Gen AI: Large Language Models and Linguistic Whitewashing

Chris Mayer

Assignments & Activities Archive

Activity Description

This multi-step lesson engages students in the critical analysis of Generative Artificial Intelligence (Gen AI) and its impact on linguistic diversity. Through assigned readings, writing tasks, and in-class activities, students will explore how Large Language Models (LLMs) can reinforce linguistic hierarchies and biases. Even older LLMs, like GPT-3, can generate content that human judges find comparable or superior to human writing in terms of fluency, coherence, relatedness, logic, and interest (Xie et al.), although they struggle to capture and express deep emotional understanding (Petroșanu et al.). This three-part mini-unit aims to familiarize students with LLMs and their relationship to linguistic diversity. First, students will engage with background readings on LLMs and linguistic diversity, followed by reflective writing. Next, for homework, they will watch a video lecture titled “Gen AI LLMs Washing Out Linguistic Features of Appalachian and Black Englishes” available on YouTube, or a similar instructor-recorded lecture. Finally, in class, students will experiment with LLMs using non-standard English dialects and discuss the social justice implications of their findings.