Contextual Challenges to Explainable Driving Automation: The Case of Machine Perception

Abstract

As happens in the case of many Artificial Intelligence systems, explainability has been acknowledged as a relevant ethical value also for driving automation. This chapter discusses the challenges raised by the application of explainability to driving automation. Indeed, designing explainable automated vehicles is not a straightforward task. On the one hand, technical constraints must be considered to integrate explainability without impairing the overall performance of the vehicle. On the other hand, explainability requirements vary depending on the human stakeholders and the technological functions involved, thus further complicating its embedment. The goal of the chapter is thus to investigate what explainability means with reference to driving automation. To this aim, we focus on machine perception and explore the related explainability requirements and challenges. We argue that explainability is a multifaceted concept that needs to be differently articulated in different contexts. Paying due attention to the contextual aspects of explainability, in particular the content of the explanations and the stakeholders they are addressed to, is critical to serve the ethical values it is supposed to support.