Social Group Differences in the Social Media Discussion about ChatGPT and Bing Chat

This paper investigates how social media discussions evolved after the release and adoption of large language models by a wider public, with a focus on ChatGPT and the integration of a GPT model into Bing. The study aims to explore how social factors impact the way in which NLP technologies are perceived in social media posts. Using the official Twitter API, we collected a dataset of English and German tweets posted between November 30, 2022, and February 19, 2023. The study employs sentiment analysis and demographic prediction, the results reveal that tweets mentioning ‘Bing’ (related to the integration of a GPT model) were more likely to be negative compared to tweets about ‘ChatGPT’, with female users relatively more likely to express negative sentiment. The sentiment of tweets varied by language and whether an account belonged to an organization or not. This study provides insights into how social factors shape the discourse around NLP technologies on Twitter.