News:

Welcome to the new (and now only) Fora!

Main Menu

Attrition in STEM

Started by jimbogumbo, October 05, 2024, 12:23:14 PM

Previous topic - Next topic

dismalist

Quote from: jimbogumbo on October 07, 2024, 09:22:07 AMfizzycist: we ar actually in complete agreement. Imo this article gives us no useful info regarding the drop off in female participation after high school or the early undergrad data. Physics and Math are essentially identical at those decision points.


Aside: I would greatly enjoy being a fly on the wall if a certain trio (picture a physicist, a mathematician and dismalist*) were passionately arguing.


* If dismalist is unavailable, substitute with a random economist.


A chemist, a physicist, and an economist are stranded on a deserted island. They were able to retrieve a can of tuna fish from the wreckage of their boat. How to open the can so that they can eat something?

The chemist suggests they start a fire and heat the can until it explodes. The physicist nods approvingly and says he'll calculate the trajectories of the pieces of tuna fish coming out of the can and put leaves on the appropriate places on the ground to catch them. Proud of their thinking, they turn to the economist, who is still lost in thought.

Well, they ask? The economist says: Let's assume we have a can opener!

I find discussion of group differences tedious in the extreme. There is no reason for two groups to be equal in anything. Men are taller than women. Men are over-represented in construction, sewer maintenance, and trucking. Blacks are over-represented in basketball.

Women's groups lobby to increase representation in all kinds of academic fields, but not in sanitation departments. The data set under discussion is used to find a problem. It's nothing more than the human proclivity to want more.




That's not even wrong!
--Wolfgang Pauli

marshwiggle

Quote from: dismalist on October 07, 2024, 10:11:39 AMI find discussion of group differences tedious in the extreme. There is no reason for two groups to be equal in anything. Men are taller than women. Men are over-represented in construction, sewer maintenance, and trucking. Blacks are over-represented in basketball.

I don't have any problem with noting group differences; my problem is with assuming that certain kinds of differences must be due to some nefarious intentions or actions of someone or other.

QuoteWomen's groups lobby to increase representation in all kinds of academic fields, but not in sanitation departments.

As noted, certain kinds of differences don't seem to need any explanation by the people who assume other differences are due to bad causes.
It takes so little to be above average.

eigen

Given that the "entry point" is 2 papers, and the "exit" point is not publishing for a year...

It's also possible that, say, the attrition in physics just happens before this study starts, and once people have "established" they're more likely to stay. There needs to be a parallel study of when someone has two papers published- for me this would be halfway through my PhD, for some fields this is before starting grad school.

I'd argue that the fact that the fields with the fewest women have the most similar attrition rates indicates that those fields lose people sooner, which fits well with the data I've seen for survival analysis of interest entering college -> graduating college -> finishing a PhD -> finishing a postdoc -> getting a faculty job.
Quote from: Caracal
Actually reading posts before responding to them seems to be a problem for a number of people on here...

Puget

It's an interesting paper, but as others have noted it seems flawed in a lot of ways. One of the biggest is that it doesn't differentiate between the reasons people stopped publishing, which is almost certainly really different between fields with a very clear non-academic career path (e.g., medicine, engineering, CS) and those without one (e.g., physics, pure math), with others (e.g., bio, neuro) somewhere in the middle. Those with clear non-academic paths are likely to have a ton of people who publish some as grad students (or even earlier) but never had any intention of continuing in academic research. That isn't really attrition to my mind.

Quote from: eigen on October 07, 2024, 11:04:09 AMGiven that the "entry point" is 2 papers, and the "exit" point is not publishing for a year...

It's also possible that, say, the attrition in physics just happens before this study starts, and once people have "established" they're more likely to stay. There needs to be a parallel study of when someone has two papers published- for me this would be halfway through my PhD, for some fields this is before starting grad school.

I'd argue that the fact that the fields with the fewest women have the most similar attrition rates indicates that those fields lose people sooner, which fits well with the data I've seen for survival analysis of interest entering college -> graduating college -> finishing a PhD -> finishing a postdoc -> getting a faculty job.

Yeah, I thought about that 2 pub bar as well -- that's going to be very sensitive to the authorship norms in different fields. There are some where literally everyone who worked on a project is listed and others where that isn't the case. There are also just different expected rates of publishing across different fields.

If I'm remembering my skim of the paper correctly, I think it was not publishing for 3 years (those with a publication in 2019 or later were considered "survivors" and data went through 2022). That is more reasonable, as gaps of a year or more can easily happen, especially during transitions from one role/job to another.
"Never get separated from your lunch. Never get separated from your friends. Never climb up anything you can't climb down."
–Best Colorado Peak Hikes

dismalist

#19
The discussion is settling on determining definitions. Once you have a name for something, you still don't know anything about it. Just observing differences constitutes measuring our ignorance. Then we attribute a cause -- unproven -- to the observed differences -- racism or sexism or disparitism or whateverism. Interest groups at work, manipulating language, nothing else.
That's not even wrong!
--Wolfgang Pauli

Hibush

I looked up a recent AIP study of change of major from Physics.

