P2P

Spring25

Peer to Peer: ILTA's Quarterly Magazine

Issue link: https://epubs.iltanet.org/i/1533864

Contents of this Issue

Navigation

Page 26 of 69

P E E R T O P E E R M A G A Z I N E · S P R I N G 2 0 2 5 27 The same applies to diversity knowledge, you never know a person's perspective or familiarity. AI can help us with many tasks we may have struggled with, like composing an email to gently explain something you have sent instructions on several times before. It can also assist with thoughtful compositional elements like inclusive language, bias detection, cultural sensitivity, and accessibility. Ask me to write a set of steps to get a user where they need to go in an application, and I will rise to that occasion – even if it is the 100th time – wink. Ask me to compose an email that is both sensitive and informative; I may struggle with that. Since we should all strive to improve communications, when possible, let us look at ways AI can help us increase our sensitivity to diversity and inclusion. INCLUSIVE LANGUAGE The American Psychological Association Inclusive step towards inclusive language, due diligence, can be accomplished through AI. Use active prompt verbs like "analyze" combined with information about the recipients to produce editing suggestions. For example, you can paste the intended text of your communication into an AI assistant and preface it with "Analyze this message for diversity and cultural appropriateness as though it were being sent to company peers in the United States." You will be surprised by how quickly and easily the suggested changes improve your message! BIAS DETECTION AI can identify and flag biased or insensitive language, offering alternative phrasings that are more inclusive and respectful. As the United Nations' Disability-Inclusive Language Guidelines mentions, "the language that we use to refer to persons with disabilities has an impact, as it shapes our perception of the world" (https://www.ungeneva.org/ Language Guide states it best: "We encourage all people to adhere to the basic principles of inclusive language, which are to choose appropriately specific terms and to show respect by calling people what they call themselves. This requires being open to continual learning and capacity building and remaining mindful that language may change. Identity is intersectional, meaning that people have multiple identities affected by interlocking systems of oppression and privilege. No group is a monolith. Make sure to use inclusive terms to acknowledge that intersectionality. People are different and may disagree on language. It is acceptable to recognize that there may be no perfect solution. Ensure you do your due diligence in the language you select." (https://www.apa.org/ about/apa/equity-diversity- inclusion/language- guidelines) Regarding the last sentence of the statement, another sites/default/files/2021-01/ Disability-Inclusive- Language-Guidelines.pdf). Closed source models (such as ChatGPT) are proprietary, and their code cannot be modified. Therefore, bias detection in closed source models may be unreliable, depending on the sources the model is trained on. Reviewing the output for bias is crucial when using a closed source model. Ways to mitigate bias in responses include using neutral language in the prompt, being mindful of your biases, and detecting stereotypical language. Example: Paste the intended message into an AI assistant and preface it with "Detect any biases in the following message. I want to send it out to staff and attorneys at a law firm in that includes people across all spectrums of gender identity, race, and religion." A diverse and inclusive workplace makes everyone feel equally involved in and supported in all workplace areas regardless of who they are or what they do for the business. The "all spectrums" part is essential. CULTURAL SENSITIVITY AI can provide insights into cultural norms and sensitivities, helping with inclusionary language A diverse and inclusive workplace makes everyone feel equally involved in and supported in all workplace areas regardless of who they are or what they do for the business.

Articles in this issue

Links on this page

Archives of this issue

view archives of P2P - Spring25