Words mean stuff: Develop a framework for removing biased language at your company
Alyssa Yell18:532023
Some biased terms need a find and replace, others need whole teams
Tech documentation and code bases still carry divisive, disparaging and racist terms, and some sit so deep in commands that removing them can take a number of teams. Alyssa Yell, Senior Staff Technical Editor at Splunk who shaped guidance on biased language in its public style guide, has also worked as the only editor making these calls alone. So how do you decide which words show bias, and which to replace first?
Get a question checklist, a three-tier severity ranking and resources for teams of one
Alyssa Yell, Senior Staff Technical Editor at Splunk, presents a framework for spotting and removing biased language in technical writing and software. She offers a checklist of questions to test whether a term shows bias, a three-tier ranking of how urgently harmful terms should be replaced, and style guides, tools and communities for people who make these decisions alone.
“And just because a term isn't offensive to you that doesn't mean that it isn't problematic for someone else.”
“On the following slides, I'll present some questions that you can ask yourself about a term, but above all, let harm reduction guide your decisions.”
What you will learn
Apply a checklist of questions to test whether a term diminishes, excludes, degrades or implies elitism or power.
Rank biased terms into first, second and third order concerns to decide what to replace first and how much effort it needs.
Distinguish surface fixes in documentation and UI text from changes embedded in code bases and commands that need several teams.
Use linters and writing assistants such as the Vale linter with Splunk style rules, Witty Works and woke to catch non-inclusive language.
Find peer support through the Inclusive Naming Initiative and the Inclusive Language Advocates LinkedIn group when working alone.
Key takeaways from the speaker
Ask yourself questions to determine whether a term shows bias
Rank-order the need to replace terminology based on severity
Locate outside references who can help if you’re a team of one or otherwise feel like you work in a silo
Details
Level
Intermediate
Format
Talk
Technical depth
Introductory technical
For
All career stages
Length
18:53
Published
June 2023
Questions this talk answers
How can I tell whether a word or phrase shows bias?
Ask whether it diminishes others' situation, puts a positive or negative connotation on color, race or people, implies otherness, degrades a group or implies elitism, and whether it makes you uncomfortable, works metaphorically or could be more literal. The more questions you answer yes or possibly, the more likely the term needs replacing.
Which biased terms should be replaced first?
First order concerns, which are egregious, overt or politically charged, should be replaced immediately even if it takes several teams. Second order concerns rooted in war, oppression, appropriation or gender should be strongly considered, and third order concerns that are idiomatic or unclear should be replaced where possible.
What can I do if I am the only person making language decisions at work?
Lean on published guides such as Splunk's, Google's, Microsoft's, Intuit's, Atlassian's and the University of Washington's, and join communities like the Inclusive Naming Initiative's Slack channel, Google group and monthly meeting or the Inclusive Language Advocates LinkedIn group.
Are there tools that flag non-inclusive language while writing or coding?
Yell points to the Splunk style guide rules for the Vale linter, the Witty Works writing assistant for Google Docs and Notion, and the woke config builder, a GitLab project for detecting non-inclusive language in source code.
If language keeps changing, is it worth trying to be inclusive?
Yes. Yell says perfection may never be reached, but that is no excuse not to check your bias, question word choices and consider who you might be excluding.
About this talk
With the increased focus on removing language that actively harms and degrades members of our global community, the tech industry has especially needed to make major changes in its written documentation and software. But how does one determine biased language when they work as a team of one or don’t have a sounding board for what is inclusive and exclusive?
By the end of this talk, you’ll develop a framework of asking yourself important questions to determine whether language shows bias, you’ll rank-order the need to replace existing terminology based on the severity of harm, and you’ll learn about outside resources to consult if you need a shoulder to lean on.
Speaker 1: Okay, I'm going to go ahead and get started. My name is Alyssa. I use the pronouns she and her, and I do most of my work as a senior staff technical editor at an enterprise software company called Splunk. Speaker 1: For those of you who can't or don't see me on your screen, I am a 40 something white woman. I have long blonde hair parted on the side, blue eyes.
