Explanation in Automated Vehicles

When it comes to automated vehicles, there are serious concerns about whether individuals will choose to employ automated vehicles (AVs). One of the most central of these concerns is the lack of trust in AVs. In this project, we explored the possibility of using explanations of vehicle actions to help drivers build trust in automated vehicles. We conducted human-subject experiments in a driving simulator to investigate three research questions:

  • How do the timing and the degree of AV autonomy influence the effectiveness of AV explanations?

  • Does the driver’s age influence the relationship between AV explanations and the driver’s effort, anxiety, and trust?

  • How does the modality influence the effectiveness of AV explanations?

We found that explanations provided before an AV acted were associated with higher trust in and preference for the AV. In addition, explanation modality and driver's age are important moderators that impact the relationship between explanation effectiveness and AV perceptions. Our results have important implications for the adoption of AVs.

Individual differences and Automated Vehicles Expectation

Expectations about AVs have been identified as one of the most important factors in understanding AVs adoption. Therefore, by understanding the public's expectation of AVs, we can better understand if or when AVs are likely to be adopted on a wide scale. To examine the research question:

  • How individual differences can influence expectations on AVs?

An online survey was conducted with 443 U.S. drivers who were recruited and divided into subpopulations by age, gender, ethnicity, census region, educational level, marital status, income, driving frequency, and driving experience. Results reveal that drivers' AV expectations differ significantly by age, gender, ethnicity, education levels, marital status, drive frequency, drive experience, and personality. The results of this study provide a foundation for future research related to expectations and have important implications on future design and development of AVs.

Demograhic Information.docx

Trust in Automation

Automated decision aids have been used in various domains, such as military operations and medical diagnosis. To facilitate appropriate trust in and dependence on automation, this study focus on the following research question:

  • How different types of likelihood information impact human operators’ trust in automation and their team performance?

To examine this research question, we designed and conducted an experiment in the context of a simulated surveillance task. The results indicate that not all likelihood information is equal in aiding human-automation team performance. Automated decision aids should avoid directly present the hit and correct rejection rates. The findings can be applied to the design of automated decision aids.


Expectation and Trust in Automated Vehicles

A lack of trust is a major barrier to the adoptions of Automated Vehicles (AVs). Given the ties between expectation and trust, this study employs the expectation-confirmation theory to investigate in trust in AVs. An online survey was used to collect data including expectation, perceived performance, and trust in AVs from 443 participants which represent U.S. driver population. Participants were presented with: 1) sunny weather and normal driving behavior; 2) sunny weather and aggressive driving behavior; 3) snowy weather and normal driving behavior; 4) snowy weather and aggressive driving behavior. Using the polynomial regression and response surface methodology, we found that higher trust is engendered when perceived performance is higher than expectation, and perceived risk can moderate the relationship between expectation confirmation and trust in AVs. Results have important theoretical and practical implications.