Who do we choose to listen to?
Curiosity
Working in a male-dominated field (and here I’m referring to natural language processing (NLP), not linguistics), means not just that the significant majority of people active in the field are men, but also that men’s voices are amplified more than others’. We see this in the existence, even in 2019, of #manels. We see it in the tendency for men’s research to be cited more frequently than others’. I was curious to see if it’s also visible on Twitter, where I have observed male colleagues who have relatively inactive Twitter accounts but surprisingly large numbers of followers, i.e. large platforms waiting for them, seemingly built on reputation alone.
To look into this further, I did a quick, informal study, as described below.
Methodology
I sampled 93 Twitter accounts (including my own) by looking at the four lists I appear on that include “NLP” in their title. My goal in doing so was to collect twitter handles that someone else determined have something to do with NLP. These lists have 12, 11, 30, and 64 members, respectively, and their union contains 93 twitter handles.
I visited the Twitter profile for each of those handles and collected the following information:
- Number of followers
- Number of tweets
- Gender: from pronouns if indicated in the twitter profile (n=6); otherwise I guessed based on name and profile picture. The possible values here are he/him, she/her, organization, and they/them, but I only chose they/them if indicated in pronouns.
Of the 93 profiles, there were two with neither pronouns, name, nor photo, so I discarded these.
Obviously, these numbers are always in flux. The specific stats I collected reflect the Twitter profiles as I visited them within the space of about an hour on August 13, 2019.
Results
First, the lists included more men (40) than women (23) or non-binary/gender-queer people (2), but not by as large as margin as one might expect (the remaining 26 accounts represent organizations). As a group, the list makers (the four Twitter users whose lists I used to create this sample) seem to do better than the average NLP Twitter users at following accounts from different genders.
Looking at the mean number of followers, the accounts belonging to the organizations have the most followers (mean = 53,179 followers), by a long shot. The next highest are the men, with an average of 10,951, more than twice that of the women (4,298) and roughly 4x that of the accounts belonging to people who indicated they/them pronouns (2,728).
I took number of tweets as a proxy for degree of Twitter engagement. This is of course indirect: among other things, Twitter users can delete tweets and the count displayed to me might not reflect tweets of users who lock their tweets, etc. In other words, some of the accounts I’m looking at might have been very active in the past and only appear relatively inactive now. Nonetheless, I think it’s interesting to use this as a proxy: specifically, looking at the number of followers/tweet says something about the degree to which people follow someone based just on reputation rather than the extent to which they post interesting content to Twitter.
Again the organizational accounts had the highest numbers here, and men outscored women 1.86 to 1.25. (This is despite the fact that the men had far more tweets on average than the women: 5,784 compared to 3,442.) More telling, I think, is that only one woman’s account had a follower-to-tweet ratio over 10 (15.71), whereas 14 men’s accounts did (ranging from 172.31 to 10.75).
Enrich your Twitter experience
Twitter, like all social media, has its downsides. But it can also be extremely valuable as a way to learn about perspectives different to our own. There are incredibly important conversations happening there, from which we all have the opportunity to learn. My own Twitter follows include many BIPOC, especially WOC in the academy, disability activists, trans activists, and others whose lived experience is vastly different from my own.
In this post, I’ve looked only at (mostly attributed) gender, but I suspect that gender is not the only dimension along which those with the most privilege are also the most heard. A corollary of this is that the average Twitter user isn’t benefitting from the platform as they could. This is to the detriment of individuals but also to fields such as NLP, as it is yet another way that marginalized voices are left to the sidelines.
If reading this has led you to be inspired to expand your Twitter follows, I encourage you to seek out hashtags like #BlackandSTEM, #CiteBlackWomen, #ActuallyAutistic to learn what else you might not have been seeing on Twitter, and start following people. (Suggestions for further hashtags welcome!)
