FairImpact

At a time when network platforms shape perceptions and drive public discourse, influence maximization strategies are capable of contributing to information asymmetry and echo chambers, leaving certain demographics out of important discussions and protecting individuals from points of view that challenge their own. Our project builds upon previous work to develop an influence maximization algorithm that acts with fairness as a constraint. Our method ensures that messaging, whether marketing, political, or public health and safety related, is spread by influencers to individuals with demographic attributes that match the original population. While fairness constraints would typically be computationally prohibitive to apply pervasively, our approach is scalable and easy to implement. With the United States’ 2024 election looming and the issue of polarization at the forefront of discussion, we present in our project an example of the potential impact of our work on political messaging and the dangers of discounting fairness.