Strategic Classification with Graph Neural Networks

TL;DR

This paper explores how graph neural networks can be used in strategic classification settings where users modify features, revealing that social relations can be exploited by strategic users to influence predictions, and proposing a robust learning framework.

Contribution

It introduces a differentiable framework for strategically-robust learning of graph neural networks, addressing inter-user dependencies in strategic classification.

Findings

Graph neural networks can be exploited by strategic users to influence predictions.

The proposed framework improves robustness in social network-based classification.

Experiments show effectiveness on real network datasets.

Abstract

Strategic classification studies learning in settings where users can modify their features to obtain favorable predictions. Most current works focus on simple classifiers that trigger independent user responses. Here we examine the implications of learning with more elaborate models that break the independence assumption. Motivated by the idea that applications of strategic classification are often social in nature, we focus on \emph{graph neural networks}, which make use of social relations between users to improve predictions. Using a graph for learning introduces inter-user dependencies in prediction; our key point is that strategic users can exploit these to promote their goals. As we show through analysis and simulation, this can work either against the system -- or for it. Based on this, we propose a differentiable framework for strategically-robust learning of graph-based classifiers. Experiments on several real networked datasets demonstrate the utility of our approach.