Meaningful Human Control over Automated Driving Systems

Abstract

High expectations rest Automated Driving Systems (ADS) to drastically transform roadway transport in the coming decades. In our project we aim at guiding a responsible transition of control toward automated driving. As a first step, we will develop a theory of “meaningful human control” over ADS, and translate the theory into design guidelines, both at the technical and at the institutional level. To maximize human safety, and avoid the creation of accountability gaps, meaningful human control should constantly be maintained over autonomous and semi-autonomous systems. The notion of meaningful human control, originated within the debate about military drones, has –as of yet– not been applied in the context of ADS. The divergent appearance of vehicles with increasing levels of automated control calls for consideration hereof. Therefore, an interdisciplinary team of philosophers, traffic engineers and behavioral scientists will work at defining the conditions for human control, and its “meaningful” elements, and mold it into a workable framework of human control. The validity of this framework will then be tested through empirical research and with the use of automated vehicle prototypes. In conducting the research, particular attention will be dedicated to ensure that the developed theoretical framework can be fully operationalized into empirical science and design recommendations. This will be facilitated by both the interdisciplinary nature of the project itself and by the active participation of a number of private and public partners, who will in turn be provided with up to date research results. In particular, designers, manufacturers and road operators will receive conclusions drawn from empirical and theoretical research they can apply in developing automated systems that achieve meaningful human control; policy-makers can use our findings to elaborate regulations that promote both innovation and human values; lawyers and insurance companies will receive original inputs for the design of liability and insurance schemes; driving licensing bodies will receive data to use for developing new procedures. In the first stage of our project, we are aiming to develop an empirically and technically usable conceptual toolbox (or framework) that can be shared across the different areas of expertise that characterize the project. We will isolate a minimal set of notions of control, based on literature from, among others, philosophy, behavioral science, and engineering. Simultaneously, we aim to identify an as clear as possible notion of “meaningfulness”. On that regard, we started from a philosophical notion of meaningful human control over ADS, to investigate whether, and to what extent, its elements could be operationalized and tested within an empirical framework. We have presented our project to our private and public partners, and to a number of other stakeholders involved in automated driving. From this presentation, we have received positive and constructive feedback, and encouraged us to continue our communication to the public. In particular, we aim to structure a bi-directional communication avenue between data production and data use. Ideally, this would benefit both sides, providing the researchers with societally relevant research questions and practical constraints, and the stakeholders with timely access to scientific results.