In the late 1950s, a team of Russian mathematicians created a ternary computer — using 0, 1 and -1, as opposed to the standard 1 and 0 — that not only worked, but provided mathematical advantages over its binary counterparts.
“It shows that even these things that we find to be fundamental are themselves sort of constructed; that things don’t have to just be ‘on’ or ‘off,’” Nikki Stevens, a software engineer and data ethics researcher, explained during this year’s DrupalCon.
The same can be said of the gender binary, Stevens continued. Many of the million-plus trans Americans who don’t identify with the gender they were assigned at birth find that binary gender options based on primary or secondary sex characteristics don’t reflect their experiences.
Much of our technology, however, is still nestled in the binary. Many applications are marketed “for men” or “for women.” User profiles often include a binary gender field, and many gender data models fail to account for the nuances of gender identity.
It doesn’t have to be that way, Stevens said to an audience of dozens of developers during the session, which covered how applications can become more inclusive for trans users — from more representative stock photos to more thoughtful gender labels.
Four questions to ask about trans inclusivity:
- How is this biased? You can be biased toward or away from inclusivity, but there’s no such thing as neutral. Think hard about your app’s target audience — who could you include who isn’t currently?
- Do you really need to collect gender data? When apps ask users to share their gender identities, they’re often looking for something else, like the users’ pronoun or content preferences. Ask for the information you actually want.
- Can you change and delete old data? Storing old names, email addresses and photos can make your application unwelcoming to trans users.
- Is it possible to model gender contextually? Many people use different pronouns and express gender differently depending on their environments and who they’re around.
How Is This Biased?
If your organization has never discussed how its products treat gender, that doesn’t mean you’re occupying some middle ground, Stevens said.
“You can be biased toward justice, inclusion, care and equity, right? Or biased away from those,” they said. “But there’s no such thing as anything that’s neutral.”
When putting together a website, an app or an API, a good place to start is: Who am I building this for? If the answer is, “for men,” dig deeper. The product could be for:
- Cisgender men.
- People with penises.
- All men, including trans men.
- People who use “he/him” pronouns.
- People who prefer a more masculine gender expression.
By defining what “men” means in this context, you can improve your user experience, product and marketing. Often, Stevens said, products we say are designed for men are designed for masculine-presenting people — and the same goes for women.
“You can be biased toward justice, inclusion, care and equity, or biased away from those, but there’s no such thing as anything that’s neutral.”
If your product is for “men and women,” you should probably include people outside of that binary as well. Even updating your copy can make an app more inclusive, Stevens said.
“If you read old science books or articles, everybody is a ‘he.’ And then we went to ‘he/she’ to be more inclusive to women. And now we understand there are people who are excluded by that language,” they added. “Switching to the neutral ‘they,’ which we all already use in English, is a very small but very significant change you can make.”
Do You Really Need to Collect Gender Data?
“I would argue that you almost never have to ask for gender,” Stevens said.
If you’re collecting gender identity data to personalize user-facing copy, try asking for preferred pronouns instead. If you’re asking because you want to make in-app content recommendations, try asking about the user’s content preferences. If you’re asking to generate a user avatar, let the user generate their own. Gender identity is a poor proxy variable — stick to asking for the information you actually want.
If you must ask about gender, proceed thoughtfully. One option is a write-in field, which comes with advantages and disadvantages, Stevens said. It likely improves user experience, but it also makes data less uniform. Another is an exhaustive drop-down list. That’s tricky as well, since our conceptions of gender are evolving, and even the lengthiest of lists could quickly become out of date.
In our previous reporting on gender data, market-research firm dscout recommended a four-option menu: man, woman, non-binary and prefer not to say, as well as a write-in option for people who prefer to self-identify. The opt-out and write-in options honor users’ identities and privacy preferences and help researchers notice if new terms are becoming widely used.
“I would argue that you almost never have to ask for gender.”
Stevens recommended site-builders check out Drupal’s gender module, which uses the list of genders from Open Demographics, an open-source effort to crowdsource and standardize gender identity questions. The project’s repository homepage lists six questions developers should ask themselves before collecting gender data:
- Do you need to address a person with pronouns? If so, ask for their preferences. Perhaps “she/her,” “he/him,” “they/them” or “custom.” Avoid using “other.”
- Do you need to address a person with a title or prefix? If so, ask for their preferences.
- Do you need to collect gender information for demographic data reasons? If so, don’t force users into gender labels that don’t fit. Consider post-processing the data to fit with asks from researchers or marketers.
- Do you need to know about a person’s health needs, clothing preferences or bathroom use? If so, ask for their preferences.
- Do you want to publicly display a person’s gender on a profile? Dating and social media sites can make this display optional.
- Do you need to know assigned gender for legal, medical or regulatory reasons? If so, clarify whether you’re asking for assigned gender at birth, legal gender marker or gender identity.
Can You Change and Delete Old Data?
Many trans people go through transitions, during which they may adopt new names, pronouns, email addresses and more. Applications should allow users to access and change all their information, and developers need a way to permanently delete that data.
“One of the things we need to do is throw data away, which is antithetical to how we’re all taught to just hoard and store data forever and ever,” Stevens said.
Stevens encouraged developers to avoid storing records of data changes and to make sure cached photos and data won’t show up in web searches.
Is It Possible to Model Gender Contextually?
What about trans inclusivity for teams that need to model gender and use it as a data point — is it possible?
“Kind of,” Stevens said. “It’s complicated.”
Gender, they said, is often contextual. For example, a non-binary person may use “she/her” pronouns and tend toward more traditionally feminine gender expression at work due to an unaccepting workplace.
In cases like these, the user’s gender identity hasn’t changed, but their needs have. Stevens explained:
“If [the application] is going to interact with me, I’m going to expect it to help me stay safe at work.”
“To be clear, it’s not that my gender in this example is one thing at home and something else at work. It’s that if [the application] is going to interact with me, I’m going to expect it to help me stay safe at work.”
This raises the need for contextual gender models. For example, developers could create an entity relationship diagram with four tables: the user, the context, the gender and an accompanying statement: “When I’m at work, people see me as a woman.”
That model, however, doesn’t account for pronouns. So, developers could add a fifth table for pronouns, and a new example statement becomes: “When I’m at work, people see me as a woman, and I use ‘she/her’ pronouns.”
In other words, this user’s gender is non-binary. In certain contexts, for any number of reasons, they may prefer that others see them as a woman — or they may choose not to correct people in certain instances, as these conversations can be difficult and draining. That leads to yet another new model with an additional table for verbs like “I use,” “I tolerate” and “I prefer.”
But even these more complex models don’t account for all the nuances of contextual gender. That makes it incredibly difficult to model well, Steven said.
“Having gone through this example, which I’ve used for some research projects and work, my conclusion is that we shouldn’t try to model gender unless we absolutely have to,” they said.