http://arxiv.org/abs/1403.3091
This all started about 6 months ago when I was attending a different astronomy conference. I observed that the gender ratio for speakers seemed well balanced, as did the audience. Both were perhaps 60%/40% (Men/Women). However, the questions mostly seemed to be asked by men!
So I decided to organize a volunteer effort to study this. We collected data using a simple web-form (that Morgan Fouesneau graciously helped me make), and asked conference attendees to record the gender of every speaker and every question asker for talks they attended.
We got over 300 submissions! I was going to be happy with 100, and figured I'd have to beg a few friends to participate. This was enough data to make some interesting plots... and also just enough data to know that we need more data!
Here are a few highlights from the study:
1. Men ask disproportionally more questions than women in talks.
 |
| FS FQ = Female Speaker, Female Questions, FS MQ = Female Speaker Male Questions, etc |
We were very glad to see that the gender ratio of all the speakers matched that of the conference participants. This also closely matches the gender ratio of astronomers under the age of ~40 as reported in the AAS Demographics survey recently.
2. Women are asked slightly more questions per talk than men
 |
| Blue is talks by men, green is talks by women |
3. The gender of the session chair has a strong impact on the gender ratio of the questioners.
 |
| FC FQ = Female Chair, Female Questions, FC MQ = Female Chair Male Questions, etc |
We need data on the format of the session to understand the origin of this result, and to make actionable suggestions/best-practices for future conferences!
The Future:
I want to conduct a more controlled follow up study! It's clear to me that there's more to learn, and maybe ways we can improve how our conferences are conducted.But I'll need help doing it!
The upcoming AAS 225 in Seattle (Jan 2015) would be a perfect time to do a follow-up study. We need to gather more detailed data from every talk. A big volunteer effort might get us there, but if the AAS is interested in helping that could be a huge shot in the arm. We did this project with $0 spent and only social media / friends to help advertise. With the AAS's help we could get this data and help make our annual meetings even better!
Lastly, a huge thanks to the wonderful volunteers who sent us data, the organizers and sponsors of AAS Hack Day, and [Morgan, Erin, Alex, Katja, Laura] for making the analysis/writing happen!
Update: this appears to be my 100th blog post on If We Assume!
Here's a sweet badge I awarded myself...
Hey James,
ReplyDeleteI'm Ben Nelson, a grad student at Penn State. We met briefly at the DC meeting this year.
Great work! I think many people, including myself, are "people watchers" and wonder about these sorts of things in general (e.g. the statistics of gender/race/affiliation under various circumstances/situations, the driving forces behind our social behavior). Although the thought is fleeting, since hardly anyone will put in the effort to develop the survey and compute the relevant statistics. This is a great first step!
I agree with the "age-effect" and think it has a major influence in these results. Having that information would be valuable, but trying to note a person's age and other parameters would be complicated in practice.
I guess my question is two-fold. What other parameters besides gender would be worth noting? If there are any at all, has anyone thought about a way to strike a balance between the amount of information gathered and the complexity of the survey process?
Hey Ben! Thanks!
DeleteYou hit the nail on the head, I think. There's so much more we need to know to unambiguously state what's going on, if it's only an age effect or if there's a weird interplay between age, sexism, and social pressure.... lots of possible causes.
One way we can narrow things down is by having complete data for an entire meeting, and by gather more "metadata" about talk sessions. This might include details about how questions are asked, if the session is hostile, if the room is full, if the chair picks questions or the speaker, if the speaker is interrupted... but course we can "boil the oceans" here as well.
I think we first need to see which sessions are "succeeding", having the best gender ratios for questions (or some other metric), and try to narrow down *why*!
Very interesting, though I wonder if the male chair/female chair difference is due to different gender ratios in subfields (e.g., theoretical cosmology tends to be more male-dominated than say, exoplanets).
ReplyDeleteI took a fair amount of data for this and I'm glad to see that I wasn't the only person (for awhile, I was thinking "maybe on one else submitted info"). My question is this, did you look at subfields or do you need more data for that? For example, cosmology vs dust vs MW galaxy science.
ReplyDeleteSecondly, about about the gender of the people who submitted the data? What if there's an implicit bias there. For example, maybe females are more likely to "forget" to write down a male questioner or vice versa. If that has no influence, at least what was the gender ratio of those who submitted data for the study?
The data to look at subfields is there if we were to broadly cluster in to, say, 6-ish categories. Given the sparseness of the data we opted to not do that. Truthfully: we discussed doing it, imagining what results we might find. I think we'd cause some waves if we published "Subfield X is worst for gender equality!"
DeleteWe have no information about who the survey respondents were. I think that level of anonymity is important, but I agree it might bias our results!
Also Katie: thank you so much for helping take data!! :)
DeleteI posted my comment on AstroBetter, but thought it might be good to copy here:
ReplyDeleteVery interesting data! Kudos for collecting and analyzing it. It mirrors what I have observed at conferences for years. I wonder also if the size of the audience could be studied, i.e., look at splinter session talks vs. large plenary session talks?
My own personal experience as a young (postdoc), female astronomer is that I find it very very hard to ask questions, especially in big talks. I am a pretty confident person in general, have no problems giving talks or fielding questions, but I get extremely nervous when I attempt to ask questions in conferences! I have had my voice falter, palms sweat, etc -- the whole nervous response. I don't understand it well myself. The result is that I usually don't ask a question unless it is a talk that is my direct specialty and I basically know the answer in advance. I think part of it is the fear of in any way asking what might seem a "stupid question" -- I think women scientists are extremely conscious of how they are perceived (and judged more harshly), and this leads to a more cautious behavior in the public sphere. And what draws every eye in the room more than a unrehearsed question?
I suspect smaller sessions have more female questioners. If I think back to all my AAS conferences, I only asked questions in splinter talks. Further, in even smaller meetings (like journal clubs, with my colleagues I know well), I have no problem asking lots of questions. So I think this is a "controlled image/caution" issue -- and it would be good to get statistics on minorities as well, as I think we might find similar results.