FEAR 2026 

Decisions always involve uncertainty. This is especially true for ethically-laden decisions. An agent who aspires to make ethical choices not only faces uncertainty about the possible consequences of her actions; she also faces "moral uncertainty" --that is, uncertainty about the correct moral view. We will discuss how she can cope with both kinds of uncertainty. This involves ideas from both decision theory and moral philosophy.

Ethics is most frequently encountered as a series of processes or checks which validate a particular field of work. But Ethics is also an academic discipline and, no less than others, is vigorously contested between divergent schools of thought. So, when we talk about ethics in the context of other realms of study or activity, including the development and application of AI, we have to ask, What sort of ethics? Today, the state of ethical thinking in the academy is out of step with what passes for ethics in Western society in general – and this poses important challenges to the ways in which AI is being rolled out. The culture of the global marketplace has fostered an assumption that consequentialist ethics is all there is, and a crude utilitarianism follows. Academic ethics has, for the last 40 years or so, turned increasingly to the nature of the human person, the idea of virtue and the practices of moral formation. Bringing the ethics of the human, and of virtue, into the ways we build AI has become an urgent necessity as the interface between technology and the human person becomes more loaded with both promise and threat. In the process, our language of AI in relation to the human person may have to change and perhaps we will come to understand better exactly what makes us human in the first place.

In this talk, I will assess the relationship between policy, law, norms and practice, and their implications for design and regulation of machine automation/autonomy. We will consider how policy, law and norms differ from one another, and the underpinning forms of reasoning that contribute to those differences. Using some examples, we will then examine how policy, laws and norms intersect are also shaped by the domain, setting and social relationships in which machine agents are designed and used over time. Considering both the underpinning reasoning and social dynamics together, we then discuss the implications for (a) how we interpret the actions of automated agents and their limits, and (b) our choices about how to enact those limits in policy, law and practice.

As autonomous agents are increasingly used in high-stakes domains like healthcare, transportation and law enforcement, aligning their behaviour with the social, legal, ethical, empathetic and cultural (SLEEC) norms of their stakeholders has become a critical engineering challenge. This short tutorial will present a tool-supported methodology for managing the SLEEC requirements lifecycle, covering elicitation, well-formedness validation, and conformance verification of agent design models.