𝗔𝗜 𝗮𝗻𝗱 𝗠𝗶𝘁𝗶𝗴𝗮𝘁𝗶𝗻𝗴 𝗕𝗶𝗮𝘀 𝗶𝗻 𝗔𝗰𝗮𝗱𝗲𝗺𝗶𝗮

Many of my colleagues in academia have shared concerns about the use of generative AI (GenAI) in academic contexts. One concern relates to GenAI producing text with bias due to biases inherited from human-generated training data. I'd like to follow up on two recent posts - one on AI and academic writing, and another on accessibility - with a few thoughts about LLMs, text, and bias.

𝗪𝗵𝗶𝗹𝗲 𝗚𝗲𝗻𝗔𝗜 𝗰𝗮𝗻 𝗶𝗻𝘁𝗿𝗼𝗱𝘂𝗰𝗲 𝗯𝗶𝗮𝘀, 𝗶𝘁 𝗰𝗮𝗻 𝗮𝗹𝘀𝗼 𝗯𝗲 𝗽𝗮𝗿𝘁 𝗼𝗳 𝘁𝗵𝗲 𝘀𝗼𝗹𝘂𝘁𝗶𝗼𝗻. Our students come from diverse backgrounds and the language we use plays a role in creating environments where all thrive. Large Language Models (LLMs like Claude, Gemini, or ChatGPT) can help identify biased language and elevate underrepresented perspectives, helping to facilitate the use of inclusive language. GenAI tools can scan academic writing (e.g., teaching materials, publications) to identify issues like gendered terms and amplify underrepresented voices.

I'm excited about AI's potential to identify subtle biases in written documents. When deployed responsibly, AI can serve as a powerful tool to support accessibility, inclusion, and diverse representation in academia.