Dominic Martin: Artificial Moral Agency: Approaches, Benchmarks and Practices
Dominic Martin is a professor of ethics at the School of Management of the Université du Québec à Montréal (UQAM). His work combines approaches from ethics, contemporary political philosophy, economics, and law, to grapple with questions of distributive justice, the role of the state – and intergovernmental regulatory bodies – in shaping market structures and the ethical obligations of business. His recent research projects deal with the new ethical issues associated with the rise of artificial intelligence (AI) and the increase usage of algorithms in society, such as questions of algorithmic accountability, the creation of artificial moral agents, and the socio-economic impact of AI.
The goal of imparting moral agency to an automated system is directly or indirectly relevant in fields ranging from computer science to psychology, health sciences, governance, management and international politics. Artificial moral agency or agents (AMA) has been a topic of interest in computer science and engineering ethics for at least two decades, but recent developments in AI led to an explosion of new contributions on the topic. While this is promising, research on the topic has also become fragmented, and even speculative in some cases, with work discussing specific aspects of AMA without the emergence of dominant views, or clear debates or research questions.
My main objective in this talk will be to take stock of the work on AMA accomplished in the last decades and examine three main questions from a philosophical perspective. First, what is the most promising approach for implementing moral agency in an AI system? Second, how to benchmark and test moral agency, especially given that systematic benchmarking does not exist for human moral agency. Third, what are industry best practices regarding the development and use of AMA?
Dominic Martin is a professor of ethics at the School of Management of the Université du Québec à Montréal (UQAM). His work combines approaches from ethics, contemporary political philosophy, economics, and law, to grapple with questions of distributive justice, the role of the state – and intergovernmental regulatory bodies – in shaping market structures and the ethical obligations of business. His recent research projects deal with the new ethical issues associated with the rise of artificial intelligence (AI) and the increase usage of algorithms in society, such as questions of algorithmic accountability, the creation of artificial moral agents, and the socio-economic impact of AI.
The goal of imparting moral agency to an automated system is directly or indirectly relevant in fields ranging from computer science to psychology, health sciences, governance, management and international politics. Artificial moral agency or agents (AMA) has been a topic of interest in computer science and engineering ethics for at least two decades, but recent developments in AI led to an explosion of new contributions on the topic. While this is promising, research on the topic has also become fragmented, and even speculative in some cases, with work discussing specific aspects of AMA without the emergence of dominant views, or clear debates or research questions.
My main objective in this talk will be to take stock of the work on AMA accomplished in the last decades and examine three main questions from a philosophical perspective. First, what is the most promising approach for implementing moral agency in an AI system? Second, how to benchmark and test moral agency, especially given that systematic benchmarking does not exist for human moral agency. Third, what are industry best practices regarding the development and use of AMA?
