## Nickel And Diming Women of Science January 9, 2007

Posted by gordonwatts in physics life, university.

Earlier this year a survey was sent around to all members of our department with the question “could diversity be better in physics?” I don’t see how any sane person could answer no — we have 20% women or so — and, indeed, the department answered as expected. So we had a department-wide (and well attended) meeting last Friday to discuss the issues. One of the introductory slides put up by Marjorie Olmstead made the point about unconscious bias in a way I’d never seen before.

Lets say when you evaluate a woman vs. a man you are only slightly biased. If the man an woman were equal you’d hope you’d go 50-50 in one or the other’s favor. But your bias means you go 52-48 in favor of the man, or 48/52 =0.92. Now, over the course of a career a man or woman is judged many times (interviews, tests, etc.). If that happens 5 times then (48/52)^5 = 0.67, or 10 times it is less than 50%! So a subtle unconscious bias applied 10 times over can make a huge difference! And it would probably be very difficult to point to a specific instance of bias in any one of those 5 or 10 decisions because the bias was fairly subtle.

1. Carl Brannen - January 9, 2007

If the effect is accumulates, and 48% is a conservative estimate for the bias, then a woman who has published 100 papers, each of which was sent to an editor and 3 peers for review, and each of whose work was considered for citation by 5 other authors, has collected about 900 biased decisions. That means she must be at least (52/48)^{900} = 19,315,040,952,540,206,369,157,345,708,234 times as good as a man. (And probably understands significant digits better than I.)

Since there are women physicists who, in fact, have published 100 papers, the only conclusion I can come to is that women have already solved the unified field theory and they are holding out on us, If they told us the secret, we’d just make weapons with it anyway.

2. sexist - January 10, 2007

There are no women physicists. A “woman physicist” is either not a woman or not a physicist, it just can’t be both. Incidentally “man physicist” is becoming more of an oxymoron as well, as the field loses its appeal. This is more of “freaks & geeks show” now.

Check out percentage of women (yeah _women_) involved in HEP in Italy. I have a suspicion that italian males just not too excieted about the salaries.

3. Carl Brannen - January 10, 2007

When I first read the post I thought the hokey statistics were bad. After I got stuck parking west bound on 520 during a windstorm, I got to thinking about it some more, and the whole thing makes me think of a tape recording of a crime lord at confession — talking about how he had spend a sinful half hour on Wednesday not thinking more about other people than himself.

Anyone at a physics department whose bias is within 10% of the fair mark deserves a Mother Theresa medal. If decisions by physicists really were within 2% of fair, then how could it be that so many papers get sent to 3 or more referees and the referees end up with completely divergent opinions on it? Why is there a shouting match going on about string theory?

And if the largest bias a physicist has is against women, well I’ll be a monkey’s uncle. There are numerous sociological studies that show that physicists don’t follow the vast majority of the papers they read. Life is too short and math is too complicated. Instead, they look at the author, and if they don’t know the author, they look at where they work or went to school. The physicist who could have done what Hardy did when he read Ramanujan’s uneducated scrawls and glimpsed genius, are few and far between.

And how the heck are you going to correct a 2% bias? It’s utterly impossible. Humans are inherently biased, they use emotions to make most of their decisions and then find arguments to back it up. Knowing your bias to a 2% level is also impossible. All you can do is feel guilty about it.

The net result will be that the next time you unfairly conclude that a woman’s paper is incorrect (without fully understanding it) you will allow one extra misunderstanding, or minor typo, before you make the same conclusion you’d have made anyway. And then feel proud for being fair.

And I’m not saying it could be any other way. If physicists read each others papers with the care that a grad student needs to exercise when learning new material, they’d never have any time to do any physics. Life is short.

4. smart arse - January 10, 2007

Well, this just doesn’t make sense. Say I have two kids – twins – and I have 0.01% bias towards one of them, and as a parent I make about 5 decisions about my kids each day (a conservative estimate), then by the time they are 10 the “favourite” shall be doing ~1500 times better than the unfortunate one. Which is simply ridiculous. Gordon, you are a scientist, aren’t you? How can you be fooled by such non-science?

