Add to Favourites
To login click here
This article discusses the need for equity and diversity in dialogue systems, specifically in the use of autoregressive language models. The authors present a framework for evaluating equity in human-AI communication and conduct a study using GPT-3 to examine its responses to different populations. They found that certain minority groups had a worse user experience and were more likely to change their attitudes after interacting with the AI. The implications of these findings for creating more inclusive conversational AI systems are discussed.
