Complier vs. Defier: A Sparse Machine Learning Approach to Understanding Bullying Influence Networks

• In Event: Advancing Crime and Criminal Justice Research with Innovative Methods

Abstract

Traditional studies on high school bullying networks rely on self-reported or nominated friendships, overlooking broader peer relations such as classmates, roommates, and schoolmates. This study introduces bullying influence networks, which capture how students' bullying or non-bullying behaviors influence one another. Using data from nine junior high schools in China and an exponential random graph model (ERGM), we examine the formation patterns of these networks. A sparse machine learning algorithm reveals two types: complier bullying networks, driven by homophily and social learning, and defier bullying networks, shaped by adversarial social interactions. Complier networks form when students share similar characteristics, such as sex or attitudes toward school punishment, leading them to imitate their peers’ behaviors. Defier networks emerge when students with stronger self-control, personal expectations, or academic ambitions resist delinquent influences from their peers. These findings offer a nuanced understanding of bullying influence networks beyond friendships, highlighting distinct social mechanisms that drive compliance and resistance to bullying behaviors.

Authors

• Shiming Hao, College of Law, Yunnan University, China

• Zhong Wang, College of Law, Hunan University, China

• Yan Zhang, Sam Houston State University