Witty analyzes texts and checks them for inclusive language. What exactly does Witty highlight? Witty Works had to develop a framework of inclusive language to define what kind of words are analyzed. In this blog post, we explain the diversity dimensions that Witty checks. Please participate. Let us know your ideas or words you would like to see highlighted in this survey.
Grammar has defined its word types for centuries. And so at school we learn what a pronoun is, what a verb is, and what a noun is. We also learn where to put commas and where not. All of this is given according to strict rules.
However, this is not the case in inclusive language, as it is still a very recent development. There are no specifications, nor is there any authority that defines what types of words inclusive language includes.
But that's exactly what we're trying to do at Witty Works. Our goal is to develop a framework of inclusive language in German and English so that we can recognize non-inclusive language more quickly and use and roll out inclusive language more easily. (We explain what inclusive language is in this blog post).
Since the topic of inclusive language is rather new to many people, we created Witty - a software that uses artificial intelligence to identify hidden stereotypes or overt discrimination in language and provide inclusive alternatives to writers.
For the development of Witty, we need a framework like the one that exists for grammar. Of course, it will probably never be as exact as grammar. Because - unlike grammar - inclusive language is very much about context. A word may well be unconsciously stereotypical in one context; but not at all in another. Moreover, language is constantly evolving. A word that is not currently inclusive may well be in a few years because society has changed.
We know that the task of building a framework is very challenging and we have to ask ourselves many ethical questions. In developing the framework, we rely on three pillars:
- We take the basic vocabulary from studies conducted by recognized universities in the fields of linguistics, psychology, and behavioral economics.
- Our users give us feedback on their own vocabulary.
- We work with organizations that specialize in the language of certain diversity dimensions and provide us with information about them.
In addition, we created an ethics committee that looks over our shoulders. They review our decisions regarding the framework, the vocabulary, as well as the algorithm.
As with any attempt to build a set of rules, our inclusive language framework is in ongoing development. I would like to show you in the following sections how we classify the word types and how they are analyzed by Witty.
In each case, Witty marks the words or word combinations that are not inclusive in some way. This will allow the writers to look through their text and make it more inclusive.
(Witty also analyzes spelling and grammar, but this will not be covered here).
We divide the words and terms into 5 top categories. In each of these top categories there are sub-categories to make the feedback in the software more granular and coherent for the user. The top category determines how severe the non-inclusivity of the term or word combination is.
||Unconscious bias / stereotypes / stigmatization|
||Gendered language (role images)|
|Severity||very severe (1)|
|Definition||These are words and word combinations that most people recognize as obvious curse words, or at least leave you grinding your teeth because you know they are very hurtful.|
In this top category, Witty identifies words that fall into the following subcategories: Racist, sexist, nationalist, homophobic or transphobic, insults towards people with an migration background or certain nationalities, as well as slurs related to religion.
Examples here are the really bad terms like (please forgive me my language): slut, cripple, the N-word, ladyboy (Again, I sincerely apologize for writing these words out. They are only here to make the explanation clear).
We don't think our user would ever write such words, but an analysis on inclusive language would not be complete if we did not mark these terms.
Unconscious bias / stereotypes / stigmatization
|Definition||In most cases, these are words and word combinations that are subtly discriminatory and/or based on outdated, unconscious stereotypes and biases that we sometimes don't even notice (anymore).|
The use of these terms has two consequences:
- Certain people do not feel addressed by the text because they are unconsciously put off by it due to their socialization or values and therefore do not feel they belong.
- The words put people in a pigeonhole. If society accepts these uncritically and continues to use them, people are stigmatized and those affected remain trapped in them. This prevents them from developing freely.
The phrases belonging to the top category B are delicate, often subliminal or even unconscious. Thus, the identification of these words is also much more difficult. In addition, it may be that words identified here are absolutely problem-free in one context; in another, however, they may well unconsciously deter or pigeonhole.Subcategories we analyze:
- Agentic language: it describes attributes traditionally attributed only to the male gender. Agentic language unconsciously does not appeal to many people (anymore); either because they do not correspond to these masculine attributes or because they unconsciously do not find these attributes attractive. This depends on one's socialization and values. Examples here are ambitious, decisive, resolute.
- Exaggerated language: In the business world, we have a tendency to exaggerate our activities in order to make an impression on our readers. However, this is often not appreciated because it does not come across as authentic and also creates a sense of pressure among underrepresented groups. Example: high performer, front runner, proven track record.
- Physical or mental impairment: We make even more detailed subdivisions here (visual impairment, health impairment, intellectual impairment, hearing and speech impairment, mobility impairment, mental impairment). People with impairments are often portrayed as "not normal" and stereotypically described or such analogies are used to describe something as "not normal." Examples here are: moron, are you insane?, turn a deaf ear to
- Culture expresses stereotypes related to a person's cultural background or makes sweeping generalizations about a group of the same cultural background. Examples: sassy, paddy wagon, small Chinese, Italians are always late.
- Migration background: These are expressions that allude to a person's migration background and therefore stigmatize those people. Examples: third world country, mother tongue, illegal immigrant.
- Sexual orientation identifies terms that represent or reinforce the image of hetero-normativity or convey an outdated and perhaps even discriminatory understanding of sexual orientation. Examples: sexual preference, saying to a woman: "your husband is also invited", "Gay men are design affine"
- Gender identity marks terms that assume the biannually or which convey an outdated and perhaps even discriminatory understanding of gender identity. Examples: "Ladies and gentlemen," (m/f), "Do you have a son or daughter?"
