BeECD: Belief-Aware Echo Chamber Detection over Twitter Stream.

Guan Wang,Weihua Li ,Shiqing Wu,Quan Bai,Edmund M.-K. Lai

DOI: https://doi.org/10.1007/978-981-99-7025-4_27

2023-01-01

Abstract:The phenomenon known as the “echo chamber” has been widely acknowledged as a significant force affecting society. This has been particularly evident during the Covid-19 pandemic, wherein the echo chamber effect has significantly influenced public responses. Therefore, detecting echo chambers and mitigating their adverse impacts has become crucial to facilitate a more diverse exchange of ideas, fostering a more understanding and empathetic society. In response, we use deep learning methodologies to model each user’s beliefs based on their historical message contents and behaviours. As such, we propose a novel, content-based framework built on the foundation of weighted beliefs. This framework is capable of detecting potential echo chambers by creating user belief graphs, utilizing their historical messages and behaviours. To demonstrate the practicality of this approach, we conducted experiments using the Twitter dataset on Covid-19. These experiments illustrate the potential for individuals to be isolated within echo chambers. Furthermore, our in-depth analysis of the results reveals patterns of echo chamber evolution and highlights the importance of weighted relations. Understanding these patterns can be instrumental in the development of tools and strategies to combat misinformation, encourage the sharing of diverse perspectives, and enhance the collective well-being and social good of our digital society.