Human and LLM Facilitation in Collaborative Design Meetings: A Computational Behavioral Comparison
The study compared human Project Manager facilitation turns with responses from Llama 3.1 8B Instruct across 199 design meeting contexts using five computational dimensions like semantic distance and topic divergence. Human turns scored higher on contextual distance and empathy/hedge density, while LLM responses excelled in topic divergence and instruction-conditioned phase-vocabulary alignment. The findings profile measurable behavioral differences rather than judging facilitation quality.