Date Published: May 2, 2019
Publisher: Public Library of Science
Author(s): Jan C. Zoellick, Adelheid Kuhlmey, Liane Schenk, Daniel Schindel, Stefan Blüher, Yan Ge.
After years of hypothetical surveys and simulator studies, automated vehicles (AVs) are now being tested in realistic traffic environments adding validity to knowledge about their acceptance. We present data from a pilot test with participants (n = 125) after experiencing a ride in an electric AV on a large clinic area in Berlin, Germany. As a first contribution, we bridge the gap between missing definitions of key constructs, confusion about their operationalisations, and a rigorous test of their statistical properties and data structure by examining scales on acceptance, trust, perceived safety, intention to use, and—for the first time applied to AVs—the emotions amusement, fear, surprise, and boredom. Tests of reliability and normality were satisfying for almost all constructs (Cronbach’s alphas ≥ .69; six of eight scales normally distributed). The vehicles were accepted (M = 1.22; SD = 0.70; range -2 to 2), trusted (M = 3.29; SD = 0.81; range 1 to 5), and perceived as safe (M = 3.29; SD = 1.03; range 1 to 5). However, factor analyses did not reflect the hypothesised data structure, and validity concerns question the suitability of some constructs for attitude assessment of electric AVs. Our open item for comments added valuable insights in qualitative aspects of user attitudes towards electric AVs regarding driving style, technical features, and (unsettling) audio-visual feedback. We thus argue for broader conceptualisations of key constructs based on interdisciplinary exchange and multi-methodical study designs.
The development of automated vehicles (AVs) presents a caesura in mobility [1–3] revolutionising travel particularly for people in old age and with disabilities [4, 5]. In this paper, we understand AVs to be shared, electrically powered, and to feature automation above SAE level 4 being able to perform at least “all driving functions under certain conditions” . These vehicles are pod-like, equipped with window fronts on all sides and opposing seats, and exhibit no obvious front and rear setting them apart from both passenger cars and public transport vehicles . Fig 1 presents a picture of the AV used in this study. With these alterations, it is unclear how people will react in encounters as co-habitants or as potential users. Research has identified several attitudes relevant for the user assessment of AVs [7, 8]. However, a variety of definitions and operationalisations [9, 10] combined with hypothetical study designs [11, 12] and rather descriptive analyses [13, 14] lead to uncertainty and confusion about people’s attitudes towards AVs.
The Charité data protection bureau (written vote 598/17/ST3) and the Charité ethics committee (written vote EA2/188/17) have approved of the study. The data security bureau and ethics committee waived the need for written consent, because obtaining written consent would have rescinded anonymity. Therefore, we only obtained verbal consent with parents/guardians verbally consenting for minors. Consent was informed based on an information sheet clarifying the topics of voluntariness, anonymity, and data processing for scientific purposes.
With our analyses, we tested various statistical properties (e.g., normality and reliability) of multiple scales regularly used to assess attitudes towards automated vehicles. These include acceptance, perceived safety, trust, and intention to use as well as four emotions with differing levels of valence and activation. This pilot test was necessary, because AV attitude research operates with a variety of definitions and measures predominantly in hypothetical study designs focusing on rather descriptive analyses. These factors lead to uncertainty and confusion about people’s attitudes towards AVs with inconsistent results. With all participants having experienced a ride in an AV directly prior answering our survey, our design differed from that of previous studies [12, 25, 32]. Given the direct experience with AVs in our study, we expect our data to have higher validity than designs with hypothetical scenarios and simulated rides.