Conversational AI has fundamentally reshaped the way we interact with technology. While large language models (LLMs) have seen significant progress in one-on-one interaction, they rarely capture the full complexity of human communication. Many real-world interactions are inherently multi-sided, including team meetings, family dinners, or classroom lessons. These interactions involve fluid changes, changes in roles, and dynamic interruptions.
For designers and developers, simulating natural and engaging multi-party conversations has historically required a trade-off: compromise for hardness Accept scripted interactions uncertainty Of a purely generative model. To bridge this gap, we need tools that blend the structural predictability of a script with the spontaneous, improvisational nature of human conversation.
To address this need, we introduce DialogLabpresented at ACM UIST 2025One open source Prototyping framework designed for writing, simulating, and testing dynamic human-AI group conversations. DialogLab provides a unified interface to manage multi-party dialogue complexity, handling everything from defining agent personas to orchestrating complex turn-taking dynamics. By integrating real-time improvisation with structured scripting, this framework enables developers to test conversations ranging from structured question-and-answer sessions to free-flowing creative brainstorming. Our evaluation with 14 end users or domain experts confirms that DialogLab supports efficient iteration and realistic, adaptable multi-party design for training and research.
