Asking AI About Your Car: AI-Augmented Vehicle Owner's Manuals Support Integrated System Understanding Beyond...

Area of Science:

• Human-Computer Interaction
• Automotive Engineering
• Artificial Intelligence

Background:

• Increasing vehicle automation complexity presents challenges for drivers understanding Advanced Driver Assistance Systems (ADAS).
• Traditional owner's manuals are effective but can be enhanced for better driver engagement via AI.
• Interactive learning experiences offer potential for improved accessibility and comprehension of ADAS.

Purpose of the Study:

• To compare the effectiveness of an AI-powered interactive manual versus a traditional static manual for understanding ADAS.
• To assess how AI enhances users' mental models and reasoning about ADAS features.
• To explore the psychological mechanisms underlying user acceptance of AI-enhanced manuals.

Main Methods:

• A between-subjects study with 38 participants comparing PDF and Retrieval-Augmented Generation (RAG) AI manuals.
• Quantitative assessment of mental model accuracy using multiple-choice questions.
• Qualitative analysis of reasoning patterns via structural topic modeling (STM) of open-ended responses.

Main Results:

• Both PDF and RAG manuals improved mental model accuracy equally.
• RAG-trained participants showed superior systems-level and feature integration reasoning.
• Technology Acceptance Model analysis indicated similar psychological mechanisms for both methods, with perceived usefulness driving acceptance.

Conclusions:

• AI-augmented manuals are as effective as traditional ones for learning ADAS, while improving feature integration reasoning.
• Interactive AI enhances, rather than replaces, traditional methods for building sophisticated mental models of automated systems.
• Findings inform automotive manufacturers on optimizing driver education for complex vehicle automation, enhancing safety and user experience.