Deomgraphics. Followed students enrolled in Intro Physics freshman year (3,917). Of those 19% were considering a Physics major. Of those, 31% graduated with a physics major. (20% graduated with a different major (mostly similar majors like engineering, math, astronomy and computer science); 50% could not be found.)

They don't break down whether women chose the same majors.

"We found that 'pul' factors, positive experiences outside of physics, were most influential in students' decisions to leave a physics major. While many students enjoyed their college physics experience, they enjoyed the activities, courses, peer interactions, and potential employment in other subjects more."

In my opinion, that is a good outcome. They find the major that fits them best.

But we get the gender difference in the push factor:

"The study also found that women were not more likely to leave the major — but when they did leave, their reasons differed.
Women who left the major were more likely to say they performed worse on assignments than their peers, and were more likely to report that they encountered discrimination."  But we don't know if that was most of the women or just a few.

marshwiggle

Quote from: Hibush on October 07, 2024, 12:42:27 PMI looked up a recent AIP study of change of major from Physics.

Deomgraphics. Followed students enrolled in Intro Physics freshman year (3,917). Of those 19% were considering a Physics major. Of those, 31% graduated with a physics major. (20% graduated with a different major (mostly similar majors like engineering, math, astronomy and computer science); 50% could not be found.)

They don't break down whether women chose the same majors.

"We found that 'pul' factors, positive experiences outside of physics, were most influential in students' decisions to leave a physics major. While many students enjoyed their college physics experience, they enjoyed the activities, courses, peer interactions, and potential employment in other subjects more."

In my opinion, that is a good outcome. They find the major that fits them best.

But we get the gender difference in the push factor:

"The study also found that women were not more likely to leave the major — but when they did leave, their reasons differed.
Women who left the major were more likely to say they performed worse on assignments than their peers, and were more likely to report that they encountered discrimination."  But we don't know if that was most of the women or just a few.


It shouldn't be surprising if students who performed worse than their peers changed career path, regardless of whether they were male or female. That's about seeing the handwriting on the wall.


It takes so little to be above average.

jimbogumbo

Quote from: dismalist on October 07, 2024, 12:32:12 PMThe discussion is settling on determining definitions. Once you have a name for something, you still don't know anything about it. Just observing differences constitutes measuring our ignorance. Then we attribute a cause -- unproven -- to the observed differences -- racism or sexism or disparitism or whateverism. Interest groups at work, manipulating language, nothing else.

Respectfully disagree. In these areas if you see a non-trivial difference by (pick your poison) which is causing qualified people to leave your field you then try to determine if something you can adjust is the cause. That just makes good sense, and to me seems remarkably similar to making reasonable workplace accommodations to attract and retain qualified workers from underrepresented fields, Especially if you need qualified workers to have a successful enterprise.

In academia, especially research and teaching fields, it's not so much having enough people; it's more that while you of course want the "best" ones, you don't want to hurt your chances of discouraging some of the very few who might really make a major difference.

marshwiggle

Quote from: jimbogumbo on October 07, 2024, 01:31:21 PM
Quote from: dismalist on October 07, 2024, 12:32:12 PMThe discussion is settling on determining definitions. Once you have a name for something, you still don't know anything about it. Just observing differences constitutes measuring our ignorance. Then we attribute a cause -- unproven -- to the observed differences -- racism or sexism or disparitism or whateverism. Interest groups at work, manipulating language, nothing else.

Respectfully disagree. In these areas if you see a non-trivial difference by (pick your poison) which is causing qualified people to leave your field you then try to determine if something you can adjust is the cause. That just makes good sense, and to me seems remarkably similar to making reasonable workplace accommodations to attract and retain qualified workers from underrepresented fields, Especially if you need qualified workers to have a successful enterprise.

In academia, especially research and teaching fields, it's not so much having enough people; it's more that while you of course want the "best" ones, you don't want to hurt your chances of discouraging some of the very few who might really make a major difference.

That is true, but the problem with many investigations into possible discrimination is that they do the opposite of what is normal scientific practice. In normal research to look for an effect, you try hard to think of any other things that might be causing the effect, and eliminate them before assuming the thing you're looking for is *responsible. But when it comes to "discrimination", it's basically the first thing assumed, and any attempt to look for other causes is suspect, and if there seem to be other factors then the "discrimination" is just assumed to be more subtle and in some other part of the process.


(*i.e. The "null hypothesis" is no effect; you have to have positive evidence for an effect beyond what can be explained otherwise.)
It takes so little to be above average.

dismalist

#24
Absolutely, Marsh! But we're not doing science. We're doing politics.

Shit, physicists have it easy! Murray Gell-Mann, the Nobel laureate who conceived the quark, famously explained the difficulty to his colleagues: "Imagine how hard physics would be if electrons could think." Given the limitations economists face in performing experiments, perhaps an even better question would have been "imagine how hard experimental physics would be if electrons had human rights." (Gell-Mann is also credited with saying that, compared with economics; the "hard" sciences are easy.) [duplicative language without appropriate attribution.]

And trust me, you don't want economists, psychologists, or -- God forbid -- sociologists doing large scale experiments! Hell, let's test monetary theories by trying to plunge half the population into a depression! [But only half.]

