Towards Accountable, Legitimate and Trustworthy AI in Healthcare: Enhancing AI Ethics with Effective Data Stewardship

Abstract

Data Stewardship is a novel governance mechanism in the context of artificial intelligence (AI) development in healthcare. This paper examines whether the conceptual tool of stewardship can remedy inadequacies of 'AI ethics' which has fundamental problems of accountability, legitimacy and trustworthiness. A modern secular conceptual explanation of stewardship involves taking a balanced account of the interests of society, and the core element of <i>answerability.</i> This conception of stewardship lends itself to legal mechanisms involving fiduciary duties, which introduces accountability mechanisms into AI development. The separation of AI development from the permanent enclosure of health data presents a useful lever to counter unethical behaviour and ensure societal engagement. Stewardship offers some promise to remedy the inadequacies of AI ethics, but there are risks that a narrow technical conception of data stewardship, without fiduciary duties and decoupled from beneficiaries, will be insufficient to drive the required fundamental change.

Ähnliche Arbeiten