Engineering Socially Intelligent Personal Agents via Norms

This thesis develops Arnor, an agent-oriented software engineering (AOSE) method to engineer social intelligence in personal agents. Arnor goes beyond traditional AOSE methods to engineer personal agents by systematically capturing interactions that influence social experience. We empirically evaluate Arnor via a developer study, and a set of simulation experiments. We find that (1) Arnor supports developers in engineering personal agents, and (2) personal agents engineered using Arnor provide a greater social experience than agents engineered using a traditional AOSE method.