Design Principles for Human-Agent Interaction

This position paper presents 14 design principles aimed at enhancing human-agent interaction, emphasizing that the effectiveness of AI agents should not be measured solely by their autonomous capabilities but also by their ability to engage with humans. The principles are categorized across four interaction stages: initial engagement, active interaction, long-term collaboration, and handling failures. By applying these principles to evaluate nine existing agent systems, the authors provide a framework for AI design teams to create agents that are more usable, trustworthy, and effective in real-world applications, addressing a critical barrier to the widespread adoption of autonomous systems.