Speaker 1: And I'm wearing a black top. I'm speaking to you today from San Francisco, California, and I'd like to just take a second to acknowledge that the land from which I'm speaking is the traditional territory of the Ohlone people of the San Francisco Peninsula. Speaker 1: And as a reminder, the process of colonization is ongoing, and it still adversely affects the indigenous people of the Bay Area.
Speaker 1: I have a number of links in my presentation to style guide pages, downloads, outside resources, and more. Speaker 1: So if you'd like to follow along and open the pages at your discretion, I've left a copy of this presentation at bit.ly/words mean stuff. Speaker 1: That link is better if you're viewing from a desktop computer, or if you prefer to check things on mobile, the QR code is a smaller scale suitable for a smaller screen.
Speaker 1: But it's the same information in both places. Speaker 1: So I'll give you a second to open the link of your choice. Otherwise, I do believe the conference shares the recording in my slides after the week wraps up, and you can take a look then. Speaker 1: So I am honored to talk with you today about my experiences developing guidance and putting into practice inclusive and unbiased communications.
Speaker 1: Since joining Splunk, I've published a public facing style guide with technical writing rules and guidelines. Speaker 1: That gets used by writers, designers, engineers, and others across the technology industry. Speaker 1: And by maintaining a style guide, I see firsthand just how much words and language are always evolving. Speaker 1: In the last couple of years, I've shaped guidance around removing exclusionary and biased language in the technology industry.
Speaker 1: Technology has deep rooted divisive disparaging and racist terms, and I'm grateful to play a role in affecting change as a result of social uprisings that are bringing more and more awareness to the power of language. Speaker 1: As a result, my experience comes from technology and technical documentation specifically. So my talk today has that lens. Although, of course, inclusive language. It really applies everywhere.
Speaker 1: So today we're going to develop a framework that you can use to help reduce harm and communications. And here are the things we're going to cover today. Speaker 1: First, we'll briefly go over why language matters. Speaker 1: Then we'll ask ourselves some questions to determine whether a term shows bias. Speaker 1: Next, we'll rank order the need to replace terminology based on the severity of harm.
Speaker 1: And finally, we will locate outside resources who can help if you're a team of one, you otherwise feel like you work in a silo, or you just want someone to bounce ideas off of. Speaker 1: So we all understand that of the power of innovation and how technology transforms the world that we live in, but only recently have we been made aware of the importance of language and communication, especially in technology.
Speaker 1: Language is the foundation of human connections. It's through language that we express thoughts and feelings and ideas. Speaker 1: It's how we connect with one another, how we build relationships, how we share knowledge. Speaker 1: So when we think about the vast array of human differences, we all play a role in making sure that everyone is on the same playing field.
Speaker 1: Language is a powerful means of interaction, and it's one of the most important tools that we have. Speaker 1: Inclusion is the intent of individuals or systems to actively include and support the array of human differences to create a sense of belonging. Speaker 1: Belonging establishes trust, and trust builds relationships, trust and respect are so important for everyone we interact with, whether it's in documentation, in a product, or even in face to face communications.
Speaker 1: Because language is always evolving and very quickly at times, I don't think we'll ever get to perfect on the topic of achieving inclusive language. Speaker 1: But I really believe that the more we think about it, consider it, strive toward inclusiveness and try to bring others in, we will always be in a better place. Speaker 1: The intention of removing bias language is to reduce harm for the communities and to create safe spaces.
Speaker 1: On the following slides, I'll present some questions that you can ask yourself about a term, but above all, let harm reduction guide your decisions. Speaker 1: When you finish answering the questions about a term, the more questions that you answer yes or possibly the more likely it is that that language in question needs to be replaced.
Speaker 1: So to determine whether a term shows unconscious bias or it harms a marginalized or underrepresented group ask. Speaker 1: Does the word diminish the historical or current situation of others. Speaker 1: Does the word place a positive or negative connotation on color, race, or people. Speaker 1: Does the word imply otherness or exclude a group.
