Here are some best practices to follow when writing about people. These apply to identity.
The goal of all UX content is to be understandable to all — not just to the people paying for something or for those in a certain industry — and to accommodate the many ways that people use products. When creating content for product experiences, think and write by centering the person you’re writing to or about in a way that’s compassionate, inclusive, and respectful. Work to grasp the perspective of underrepresented groups, and avoid writing in a way that may view or treat someone as intrinsically different from yourself. You can use methods like co-designing and UX research.
Keep the following best practices in mind when writing:
Only include personal qualities if they’re relevant and important. Write what you mean, then look back at what you wrote and think about whom you’re centering with your words. Doing this can reveal which people you’re leaving out. What’s the sentiment behind your words?
Be cautious of appropriating terms from marginalized communities. In this guide, we say “underrepresented groups.” You can also reference 3rd-party sources such as Wikipedia’s list of which words to use and which to avoid.
Be on the lookout for proxy questions and statements, which appeal to generalizations and stereotypes. For example, saying, “just buy more storage” is a proxy statement on economic status, while “view additional storage options” doesn’t make those assumptions. Communicate from a place of equality, not condescension, and think about the worst-case interpretation of your words. Clear intent excludes fewer people and reduces bias.
When collecting user data in app or web experiences, first think about whether that information is actually needed, and then if it really is, communicate why. Allow for both common and custom responses, self-identification, multiple selections, and the option to opt out of responding. Artificial intelligence learns only from the information we provide to it, so our inherent biases can easily become included in training data. If content allows for variable and AI-provided information, consider the ways that may affect any copy.
Person-first language centers the person, not their qualities, by using those qualities as modifiers: “Design Adobe apps for people who use assistive technology.” But for identity-first language, which some communities and individuals prefer instead, language highlights the disability: “Design Adobe apps for deaf people.” No group unilaterally chooses one over the other, so when you’re writing about someone, ask them how they want to be identified. Avoid euphemisms like “differently abled,” which are regarded as condescending, and descriptors used as nouns, like “the disabled” or “the blind.” These tend to present a group of individuals as a monolith and suggests a lack of individual diversity within the group.
Preferred | Avoid |
---|---|
Disabled person or person with disabilities | Differently abled or the disabled |
Blind person or person who is blind | The blind |
Some phrases in common parlance that imply negativity are based on slurs against people with disabilities, such as “crazy” or “lame.” Never imply that a person is “suffering” from a disability or is a “victim” of a condition. Avoid appropriating terms from the disability community.
Preferred | Avoid |
---|---|
Ridiculous or unpredictable | Crazy |
Inconsiderate | Tone-deaf |
Incompetent or bad | Dumb or lame |
Keyanna has autism. | Keyanna is suffering from autism. |
Placeholder variable | Dummy variable |
Amir uses a wheelchair. | Amir is confined to a wheelchair. |
With imagery and language, avoid implying that a person has to look a certain way, be a certain size, or have a certain cognitive ability to do something. Depict more types of people as typical.
Preferred | Avoid |
---|---|
This tutorial teaches cropping and usually takes 5 minutes. | Follow this fast, easy tutorial. |
This feature works best when you zoom out to 75%. | This feature isn’t for the vision-impaired. |
Be aware of how words that are often associated with physical and mental health are often used as metaphors to describe interactions and product functionality.
Preferred | Avoid |
---|---|
Unavailable or locked or turned off or deactivated | Grayed out |
Coherence check | Sanity check |
Organize or organized | OCD |
Enter metadata with caution. For example, don’t tag a photograph of a child with words like “crazy” or “weird.”
Let’s say you’re writing a persona. When describing a person's country of origin or race, be as descriptive as possible as to not generalize any race or ethnicity. Race is only pertinent to biographical and announcement-related content that involves significant, groundbreaking, or historical events. For capitalization, Adobe follows AP Stylebook guidelines: capitalize nationalities, peoples, races (all except white), and tribes.
Adobe avoids using software terms such as “whitelist,” “blacklist,” “master,” and “slave.” Don’t use terms assigning value to racial characteristics, such as “dark pattern.” (Terms like "dark mode," "light theme," or "black screen" literally refer to color and brightness and don't assign good or bad values, so continue using them.)
