A multilayered graph-based framework to explore behavioural phenomena in social media conversations

Area of Science:

• Social media analysis
• Health communication
• Computational social science

Background:

• Social media platforms are integral to modern health communication.
• Understanding the dynamics of health opinion formation on these platforms is crucial.

Purpose of the Study:

• To investigate the influence of social contagion, biased assimilation, and homophily on health opinions shared on social media.
• To analyze how these factors shape and alter user stances regarding health topics, using COVID-19 vaccination as a case study.

Main Methods:

• Utilized a multilayered graph-based framework to analyze English and Spanish Twitter conversations about COVID-19 vaccination.
• Applied deep learning models to infer user stances and employed graph centrality and homophily scores to interpret information reproduction patterns.

Main Results:

• English posts showed higher stance similarity (r=0.51) compared to Spanish posts (r=0.38).
• Stance homophily was observed for specific vaccines (Pfizer, AstraZeneca) in English and to a lesser extent in Spanish conversations.
• Highly connected users were a minority and lacked significant social influence, while Spanish conversations predominantly promoted vaccination, unlike English ones which presented more contrasting views.

Conclusions:

• The developed methodology effectively quantifies social behaviors' impact on health information dissemination across social platforms.
• Demonstrated effectiveness through case studies in English and Spanish, highlighting demographic and sociocultural differences in online health discourse.