Emotional Dimensions of Human–AI Interaction: An NLP Analysis of ChatGPT Discourse

Abstract

The rapid progress made in artificial intelligence (AI) has significantly affected and altered human technology interactions. Out of these recent innovations and advancements in AI, OpenAI’s ChatGPT is an extremely significant breakthrough in natural language processing. Despite its popularity and rapid adoption across various sectors, it has triggered some conflicting feelings, and these include admiration for its smooth and creative conversations, as well as worries about spreading wrong information and plagiarism.To better evaluate these worries and opportunities, we conducted an analysis of 20,000 English-language X (formerly Twitter) postings about ChatGPT from 2023-2024 using a BERT-based emotion classifier that incorporated Ekman’s basic emotions and Latent Dirichlet Allocation, a topic modeling technique. We then conducted Chi-squared tests on a set of emotions and associated discussion topics.The findings showed that surprise with 37% followed by joy with 22%, are mostly associated with conversations regarding novelty, playfulness, and creativity, while fear, anger, and sadness are grouped together with conversations involving misinformation, system transparency, and job disruption. The implications here are apparent: there are some patterns in the conversations about ChatGPT, and an emotion-centric NLP analysis would be useful for understanding user interactions with AI systems.

Ähnliche Arbeiten