A Solution to the Attitude-Behavior Puzzle in Bullying Research – A Computational Model of Group Dynamics in School Classes

• In Event: 2-037 - But What Do *They* Say? The Role of Group Dynamics in Bullying Behaviors and Bystanders Responses to Bullying

Integrative Statement

There is a puzzling mismatch between students’ negative attitudes towards bullying and their actual behavior (e.g., Salmivalli & Voeten, 2004). A majority of school-aged children consider bullying undesirable, yet it undeniably continues to occur. Although we know that group-level dynamics such as social norms play an important role in bullying behavior, we lack understanding on how bullying norms can emerge through the repeated interaction of individuals. This involves a feedback process, where individuals simultaneously are influenced by the norm and contribute to it. In order to model such feedback loops and their potentially unforeseen consequences, we propose an agent-based model (Bonabeau, 2002). An agent-based model is a theoretical tool that allows researchers to implement basic assumptions on how individuals interact in a computer program. Through computer simulations, we are able to study the group-level outcomes of these assumptions, and examine how bullying behavior can emerge in an environment characterized by anti-bullying attitudes.

Based on the current literature, we model two feedback loops that are presumed to be decisive in bullying dynamics. The first process is norm emergence. Here, we assume that if a popular child bullies a classmate, it will become more acceptable to bully that classmate. This in turn makes other children more likely to bully that individual, which makes victimization even more normalized. The second process is dominance reinforcement. Here, we assume that people who bully gain dominance. The larger the difference in dominance between two individuals, the more likely bullying becomes. Bullying again leads to a gain in dominance. Figure 1 shows an illustration of this process for a group of 20 individuals (agents) with a high average of anti-bullying attitudes. The model is able to replicate the empirically observed mismatch between attitudes and bullying behaviour. Very rapidly, a status hierarchy forms which leads to the bullying of five individuals (also see Figure 2). However, the number of victims rapidly declines to three and then two structural victims. Thus, initially a large number of individuals is bullied, and later the bullying rate declines. This is consistent with the empirical observation that at the start of a new group (e.g., transition to middle school) there are high rates of bullying, which then ebb away to leave just a few persistent victims.

As a next step, we perform simulation experiments, where we compare different classroom compositions. We investigate how both the mean level and the variance in bullying preferences in a group can determine levels of bullying. We furthermore examine whether one student with low anti-bullying preferences can set into motion the group-level processes that lead to the normalized victimization of another individual.

Ultimately, the proposed agent-based model can contribute to deepening our understanding of how the normative and persistent bullying of a few individuals can emerge. Moreover, we can systematically investigate different parameter values and group compositions. This enables us to formulate new testable hypotheses about group-level outcomes which can in the future be tested using empirical data.

Authors

• Julia Torres, University of Groningen
  Presenting Author

• Michael Mäs, University of Groningen
  Non-Presenting Author

• Gerine Lodder, University of Groningen
  Non-Presenting Author

• Rene Veenstra, University of Groningen
  Non-Presenting Author