Modeling the Impact of Timeline Algorithms on Opinion Dynamics Using Low-rank Updates
Keywords: opinion dynamics, opinion-formation, Friedkin--Johnsen model, social networks, polarization and disagreement in social networks
TL;DR: We present a new model and methods for integrating aggregate statistics from timeline algorithms with opinion-formation models. We show how to make small changes to the users' timeline compositions to minimize polarization and disagreement.
Abstract: Timeline algorithms are key parts of online social networks, but during recent years they have been blamed for increasing the polarization and disagreement in popular social networks. One of the key obstacles to explaining these phenomena is that polarization and disagreement appear in a *global network-level*, whereas timeline algorithms operate on a *local user-level*. Bridging between these two levels of abstraction is a major challenge. In particular, while network-level polarization and disagreement have been successfully studied using opinion-formation models, it has remained an open question of how these models can be augmented to take into account the fine-grained impact of user-level timeline algorithms.
We make progress on this question by providing a way to model the impact of timeline algorithms on opinion dynamics. Specifically, we show how the popular Friedkin--Johnsen opinion-formation model can be augmented based on *aggregate information*, extracted from timeline data. Our idea is to combine the underlying follow-graph of the online social network with a graph that is induced by data from a timeline algorithm. The aggregate information that we consider are the topics that are discussed in the social network, as well as the users' interests and influence on these topics. To the best of our knowledge, this is the first work that allows to obtain theoretical guarantees for combining an opinion-formation model with a graph induced by a timeline algorithm.
We use our model to study the problem of minimizing the polarization and disagreement; we assume that we are allowed to make small changes to the users' timeline compositions by strengthening some topics of discussion and penalizing some others. We present a gradient descent-based algorithm for this problem, and show that under realistic parameter settings, our algorithm computes a $(1+\varepsilon)$-approximate solution in time $\tilde{O}(m\sqrt{n} \lg(1/\varepsilon))$, where $m$ is the number of edges in the graph and $n$ is the number of vertices. We also present an algorithm that provably computes an $\varepsilon$-approximation of our model in near-linear time. We evaluate our method on real-world data and show that it effectively reduces the polarization and disagreement in the network. We also show that our algorithm is orders of magnitude faster than a non-optimized black-box optimization approach. Finally, we release an anonymized graph dataset with ground-truth opinions and more than 27,000 nodes (the previously largest publicly available dataset contains less than 550 nodes).
Track: Social Networks, Social Media, and Society
Submission Guidelines Scope: Yes
Submission Guidelines Blind: Yes
Submission Guidelines Format: Yes
Submission Guidelines Limit: Yes
Submission Guidelines Authorship: Yes
Student Author: Yes
Submission Number: 2502
Loading