Reminds me of this GE ad:
It’s a decent theory, but unfortunately I don’t see how the numbers could possibly add up.
20/80 seems like a decent guess at the ration of women/men in CS programs, and by this theory we should be seeing the same ratio in the competition. It’s also a given that success is totally independent of gender. 26 people made it to the finals, so the odds of an all-male finalist group should be 0.8^26. That’s about 0.3%. The odds of that happening 14 times in a row is about 0.0000000000000000000000000000000005%.
Could the problem be with the 20/80 guess? Not really. Let’s take some absurdly low number, like just 5% of CS students being women. You’d still have just 26% odds of a single all-male finalist group, and 0.0000008% odds of having it happen 14 times in a row.
Only if contestants were randomly selected from all computer science majors. Which isn’t the case at all.
Sure, but I can only respond to the argument you actually made, not to all possible ones. So just to be clear, you’re switching your theory from “women are less likely to learn CS” to “women CS majors are less likely to enter this competition”?
So the competition gets some money, bragging rights and lots of attention from Google along with a free amount of endless harassment. So if a presumably young lady doesn’t want to work at Google and would rather not have a target on their back is there really is a reason to enter the competition? So of the women in CS you’re actually only pulling the women who want the opportunity and the shit that would come with it. That’s going to be a smaller sample size. How many actually enter?
I think it’s self-evident that they are not taking such a broad view of diversity. I mean, look at their own organizational efforts. How many are aimed at promoting diversity of political thought? You don’t have to be as smart as a Google engineer to put two and two together.
Maybe it’s incumbent on the board of directors. In private discussions.
But a low level employee making a public announcement that he doesn’t think Chinese customers are worth the effort, right after the company has started a massive PR campaign to increase its Chinese customer base? Yeah, that guy might easily get fired.
I mean, where do you draw the line? Is any employee with a brilliant idea permitted to blurt it out in public? Suppose a Chipotle line worker sends out his cost-benefit analysis of food safety, and argues that way too much time is spent on hand-washing and gloves given the very low likelihood of transmitting dangerous strains of E. coli by ordinary human fecal-food contact. Forget about whether his science is right or not. He works in your home town and his manifesto just made the evening news: “Local Chipotle worker says poo is not so dangerous to eat.” What do you think is gonna happen to him?
More women CS majors means that more women are likely to enter the competition, which means more women are likely to win the competition.
It doesn’t mean that women will achieve parity with men among the contestants or among the winners. There are obviously other factors at play. I mean, the US has zero first place prizes in the history of the contest. Belarus has five.
But I already showed that this can’t be explained just by the number of CS majors of each gender. You seemed to accept it, but now we’re back to the “more women CS majors” as the fix.
Sorry to ask again, but you’re actually not giving me a straight answer on a direct question. And I need to understand what argument I’m replying to. Is your proposed explanation that a randomly chosen female CS major is less likely to enter the competition than a randomly chose male one?
But surely they must, if there’s as big a pool of qualified men an women, they’re all equally likely to enter, and if success in the competition is totally uncorrelated with gender?
(Incidentally, this isn’t about “winners” even though you keep using that word, but about the finalists. Looking just at the winners wouldn’t be a great way to do this experiment. It ends up throwing away most of the data, and gets even more distorted by exceptional individuals. There’s only been 10 distinct winners.)
Ok, fear of harrasment is a good idea on a cultural reason for why women could be less interested in entering this competition. My main worries with that as an explanation are:
- Would doing well in the Code Jam really put a target on your back? I’ve never heard of any of these winners before now, let alone one of the people who “just” got to the finals.
- There are clearly women in this field who are willing to face the risk of being in the public eye in other cases, e.g. by presenting at conferences.
As for why you’d enter if you’re not interested in the rewards… I’ve done a lot of programming competitions over the year. It’s never been due to money or opportunities. It’s been since programming is fun, competing is fun, so combining them must be even more fun. In fact, doing these kinds of competitions just for the rewards would have been insane from an ROI perspective. (I’ve spent thousands of hours on various programming competitions. Most of them didn’t even have any reward other than bragging rights. Out of the ones that did, I think my total haul was $300 and an O’Reilly book of my choice).
Also, given the number of people who have repeatedly done well in the competition, it seems safe to say most of them aren’t in it just for the job interview…
I’d be surprised if Google released those numbers (or even had them). But even without those numbers, there is a calculation we can do to see if this explanation might work.
To have 50% odds of no women in the finals 14 years in a row, you’d need to have 0.2% of the contestants be women (that is, (0.998^26)^14 =~ 0.48). Given our earlier guess of 20% CS majors, that’d mean men are 100 times more likely to enter the competition. Could an extremely theoretical fear of harrassment really explain that kind of discrepancy? (“Extremely theoretical”, since even qualifying in this competition is so unlikely). I could believe a factor of 2; could be conviced of a factor of 5. But a factor of 100 seems pretty tough to explain.
How about this alternative theory, fairly consistent with Damore memo.
-
While on average men and woman have equal intelligence, with perhaps men being slightly superior in math, and women slightly superior in verbal skills, the distribution of these skills is significantly different with men having a longer tail, i.e. a higher standard deviation than women.
