AI Chatbots Judge You by Your Dialect, Study Finds

New research reveals that popular AI chatbots show "alarmingly" high levels of bias against people who do not speak perfect, standard English. A study found that large language models (LLMs) like those behind ChatGPT can discriminate based on dialect. Users who type in non-standard English, such as African American Vernacular English (AAVE) or a Scottish dialect, may receive lower-quality, less polite, or even less accurate responses. This bias is embedded in the AI's training data, which is often sourced from the internet and over-represents formal, standard English. Experts warn this could create a new digital divide. People who speak dialects could be unfairly denied good job advice, healthcare information, or customer service from AI tools. The researchers suggest one solution is to create customized AI models. These would be trained specifically on diverse dialect data to ensure fair and accurate responses for all users.