Tracking sex bias in neuroscience conferences
August 31, 2015
A Tweep directed my attention to biaswatchneuro.com of which the About page says:
The progress of science is best served when conferences include a panel of speakers that is representative of the field. Male-dominated conference programs are generally not representing their field, missing out on important scientific findings, and are one important factor contributing to the “brain-drain” of talented female scientists from the scientific workforce. As a group, BiasWatchNeuro has formed to encourage conference organizers to make every effort to have their program reflect the composition of their field.
Send information about conferences, seminar series or other scientific programs to biaswatchneuro@gmail.com
Check it out.
August 31, 2015 at 6:59 pm
(1) These fucken azzholes need to spread their scores. Giving a zero to any conference in which the number of women is below the base rate means that every conference gets a zero. Do they really believe that a conference with a 0:22 F:M ratio is equally as bad as a conference that is 9:16?
(2) These fucken azzholes need to enable comments on their blogge.
LikeLike
August 31, 2015 at 7:05 pm
I see what you mean. Yeah, this scale need tweaking. Too easy to attack as it stands.
LikeLike
August 31, 2015 at 7:13 pm
What does “first PI” mean? Last author?
LikeLike
August 31, 2015 at 7:34 pm
So, while I support the effort, their method of estimating the female percentage of the field is almost nefariously simplistic.
For one, they’re not adjusting for the search’s alphabetical order bias. I got the same percentage as them (8/25, or 32%) by NIH reportering computational + neuroscience, without adjusting anything else and picking the first page. However, depending on which page of 25 results you pick, you get 3/25, 2/25, 5/25, or 4/25 for female representation in those search terms (+/- some ambiguous names). Likewise the “orbitofrontal cortex” search returns 45% female in the first half of the results (if you treat some of the ambiguous names as female, e.g Leslie), but only 28% in the second half (13/46). In this case, this would mean the conference, with 9/25 speakers being women, was actually above the representation (25% actual vs 36% represented).
There might be an over-representation of females at the beginning of the alphabet in science (and so in this methodology). This seems plausible, given ethnic distributions of first letter of surnames (e.g. 10% of the 100 most popular names in the US start with A or B, versus none of the Chinese surnames) and the correlates this might have with amount of women in science.
They should at least randomize the page they select, if they are in fact picking the first one (which seems like the case).
Obviously this doesn’t really matter in the case of conferences with 1 female out of 20 speakers, but it is important to not impugn conference organizers making an honest effort.
LikeLike
August 31, 2015 at 7:41 pm
^third reviewer
LikeLike
August 31, 2015 at 7:54 pm
Thanks for spreading the word, DrugMonkey.
@Comradde PhysioProffe — We’d love to give higher scores to conferences, if only they actually came within a margin of error (+/-5%?) of the base rate. As scientists, we all know what a biased sample is, and if a conference is below base rate (even if it has 36% females), it is biased. It is hard to sugar-coat that.
Please suggest conferences that have base rate or above to biaswatchneuro@gmail.com. That is also the address for alternative measures of base rate etc.
@Selerax — first PI is the first PI that is named on a grant (not drilling in to the “et al.” in the search results)
@jmz4gtu — Thanks for pointing out that the order in NIH RePORTER is alphabetical rather than by date. But note that we did not assume anything about ambiguous names. We searched for these people’s websites one by one. Yes, it takes time and work — and we’d appreciate any help we can get. If you have better estimates, please send them along.
As mentioned on biaswatchneuro.com — please send alternative estimates of base rates in different subfields to biaswatchneuro@gmail.com. This effort is in it’s infancy, and we are trying to get the most scientifically accurate data out there, but it is not easy.
LikeLike
August 31, 2015 at 8:04 pm
At a connectomics conference last year it was a bit eyebrow-raising to see only 3 women (iirc) out of 30 people who spoke or presented. The one female PI to speak even dryly referenced feeling like a token invitee (spoken tongue in cheek as a reference both to being a woman and as a representative of the kind of work her lab does). The craziest part is that 10% actually was representative of the number of women at that conference. Not sure about the field overall, just something I noticed.
LikeLike
August 31, 2015 at 8:51 pm
“We’d love to give higher scores to conferences, if only they actually came within a margin of error (+/-5%?) of the base rate. As scientists, we all know what a biased sample is, and if a conference is below base rate (even if it has 36% females), it is biased. It is hard to sugar-coat that.”
For the love of fucken godde, are you really this fucken oblivious? You made up your own fucken scoring scale. It wasn’t handed down to you like the ten motherfucken commandments.
Answer me this. Do you really consider 0:22 and 9:16 conferences equally bad? Because if you don’t, and you consider the former worse than the latter, then you need to rethink your fucken scoring system.
LikeLike
August 31, 2015 at 9:16 pm
“But note that we did not assume anything about ambiguous names. We searched for these people’s websites one by one.”
-Sorry, I meant *I* was assuming, based on the numbers you got, that some of the ambiguous names were, in fact, female.
