Towards misinformation mitigation on social media: novel user activity representation for modeling societal acceptance

Ahmed Abouzeid,Ole-Christoffer Granmo,Morten Goodwin,Christian Webersik

DOI: https://doi.org/10.1007/s42001-024-00256-9

2024-03-22

Journal of Computational Social Science

Abstract:Abstract Intervention-based mitigation methods have become a common way to fight misinformation on Social Media (SM). However, these methods depend on how information spreads are modeled in a diffusion model. Unfortunately, there are no realistic diffusion models or enough diverse datasets to train diffusion prediction functions. In particular, there is an urgent need for mitigation methods and labeled datasets that capture the mutual temporal incidences of societal bias and societal engagement that drive the spread of misinformation. To that end, this paper proposes a novel representation of users’ activity on SM. We further embed these in a knapsack-based mitigation optimization approach. The optimization task is to find ways to mitigate political manipulation by incentivizing users to propagate factual information. We have created PEGYPT, a novel Twitter dataset to train a novel multiplex diffusion model with political bias, societal engagement, and propaganda events. Our approach aligns with recent theoretical findings on the importance of societal acceptance of information spread on SM as proposed by Olan et al. (Inf Syst Front 1–16, 2022). Our empirical results show significant differences from traditional representations, where the latter assume users’ exposure to misinformation can be mitigated despite their political bias and societal acceptance. Hence, our work opens venues for more realistic misinformation mitigation.