Regulating for trustworthy autonomous systems: exploring stakeholder perspectives on answerability

Abstract

The proliferation of autonomous and semi-autonomous systems (AS) poses important and pressing regulatory challenges. Underpinning these is the recognition that many different stakeholders will need to trust systems to ensure their effective adoption and implementation. However, research on stakeholder perspectives is lacking, and assessing trustworthiness is difficult due to ‘responsibility gaps’ where it is unclear where responsibility for harms arising from AS ought to lie. Bridging these gaps is important because holding one another responsible is how social trust is maintained, and trust is vital to unlocking the promise that AS hold. This article considers how the concept of answerability could provide a useful framework for boosting the trustworthiness of AS. We present findings from a series of stakeholder interviews identifying what answers different stakeholders need to trust AS in health, finance, and government applications, and consider the implications of our research for current proposals for regulating AS in the United Kingdom and the European Union.