Theo, take a right… uh… left: Conversational Route Negotiations with Autonomous Driving Assistants
Published in Adjunct Proceedings of the 10th International ACM Conference on Automotive User Interfaces and Interactive Vehicular Applications (AutomotiveUI ’18), ACM 2018
Abstract
We foresee conversational driver assistants playing a crucial role in automated driving interactions. In this video we present a study of user interactions with an in-vehicle agent, “Theo”, under SAE Level 4 automated driving. We use a remote Wizard-of-Oz setup where participants, sitting in a driving simulator, experience real-life video footage transmitted from a vehicle in the neighborhood and interact with Theo to instruct the vehicle where to go. We configured Theo to present 3 levels of conversational abilities (terse, verbose and helpful). We show the results of 9 participants tasked to negotiate destinations and route changes. Voice interaction was reported as preferred means of communication with Theo. There was a clear preference for talkative assistants which were perceived more responsive and intelligent. We highlight challenging interactions for users such as vehicle maneuvers in parking areas and specifying drop off points and interesting associations between the agent performance and the automated vehicle abilities.
Key Contributions
- Automated driving system development
- Pioneers a remote AI Wizard-of-Oz setup
- Tests three conversational agent styles
- Finds user preference for talkative agents
- Identifies challenging user interaction scenarios
- Links agent performance to perceived safety