-
It is very rare for a smart woman to be only good in math. It is not uncommon for a man to have exceptional math ability but only moderate verbal skills. This constrains career choices for some men.
-
On average, men are more likely than woman to become focused/obsessed/single minded. Woman are better at multitasking and maintaining life balance.
-
Men and woman have difference preferences due to both biological and cultural factors, Men have a preference for working with things, women have a preference for working with people and helping.
-
As society has become more equalitarian and wealthier, preferences (e.g. " follow your passion") have become a more important factor in career choices for developed countries, than dollars.
So let’s see how these assumptions explain the work world of smart people.
The code jam winners are at the very far end of the ability top .01%, there are roughly 4 men for each woman at that level. The top prize is only $15K even for young people there probably better ways to earn money or scholarships. So even less woman are likely to compete.
In 1960 90% of vets were male, by 2010 female vets are a majority, today 80% of new vets are a woman. Vets earn more than a programmer, but less than doctors althought it is roughly as tough and expensive to become a vet as a doctor. A vet is consistently a dream job for young girls but not boys.
(Most dream jobs for both sexes, Actor, pop star, astronaut, pro athletes are very limited, most dream jobs are competitive pilot (boys), scientist (girl), engineer (boy) doctor, CEO, choreographer (girls), firefighter (boys),
At its peak in the mid-1980 woman were 37% of CS major today that number has been cut in half to 19%. (Harvey Mudd mentioned by the Stanford lecturer is an outlier and while Stanford is 30%, Cal an even bigger supplier to Google is at 19% for CS majors and 12% for EECS majors). In the 1980s, programming was decent paying but not particularly high status. I think two things have changed the status, tech billionaire and computer games. One of the few changes to dream jobs for young boys over the last 30 years, is Video game programmer/designer. Anecdotally, I have yet to meet a smart 11-17-year-old boy who doesn’t want to become a video game designer/programmer.
So I think what we’ve seen is not that girls are less interested in CS, (in absolute numbers the number of woman CS grads is flat). But rather a big increase in boys driven by video games and dreams of being the next Zuckerberg.
What I really wonder about is why do we care? I don’t think my cats care is hurt by the lack of men in the field. Are we really worse off that programmers are primarily written by men?
I am not sure if it’s a valid reason so much as it could be a reason. Assuming the programming circles overlap the gaming circles on some level, it seems like every time a lady achieves any sort of ranking in these professional epsorts arenas she experiences high level of harassment from both the participants and the viewers, and a few have been accused of cheating too. More than that, we know women game, we game a lot more than most our peers will admit despite the data that proves that people still doubt it… and you don’t see as many women in gaming competitions either. So why is that? Similar question right?
Yes, you wrote that the number of CS majors of each gender is insufficient to account for the differences in outcomes at the Code Jam. But that’s irrelevant, because the problem is most likely multifactorial.
So it doesn’t matter to me whether achieving a 50/50 gender ratio among CS majors would result in a 50/50 gender ratio among finalists. It would almost certainly increase the likelihood of seeing women finalists. Even if they only increased from 0% to 1%, it would be a worthy goal.
There are plenty of other indications that Americans underperform in this contest. This year, of 3280 American participants, 2293 qualified and 44 obtained a perfect score. In contrast, there were 932 Russian participants, of whom 837 qualified and 56 obtained a perfect score. Go ahead and run some statistics on that.
In short, I don’t think it’s reasonable to use Code Jam results to support a theory of innate differences in programming ability.
There are plenty of alternative explanations. Jsnell asked for just one, and one is all Google needs to justify its course of action. I bet they like mine better.
Everyone will answer this question differently. And long ago, people used to have different answers when asked, “Are we really worse off that all major league baseball players are white?”
But Google has clearly decided that they are indeed worse off, and I think I see why. Monocultures cannot adapt.
So they are perfectly justified in trying to change. You don’t have to participate in their efforts if you don’t care.
After accounting for the 20/80 ratio, we were still left with a sequence of events that appears to have a probability of 0.0000000000000000000000000000000005%. Clearly the number of CS majors of each gender has so little explanatory power that we can totally discount it. Getting a 50/50 gender ratio in CS would not put any kind of dent into that number.
It wouldn’t have that effect. But also, among all the reasons for trying to increase the number of women in CS, this seems like one of the least worthy possible ones. Would it be worth aiming for get a 99/1 ratio of CS majors in the hopes of raising the number of women in the finals to 50%? Of course not.
Google is trying to hire more women engineers because they think it’s best for the company. They’re probably right to do that, and I’m sure that their employees are on average as talented regardless of gender. But it’s totally absurd to suggest that they’re doing this in order “fix” the Code Jam,
Let’s remember that where this started was Menzo making the claims that Google was trying to fix women not having done well in that competition, which suggests some form of discrimination. Except he didn’t explain what form that discrimination could be taking. (Thanks to you and Nesrie for proposing explanations though; I just don’t think the numbers add up on them. To explain the results purely by participation rates, the difference would have to be enormous).