Speaker 1: Does the word harm degrade or insult anyone or any group, whether it be a mental physical sexual functional or criminal. Speaker 1: Does the word imply elitism or a position of power. Speaker 1: Then these final questions are a little more general in nature, does the language make you uncomfortable, even if you can't quite articulate the reason why.
Speaker 1: Is the language working metaphorically what are the implications behind that metaphor. Speaker 1: Can the language be substituted for something clearer or more literal. Speaker 1: So there's a lot of nuance and complexity in the evolving English language so if a word doesn't fit one of these questions and it makes you uncomfortable don't use it.
Speaker 1: Likewise, if someone tells you that a word makes them uncomfortable take their opportunity that take an opportunity and listen to their opinion as an opportunity to learn. Speaker 1: And just because a term isn't offensive to you that doesn't mean that it isn't problematic for someone else. Speaker 1: We might think we know who our audiences and what we can get away with, but we don't know everyone's story and it's just not safe to assume so again.
Speaker 1: The aim is to be aware of our own biases and writing and to reduce harm to communities. Speaker 1: After we've answered the previous questions and we've determined that we want to replace a word because it's causing harm to someone or to a community we can rank the severity of that term into one of three orders of concern.
Speaker 1: So some changes are more surface level, such as doing a find and replace and documentation or updating some customer facing ui text. Speaker 1: But others are deeply embedded in code code bases and commands so it might take a team to implement changes you can use these orders of concern to help you determine how much involvement might be needed to make a change and to prioritize what to update first.
Speaker 1: So first order concerns are language where the harm is egregious overt and clearly problematic and should be replaced immediately, it might take a lot of effort from a number of teams to make that happen, especially if it's on the back end. Speaker 1: So here are the criteria, you can use to determine whether a word or a phrase is a first order of concern and it needs to go now.
Speaker 1: The term perpetuates bias or stereotypes in a manner that's harmful to any historically underrepresented group of people. Speaker 1: The term is overtly disparaging toward any person or any group. Speaker 1: The term is overloaded problematic or politically charged. Speaker 1: In the case of first order concerns. Speaker 1: it's good to research, whether a term refers to a group of people or is disparaging and other contexts, even if the language itself is etymologically neutral.
Speaker 1: So the examples on this slide and my following slides are by no means exhaustive but instead I put them here to demonstrate the levels of concern. Speaker 1: Again, many of my examples are viewed with technology as the lens, but the questions that we asked ourselves in the previous section might result in other words also needing immediate replacement.
Speaker 1: So I won't go into the full historical background of these words, but as a result of the questions we asked ourselves paired with research we do on etymology of the words and their usage. Speaker 1: These are the terms that are not welcoming and they cannot be used casually in technical contexts. Speaker 1: Second order concerns are language that is problematic, but with less definite impact, and by less definite impact, I mean that it might not be outright disparaging like a first order of concern, but it has underlining meanings that stem from history, war, genocide or gender.
Speaker 1: These are words that you'll want to strongly consider replacing language that is unnecessarily violent oppressive or elitist. Speaker 1: Language that originates from a history of injustice oppression or identity based discrimination. Speaker 1: Language that is culturally appropriated. Speaker 1: Or language that is unnecessarily gendered. Speaker 1: So in the case of second order concerns it's good to look into whether a phrase has militaristic connotations or it arises from a context of war classism segregation or gender.
Speaker 1: And these are a few examples of second order concern language which is again language to strongly consider replacing. Speaker 1: In these cases, the terms are unnecessarily violent they relate back to war are culturally appropriated or our term taken out of a historical context and placed somewhere else inappropriately. Speaker 1: Third order concerns our language that can use improvement, but does no demonstrable harm so this language is idiomatic or understanding it is limited to those in a particular culture or an in group.
Speaker 1: These terms are unclear or they use metaphor when a more precise term is available. Speaker 1: Or the terms ascribed human traits to non human objects when categorizing third orders of concern think about whether the term uses a literary device to communicate a point or if understanding a term is conditional on belonging to a certain group.
