I regret using this package name, and I regret publishing this package carelessly. I unpublished past versions of the package. The current version only contains this README file.
In 2016, for a school project, I analyzed the gender breakdown of bylines on the New Yorker website. This project used a classifier based on baby name and sex data reported to the Social Security Administration to guess the gender of writers based on their first name. When my teacher taught us this strategy, he gave the caveat that it was a "dirt-cheap super naive and ethnocentric algorithm."
In 2018, on my own time, I revisited this technique to explore the bylines of other media outlets. I was interested in further comparing and visualizing gender inequities in the media. Since the "gender guessing" algorithm was a self-contained part of the new project, I organized it as its own JavaScript package so that I (and anyone else) could reuse it.
I knew the classfier was unsophisticated and prone to misgender. I also knew
the classifier bought into the false and harmful concept of gender binary. I
published the package anyway because I thought it was a useful tool for
exploring gender inequity despite these limitations, and I planned to
acknowledge the limitations if I ever published gender analysis work that used
the tool (I never published this work). I named the package misgender in what
I thought was a self-deprecating, oblique acknowledgement of these problems.
While thinking about the disturbing and catastrophic male supremacy in open source software, I remembered that I had created this tool. It seems that almost nobody has ever come across the package, and I'm not so arrogant to assume that those who did cared enough to give it space in their mind. But if even one person encountered this package name and felt a little more excluded from a community—or conversely, if even one person encountered this package and got the impression that misgendering is a concept to be used lightly—then I've done harm, to an individual and to the the world. I unpublished the package to avoid causing this harm in the future.
In 2018, I understood that misgendering was a bad and hurtful behavior. But I must not have given serious consideration to the social and psychological harm it can inflict on people, especially trans* folks. And I didn't seriously consider the weight of throwing a word around, even if this tool was never intended to misgender people or somehow promote misgendering. I feel now that the name choice made misgendering seem like a light thing, especially since I didn't provide any context about the name. It was wrong to use the word in such a silly and vacuous way, just as it would have been wrong to name the package after any other horrible thing that people to do each other.
Although I still believe that name-based gender inference can be a useful, if flawed, tool to study gender inequity,1 I don't think it was responsible to publish code that could so easily be misused or misinterpreted. Without purposeful contextualization, this tool's benefits can't outweigh its capacity to misgender people and its perpetuation of gender binary.
Given that I'm not interested right now in updating, contextualizing, and maintaining this code to mitigate its potential for harm, it's better that the package doesn't remain available.
-
I don't want anyone else to use this package name (unclear whether that's possible).
-
Selfishly, embarrassingly, I have a small worry that even if I entirely unpublish the package, someone might eventually find a record of it and criticize me. This would be fair use of social pressure to hold me accountable for my actions! I don't want it to seem like I tried to hide anything; frankly, I think being transparent reflects better on me.
-
Again, I'm not so conceited to believe this piece of software mattered to anyone. Similarly, I don't presume that this document is a particularly productive form of reparation. But if this package has hurt anybody, I hope that the document serves as a not entirely meaningless apology. I'm sorry.
1 "Why it is sensible to examine the reliability of gender inference methods" by Helena Mihaljević and Lucía Santamaría summarizes the utility and the problems of this kind of tool. Their associated paper, "Comparison and benchmark of name-to-gender inference services" furthers this discussion.