In this month’s research spotlight, we highlight recent research from COSMOS that focuses on online toxicity diffusion, ways to computationally model its spread, and mitigation methods. These studies were published and presented recently at the 2024 Complex Networks & Their Applications (CNA) conference, which took place from December 10 to 12, 2024, in Istanbul, Turkey. Specifically, the studies we spotlight here are titled,

• “Combating Toxicity: A Systematic Approach to Model Quarantine Intervention for Varied Toxicity Levels,” and
• “Modeling Toxicity Propagation on Reddit using Epidemiology.”

Each of these studies focused on how online toxicity propagates on social media platforms, as well as how it can be mitigated. The former studied the impact of quarantining users in the novel SEIQR model, while the latter studied exhaustively the accuracy of five separate toxicity models.

“Combating Toxicity: A Systematic Approach to Model Quarantine Intervention for Varied Toxicity Levels” applied the SEIQR epidemiological model to analyze toxicity spread on social media platforms. Researchers examined datasets from COVID-19 and social movement discussions on X (Twitter), categorizing toxic content into moderate and high intensity levels. By splitting datasets, they achieved lower error rates in modeling toxicity propagation. A sensitivity analysis revealed key parameters influencing toxicity spread, including positive factors like infection rate and recruitment rate, and a negative factor in user exit rate. The research recommends various strategies for managing online toxicity based on the model’s findings.

“Modeling Toxicity Propagation on Reddit using Epidemiology” examined how epidemiological models can be applied to measure toxicity propagation on Reddit. The researchers tested five established models (SIS, SIR, SIRS, SEIR, SEIRS) on two datasets from Reddit. Results showed these models can achieve less than 2% fitting error when tracking toxicity spread. More complex models like SEIRS and SIRS performed significantly better than simpler ones like SIS, though SIRS worked well for linear data patterns. This research demonstrates epidemiological modeling’s effectiveness on Reddit and could inform future toxicity prevention strategies.