Another thing to consider is that there is a much higher proportion of females at the trainee level, so for some conferences (e.g. large ones that let trainees talk) you might want to just do really broad terms and focus on trainee level awards, since this cohort might have different demographics. If it is PIs, you might want to choose to set your grant search back a couple years to make sure you’re examining equivalent populations (i.e PIs with standing in the field).
But yeah, it is generally a problem getting good data on workforce demographics in the sciences. Best of luck.
@DM
Reviewer 3? But I didn’t even once call for them to be drummed out of science and shunned by all mankind!
LikeLike
August 31, 2015 at 10:22 pm
Why do they calculate the SD as 2*sqrt(p(1-p)/n). I’ve never seen the factor of 2 before in the SD of a proportion.
LikeLike
August 31, 2015 at 10:37 pm
Did the base rate exclude PIs of fellowship grants? If not, then for conferences that have only faculty-level speakers, the bias index would be, well, biased.
LikeLike
September 1, 2015 at 7:17 am
You should export all of the results to a csv table, use your stat tool of choice to randomly sample as many rows as you can handle to manually evaluate. Buy pizza for a friend to help for an hour or two to double your sampling. Not hard.
LikeLike
September 1, 2015 at 9:55 am
I agree with Physioprof that the perfect should always be the enemy of the good.
LikeLike
September 1, 2015 at 10:09 am
This is “good”? Where everything below 5% above the base rate gets the same score? Cmon.
LikeLike
September 1, 2015 at 10:20 am
My contention is that the useful should be the enemy of the useless. Any scoring rubric that in practice gives the exact same score to everything is useless. Why give scores at all if everything gets the same one?
BTW, my guess is that this blogge is run by well-meaning obsessive absolutist clueless dudes. Think MBEisen crossed with Perlstain. Because women would not be so stoopid.
LikeLike
September 1, 2015 at 11:05 am
Sure, their scoring metric is shitty. And I do actually think that the suggestions provided here will serve to improve the scoring, and the estimates of female/male participation in the field.
But I think that the main utility of this website is the breakdown of women/men at actual meetings, pointing out the ones that are doing better than others (which I think is CPP’s main point, that 9/16 should be honored, not found wanting, which there is a decent argument for.)
Should we be striving for the precise representation of speakers that there is at the PI level? Only at the level of tenured professors at the meeting? Or at the level of all attendees of the meeting? All of those will produce different women/men ratios, and there are arguments to be made for all of them being the “correct” speaker representation.
Maybe at first pass, the best thing would be to put the ratios for as many neuroscience conferences as possible on there, and rank order them. How many GRC this summer could we have data for? How many SFN symposia, or ACNP symposia? Would we find that GRC tend to do a better job with this than ACNP (knock me over with a feather)? Would that data help sexist orgs clean up their act with a quickness?
All I know is, we’re not gonna find out without a database like this.
LikeLike
September 1, 2015 at 11:09 am
Five point scale with 3 anchored to the base rate. 0 reserved as a N/A off the chart score for no women included.
LikeLike
September 1, 2015 at 11:14 am
why restrict such tools only to neuroscience conferences. The exercise can be extended to other fields as well.
LikeLike
September 1, 2015 at 11:54 am
which I think is CPP’s main point, that 9/16 should be honored, not found wanting
That is not my point at all. I have not asserted any position whatsoever on whether 9:16 is good, bad , or neutral. My only point is that if one considers there to be a difference between 0:22 and 9:16, then a scoring rubric that gives them the same score is fucken absurd.
LikeLike
September 1, 2015 at 1:17 pm
I think it is mistaken to interpret the ‘bwn’ rating as a numerical score – instead it is a numerical label for a categorisation: parity (3), above base rate (2), around base rate (1), below base rate (0). I agree that it might have been better to have a separate category for no women at all. But given the raw data (the actual ratios) are also given, is it really necessary to focus on complaining about the ‘scoring rubric’? If being really fussy, one might complain that numeric labels should not be used for categories (e.g. it might be a problem if someone started reporting the ‘average’ bwn or similar…). Note there are already conferences listed that refute PP’s contention that this rubric “in practice gives the exact same score to everything”…
LikeLike
September 1, 2015 at 1:35 pm
I think they changed their scale in the last 24 h, no? I could swear it used to be a zero all the way up to slightly above the base rate.
LikeLike
September 1, 2015 at 2:17 pm
They changed their scoring rubric without even mentioning it. You know what? The goal is obviously commendable, but fucke these hamhanded assholes. This is just goddamned idiocy.
And BTW, what is with this “females”/”males” business? We’re not talking about fucken laboratory animals; we’re talking about “men” and “women”. The only people who seem to refer to women as “females” are fucked up MRA types.
I’m done with this idiocy.
LikeLike
September 1, 2015 at 3:25 pm
Not trying to nickpick here, but I don’t understand how this new scoring system would work. Aren’t there ambiguities?
3 – Gender parity — ok, 50% female, 50% male
2 – 5% or more above base rate. So if a field had a base rate of 47%, then a 50/50 conference would score a 3, but a 53/47 conference would be a 2. What?
