Enhancing Suspicion Drivers’ Trust in Autonomous Vehicles: The Impact of Communication Modality on AV Explanation Effectiveness

This post summarizes the paper “The Impact of Modality, Technology Suspicion, and NDRT Engagement on the Effectiveness of AV Explanations” which was co-authored by Qiaoning Zhang, Connor Esterwood, Anuj K. Pradhan, Dawn Tilbury, X. Jessie Yang, and Lionel P. Robert Jr. This paper has been published in IEEE Access. You may access the paper here.

As we inch closer to a future filled with automated vehicles (AVs), it becomes crucial to ensure not only their functionality but also the trust and acceptance of these cutting-edge technologies. One of the key elements in building this trust is the ability of AVs to explain their actions and decisions effectively. These explanations serve as the key to making AV actions transparent and predictable, ultimately reducing fears and concerns associated with this groundbreaking technology. In a world where AVs at levels 4 and 5 of driving automation take on full driving responsibilities, it’s paramount that humans can understand and predict their behavior.

This study delves into the realm of AV explanations and their modality, seeking to answer vital questions about how the way we receive information from AVs can impact our understanding, engagement, and safety. Additionally, it explores the role of an individual’s level of technology suspicion in this equation.

The Power of Explanations and Communication Modality

Explanations are the linchpin in fostering transparency and trust in AVs. They bridge the gap between the machine’s actions and the human understanding of those actions. By making AV behavior predictable and understandable, explanations have the potential to alleviate fears, concerns, and driving-related anxiety, ultimately nurturing trust.

However, the mode of delivering these explanations, or modality, is a pivotal factor that has been largely unexplored. Typically, AV explanations are conveyed through two unimodal modalities: auditory and visual. Auditory cues can promote safety by alerting drivers to potential dangers, but they can also be annoying and startling. In contrast, visual cues allow for faster information recognition but can divert attention away from the driving environment.

Understanding the Impact of Communication Modality

To comprehend the impact of modality on AV explanations, we conducted an experiment involving 32 participants using a high-fidelity driving simulator. We exposed participants to four AV explanation conditions: (1) auditory explanation with NDRT, (2) auditory explanation without NDRT, (3) visual explanation with NDRT, and (4) visual explanation without NDRT.

Key Findings

Our study’s insights into the world of AV explanations:

Modality Matters: The choice of modality significantly affects the effectiveness of AV explanations. It’s not a one-size-fits-all scenario, and the mode of communication can make or break the user’s perception of AV actions.

Technology Suspicion: An individual’s level of technology suspicion plays a vital role in how they perceive and accept AV explanations. Suspicion can stem from concerns about data privacy, algorithmic biases, or even the underlying intentions of the technology.

Non-Driving-Related Tasks: Engaging in non-driving-related tasks (NDRTs) while using AVs can impact how users respond to explanations. It can either alleviate the burden of visual attention or, conversely, exacerbate distraction.

Implications and Recommendations

Our findings offer several important takeaways for the development and deployment of AVs:

Tailored Explanations: Developers should consider tailoring explanations based on the modality that best suits the situation. The context of use, user preferences, and task demands should inform this choice.

Addressing Suspicion: To build trust in AV technology, it’s crucial to address and alleviate technology suspicion. Transparency in data usage, algorithmic decision-making, and clear intentions can help ease users’ concerns.

Managing NDRTs: When designing AV systems, the potential impact of non-driving-related tasks on user attention should be carefully managed. Striking a balance between providing information and minimizing distraction is key.

As the automotive world hurtles toward an autonomous future, the effectiveness of AV explanations cannot be underestimated. It’s a powerful tool for building trust and acceptance among users. However, the choice of modality and an individual’s level of technology suspicion can significantly influence the impact of these explanations.

To learn more about this work and engage with our MAVRIC, take a look at the paper here, and find us on Twitter (X), on the web, and in person at the University of Michigan Robotics. We welcome collaborators and interest from diverse disciplines. Hope to talk to you soon.