The new Medium audio feature is biased, and harmful to transgender authors.
A few small changes to the implementation would improve this immensely.
Medium is currently testing a new audio feature where an AI voice reads aloud stories. I love the idea of this, but I want to raise a concern that I'm not certain I've heard voiced yet. I'm a transgender woman, and I noticed today that this feature was available to me when looking at my one of my own stories. I clicked the "listen" button out of curiosity, and was unpleasantly surprised to hear a very masculine voice reading my words. It was jarring and upsetting to hear my words presented in what sounded like a man's voice. I investigated the player and learned that I could change the "default" voice to one of two feminine voices with American accents, or one with a British accent, but the implication -- that any listener who didn't take this step would hear a male-coded voice -- is actually pretty upsetting to me.
Trans people already face a lot of default skepticism and discrimination from people who don't recognize us as our affirmed and lived genders. This is exacerbated by a media landscape that platforms anti-trans hate far more frequently than the perspectives and stories of actual transgender people. Having the apparent gender of the voice reading an article misaligned with the gender of the author might be a minor annoyance for a cisgender author, but for a trans author it reinforces the pervasive misperception that we are "pretending" to be our affirmed gender.
Defaulting to a masculine voice suggests that white, American, men, are the default -- the norm -- while everyone else is just an afterthought. There's also been much research into how often people actually change the default settings on software that suggests that as many as 53% of software users never deviate from the supplied defaults. The audio features on Medium are a textbook example of implicit bias by design.
At the minimum, authors should be allowed to specify the voice that reads their stories by default. For stories where authors have not designated a voice, the system should cycle through the available voices. While this wouldn't solve many of the issues with this system (it currently doesn't have any BIPOC voices, for example) it would at least be a step towards reducing a source of implicit patriarchal bias in the design of this feature.
So, while I love the idea of a scalable audio feature for listening to Medium articles, the current implementation leaves much to be desired.