So let’s look at the interview example. Say we have two candidates, a male and a female, who both visit 10 places. We assume that the male candidate has a 10% chance of getting an offer from any individual institution. Given the 48/52 bias, the female would have a 9.2% chance. Splendid. Now, the probability that the male candidate will get offers from ALL of the places (we assume, and hope, that their decisions are uncorrelated) is 0.10^10 = 10^-10. Small, isn’t it? For our female candidate we shall have (0.092)^10 = 0.43*10^-10, or less than one-half that of male’s (as advertised in Gordon’s example). But is it relevant? No. A more relevant question to ask: what is the probability that our candidates get at least one offer each. And that one is easy, too: 1-0.9^10=0.65 for the male and 1-0.908^10=0.62 for the female. 65% vs 62%, which, by the way, is a ratio of about 0.95, better than the original odds. If they interview at 100 places, their chances of getting a job are almost equal!

The bottom line – a really bad argument, will only work well on the uneducated.

5. Ann Nelson - January 10, 2007

Smart Arse is right when it comes to parallel decisions and judgements (judgements which dont influence each other). However in the career of a physicist there are many sequential deicision which do influence each other–i.e. whether you get into a top school influences whether you get the best postdoc which influences whether you get the best faculty job which influences whether you have the opportunity to do the best research which influences whether you win awards etc. Women and men do seem to to about equally well at landing facutly jobs according to the AIP statistics, but the statistics also show that the more prestigious jobs go more often to men, with women ending up disproportionately at small teaching schools and lower ranked colleges. I dont know how to actually calculate the accumulated disadvantage since we dont actually know what the amount of unconcious bias is. We also dont know how well the current system of affirmative action does in counteracting it. I wonder whether affirmative action actually increases the bias since many people will assume conciously or unconciously that any success a minority has had must be due to affirmative action.

In conclusion the calculation Marjie gave has to be regarded as a toy model which is not implausible.

I conclude that it is important that we scrutinize and analyze all our judgements and decisions carefully. One game I play when refereeing a paper is to imagine that the author is someone well known and respected and see whether that changes my opinion. When evaluating a candidate i try and use multiple metrics and also try imagining that the application came from a different institution, that the gender and race were different etc. I also look at the research accomplishments independently and see it the letters of recommendation seem fair when compared with what the person has actually accomplished. It is hard work but I think we all have to do it.

I have seen many lazy and bad decisions, eg where a decision was made between 2 very comparable faculty candidates, one male and one female, based on the committees’ gut reaction that the male ‘seemed like more of a leader’, and ‘seemed like a good guy’. That decision turned out to be a clear mistake.

6. gimme a break - January 10, 2007

Lack of (gender) diversity in physics does not necessarily mean that the “old white male establishment” is biased against women. It could very well be that. But it could also be a *combination* of many other things. It could mean e.g. that women are *less interested* in physics. Maybe for cultural reasons. Maybe for biological reasons. Women get as many PhDs as men do in biology, biophysics, medical schools. Who do you think the establishment in those fields is? Young black females? Jeeezzzz….

7. Gordon Watts - January 10, 2007

Wow. This post generated a lot of heat.

Smart Arse — Ann has already responded, I think. Her statement “The calculation Marjie gave has to be regarded as a toy model” has got to be the case. As Ann points out (and others make a reference to) different decisions will have different bias built in.

The only data we have is the final outcome. I don’t know of any way of measuring individual bias — especially if it was as small as 2% — it would be hard. Studies have been done: for example, referees reading a paper with the names changed (I’ll try to get that reference) — where it made a clear difference in the out come.

This toy model is simply there to make a point. And the point it makes is impressive (I think): It doesn’t take much!

8. Gordon Watts - January 10, 2007

gimme — I’m sure it is a combination. Perhaps cultural (or why is it that other contries have better fractions of women and minorities). Whatever the reason — I want to know them all!

9. gimme a break #2 - January 11, 2007

This is an article you may find interesting:
http://www.usatoday.com/news/opinion/2005-04-11-girls-math-forum_x.htm

(probably getting off-topic here)