Speaker 1: Here are a couple examples of third order concerns which is language to replace where possible. Speaker 1: In these cases using words associated with mental health or physical abilities is problematic because it perpetuates untrue stigmas. Speaker 1: These terms are also colloquial in nature, there are clearer more precise alternatives that all users can understand and relate to without feeling undermined.
Speaker 1: So I am fortunate to work at a company that has employee support groups diverse team members and I have a supportive manager who's good at helping clear obstacles for me. Speaker 1: But I've also worked at places where I was the only editor and I was the one making language decisions and it was hard to know if what I was doing was right or wrong.
Speaker 1: So if you're in that latter latter category and you don't have support systems at work or you work solo or in a silo there are resources to help you. Speaker 1: And even if you do work at a company that has these systems in place it's just always good to join in converse conversations and share your experiences and learn from others.
Speaker 1: So the examples that I showed on the order of concerns slides were a very small sample of a larger collection that many sources have already identified for replacement. Speaker 1: In a technology context here are some of my favorite sources. Speaker 1: So i'm biased when I tell you to refer to splunks right unbiased documentation page because I had the heaviest hand in it.
Speaker 1: But it's got more terms to consider than I discussed here today. Speaker 1: The inclusive naming initiative or I and I is a community of volunteers, mainly in the technology sector who've come together to try to a Polish racist and demeaning terminology in the tech industry. Speaker 1: I participate in the monthly meetings I help do research i've come up with replacement terminology suggestions and I encourage everyone to join.
Speaker 1: The University of Washington's inclusive language guide is quickly growing as one of my go to guides because it includes replacements as well as the rationale and history behind the language and I just like having that research. Speaker 1: And that background learning as I go. Speaker 1: And then Google's right inclusive documentation and Microsoft spice free communications are also excellent sources.
Speaker 1: So for help developing inclusive language outside of the tech industry here are a couple sources that I also look to. Speaker 1: I do use these for technology research too, but they're just a little more broad in scope. Speaker 1: There's into its inclusive language sorry into its inclusive content page. Speaker 1: Atlassian's inclusive language guide and the responsible communication style guide, which is available in a number of formats, including print and ebook.
Speaker 1: So if you're looking for help while you're in the throes of writing and you want some tools to help you keep inclusive language top of mind, I urge you to look at these resources. Speaker 1: The splunk style guide for the veil linter is a repository of rules ported from the splunk style guide it has rules for inclusive language device agnostic language accessible language and so much more.
Speaker 1: The veil linter can process docs and most markup languages, such as markdown and ascii doc and it has a number has a number of integrations with code editors. Speaker 1: witty works is a writing assistant add on for inclusive language it's really similar to grammarly but with a very specific lens and it works for Google docs and notion.
Speaker 1: And then there's the woke config builder, which is a get lab project that you can use for detecting non inclusive language in source code. Speaker 1: So these are just a couple that I know of, but if you have any others that you can eat that you use please share them with me you can email me later or put them in the chat.
Speaker 1: And then I also understand that doing research and using tools doesn't always feel the most satisfying sometimes you want to talk about things you want to converse with others and you want to get more opinions. Speaker 1: This is especially true if you work as a team of one or you don't feel you have the resources to get other opinions and perspectives.
Speaker 1: So the inclusive naming initiative, like I mentioned, has a slack channel where you can get the full brain power of multiple perspectives and roles and backgrounds and levels of experience. Speaker 1: there's also a Google group where you can start new discussions and you can chime in on ongoing conversations and there's also the monthly meeting that you can add agenda items to and talk it over with the group.
Speaker 1: And then i'm also part of a linkedin group called inclusive language advocates that has nearly 2000 members and we are dedicated to gathering online to exchange knowledge and. Speaker 1: support each other and help each other use inclusive language, so I suggest checking that out and again if if you have other resources that you use I would love to know, please put them in the chat or share them with me personally, I would love to check them out and learn from others.
