Comfort with varying levels of human supervision in self-driving cars: Determining factors in Europe

Abstract

While numerous studies have investigated attitudes towards self-driving cars in general, less research attention has been focused on individuals' comfort with the presence (or absence) of third-party human supervision of this automation, and its potential correlates. In the present study we perform a secondary analysis of pre-existing data from The European Commission’s Eurobarometer 92.1, a large-scale European survey ( n = 27565) of expectations and concerns of connected and automated driving. By comparing responses to three levels of human supervision in self-driving cars, we aim to identify changes in the importance of predictors of comfort with automation. We find considerable heterogeneity in both individual attitudes, as well as in country-level attitudes in our descriptive analysis. We find a trend of decreasing comfort as external human supervision is reduced, although this effect differs between countries. We then investigate potential drivers of self-reported comfort with varying levels of external human supervision in a regression framework. Gender differences get stronger with decreasing supervision, suggesting a possible resolution to conflicting evidence in previous studies. Following this, we fit an ordinal random forest model to derive variable importance metrics, which enable us to compare the changing role predictor variables might play in shaping self-reported comfort, depending on varying levels of third-party supervision. Data privacy is highlighted as an important variable, regardless of level of supervision. Our findings provide confirmation for previous literature in a large sample, while also uncovering a number of novel associations, providing guidance for future policy-making and research efforts.