- Racist Source: Here we look for terms in the text that have a racist root or words that use skin color in a negative context. Examples: master bedroom, Eskimo, chinaman
- Classism: terms that communicate a top-down perspective, i.e., readers notice that people are not speaking at eye level and a sense of social class differences is expressed. Examples: social climber, meritocracy
- Faith/Religion/Belief: identifies terms that express a discriminatory root or stereotypes related to faith or religion, generalize a group of the same religious background, or place certain religions above the others. Examples: in God we trust, man is the measure of all things, moslem
- Pressure: These terms communicate pressure, i.e., readers understand that it is up to them to submit, withstand pressure, and conform. Examples: "I'm counting on you.", "Work hard, play hard", fit to the culture.
- Age: Words and word combinations that are off-putting to either people over 50 or people under 25. In addition, if age is asked in any way, it is also flagged because in many cases it is not relevant but can lead to stereotype threat. Examples: Minimum age, young and dynamic, dried up
The style in which we write can also have an inclusive or non-inclusive effect. So here we identify words and word combinations that make the language more difficult to understand or that prevent closeness and emotionality.
Subcategories we analyze:
- Americanism: Today we use American or English terminology in the German language. For people who did not have an English-oriented education or who do not come into contact with such expressions in their professional life, this is often a stone in the path of communication. It is also perceived by these people as an expression of "being better". Examples: "What were your highlights and lowlights?", Minimum Viable Product, Financial Plan.
- Empty words: Examples here are "really", "actually", "finally", etc.. They add nothing to the content. For people who do not speak German, these words complicate things. It is more difficult for them to understand and therefore they are more likely to hang off. Examples: innovative, fair wages, flat hierarchies.
- Formality: If a company wants to be inclusive, it is also a matter of creating a closeness between superiors and the workforce or even between the official functions (such as works councils, administrative boards, etc.) and the workforce. The language must express that everyone is on eye level. Overly formal language prevents this. Examples: “Honourable ... ”, doctor or professor titles, in German: the polite version of "you".
- Generation Boomer: Boomers and Generation X sometimes need expressions that younger generations no longer understand. Therefore, these are not inclusive. Examples: "corner office" (for managers/executives), “You should know a man seven years before you stir his fire."
- Generation Z: Generation Z sometimes needs expressions that older generations do not understand. Many of these are, of course, Americanisms. These are therefore not inclusive. Examples: cringe (youth word 2021), woke, RT
- Filler words: These are usually small words that we use to emphasize something in particular or to round off a sentence. But they add no value in meaning and complicate the language. Examples: really, actually, etc.
Gendered language (role images)
In a gendered language like German, but also in a gender-neutral language like English, certain words hide images that represent an outdated conception of gender or gender roles.
Subcategories we analyze:
- Binary pronouns. When we make an example in a text and describe the example person in the third person, we have a tendency to use the masculine form in the continuation of the sentence, which is then supposed to be representative of everyone. E.g. the text reads: "In the literary genre of fantasy, an author needs different skills in writing. He must have a lot of imagination himself.... "
- Titles: When using nouns to describe people often images of men pop up behind these titles without us realizing it. Many of these can be neutralized to circumvent these automatisms. Example: users, developers, doctors.
- Functions: Certain expressions of functions are more masculine than others for historical reasons. We highlight these specifically. Examples: Expert, engineer, surgeon.
- Leadership: The subject of leadership mostly has the connotation of stereotypically male and traditionally hierarchical . Examples: Boss, supervisor, CEO.
- Hidden gender: In certain words there are imperceptibly male terms. Here, too, a male image is created in the mind's eye. Example: user manual, developer guide, astronaut suite.
- Endings (in German only): To include the non-binary gender, a special form is needed. Witty suggests the use of the gender asterisk or gender colon. Examples: Kund*innen, Frauen: und Männer, Gastgeber:innen
- Female stereotype: Word combinations that convey an outdated image of the role of women or discriminate against them. Examples: Housewife, career woman, "that's so easy, your mother could do it".
- Male stereotype: word combinations that convey an outdated image of men's role or discriminates against them. Examples: Master of the situation, father of the family, brogrammer
None. Positive impact on inclusion
Here Witty marks terms that have an inclusive character. Therefore, no alternatives are offered, but by marking them, writers learn which words they have written are already inclusive.
Subcategories we analyze:
- Communal language: it expresses cooperation, support and teamwork. All people recognize themselves in this language, as they are picked up on the level that defines us as human beings: working something out together, achieving something together, emphasizing meaningfulness behind activities. Generation Z in particular also expects this culture in the workplace. Examples here are: cooperation, coach, support.
- Diversity & inclusion as a company policy: It is increasingly important for many employees that they encounter diversity in the workplace and that an inclusive culture is lived where all people can contribute their full potential. If a company emphasizes this in communication (and of course lives it), it acts as an invitation for people with diverse backgrounds. Examples: diverse backgrounds, equality, intersectionality.
Please share your opinion
It is our desire to expand the framework with you. Do you think of any words or word combinations that fit into one category or another? Or do you think there are categories we have forgotten? We know that a framework is not set in stone and needs to remain agile. Also, we don't think of ourselves as doing everything right from the start. That's why we are super happy if you give us feedback. We look forward to your opinions and ideas.
Please add your ideas in this survey.