Speaker 1: So the final point of my talk is this, even though I don't think we'll get to 100% perfect because language is always changing and sometimes really rapidly. Speaker 1: I don't think that's an excuse to not check your bias at all times question word choices and use this framework to open your mind and consider new perspectives.
Speaker 1: Make a checklist if you need to keep it on your desktop or download these tools to highlight ideas and concepts that you don't consider naturally. Speaker 1: Think about who you're excluding whether it's a new or an experienced user someone from another culture or religion. Speaker 1: Someone from a different age group or gender recognize the responsibility we all have to fostering inclusive environments.
Speaker 1: Remember that inclusive language prevents frustration and demeaning experiences, whether it's online in person in documents or in a product. Speaker 1: Inclusive language builds trust and relationships and it makes people feel welcome safe and valued. Speaker 1: And those are the best feelings that any of us can ask for. Speaker 1: So thank you so much for caring about inclusive language and for attending my talk, you can feel free to connect with me on linkedin or head over to my website words mean stuff.com.
Speaker 1: Again, I would love to know what you're doing to foster inclusive language if you have other resources that you refer to please share them with me, I would love to know so thank you again and enjoy the rest of your conference. Speaker 1: Thanks Madeline.
Transcribed automatically; names checked against the speaker record. Select a time to jump there.
Article based on this talk
How to decide which biased words to remove first, even when you work alone
Splunk technical editor Alyssa Yell shares a question checklist, a three-tier severity ranking and the resources she relies on for inclusive language.
In short
Alyssa Yell, Senior Staff Technical Editor at Splunk, argues that anyone who writes for technology products can find and remove biased language systematically, even without colleagues to consult. Her framework combines a set of self-check questions, a three-tier ranking of how urgently a term must go, and a list of style guides, tools and communities for people working in a silo. Harm reduction, not perfection, is the guiding principle.
Key takeaways
Ask yourself questions to determine whether a term shows bias.
Rank-order the need to replace terminology based on severity.
Locate outside references who can help if you're a team of one or otherwise feel like you work in a silo.
Why words in technology carry weight
Yell has published a public facing style guide at Splunk that is used by writers, designers and engineers across the technology industry, and maintaining it has shown her how quickly language evolves. Over the last couple of years she has shaped guidance on removing exclusionary and biased language, noting that technology has deep-rooted divisive, disparaging and racist terms. Her lens is technical documentation, although she stresses that inclusive language applies everywhere.
For Yell, language is the foundation of human connection, and inclusion is what turns it into belonging. "Belonging establishes trust, and trust builds relationships, trust and respect are so important for everyone we interact with, whether it's in documentation, in a product, or even in face to face communications." (3:56)
She is realistic about the goal: "Because language is always evolving and very quickly at times, I don't think we'll ever get to perfect on the topic of achieving inclusive language." (4:28) The point is to keep striving and to reduce harm for communities.
Questions that reveal bias in a term
The first part of the framework is a checklist. Yell suggests asking whether a word diminishes the historical or current situation of others, places a positive or negative connotation on color, race or people, implies otherness or excludes a group, degrades or insults anyone, or implies elitism or a position of power. More general questions follow: does the language make you uncomfortable, is it working metaphorically and what does that metaphor imply, and could it be replaced with something clearer or more literal? The more questions answered yes or possibly, the more likely the term needs to go.
"On the following slides, I'll present some questions that you can ask yourself about a term, but above all, let harm reduction guide your decisions." (4:58)
She also asks writers to listen when someone says a word bothers them, and not to assume they know their audience. "And just because a term isn't offensive to you that doesn't mean that it isn't problematic for someone else." (6:37)
Ranking terms by severity of harm
Once a term is flagged, Yell sorts it into one of three orders of concern. Some fixes are surface level, such as a find and replace in documentation or a change to customer facing UI text, while others are embedded in code bases and commands and need a team to implement. The ranking helps decide how much involvement a change needs and what to update first.
