An Intelligent Multi-Agent System Approach to Automating Safety Features for On-Line Real Time Communications: Agent Mediated Information Exchange

Child safety online is a growing problem, governmental attempts to highlight and combat this issue have not been as successful as it was hoped, and still there are highly publicised cases of children, young people and vulnerable adults coming to harm as a result of unsafe online practices. This thesis presents the research, design and development of a prototype system called SafeChat, which will
provide a safer environment for children interacting in online environments.

In order to combat such a complex problem, it is necessary to integrate various artificial intelligent technologies and autonomous systems. The SafeChat prototype system discussed within this research
has been implemented in Java Agent Development Environment (JADE) and utilises Protégé Ontology development, reasoning and natural language processing techniques. To evaluate our system performance, comprehensive testing to measure its effectiveness in detecting potential risk to the user (e.g. child) is in constant development. Initial results of system testing are encouraging and demonstrate its effectiveness in identifying different levels of threat during online conversation.

The potential impact of this work is immense, when used as a plug-in to popular communications software, such as Facebook Messenger, Skype and WhatsApp. SafeChat provides a safer environment for children to communicate, identifying potential and actual threats, whilst maintaining the privacy
of their discourse. The SafeChat system could be easily adapted to provide autonomous solutions in other areas of online threat, such as cyberbullying and radicalisation.