In this month’s research spotlight, we highlight recent research from COSMOS that focuses on network analysis and community dynamics during polarized discourse. These studies used social media analysis to examine online interaction patterns. Specifically, the studies summarized are titled,

• “Analyzing Key Network Structures of 2022 Malaysian General Elections from the Lens of Instagram,” and
• “Examining the Impact of Toxicity on Community Structure in Social Networks,” which also won the Best Paper Award.

Each studied how social networks form and evolve during specific periods of heightened or polarized public discourse, such as elections and pandemic. However, they differed in the specific platforms studied (Instagram vs. Twitter/X). These studies were published and presented recently at the 14th International Conference on Social Media Technologies, Communication, and Informatics (SOTICS 2024), which took place from September 29 to October 3, 2024, in Venice, Italy.

“Analyzing Key Network Structures of 2022 Malaysian General Elections from the Lens of Instagram,” analyzed social media’s influence during the 2022 Malaysian general election, focusing specifically on Instagram. The study examined 53,116 Instagram posts using two main analytical approaches: Contextual Focal Structure Analysis (CFSA) and topic modeling. The researchers collected data using specific election-related hashtags. Their analysis revealed complex networks of interaction between journalists, media houses, politicians, and political parties on Instagram during the election period. 

The authors further identified 11 key network structures, with particular attention given to one of these structures, which exerted the most influence. This structure highlighted the intricate relationship between media professionals (including figures from Astro AWANI and OnAirTalentManagement) and political actors (such as Friends Of Harapan Selangor). Their analysis demonstrated how these groups shaped political discourse during the election.

The findings suggest a significant blurring of lines between media and political sectors in Malaysia’s digital sphere, raising questions about media independence and narrative framing during elections. This research contributes to understanding digital democracy in Malaysia and offers insights into how social media shapes modern political communication.

The second study, titled “Examining the Impact of Toxicity on Community Structure in Social Networks,” examined how toxic content affects community structure on social media, focusing on pro-vaccine and anti-vaccine communities during the COVID-19 period. The study analyzed data from X (formerly Twitter) between January 2020 and June 2021, using metrics like toxicity scores, modularity, and clustering coefficients to understand community dynamics. Key findings revealed that increased toxicity generally leads to community fragmentation, but with different impacts across groups. Pro-vaccine communities showed more significant fragmentation in response to toxic content compared to anti-vaccine communities, possibly due to greater opinion diversity in pro-vaccine discussions. The researchers used various analytical methods, including Granger causality tests. Community dynamics were visualized through the Sankey diagram.

Notably, when examining community evolution, 15 out of 20 toxic communities eventually transformed into non-toxic communities by the end of their observed periods. The study also found that toxicity’s impact often appears with a time lag, i.e., toxicity’s impact on community fragmentation is not felt instantaneously. 

Prof. Agarwal said, “These findings have important implications for social media platform moderation and community management strategies, offering insights into how toxic behavior shapes online discourse and community stability.”