Topics, Temporal Patterns, and Network Characteristics of AI-Related Discourse on Reddit

Pingjing Yang, Kanyao Han, Jana Diesner

Abstract: How does the public perceive Artificial Intelligence (AI)? We present a sliver of an answer to this question by analyzing discussions on 26 AI-related subreddits from 2005 to early 2023. We apply (1) topic modeling to find latent topics that represent the gist of subreddits as collections of representative keywords, (2) network analysis to identify interaction patterns around these topics, and (3) statistics to test for correlations between AI-related topics and how they are discussed. The identified topics range from high-level AI concepts to applications and societal impacts. The temporal analysis revealed four types of topics: vanishing, resurgent, ephemeral, and emerging ones; representing the dynamic nature of public interest and concerns. We found discussions of vanishing (e.g., self-driving) and ephemeral (e.g., Metaverse) topics to lack dispersion, depth, and controversiality, while most emerging (e.g., Workplace replacement) and resurgent (e.g., Civilization and AI) topics are positively correlated with at least one of these measures. This research advances the understanding of the substance and evolution of AI-related discourse through based on Reddit data.