Preferred | Avoid |
---|---|
Use this format to provide contextual clarity: (Result in past participle form) (object) Examples: Shared domains, approved people, targeted sites For coding constructs: Allowlist | Whitelist |
Use this format to provide contextual clarity: (Result in past participle form) (object) For coding constructs: Blocklist | Blacklist |
Legacy | Grandfather clause |
Futile undertaking or a project destined to fail | Death march |
Primary or main or source (e.g., “main track”) | “Master” descriptors |
Primary/secondary | Master/slave |
If you want to use a certain idiomatic or casual phrase, research its history before doing so. For example, imperfect spellings or pronunciations of words can imply pejorative associations with an accent. Be on the lookout for proxy questions, such as relating postal codes to ethnicity in rejecting job candidates, or making pricing or marketing decisions based on the average income of postal codes.
Since English isn’t everyone’s first language, it’s best to write using clear, plain language — as well as avoid idioms and phrases that might be complicated for non-English speakers to understand. Plain language is more widely understood and, therefore, avoids alienating people. It especially avoids alienating people in ways that specifically belittle non-English speakers. For example, the conversational and casual phrases “long time no see” and “no can do” were originally used to belittle Native Americans.
Preferred | Avoid |
---|---|
Welcome back | Long time no see |
2-step process | Wax on, wax off |
Sorry, something went wrong | No can do |
We must focus on building successful experiences for all users. That means writing and designing in a way that depicts all skin types, names, and cultures as typical. We cannot keep centering white-skinned, Western cultures in our designs.
Preferred | Avoid |
---|---|
Dark brown or beige or tan or peach, etc. | Skin or flesh or nude (referring to color swatch) |
Critics | Peanut gallery |
A broad range of name examples within a product experience (e.g., Ayesha, Ibrahim, Vignesh, Quynh) | Only culturally white name examples within a product experience (e.g., John, Bill, Karen, Amy) |
Here’s a list of preferred words that are alternatives to common technology industry jargon.
Preferred | Avoid |
---|---|
Native to the operating system or built-in feature | Native |
Meeting | Pow wow or circle the wagons |
Vision statement or strategic statement or value proposition | Zen statement or Zen garden |
Role model or kindred spirit | Spirit animal |
Guide | Sherpa |
Authority or expert | Guru or ninja |
Rather than “he” or “she,” if you don’t know a person’s pronouns, make the phrase plural and use “they” instead. Use of “they” to describe one person is also accepted, although the syntax remains plural (e.g., “they are” = “that person is”). It’s also best to avoid using roles or stereotypes that have gendered roots (e.g., “businessman" or “waitress”).
Preferred | Avoid |
---|---|
Server | Waitress |
Businessperson | Businessman |
Flight attendant | Stewardess |
They | He/she or (S)he |
A group of people or a group of women | Guys or girls or ladies |
Parents | Moms |
Use gender and sexuality descriptors as modifiers, not nouns (e.g., “transgender woman” rather than “a transgender,” “bisexual person” rather than “a bisexual”). A person’s pronouns are not opinion or preference, even if they may change over time (view Spectrum’s guidelines on pronouns). All of this helps us emphasize every person’s humanity, and keeps us from alienating people who aren’t cisgender and heterosexual.
Preferred | Avoid |
---|---|
Transgender people or trans people | Transgendered people or transgenders or the transgendered or transexuals |
Trans women | Trans-women |
A transgender man | A transman |
Alejandra, a lesbian woman | Alejandra is a lesbian |
Jing, a non-binary person | Jing is a non-binary |
Saadi is cis | Saadi is CIS |
What are your pronouns? | What are your preferred pronouns? |
Jamal’s pronouns are he/him/his. | Jamal prefers he/him pronouns. |
Wholehearted or impassioned | Hysterical |
Know the difference between sex (male/female) and gender (man/woman). When collecting personal data from users, consider if it is really necessary to ask for a person’s gender. Data collection and forms, while useful to product builders, can feel intrusive when asking about gender. When you really do need the information, allow for both common and custom responses, self-identification, multiple selections, and the option to opt out of responding. Avoid asking proxy questions, for example, asking for someone’s gender when the information that is actually needed is their bike size.
Preferred | Avoid |
---|---|
Prefer to self-describe and Prefer to not respond | Other |
People globally identify with many genders and sexualities, so it's important to teach AI exactly that. It wasn’t until June 2018 that the World Health Organization (WHO) declassified being transgender as a mental illness, so even though humans have adjusted this perspective, machine learning and AI can still perpetuate these biases. Don’t use AI or machine learning to guess genders based on image recognition, text analysis, or anything else.