Social media's increasing global importance has expedited the spread of harmful practices. The proliferation of digital platforms and interactive websites has turned the Internet into a place where individuals with harmful behaviors can torment others. Computational algorithms are being actively developed by researchers and network providers to detect and eliminate harmful content from text provided by users. This unprecedented situation calls for rapid and accurate epidemiological models to support network providers in developing successful strategies. In this work, we adapted the SEIQR (Susceptible, Exposed, Infected, Quarantine, Recovery) epidemiological model to study toxicity spread on social media platforms. To increase the precision of the model fit, we also incorporated memory effects. Most of the epidemiological models exclude the past of an individual, which is critical for controlling social media behavior. We have evaluated the toxic post-free equilibrium point, the reproduction number $(R_0)$, the existence-uniqueness solution, and the stability solution. Our findings of this study reveal that the index of memory and the quarantine rate can be utilized as preventive measures for the toxicity spread on social network platforms. The key contribution of this study lies in the model's dual effect: not only does it effectively reduce the error rate to 0.003, but it also accomplishes this with a reduced number of users needing removal from the network.

PDF