First order concerns perpetuate harmful bias or stereotypes, are overtly disparaging, or are overloaded and politically charged. "So first order concerns are language where the harm is egregious overt and clearly problematic and should be replaced immediately, it might take a lot of effort from a number of teams to make that happen, especially if it's on the back end." (7:53)
Second order concerns are problematic with less definite impact: language that is unnecessarily violent, oppressive or elitist, that comes from a history of injustice or identity-based discrimination, that is culturally appropriated, or that is unnecessarily gendered. These should be strongly considered for replacement. Third order concerns do no demonstrable harm but can be improved: idiomatic phrases understood only by an in group, vague metaphors, or terms that give human traits to objects. Yell notes that casual words drawn from mental health or physical ability also belong here: "In these cases using words associated with mental health or physical abilities is problematic because it perpetuates untrue stigmas." (11:34)
Help for the team of one
Yell now works with employee support groups, diverse colleagues and a supportive manager, but remembers working without them. "But I've also worked at places where I was the only editor and I was the one making language decisions and it was hard to know if what I was doing was right or wrong." (12:03)
Her go-to references include Splunk's Write unbiased documentation page, the Inclusive Naming Initiative, the University of Washington's inclusive language guide, Google's Write inclusive documentation and Microsoft's bias-free communication guidance. Beyond tech, she uses Intuit's inclusive content page, Atlassian's inclusive language guide and the Responsible Communication Style Guide. For writing tools, she recommends the Splunk style guide rules for the Vale linter, the Witty Works writing assistant for Google Docs and Notion, and the woke config builder, a GitLab project for detecting non-inclusive language in source code. For conversation, the Inclusive Naming Initiative offers a Slack channel, a Google group and a monthly meeting, and the Inclusive Language Advocates group on LinkedIn has nearly 2,000 members, as the speaker cited.
Keeping inclusion top of mind
Yell closes by urging viewers to keep a checklist on their desktop, use tools that surface what they would not notice naturally, and think about who they might be excluding, whether by experience, culture, religion, age or gender. "I don't think that's an excuse to not check your bias at all times question word choices and use this framework to open your mind and consider new perspectives." (17:05)
"Inclusive language builds trust and relationships and it makes people feel welcome safe and valued." (17:54)
Watch the talk
Words mean stuff: Develop a framework for removing biased language at your company, by Alyssa Yell (19 minutes): https://www.womentech.net/video/words-mean-stuff-develop-framework-removing-biased-language-your-company
About the speaker
Alyssa Yell is Senior Staff Technical Editor at Splunk, based in San Francisco, and describes herself as a seasoned editor on the outside and an insatiable adventuress on the inside. A proud global citizen who has made multiple trips around the world, she strives to make sure everyone feels included, important and like they belong, both inside and outside of work.
Questions this talk answers
How can I tell whether a word or phrase shows bias?
Which biased terms should be replaced first?
What can I do if I am the only person making language decisions at work?
Are there tools that flag non-inclusive language while writing or coding?
If language keeps changing, is it worth trying to be inclusive?
Drafted with AI from the talk transcript. Editor review pending.
Frameworks and ideas mentioned
Three orders of concern (first, second and third order)
Bias self-check question checklist
Resources mentioned
Splunk Style Guide: Write unbiased documentation
Inclusive Naming Initiative (Slack channel, Google group, monthly meeting)
University of Washington inclusive language guide
Google's Write inclusive documentation
Microsoft's bias-free communication guidance
Intuit's inclusive content page
Atlassian's inclusive language guide
The Responsible Communication Style Guide
Splunk Style Guide rules for the Vale linter
Witty Works
woke config builder (GitLab project)
Inclusive Language Advocates LinkedIn group
Speaker's slides via bit.ly link
Speaker's website, Words Mean Stuff
Cite this talk
Alyssa Yell. "Words mean stuff: Develop a framework for removing biased language at your company." WomenTech Network, 2023. https://www.womentech.net/video/words-mean-stuff-develop-framework-removing-biased-language-your-company