Adversarial Manipulation of Human Social Sensing

The consequences of inaccurate collective perceptions of what a population is doing can be devastating. Social media users’ overestimating other users’ expressions of outrage on their favorite platforms can heighten intergroup conflict. Stock traders’ overestimating the number of other traders selling an asset can result in panic selling. In this paper, we take on the role of an adversary wishing to induce these kinds of consequences for a target nation. We assume that the adversary wishes to do this by inducing collective misperceptions about the behavior of the nation’s population. We propose that the adversary can achieve this by changing the behavior of relatively few well-positioned individuals within the social network structure of the population. To achieve its goal, the adversary can then rely on individuals’ inaccurate inferences about the population based on their ability to locally “sense” their social environments. We frame the problem of selecting which subset of individuals’ behavior to change to maximally skew collective perceptions as a combinatorial optimization problem. We then provide an algorithm that optimally solves this problem.