Presentation
User-Centered Explainable Artificial Intelligence in Partially Automated Vehicles
DescriptionArtificial Intelligence (AI) has been integrated into emerging technologies, such as automated vehicles (AVs), to help the vehicle sense the driving environment and make maneuvers. However, AI is imperfect and susceptible to errors. In partially AVs, the human driver is still needed to oversee the vehicle’s actions and take over when AI makes an incorrect identification. Explainable AI is needed for drivers to better calibrate their understanding of AI’s capabilities. The goal of the current study was to determine what AI explanation best supports drivers’ mental models and understanding of the partially AV when encountering various driving actions and road signs. Thirty-six participants were included in this 6 (driving scenario) x 3 (timing of the explanation) within-subjects design. Drivers’ understanding and subjective evaluations of the explanations were collected for 18 trials presented visually and audibly through E-Prime. Results suggest that drivers prefer explanations that occur before or during the action rather than after the action happens. Future studies can include driving scenarios where the partially AV does not perform as expected to determine if drivers’ preferences regarding the explanations change. This study has significance in the fields of emerging technology in transportation, human understanding of AI, and cybersecurity.
Event Type
Lecture
TimeTuesday, October 14th1:30pm - 1:50pm CDT
LocationGrand Hall J
Surface Transportation
Similar Presentations