This study extends uncertainty reduction theory beyond dyadic interaction by introducing communal uncertainty reduction strategies in human-AI socio-emotional communication, wherein users navigate AI-related uncertainty by engaging with both AI chatbots and online communities. Through the content analysis of 1772 posts and 3021 comments extracted from 35,579 conversation episodes in the Replika subreddit, we identify five community practices (e.g. anchoring, help-seeking), five peer response types (e.g. collaborative interpretation, group identification), and four uncertainty reduction outcomes (e.g. behavioral pattern recognition, predictive understanding), demonstrating that uncertainty reduction is a triadic process involving users, AI, and communities. The findings illustrate how communal uncertainty reduction transforms AI opacity into shared knowledge and solidarity, offering a new framework for understanding uncertainty in human-AI relationships.
Gan et al. (Mon,) studied this question.