The integration of automated vehicles promises several benefits for urban mobility, including increased safety, reduced traffic congestion and increased accessibility. Automated vehicles also allow drivers to engage in non-driving tasks (NDRTs), such as relaxing, working or watching multimedia while on the road. However, widespread adoption is hampered by limited passenger confidence. To address this, explanations for automated vehicle decisions can foster trust, providing control and reducing negative experiences. These explanations must be informative, easy to understand and concise to be effective.

Existing explainable artificial intelligence (XAI) approaches are mainly aimed at developers, focusing on high-risk scenarios or comprehensive explanations, potentially inappropriate for passengers. To fill this gap, passenger-centric XAI models need to understand the type and timing of information needed in real-world driving scenarios.

Addressing this gap, a research team led by Professor SeungJun Kim from the Gwangju Institute of Science and Technology (GIST), South Korea, investigated the explanation requirements of autonomous vehicle passengers under real road conditions. They then introduced a multimodal dataset, called TimelyTale, which includes passenger-specific sensor data for timely and context-relevant explanations. “Our research shifts XAI’s focus in autonomous driving from developers to passengers. We developed an approach to collect actual passenger demand for in-vehicle explanations and methods to generate timely, situationally relevant explanations for passengers. Prof. Kim explains.

Their findings are available in two studies published in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies on September 27, 2023and 09 September 2024. The authors received the “Distinguished Paper Award” at UbiComp 2024 for their pioneering study entitled “What and When to Explain?: On-road Evaluation of Explanations in Highly Automated Vehicles”.

The researchers first studied the impact of different types of visual explanations, including perception, attention and a combination of both, and their timing on passenger experience under real driving conditions using augmented reality. They found that the perceived state of the vehicle alone improved confidence, perceived safety, and situational awareness without overwhelming passengers. They also found that the likelihood of traffic risk was most effective in deciding when to provide explanations, particularly when passengers felt information overload.

Based on these findings, the researchers developed the TimelyTale dataset. This approach includes exteroceptive (about the external environment, such as sights, sounds, etc.), proprioceptive (about body positions and movements), and interoceptive (about body sensations, such as pain, etc.) data collected from passengers using a variety of sensors in naturalistic driving scenarios as key features for predicting their explanation requests. In particular, this paper also incorporates the concept of disruption, which refers to the shift of passengers’ focus from NDRTs to driving-related information. The method effectively identified both the timing and frequency of passenger requests for explanations and the specific explanations passengers want during driving situations.

Using this approach, the researchers developed a machine learning model that predicts the best time to provide an explanation. Additionally, as a proof of concept, the researchers performed city-level modeling to generate textual explanations based on different driving locations.

“Our research lays the groundwork for increased acceptance and adoption of autonomous vehicles, potentially reshaping urban transportation and personal mobility for years to come.” Prof. Kim remarks.

***

Reference 1

TWO: 10.1145/3678544

Reference 2

TWO: 10.1145/3610886

About Gwangju Institute of Science and Technology (GIST)
Gwangju Institute of Science and Technology (GIST) was established in 1993 by the Korean government as a research-oriented graduate school to help ensure Korea’s continued economic growth and prosperity through the development of advanced science and technology, with an emphasis on collaboration with the international community . Since then, GIST has pioneered in 2010 a highly regarded undergraduate science curriculum that has become a model for other science universities in Korea. To learn more about GIST and its exciting opportunities for both researchers and students, please visit: http://www.gist.ac.kr/.

About the author
Dr. SeungJun Kim is a professor and director of the Human-Centered Intelligent Systems Laboratory at Gwangju Institute of Science and Technology (GIST). Prior to joining GIST, he was research faculty at Carnegie Mellon University’s Human-Computer Interaction Institute, leading interdisciplinary research projects exploring XR, robotics, and human-AI interaction. Prof. Kim holds a BS in Electrical Engineering from KAIST and an MS and Ph.D. in Mechatronics from GIST, respectively. His research focuses on sensory intelligence and augmentation through multimodal XR in ubiquitous computing environments. He has been honored with several paper awards from UbiComp, ACM CHI, IEEE ISMAR, and ACM AutoUI.