AbstractAutomated vehicles (AVs) are one critical application area of intelligent automation. However, a lack of appropriate trust can be a major barrier to successfully introducing AVs into the market and permitting safe and effective interaction between pedestrians and AVs. Given that the topic of pedestrian-AV trust is relatively new, much remains unknown about the factors and strategies required to calibrate pedestrians’ trust in AVs. Before seeking design solutions to the problems of misaligned trust, it is essential to first understand how pedestrian-AV trust may develop and how to effectively assess and predict it in practice. Hence, this research project employs a mixed-methods approach to address these questions mainly from the perspectives of human factors engineering and human-computer interaction (HCI).
In study one (Chapter 4), the systematic reviews of the literature were conducted to develop a theoretical model that comprises three layers of variability in pedestrian-AV trust, including dispositional trust, situational trust, and learned trust. The results revealed that there are three broad categories of factors influencing pedestrian-AV trust, namely the human (pedestrians), the automation (automated driving systems), as well as the environment.
In study two (Chapter 5), the immersive virtual reality (VR) and scenario-based interviews were employed to examine pedestrians’ trust in AVs in a qualitative manner, based on the attributes of trust and trustworthiness. A hybrid approach of inductive and deductive thematic analysis of the responses of 36 participants was undertaken. The results revealed eight attributes constituting pedestrian-AV trust, including statistical reliability and dependability, competence, predictability, familiarity, care/harm, authority/subversion, liberty/oppression, and sanctity/degradation. The findings of this study provide an empirical grounding for trust theories. Specifically, the importance of subjective qualities (related to automation morality) is highlighted.
Study three (Chapter 6) described the development of a questionnaire-based method for assessing and predicting the initial trust levels that pedestrians have toward AVs. The questionnaire was validated by the use of the partial least squares structural equation modelling (PLS-SEM) and machine learning algorithms, based on the data collected from 436 pedestrians. The results of the PLS-SEM and machine learning model indicated that seven constructs, including propensity to trust, statistical reliability, dependability and competence, predictability, familiarity, care/harm, authority/subversion, and sanctity/degradation, play a critical role in predicting pedestrians’ initial trust in AVs.
In conclusion, the thesis investigates the development and assessment of pedestrian-AV trust and identifies relevant research frameworks. The findings have direct implications for the design of AVs and the calibration of trust between pedestrians and AVs.
|Date of Award
|Xu Sun (Supervisor), Bingjian Liu (Supervisor) & Gary Burnett (Supervisor)
- Automated Vehicles
- human-machine interaction