Mathematical modeling of disinformation and effectiveness of mitigation policies

Abstract

AbstractDisinformation is spread to manipulate public opinion for malicious purposes. Mathematical modeling was used to examine and optimize several strategies for combating disinformation—content moderation, education, and counter-campaigns. We implemented these strategies in a modified binary agreement model and investigated their impacts on properties of the tipping point. Social interactions were described by weighted, directed, and heterogeneous networks. Real social network data was examined as well. We find that content moderation achieved by removing randomly selected agents who spread disinformation is comparable to that achieved by removing highly influential agents; removing disinformation anywhere in a network could be an effective way to counter disinformation. An education strategy that increases public skepticism was more effective than one that targets already biased agents. Successful counter-campaign strategies required a substantial population of agents to influence other agents to oppose disinformation. These results can be used to inform choices of effective strategies for combating disinformation.