Acceptance of Self-Driving Vehicles in the United States: Expectations, Willingness to Pay, Supervision Capability, and Funding Oversight

Abstract

As technology has advanced to a level that supports a potential transition of the private and corporate vehicle fleets from human-operated vehicles to vehicles with self-driving capability, research and development efforts have typically focused on improving the technology and system integration of self-driving vehicles (SDVs). This emphasis has often come at the expense of attention to other systems within the broader self-driving vehicle ecosystem. One of these key systems is the potential SDV consumer, who would serve as the catalyst of any potential vehicle fleet transition. Research gaps currently exist regarding reliability and safety expectations for self-driving vehicles, how potential juries may assign blame for negative events involving SDVs, and how safety oversight and SDV-required infrastructure improvements can be funded with the least-opposed new or increased taxes and fees. Other research regarding human interaction with SDVs, such as willingness to pay, has also become outdated as these values change over time. Addressing these research gaps and refreshing relevant data is critical to supporting a transition to SDVs. To mitigate these gaps and provide updated SDV-relevant data, several experiments were conducted with participants residing in the United States. In one experiment, data was collected regarding how participants would assign blame for traffic collisions and other negative incidents involving SDVs, and how manufacturer advertising claims would impact their blame assignments. This research found that a majority of the sample population was primarily inclined to blame the driver of the SDV, though claims by the manufacturer about human oversight not being needed led a majority of participants to assign additional blame to the manufacturer. In another experiment, participants were asked to select between competing feature packages that addressed safety and reliability architecture elements. Results from this experiment found that rate of takeover requests and methods of alerting the driver to system failures were primary decision drivers, while the vehicle’s failsafe behavior was not a primary factor in decision making. On economic issues, experimental data indicated that the percentage of the population willing to pay a premium for self-driving technology was increasing, as was the mean premium participants were willing to pay. However, the median premium value was lower than what was found in previous studies. The data also showed that increases in vehicle purchase prices or subscription fees for self-driving capability were the preferred methods of funding safety oversight and SDV-supporting infrastructure, compared to increases in fuel, mileage, or vehicle taxes. In a final experiment, human oversight of autonomous systems was found to be ineffective, with participants detecting an average of only 15% of induced system errors, regardless of how system capability was framed or how errors were distributed. The experimental results contained within this dissertation provide support for systems engineering efforts in support of the development and integration of SDVs within the ground transportation system-of-systems. Specifically, this research better defines potential consumer expectations for safety and reliability architecture elements, which can be translated into system design requirements for future SDVs. In parallel, this work finds that human oversight of SDVs is ineffective, suggesting that assumptions about humans being able to observe and correct SDV errors may be unrealistic. Updated willingness to pay values provide insight into the cost constraints for future SDV products relative to otherwise equivalent human-driven vehicles, while funding preference information provides policy guidance on which fees or taxes authorities could implement to fund oversight and infrastructure improvements. Finally, this dissertation provides a mapping of predictors that can be used to estimate willingness to pay, the impact of manufacturer advertising, primary blame assignments for SDV incidents, and preferences for safety and reliability architecture features, while also identifying predictors that have second-order effects on other predictors.