Estimating the impact of coordinated inauthentic behavior on content recommendations in social networks
Year: 2022
Type: conference-paper
Abstract: Ranking and recommendation algorithms drive user interactions on social networks.However, these algorithms are vulnerable to attack from coordinated campaigns spreading misleading information online.In the absence of proprietary data from platforms in general it is not possible to estimate the true impact of harms arising from such coordinated inauthentic behavior.We employ simulations to estimate the relative effects of coordinated inauthentic behavior on content recommendations on social networks.We develop a multiagent simulation of a popular social network, Reddit, and use millions of posts from users on the platform to model content-sharing on the platform.Quantifying the harms that arise as a result of coordinated behaviors helps make informed design choices for recommendation algorithms.We provide a general framework to employ simulations to estimate effects of coordinated activity, grounded in empirical evidence from real-world social networks.
Language: English
Subfield: Sociology and Political Science
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7225207378
