Enhancing social science research on cyberbullying through human machine collaboration

Area of Science:

• Social Sciences
• Computer Science
• Adolescent Psychology

Background:

• Cyberbullying (CB) affects nearly half of European children, posing a significant challenge due to digital anonymity.
• Traditional statistical methods struggle to identify causal links in CB due to reliance on correlations.
• Ethical constraints prevent experimental studies on cyberbullying, necessitating advanced analytical approaches.

Purpose of the Study:

• To introduce a novel human-machine consensus framework for causal discovery in cyberbullying research.
• To support social scientists in understanding the complex causal mechanisms underlying cyberbullying.
• To leverage data-driven causal inference and expert knowledge for robust analysis.

Main Methods:

• Utilizing Directed Acyclic Graphs (DAGs) for identifying causal relationships from observational data.
• Integrating automatic causal discovery algorithms with expert knowledge to mitigate limitations of each approach.
• Employing Probabilistic Graphical Causal Models (PGCMs), or Bayesian Networks, for enhanced interpretability and probabilistic reasoning.

Main Results:

• The proposed framework enables a hybrid methodology combining algorithmic and expert-driven insights.
• It facilitates the creation of model ensemble estimations for causal effects.
• The approach mitigates cognitive biases and enhances transparency in model construction.

Conclusions:

• The human-machine consensus framework offers a powerful tool for causal discovery in cyberbullying research.
• Interpretable, expert-informed causal models are crucial for guiding policy and intervention strategies in ethically constrained research areas.
• This methodology advances the understanding of complex social phenomena like cyberbullying.