Comments on: Bias against women persists in science

By: Phillip Helbig

Phillip Helbig — Mon, 24 Sep 2012 13:53:12 +0000

Here’s the Spiegel link: http://www.spiegel.de/karriere/ausland/diskriminierung-bei-bewerbungsfotos-a-826089.html

Summary: Pretty women who included photos in their application had worse chances because women in the personnel department wanted to keep them out. (Handsome men, in contrast, had an advantage.) This was an Israeli study. (Note that Israel has military conscription for women, a degree of equality rare elsewhere.)

After about 20 attempts this comment got through!

By: Phillip Helbig

Phillip Helbig — Mon, 24 Sep 2012 13:52:26 +0000

Here’s the Spiegel>/I> link: http://www.spiegel.de/karriere/ausland/diskriminierung-bei-bewerbungsfotos-a-826089.html

============

By: Phillip Helbig

Phillip Helbig — Mon, 24 Sep 2012 09:48:36 +0000

From the first link above:

“The study was based on evaluations by scientists of hypothetical student applications for a lab manager position, with the application materials identical in every way, except that half of the pool received applications with a male name and the other half received applications with a female name.”

and

“The scientists evaluating these applications (which were identical in every way except the gender of the “submitter”)”

See the contradiction?

As a control, one should a) use gender-neutral names (e.g. Tracy, Stacy, Jamie, Lee, Lynn, Sasha,…) or b) no names on the application. The latter has been tried with real applications in some countries, though primarily to eliminate racial bias (e.g. a Frenchman might prefer to hire Jean-Pierre to Mohamed).

By: Phillip Helbig

Phillip Helbig — Mon, 24 Sep 2012 09:40:59 +0000

“it would be better to have more women in physics and would never deliberately discriminate against women”

Does either one of these statements imply the other? I think all sensible people must agree with the second statement. However, in what sense is the first statement true, independent of the second statement?

…

By: Phillip Helbig

Phillip Helbig — Mon, 24 Sep 2012 09:40:02 +0000

I read about a similar story recently (in the online version of Der Spiegel, a German weekly news magazine, generally considered to be “serious”) concerning the fact that, as mentioned above, women tend to prefer to hire men, though this other example was not in the sciences. The conclusion of the researchers was that those women who had made it wanted to keep away the competition, and particularly opposed the hiring of young and pretty women.

By: Ted Bunn

Ted Bunn — Mon, 24 Sep 2012 03:01:46 +0000

Of course I agree that Bayesian inference is the right way to think about this — I was mostly playing around when I mentioned Mr. Jorgensen. But I don’t think that that changes things. You’re right that if P provides evidence for Q, then not-P must provide some amount of evidence for not-Q, but you can’t conclude anything about the amount of evidence. As the example I gave indicates, the amount of evidence may be arbitrarily small. So the answer to the question in your original comment (“Should we conclude that … ?”) is an unambiguous No.

I won’t attempt to read the authors’ minds, so I’ll speak for myself. I find the following proposition quite plausible:

(A) the fact that males and females both show bias is strong evidence that the bias is not largely conscious.

On the other hand, I find the inverse proposition, namely,

(B) if males and females did not both show bias, that would be strong evidence that the bias is largely conscious,

to be quite implausible.

For (A) to be incorrect, one would have to assume that women harbored large amounts of conscious bias against women — not impossible, but in my opinion unlikely. To conclude (B), on the other hand, one would have to believe that *unconscious* bias was more or less equally distributed across men and women (so that the unconscious-bias hypothesis would have low probability given the hypothesis). Since I have no reason to believe this extra assumption, I have no reason to draw conclusion (B).

By: Allen Downey

Allen Downey — Sun, 23 Sep 2012 23:36:08 +0000

How surprising to hear my Bayesian friend spouting Aristotelean nonsense! If you told me your cousin did not play in the NFL, that would in fact be evidence that she was female. In this example it’s very weak evidence, since you eliminated such a small fraction of the male population, but it would still be evidence.

In the less contrived example from the paper, the authors accept widespread bias as evidence against the hypothesis of malice. So I assume they believe that malice is more likely to produce bias among more focused groups, and less likely to produce widespread bias. Since they found the latter, they take it as evidence against the malice hypothesis. If they had found bias among men only, the authors would be consistent, if not correct, to take that result as evidence of malice.

Or at least I think it is fair for me to infer they would, based on their discussion.

By: Ted Bunn

Ted Bunn — Sun, 23 Sep 2012 22:41:14 +0000

That certainly doesn’t follow. As Mr. Jorgensen taught us in geometry, there’s no necessary logical relationship between a conditional statement (if P then Q) and its inverse (if not-P then not-Q).

If you told me your cousin played in the NFL, I’d infer he was male. If you told me your cousin did not play in the NFL, I would not make the opposite inference.

By: Allen Downey

Allen Downey — Sun, 23 Sep 2012 20:01:25 +0000

Interesting stuff, and important. But I have a beef with “The fact that faculty members’ bias was independent of their gender, scientific discipline, age and tenure status suggests that it is likely unintentional, generated from widespread cultural stereotypes rather than a conscious effort to harm women.” Does that mean that if the bias were limited to male faculty, we should conclude that it _is_ generated by a conscious effort to harm women?

(Note that this beef is with the PNAS article, not The Chronicle or Ted.)