1 – within 95% confidence interval of the base rate. What happens when the 95% CI is at least 5% above the base rate?
I think I understand what they’re trying to do, but why not just say
3 – 50%+ female
2 – >(base rate + 5%)
1 – base rate +/- 5%
0 – <(base rate – 5%)
and not try to get all fancy.
LikeLike
September 1, 2015 at 3:37 pm
” that 9/16 should be honored”
That’s also confusing. There’s an actual conference listed on the site that mentions these exact values (“Invited speaker gender ratio: 9 Females / 16 Males (36%)”). The proper terminology would be 9 Females:16 Males or 9 Females/25 Total. 9/16 is just incorrect enough to be confusing.
And I’m with CPP: we’re talking about men and women here, not males and females.
LikeLike
September 1, 2015 at 5:42 pm
Maybe they are Ferengi?
LikeLike
September 1, 2015 at 10:19 pm
I’d be curious to see how the composition of the organizing committees reflect the composition of the invited speakers.
LikeLike
September 2, 2015 at 2:23 am
damn cpp is quite over-rustled about this one…
seriously though the rubric isn’t even necessary – the ratio of ratios as it were is enough information for the reader to draw their own conclusions.
LikeLike
September 2, 2015 at 1:01 pm
We actively go out of our way to ensure as equal a M:F speaker/ attendee ratio as we possibly can for our relatively small society meetings. However, the one issue we’ve never been able to address is how mighty white are our conferees. Often only a single black person out of 200+ attendees! I suspect this is a reflection of how few black people have been hired into the field rather than any active bias about not inviting them, or them not being invited by their lab heads.
LikeLike
September 2, 2015 at 2:50 pm
A bit off topic, but since we are talking about conferences . . .
This guy is a top US counterintelligence official. He exchanges documents with a Chinese national and thinks nothing of it . . .
But then, a week ago, he says, he got another message from China via his email account at George Washington University, where he has lectured on national security law since 2003.
“It was apparently from a university in China asking me come to speak at a conference on the environment”—not even remotely one of his areas of expertise, Bowman says. He called the FBI.
http://www.newsweek.com/china-hackers-fbi-marion-bowman-367451
LikeLike
September 2, 2015 at 3:29 pm
That is…..amazing.
LikeLike
September 2, 2015 at 4:03 pm
Pardon the obtuseness, but what? He got an email inviting him to a conference and so he called the FBI? I must be missing something.
LikeLike
September 2, 2015 at 4:17 pm
If that isn’t clear evidence of being hacked, what is?
LikeLike
September 2, 2015 at 4:18 pm
The fact I am not encouraging everyone to report their China conference Dear Esteemed Scientist emails to their local FBI office shows that I do have a line.
LikeLike
September 2, 2015 at 7:27 pm
Jeezus fucke! I get several of those per motherfucken DAY!
LikeLike
September 2, 2015 at 7:50 pm
But what’s funny about it is that in his case, he actually does appear to have had malware. Maybe we are all sending our incredibly boring emails to the Chinese. It is like having online notebooks, except without all the hassle.
LikeLike
September 2, 2015 at 9:04 pm
Correlation / causation error.
LikeLike
September 3, 2015 at 3:59 am
A great consciousness raiser and thanks for link!
What I’ve been doing is when I decline an invitation is to recommend at least two women to speak instead of me. Surprisingly I received a symposium proposal yesterday from a woman who invited 4 men..and no women (there are many who could have been invited)
LikeLike
September 3, 2015 at 10:50 am
That is as unsurprising as black cops who brutalize black community members. Systematic bias is systematic, yo.
And look… When you try to organize a seminar/symposium schedule, sometimes the organizer is using that to their own advantage. Namely to meet VIPs. If the perception is that 4 men are more likely to enhance the organizer’s career then…..
LikeLike
September 5, 2015 at 9:27 pm
Thanks again for all the comments and feedback here and elsewhere. Biaswatchneuro is (still) a work in progress, and in the past few days has undergone an overhaul. In particular, the ratings (0-5) now reflect distance (in standard deviations) from the base rate, they are implemented as categories (to allow one to pull up all conferences with a certain rating), and there are new pages with best practices and how to help suggestions.
We hope to hear more feedback, preferable directly to biaswatchneuro@gmail.com, but we also follow the comments here. And please spread the word and send conferences our way! Thanks.
LikeLike
September 6, 2015 at 3:19 am
Thank you for this.
LikeLike
September 6, 2015 at 8:07 am
So far, based on an admittedly small sample, the median conference reported on this site has more women than the base rate. And instead of being azzholes and exploiting the comments here, why don’t these azzholes open the comments on their blogge?
LikeLike
September 9, 2015 at 3:15 pm
[…] were just talking about a site which tracks conferences in the neurosciences to determine how well their speaker list […]
LikeLike
September 9, 2015 at 3:20 pm
current biaswatch rating scheme:
LikeLike