Operationalizing the Ethics of Connected and Automated Vehicles: An Engineering Perspective

Introduction

Connected and Automated Vehicles [CAVs] are arguably one of the most researched and discussed applications of Artificial Intelligence [AI] technologies. Advances in design and development fuel the anticipation of a future where our roads will be populated by both regular and automated vehicles. Concurrently, social, ethical, and legal issues surrounding the impacts of CAV technologies have been raised (Nyholm, 2018a, 2018b; Taeihagh & Lim, 2018). This kindled a lively interdisciplinary debate and highlighted the necessity of shared normative frameworks to steer innovation towards ethically desirable and socially sustainable directions. In line with this trend, the European Union [EU] has recently presented its ethical framework to promote responsible innovation in CAV technology (Horizon, 2020) and asked stakeholders to contribute to its operationalization.

This paper responds to the call by presenting some methodological suggestions on how to ease the translation of the EU recommendations into practice from the viewpoint of engineering.¹ In what follows, we elaborate on a bottom-up, function-based working approach for the development of flowcharts, checklists, and similar methodological tools supporting the exercise of moral judgment aimed at aligning CAV design to the EU normative framework². By focusing on given functions, determining which ethical challenges they pose vis-à-vis the EU framework, and devising ad hoc methodological tools to discuss them, the gap between principles and design practices can be narrowed down and the need for further conceptual refinements of the normative framework can be better specified.

The paper is structured as follows. In Section I we present the EU ethical framework, while in Section II we further clarify our aims and sketch the main features of the function-based working approach. In the remaining sections we show how this approach can be applied to outline tools for bridging gaps between recommendations and practice. In particular, Section III focuses on problems revolving around the principle of personal autonomy; Section IV considers challenges posed by explainability; finally, Section V takes a closer look to privacy issues. In each case we map a methodological tool aimed at further operationalizing the EU guidelines. However, due to the preliminary stage of our research, the suggestions we advance are to be read more as evidence in support of the function-based approach than as refined tools for inquiry. Therefore, for each tool we underline what aspects are still in need of further clarification, thus setting the agenda for future research. Notwithstanding this limitation, we believe that our work might already demonstrate the productivity of the function-based approach and foster its adoption in the CAV scientific community.