Echo chamber effect within and across communities in rumor spread during public health crisis

Purpose During public health crises, multiple rumor events emerge simultaneously on social media, forming interconnected discussion communities. Echo chamber effect within these communities may reinforce misinformation and weaken rumor-refutation efforts. This study aims to quantify rumor and anti-rumor echo chamber effects, examine their cross-community contagion patterns and explore factors related to these patterns. Design/methodology/approach We analyzed data from 49 highly discussed rumor events on Sina Weibo during the COVID-19 pandemic using content analysis, sentiment computation, network mining and generated statistical models. Indicators along two dimensions, standpoint-distribution homogeneity and standpoint-interaction homogeneity, were developed to measure echo chamber effect in user-level interaction networks. Community-level networks based on shared users were then constructed, and Exponential Random Graph Models were employed to examine contagion patterns, supplemented by similarity analysis of event topics and sentiments. Findings Retweeting networks exhibited strong homogeneity in both standpoint distribution and interaction, whereas commenting networks showed strong interactional but weaker distributional homogeneity, forming polarized cliques. Importantly, cross-community similarities in echo chamber effects, particularly standpoint-distribution homogeneity, were significantly associated with users' cross-community interactions, which were closely related to whether interconnected communities discussed shared topics or expressed similar sentiments. Originality/value This study extends echo chamber research from single communities to interconnected rumor ecosystems. The findings provide data-driven insights for identifying clusters of highly homogeneous rumor communities and supporting coordinated rumor-refutation and infodemic management.