidate found that certain patterns of messages correlated with certain movements, and by arranging these messages in a certain order he was able to actually move his legs and walk across a room. Obviously the man is a genius, and he somewhat humbly brags that he was able to cross the room in only 50 baby steps, after only six months of research and effort.

RKN said...

Fifty dates...pathetic. Do regular OK Cupid users even take that long?

I thought something similar. Can't say for sure w/o knowing the relevant distributions, but 1 in 50 doesn't strike me as much better than random chance, leaving me incredulous as to the predictive power of his analytical approach.

I did find the pie chart display of answers across four questions and seven groups quite interesting.

Power Child said...


The average man dates far fewer than 50 women before getting married, so 1 in 50 is considerably worse than random chance.

This is why the whole story is so ludicrous to me: the "math genius" did worse using his hacked saber-metric model than an Average Joe would do using his semi-literate "I dunno...long walks on the beech [sic]?" approach.

The headline should have read "Egghead Dweeb Uses Egghead, Fails At Dating" but that would be Dog Bites Man.

Actually, this story is further evidence for my observation that journalism is just the studied art of lying.

jimbino said...

It's more than discouraging to live in an age that still considers marriage to be the Big Prize. I'd rather have a beer with Ayn Rand, George Eliot or Madalyn Murray O'Hair than date a woman keen on marriage, especially one who has a cat or dog.

Thank Darwin we still have male bondage!

Tibor said...

hudebnik: I would drop the word "American" and agree with you :) But in this sense, the word "genius" is synonymous to "weirdo".

Russ, David: (While I am pretty sure Russ is just trolling, I will react anyway) What about Emmy Noether?


George Haley said...

David: I think in colloquial, casual language 'genius' doesn't mean much more than 'moderately smart person.' I agree that beggars the term, to conflate qualifying for Mensa or finishing the NYT crossword puzzle or being the guy people call to fix their computer, with inventing the calculus or composing The Magic Flute or hiding secret codes about the lineage of Christ in the backgrounds of your paintings and doodles.

David Friedman said...

Several people argue that the number of dates it took him is evidence he was doing a bad job. But the claim was not that he was good at persuading women to fall in love with him or that he was a person who could easily find a match. It was that he had done a good job of using OK Cupid to present himself in a way that would make many women who were possible matches want to date him--which he apparently did.

Power Child said...


You said, "It took more than fifty dates to find one that worked." [emphasis mine]

From this I got the impression that his goal was to find a match, and that attracting women he was likely to find of interest was just a means to that end.

If he was really doing all this data analysis and applying his expert math skills, a huge proportion of (and thus one of the first few) women he attracted would fit the bill, without a need to continue with the "casting process."

Otherwise he might as well just say he likes long walks on the beach and so forth.


Marriage isn't the big prize, unless you're interested in a single act that is statistically likely to dramatically raise your chances of being happy, living longer, earning more, abstaining from crime and other antisocial activity, and a host of other desirable things.

But by all means, go have beer with Ayn Rand instead. Just don't pass on your genes; Darwin will thank you.

RKN said...

Amassing a long list of women who would likely date him was merely a means to an end, which pretty clearly was finding someone who would date him multiple times that would in turn lead to a long term relationship.

Seems to me the relevant question for judging the merits of his analytical approach is this: Is the probability of finding a woman like that among thousands who matched him at 99% higher than among the 100 (or so) who had matched him at 90%(pre-analytics)? It could be his method suffers from a high false positive rate, especially given that the woman he's now engaged to matched him at 91%.

David Friedman said...


If I understand how it works, the closeness of match depends largely on what questions you choose to answer, so it isn't a matter of a better match so much as of finding more of the women who match.

Anonymous said...

These criticisms are a little strange. To most people, a UCLA math grad student is a genius. And yes, some people find a math without ever going on dates or going online. How is that relevant to the story?

RKN said...

Yes, his analysis identified the right questions to ask to increase the size of the candidate list. Agreed.

But that wasn't his ultimate goal. His ultimate goal, I think, was to find a woman for a long term relationship, most likely one who would date him multiple times. I'm just saying it's not clear to me his analysis bought him much in terms of improved sensitivity. Who knows, he might have found a would-be fiancee among the women he matched (~100) on his original list, based on the questions he originally chose to answer before his analysis. And possibly in fewer dates.

Power Child said...

I'm confident that a survey of married men who had not hacked any databases, etc., to help them attract potential spouses would find that these men went on an average of far fewer than 50 dates before meeting the women who would become their wives. My confidence grows if the survey can be extended backward in history.

RKN says "it's not clear to me his analysis bought him much in terms of improved sensitivity."

I think it's in fact very likely that his analysis actually ended up costing him a lot of wasted time and effort.

Though, at least he got called a genius in Wired.

Rebecca Friedman said...

RKN, Power Child:

You're missing the point. The cost of finding a potential spouse consists of two parts*; the cost of finding a potential and persuading her you're worth getting to know, then the cost of getting to know her and determining whether it will work out long-term. What he short-circuited was the former; not the cost of going on a date, but the cost of getting a date. He said quite clearly that he wasn't getting women to return his messages initially, and was having trouble setting up any dates at all; once he'd adjusted his profile to be more optimal, he had no trouble at all getting dates. That left only the second part of the task, getting to know the potentials and figuring out which of them, if any, might work out long-term. The fact that this took him fifty dates when most people who succeed in finding a spouse take many fewer does not mean that he did "worse than average." After all, how many people who would like to be satisfactorily married aren't... because they never found the right person at all? The cost on that first part can really be a killer.

*One could technically divide it up lots of ways, depending on what one was discussing. For this context, though, this specific one works better than any other I can think of.

Power Child said...

@Rebecca Friedman:

Most people who successfully get dates from sites like OK Cupid have not hacked any databases or used any special math skills to optimize their profiles.

Even if a significant portion of OK Cupid users get zero dates, a significant portion also clearly get more than a few, and a few is (or should be) enough for most people to find a satisfactory spouse--especially when those few have already been vetted by a matchmaking algorithm rather than by simple meatspace proximity, e.g. walking up to someone attractive at a bar.