Content

Social agents

Kevin Nam, Mark Ackerman

Keywords: social agent system, virtual community, blackboard architecture,
latent semantic indexing

Virtual communities, like any community of people, often have trouble regulating their participants. These problems can range from people being
disruptive, people attacking others verbally, to people abusing group norms.
It may even include people misposting, placing objects in the wrong places,
or otherwise acting erroneously.

This project’s approach includes the design and implementation of an agent-based system written in Java that helps facilitate and regulate online view2social spaces appropriately and also helps maintain a pleasurable environment for users. The system comprises of a number of agents that work collaboratively through a blackboard architecture. Each agent looks for a specific problem,  and new agents can be built and added to the system as required. This  system as a whole is not designed to completely eliminate the need for human regulator, but rather to help reduce human intervention in regulating online communities.

The agent system is written in JAVA.  It utilizes the SlashDot web forum as a test bed.  At its core is a blackboard architecture that acts as a central repository for knowledge gathered by many agents, with each agent trying to contribute to a solution to a given problem.  It also implements and utilizes the latent semantic analysis of the posting contents.