Health Misinformation Vs. Facts on Social Media: Co-Occurrence Network Analysis in Bangladesh


  • Parinda Rahman University of British Columbia
  • Ifeoma Adaji University of British Columbia



Health, Social media, Misinformation, Social Network analysis, Word Co-occurrence Network


The increased usage of social media provides a way to disseminate health-related information more quickly. Alternatively, sharing health content on social media poses risks due to unrestricted posting, enabling misinformation to spread. Regional social and cultural contexts influence themes in social media posts, underscoring the importance of understanding content and prevalent misinformation themes. This insight is crucial for tailoring interventions, resource allocation, misinformation detection algorithms, and policy formulation. We conducted word co-occurrence network analysis, creating and analyzing two networks for valid information and misinformation in Bangladesh. The prevalence of misinformation regarding natural ingredients and treatments in Bangladesh underscores the need for targeted efforts to combat health misinformation on social media. For each network, we computed metrics such as betweenness, Katz centrality, out-degree, and degree distribution. Furthermore, we computed the Louvain clustering algorithm to identify word clusters. A comparative analysis of both networks suggested that the context of words used in sentences was important and that both networks contained information about natural remedies or ingredients for health benefits. The misinformation network contained the word raw turmeric with the highest bigram frequency of 162. These natural remedies were stated as cures, and there was much misinformation and valid information surrounding common health conditions such as blood pressure. This was depicted through the word blood having an outdegree of four and seven in the misinformation and valid information networks, respectively. The valid information network emphasized the beneficial properties of natural ingredients rather than their supposed ability to cure diseases. This study provides insights into the distinctions and parallels between valid health information and misinformation on social media, considering their social and cultural context. It underscores shared semantics and bigram words between them, suggesting that understanding these differences can aid in addressing region-specific challenges.

Author Biographies

Parinda Rahman, University of British Columbia

Parinda Rahman is a graduate student studying at the University of British Columbia. She is a graduate research assistant working in the persuasive technologies and social computing lab. She is passionate about exploring the social implications of technology. Her research focuses on misinformation and the ethical implications of persuasive technologies.

Ifeoma Adaji, University of British Columbia

Dr. Ifeoma Adaji is an assistant professor at the University of British Columbia. She leads the Persuasive Technologies and Social Computing Lab and her research interests focus on designing behavior-changing apps/games, analyzing social media user behavior, and exploring trust and ethics in persuasive technology development.