Liability implications of artificial inteligence use in health

: fault and risk in public sector healthcare

Abstract

Raising healthcare systems’ sustainability challenges propel health care providers and payers to urgently seek ways to deliver care that is faster and more cost-efficient, while maintaining quality. Smart healthcare technologies, such as Artificial Intelligence (AI) Systems are expected to help, but technology can also cause harm. AI is no exception, yet in Health, that can mean disease, suffering or even death. Responsibility is key for justice, acceptability and personal support, especially in public healthcare. This study explores liability implications on the use of AI in health, in the case of public institutions. Focusing on the question of whether fault-based, risk or sacrifice liability regimes are applicable and, in their capacity, to assert just and fast compensation when damage is unavoidable, but also in prevention, mitigation and compensation of damage to citizens’ health or loss of life by public administration. The analysis of the case of AI systems usage for Triage and Referral of acute patients in the case of Hospital Emergency Rooms is used to exemplify and explore liability implications: contractual, extra-contractual, and fault, risk and sacrifice regimes. EU-applicable legal regimes, but specially the Portuguese law, was taken into consideration. Administrative law, both national and European, must be interpreted and adapted to better fit AI into the practices of public administration, not just in public sector health, but more broadly. This will provide benefits from new concepts and rethinking and expanding the role of administrative law in securing and safeguarding citizens’ rights facing evolving AI technologies. De iuri condendo proposals are advanced for slight changes in the law, to better frame: the concepts of adequate diligence and care, with regards to “unknown” or “exploratory” technologies; the consequences of a “faut-duservice” sentence with regards to continuous or repetitive action, and as a preventive mechanism; the concept of multiple levels of dangerousness in the same activity, hence impacting differently on different citizens; and, the idea that the degree of dangerousness needs to be analysed also in face of previous experience and maturity of the state and public administration, particularly in innovative technologies such as AI. This thesis is a contribution to thinking about AI use in health and in public administration, the risks and potential harms associated, and the regimes that can hold public administration liable in the case of AI mal-use or mal-functioning.

Date of Award	25 Jan 2021
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Rui Medeiros (Supervisor)