Liangxing Shi, Huani Shi, Feng Guo
9 min
As automated driving vehicles (ADVs) move from conceptual technology to potential market reality, their adoption remains lower than initial industry predictions. This study investigates the psychological drivers of customer willingness to travel in ADVs, moving beyond traditional technology acceptance models to incorporate the Cognitive Appraisal Theory (CAT) of emotions. The researchers examine how specific cognitive appraisals—the internal mental evaluations people make—trigger trust, and how that trust ultimately influences behavioral intentions.
The authors developed a structural model to map the decision-making process for potential ADV users. Based on the CAT framework, the study posited that four cognitive dimensions serve as the foundation for trust: performance expectancy (perceived benefits over manual driving), effort expectancy (ease of use/transparency), perceived enjoyment (hedonic value), and anthropomorphism (the perception of human-like traits in the vehicle). Using survey data from 349 participants in China, the team employed hierarchical polynomial regression to test how these appraisals influence trust, how trust leads to adoption intention, and how perceived cost moderates these relationships.
The study confirms that performance expectancy is the strongest cognitive driver of trust. All four appraisal factors were found to have a significant, positive relationship with trust, and trust subsequently acted as a critical mediator between these cognitive appraisals and the final willingness to adopt ADVs. Notably, while trust is a strong motivator for adoption, the study revealed that perceived cost acts as a significant barrier; when customers perceive the financial or incidental costs of ADVs as high, the positive impact of their trust on their willingness to travel in the vehicle is significantly reduced.
The findings suggest that for manufacturers and marketers, success depends not only on proving the safety or efficiency of the technology but also on humanizing the experience. Enhancing anthropomorphic design elements and focusing on the hedonic/fun aspects of the ride can help build the necessary trust. Furthermore, because cost significantly dilutes the conversion of trust into adoption, industry stakeholders must address price concerns or provide better communication regarding long-term cost-savings to bridge the gap between user trust and actual market adoption.
Sam: And what evaluations are supposed to create that confidence? Give us something concrete rather than just the model's labels.
Alex: Performance expectancy means expecting practical benefits from using the vehicle. A questionnaire item asks whether it would increase the respondent's productivity. Effort expectancy means expecting it to be easy to learn and use. One item asks whether interactions would be clear and understandable.
Sam: So ease of use includes understanding what the system is doing, not just operating it. What else enters the model?
Alex: Perceived enjoyment covers whether interacting with the vehicle feels fun or pleasant. Anthropomorphism means attributing human-like qualities to it. The authors discuss names and driving styles, but their questionnaire asks about consciousness, free will, emotions, and intentions.
Sam: Those are quite different ways of being human-like. A personalized driving style is not the same as believing a vehicle has consciousness.
Alex: That distinction matters when interpreting the result. The measured association concerns those attributed mental qualities. It does not directly test whether adding a name or changing driving style increases trust. Those design suggestions belong to the authors' interpretation, not a comparison they ran.
Sam: How did participants encounter the vehicles? I would trust an experience-based survey differently from one asking people to imagine an unfamiliar system.
Alex: The team recruited 411 volunteers through email and social media. Participants were invited to experience automated vehicles in various traffic scenarios. They also received explanations intended to increase their understanding, then completed the questionnaire and received a small monetary reward.
Sam: Was everyone recruited included in the analysis, and was this a controlled test of those explanations?
Alex: The final analysis included 349 participants after incomplete responses and missing information were excluded. Although the authors call the experience a field experiment, the described analysis is based on survey associations. The excerpt describes no randomized comparison of explanations, vehicle features, or prices.
Sam: That makes the experience useful context, but not an intervention estimate. How did they measure the proposed pathway?
Alex: They used agreement ratings on a seven-point scale, mostly adapted from earlier studies. Their models accounted for demographic characteristics and daily commuting mode. They checked whether the questionnaire items supported distinct constructs, then estimated relationships among evaluations, trust, willingness, and perceived cost.
Sam: What held up in those relationships?
Alex: All the proposed evaluations were positively associated with trust. Expected practical benefits had the strongest association in their main analysis. Trust was also positively associated with willingness. The reported coefficient was about zero point six before the appraisal variables entered the willingness model.
Sam: A coefficient is hard to translate into a travel decision. Does that tell us how many people would actually get into a vehicle?
Alex: No. It summarizes a relationship between questionnaire measures, not a change in the proportion taking a ride. When the evaluations were included, the trust coefficient was about zero point two. Trust still contributed, but the evaluations also retained their own associations with willingness.
Sam: So trust is not the whole pathway. What did the authors conclude from that adjustment?
Alex: They concluded there was partial mediation: trust accounted statistically for some, but not all, of those associations. A second modeling approach also supported indirect links through trust. Bootstrap confidence intervals for those indirect effects excluded zero.
Sam: That supports the model statistically. Does it establish that the evaluations happened first and trust then caused willingness?
Alex: It does not. The proposed order comes from the theory; these survey responses do not establish its causal direction. I would read this as evidence consistent with the pathway, not proof of it. And willingness is a stated intention, not observed adoption.
Sam: Bring price back in. Did they test willingness at different actual prices?
Alex: They measured perceived expense, including whether vehicles seemed costly or more expensive than expected. At higher perceived cost, trust had a weaker positive relationship with willingness. That pattern appeared in both modeling approaches. It is not an estimate of what a particular price reduction would achieve.
Sam: The practical message is that confidence and affordability are separate obstacles. How far should we generalize it?
Alex: The authors explicitly limit generality because all data came from China. They call for cross-national testing, research on shared automated vehicles, and attention to social networks. The sample also leans toward younger and highly educated respondents, so it is not a universal picture of potential passengers.
Sam: Who should read the full document, and where should they start?
Alex: Researchers in technology acceptance and human–automation interaction should start with the methodology and measurement appendix, then the regression analysis and limitations. Check what each construct actually measures. Also check reporting consistency: some coefficients are assigned differently across passages, and the trust-item count differs between the methods and appendix.
Sam: And for someone who only needs the main lesson?
Alex: In this sample, trust helped explain willingness, but perceived expense weakened the connection. Building trust and making adoption affordable are not the same task.
Sam: Keep those two questions separate.