Gender disparities exists in many aspects of science: a fact. The question--or argument--is why. Many causes have been proposed. Almost all of them are non-mutually exclusive, and many of them are difficult if not impossible to separate. /1
Let’s say for a given cause, the weight of that cause on disparate outcomes can be anywhere from 0 to 100. What we are ostensibly arguing about is these weights. /2
There are some causes that I think have run their course and zeroed out, like the “early pipeline” explanation—it states that girls and/or young woman aren’t interested in or don’t pursue science for some reason. Its primary appeal was that research institutions can't fix it. /3
It turns out it is a bad explanation because disparities persist even in academic subjects where girls get better grades and women obtain undergraduate and graduate degrees at higher rates. /4
Most causes can’t be isolated. For example, women not choosing a specific discipline, or “choosing family over career,” cannot be separated from hostile environments or other factors. Encouragement and opportunities are important determinants for everyone. /5
But the most important complication, in my view, is that careers are non-Markovian. Advantages and disadvantages accumulate over time. When you try to test causes at one time point, all past causal factors come to bear as well. This is inherent in scientific assessment. /6
One of the most dramatic examples of this I’ve seen recently is a study of an early-career award. PIs who received scores either *just* above or *just* below the payline were compared for career outcomes. /7 pnas.org/content/115/19…
These 2 groups were indistinguishable in their “worthiness” for this award as ESIs—getting it or not was determined arbitrarily by the payline, which is just a function of how much total money was available that year. /8
Yet during the following 8 years, those just above the payline won twice the funding as those just below. Some of those just below left science. If you examined CVs at the endpoint, obviously the former group would exhibit more of that special something we call “excellence.” /9
This is a good example of the pervasive Matthew Effect, but it is unsurprising—positive feedback and amplification of small advantages is a given in a system where competing for new inputs is judged according to past outputs. /10
Having resources positions you to get more resources. Funding is a cause, not an effect, of “excellence.” /11
Small—even negligible, as in this case—differences can matter a lot, and matter more and more over time. Not all things are like this. Runners don’t get faster every time they win a race, but scientists effectively do. /12
The pattern we see with women in science—early parity with steadily increasing disparities at later career stages, is what is predicted by cumulative disadvantage—it doesn’t take much at each point—at play. /13
This is where ethical choices come in. If you believe that gender disparities are 100% explained by factors intrinsic to the qualities and choices of individual women, fine. (Well not fine, but it is the only way to get yourself logically off the hook.) /14
On the other hand, if you believe, that some—I don’t care how much—gender bias exists in scientific assessment, by any ethical standard you should see it as a serious problem and be trying to end it. /15
“Activists” don’t have to demonstrate that the disparity is entirely caused by bias, only that bias exists, because even small biases will have large, amplified effects in scientific careers for many people. /16
I will add that my “objective” beliefs about how bias and the Matthew Effect intersect in scientific assessment are reinforced by first hand observation of pretty much every flavor of gender bias, ranging from casual sexism to overt discrimination. /17
What are the chances that me, a white guy just minding his own business, would routinely see these things happening if they were rare? /18
That’s obviously anecdotal, but since the first-hand observations of others are in agreement, I trust my perspective. Also: patriarchy. The claim that gender bias would be completely absent in any aspect of our culture is an extraordinary one. /19
I find the explanation that bias is the cause of largest effect on disparate outcomes to be the most parsimonious and best-supported explanation. I realize people disagree. That’s fine, but again, it doesn’t get you off the hook. /20
In practical life, it is not required or even necessarily desirable to have a complete quantitative explanatory model for a phenomenon in order to act, especially where there is harm. In fact, it is often counter-productive. /21
We don’t need--and do not have--complete causal explanations or perfect predictions regarding climate change or cancer risks to know we should do something about them. We only need partial understanding. /22
This is why people who, in the name of “rigor,” quibble, question, flog alternative hypotheses, or demand some arbitrary standard of proof, are perceived as disingenuous. /23
Especially if you don't seem to understand the work you cite and just keep writing the same thing over and over and mailing it to different journals. Just as an example. /24
Bogging down in the minutiae of causal complexity is a distraction from our practical, ethical obligations. Even if you believe the biases are small, the effects can be large. There are credible reasons to believe harm is being done. We are obliged to try and prevent it. /fin
P.S. If your standard of proof/evidence in making an ethical judgement about preventing harm in practical situations in daily life is the same as that you would use for a clinical trial or theory of physics, you are definitely a dissembling, do-nothing, time waster.
