Research Summary: Teachers' perspective on AI-based Multi-Agent Simulation Design to Combat School Bullying

Research has explored the co-design of an AI-based multi-agent simulation system with K-12 teachers to address school bullying. This system, powered by Large Language Models (LLMs), simulates realistic scenarios to foster anti-bullying skills in students, aiming to overcome limitations of current interventions.

Why it matters

The development of AI-driven multi-agent simulations presents a novel strategic avenue for addressing complex social issues like school bullying. This approach could lead to more effective and scalable interventions, potentially transforming how institutions prepare individuals to navigate challenging social dynamics.

Key insights

• Existing anti-bullying interventions, both in-person and digital, often do not adequately address the complex social dynamics of bullying.
• A multi-agent anti-bullying system, co-designed with K-12 teachers and driven by LLMs, can simulate authentic scenarios.
• The system's objective is to enable students to develop effective anti-bullying skills.
• The research identifies critical design parameters for such an LLM-driven simulation system.
• This approach offers insights for developing more effective and scalable anti-bullying tools.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2609.35776