All of this is just the grown up two year old announcing "I want more", and the means are irrelevant. Politics, using language as an instrument.
That's not even wrong!
--Wolfgang Pauli

jimbogumbo

Quote from: marshwiggle on October 07, 2024, 01:43:31 PM
Quote from: jimbogumbo on October 07, 2024, 01:31:21 PM
Quote from: dismalist on October 07, 2024, 12:32:12 PMThe discussion is settling on determining definitions. Once you have a name for something, you still don't know anything about it. Just observing differences constitutes measuring our ignorance. Then we attribute a cause -- unproven -- to the observed differences -- racism or sexism or disparitism or whateverism. Interest groups at work, manipulating language, nothing else.

Respectfully disagree. In these areas if you see a non-trivial difference by (pick your poison) which is causing qualified people to leave your field you then try to determine if something you can adjust is the cause. That just makes good sense, and to me seems remarkably similar to making reasonable workplace accommodations to attract and retain qualified workers from underrepresented fields, Especially if you need qualified workers to have a successful enterprise.

In academia, especially research and teaching fields, it's not so much having enough people; it's more that while you of course want the "best" ones, you don't want to hurt your chances of discouraging some of the very few who might really make a major difference.

That is true, but the problem with many investigations into possible discrimination is that they do the opposite of what is normal scientific practice. In normal research to look for an effect, you try hard to think of any other things that might be causing the effect, and eliminate them before assuming the thing you're looking for is *responsible. But when it comes to "discrimination", it's basically the first thing assumed, and any attempt to look for other causes is suspect, and if there seem to be other factors then the "discrimination" is just assumed to be more subtle and in some other part of the process.


(*i.e. The "null hypothesis" is no effect; you have to have positive evidence for an effect beyond what can be explained otherwise.)


Pretty sure I know what a null hypothesis is (statistics teacher here). Of course you are correct about what is often done; I was describing what I think should be done, and explaining in response to dismalist why I think it's important.


And yes d, I want economists to do what they've always done. Imagine a curve, come up with a cause-effect model prior to gathering data, and then have political parties or various isms glom on to the model as fact. Only slightly less horrible than what you describe.


The sentence above is only slightly sarcastic. There doesn't seem to be a happy medium between the two situations you and I describe.

Hibush

#26
Quote from: dismalist on October 07, 2024, 02:24:41 PMAnd trust me, you don't want economists, psychologists, or -- God forbid -- sociologists doing large scale experiments! Hell, let's test monetary theories by trying to plunge half the population into a depression! [But only half.]

Given the authoritarian governments around the world these days, there should be opportunities to assign people to treatments randomly. The conclusions would be uncommonly robust, albeit only applicable to societies under authoritarian regimes. What an opportunity to get some good pubs!

Seriously though, there must be "natural experiments" happening around the world that allow more insightful economic and behavioral research. They'd be a lot more informative than the ones that put undergraduates in artificial situations and claim they mean something about people in general. If I was stuck doing that kind of research in my REU, I'd switch to a major where I could do destructive sampling.

marshwiggle

Quote from: Hibush on October 07, 2024, 03:21:11 PM
Quote from: dismalist on October 07, 2024, 02:24:41 PMAnd trust me, you don't want economists, psychologists, or -- God forbid -- sociologists doing large scale experiments! Hell, let's test monetary theories by trying to plunge half the population into a depression! [But only half.]

Given the authoritarian governments around the world these days, there should be opportunities to assign people to treatments randomly. The conclusions would be uncommonly robust, albeit only applicable to societies under authoritarian regimes. What an opportunity to get some good pubs!

Seriously though, there must be "natural experiments" happening around the world that allow more insightful economic and behavioral research.
They'd be a lot more informative than the ones that put undergraduates in artificial situations and claim they mean something about people in general. If I was stuck doing that kind of research in my REU, I'd switch to a major where I could do destructive sampling.

Here's a suggestion: Survey students in a course about how much they enjoyed it, and how likely they are to take a followup course in the discipline. THEN see how many people actually take followup courses in the discipline, and look for variations.

My hunch is that you'd see variations in the followup rate based on various factors. If the followup rate for any group is below what would be predicted by the data on how likely they were to take followup courses, then it's much more likely that it's due to something other than being turned off the subject between one course and the next.
It takes so little to be above average.

dismalist

Oh, my God, self reported studies of motivation!

What sounds better to the world and especially to oneself: I flunked physics [or econ, for that matter] or I've been discriminated against?

"Loser, loser, double loser, as if, whatever, get the picture, duh!"
That's not even wrong!
--Wolfgang Pauli

marshwiggle

Quote from: dismalist on October 07, 2024, 04:05:31 PMOh, my God, self reported studies of motivation!

What sounds better to the world and especially to oneself: I flunked physics [or econ, for that matter] or I've been discriminated against?

"Loser, loser, double loser, as if, whatever, get the picture, duh!"

My point is that even between groups self-reporting the same level of motivation, there are likely to be different outcomes. All people aren't the same in level of follow-through, and that has nothing to do with whether they feel discriminated against or not.
It takes so little to be above average.