Presentation
“What Was That?” Retrieval-Augmented Generation to Aid Understanding of Vehicle Owner’s Manuals
DescriptionWe present a retrieval-augmented generation (RAG) system that transforms static vehicle owner’s manuals into interactive, context-aware resources for drivers. Combining natural language queries with real-time vehicle sensor data (OBD-II), our system retrieves and synthesizes relevant manual content to deliver situation-specific responses.
We evaluate system behavior through simulations of sensor error scenarios and develop human-centered metrics tailored to driver needs. Success is defined by responsiveness, contextual appropriateness, and actionable relevance, promoting the evaluation of RAG systems through human-centered design principles.
Our work exemplifies the expansion of human factors research from interface design toward the evaluation of adaptive and context-driven systems. By grounding both generation and assessment in operational context, we offer a framework for building reliable driver-assistance systems that prioritize meaningful support.
We evaluate system behavior through simulations of sensor error scenarios and develop human-centered metrics tailored to driver needs. Success is defined by responsiveness, contextual appropriateness, and actionable relevance, promoting the evaluation of RAG systems through human-centered design principles.
Our work exemplifies the expansion of human factors research from interface design toward the evaluation of adaptive and context-driven systems. By grounding both generation and assessment in operational context, we offer a framework for building reliable driver-assistance systems that prioritize meaningful support.
Event Type
Lecture
TimeTuesday, October 14th1:30pm - 1:50pm CDT
LocationGrand A
Cognitive Engineering & Decision Making