Description

Title A Normative Agent System to Prevent Cyberbullying
Abstract Automated approaches to prevent the negative effects of cyberbullying mainly focus on affective agents that provide support for victims. The current paper takes a complementary approach, which attempts to minimise the amount of occurrences of cyberbullying in the first place. The approach consists of a system of normative agents, which are physically present in a virtual society. The agents, which reason based on a BDI-model, use a number of techniques to detect various norm violations, including insulting and following. By using rewards and punishments, they try to reinforce the desired behaviour of the users. The system has been implemented and tested within a virtual environment for children between 6 and 12 years old, called Club Time Machine. In a real world experiment, the behaviour of the users of the virtual environment has been logged and analysed by means of a logic-based checking tool. The results show that the normative agents have the potential to reduce the amount of norm violations on the long term.