Could we explain this in some other way? Well, Menzo felt that any alternative was “suggesting that women can’t win the Code Jam because their brains aren’t wired for it or something”. Clearly nobody’s going to say that, or anything remotely equivalent to that. So the discussion has been poisoned now, it’s impossible for anyone to suggest any kind of biological differences as the explanation, no matter how subtle or innocuous.
I do not take “That women haven’t won the Code Jam is something they’re trying to fix” to mean that Google’s efforts are primarily directed at influencing Code Jam results. They are trying to improve women’s representation in tech at all levels, from CS majors to the workplace. To the extent that Code Jam results reflect women’s representation, that may also improve. You have made up some estimates as to how Code Jam results will change if Google succeeds, but I have no idea how you came up with them.
You can explain this in any number of ways, none of which are well validated.
And given the choice between an unvalidated hypothesis that improvement is possible, and an unvalidated hypothesis that improvement is impossible, I will generally prefer the former. If you want to be a pessimist, that’s your business. But please keep it to yourself while the rest of us try to make a difference. See also: climate change.
If a ship has a ten meter long tear in the hull, and you patch one meter of it, you have not fixed the ship. If these results are considered a problem, they clearly can’t be “fixed” by getting parity in CS enrollment. Hell, it seems incredibly unlikely that the results could even be perceptibly shifted.
I have not “made up” anything in general, just made back of the envelope calculations with assumptions that nobody seemed to find problematic. The math is just elementary probability and is already in my posts. But if you’re genuinely interested, I’d be happy to go through it step by step. Who knows, maybe I’ve made a mistake somewhere.
Wow. You want this forum to be such a total echo chamber that even things that are clearly incorrect can’t be corrected, and nothing can be debated?
- “Google is aiming to achieve world peace through their diversity program.”
- “Um, I don’t see how that could possibly work”
- “We don’t want pessimists here. World peace, or get the fuck out.”
It seems particularly inane to make this kind of argument since discussions on this forum will have literally zero effect on what Google does or does not do. You’re not making a difference here, nor is anyone else.
That’s true, because a ship with a nine meter hole is no more seaworthy than a ship with a ten meter hole.
On the other hand, a world in which precisely one woman is better off, but is otherwise exactly the same, is still a better world.
Your calculated value (0.0000000000000000000000000000000005%), is used to test the hypothesis that Code Jam finalists are a random sample of CS majors. That probability is low, so the hypothesis fails. Women are underrepresented. But so what? I presented a similar challenge: test the hypothesis that Code Jam qualifiers are a random sample of programmers worldwide. That hypothesis also fails (you can calculate the p value yourself). Americans are also underrepresented (or Russians are overrepresented).
You cannot conclude that women are innately worse at the Code Jam from your work, just as you cannot conclude that Americans are innately worse. You can only conclude that Americans, as well as women, have room for improvement.
You cannot conclude that increasing women’s representation in CS majors will not significantly change the composition of Code Jam finalists. Why? Because you haven’t shown why women are underrepresented. It could easily be a nonlinear function of their representation CS majors, e.g. you might very well start seeing big improvements at 25%, or maybe 30%, or whatever. Or maybe not. Nobody knows for sure.
Hmmm, maybe I didn’t put that well. I apologize, I was channeling Google but I didn’t make that clear at all.
If you are pessimistic in an organization dedicated to change, you will be asked to GTFO. QT3 isn’t Google and doesn’t have the same mission, so of course debate as you will.
I’m not sure that’s true. I’m guessing we don’t have these stats for the finalists as a whole anywhere, but the Wikipedia page shows Americans as having 3/42 of the top 3 finishes. That’s about 7%, which is pretty reasonable for a country with 5% of the world’s population.
Of course there’s going to be discrepancies, but they’re much smaller than in the gender case, and thus easier to manage. It’s not hard to accept e.g. that a programmer from Russia could be twice as likely to enter the competition than one from the US due to cultural or economical reasons. But 100x differences in participation are a lot harder to explain.
Sure. But what can be shown is that explanations based on participation rate need incredibly high differences in the rate. The difference would be so high that it should be easy to find the mechanism, rather than just hand wave it away as possibly being nonlinear somehow.
NPR had an interesting piece, where they pointed out some successes in bringing more women into CS.
Carnegie Mellon, one of the best CS programs in the world, managed to get the number of women in their program up to 40% of the population with some fairly minor changes to the way the classes were presented.
I was referring to the numbers for elimination, qualification, and perfect scores. As you recall, this year, of 3280 American participants, 2293 qualified and 44 obtained a perfect score. In contrast, there were 932 Russian participants, of whom 837 qualified and 56 obtained a perfect score.
By chance alone, this discrepancy would occur less than 10^-13 % of the time. That’s a lot of leading zeroes, yet nobody is seriously suggesting that Americans have genetic tendencies to be worse programmers, or that their performance cannot be improved.
The probability that all 26 finalists this year were male is p^26, where p is the fraction of male participants. If 90% of participants were male, which is plausible, then the probability of all male finalists is 6%. This is low, but not that low. Statistically it would still be regarded as a fluke rather than evidence of gender effect. I certainly don’t see support for your claim that gender effects are far greater than other effects.