Brian Ness

Genomic research results and incidental findings with health implications for a research participant are of potential interest not only to the participant, but also to the participant's family. Yet investigators lack guidance on return of results to relatives, including after the participant's death. In this paper, a national working group offers consensus analysis and recommendations, including an ethical framework to guide investigators in managing this challenging issue, before and after the participant's death

Research biobanks are heterogeneous and exist to manage diverse biosample types with the goal of facilitating and serving biomedical discovery. The perspectives of biobank managers are reviewed, and the perspectives of two biobank directors, one with experience in institutional biobanks and the other with national cooperative group banks, are presented. Most research biobanks are not designed, nor do they have the resources, to return research results and incidental findings to participants or their families

This thesis is concerned with the fusion of two deeply human innovations – one very ancient (moral thought), the other very modern (artificial intelligence – or AI). AI systems are rapidly penetrating and influencing our personal, social, political and intellectual lives. As with all major technological developments through history, this one has the potential to both damage and enhance human wellbeing – not least in the moral realm. My aim here is to chart one pathway towards enhancement. I show that an AI bot, grounded in virtue ethics, can provide constructive guidance to human agents who find themselves facing challenging ethical problems in specified moral domains. I argue that this guidance can not only improve the ethical outcomes of particular situations but also help users to improve their skills in moral reasoning. I have named this bot Phronimo. There are seven stages in my argument: I begin in Chapter 1 with an account of, and preliminary justification for, Christine Swanton’s theory of virtue ethics (called Pluralistic Virtue Ethics – PVE) as the grounding for an AI-based ethical assistant. I then give an overview of the theory with an emphasis on its account of practical moral reasoning. PVE offers a comprehensive account of virtue-guided moral reasoning – in contrast to the kinds of principle-driven procedural guidance that alternative deontological and consequentialist theories are compelled to offer. In addition, as against monistic forms of virtue ethics, Swanton’s endorsement of pluralism accommodates a more comprehensive account of the demands, constraints and considerations which inevitably bear upon the agent who is grappling with an ethical problem. In Chapter 2 I begin with additional arguments in support of virtue ethics, and PVE specifically, as a grounding for Phronimo. I then describe, in principle, how Phronimo’s capabilities can be assembled around a basic reasoning structure which closely aligns with the systematic elements of Swanton’s theory. I explain why and how this rudimentary reasoning system must be enhanced by the addition of tailored AI software systems to provide Phronimo with appropriate analytical capabilities and user-friendly dialogical capabilities. I also introduce a central aspect of Phronimo’s mode of operation – a Socratic style of dialogue with which it engages its users. Chapter 3 provides a more comprehensive account of Phronimo’s virtue-orientated reasoning capabilities that were introduced in the previous chapter. To do this I imagine a Phronimo developed to assist teachers in the specific moral domain of the school classroom. I show how this Phronimo can be given comprehensive virtue-orientated reasoning expertise which is tailored to this domain. I also introduce the reader to an actual secondary school example which I utilise through the remainder of my thesis to demonstrate Phronimo’s capabilities. In Chapter 4 the reader is introduced to the AI world. Phronimo’s basic classroom-focussed reasoning capabilities must be refined and developed by the inclusion of a number of AI-based systems. I show how this can be done – in a way which I hope can be grasped by readers with no expertise in AI systems. The outcome of this process is my account of an integrated set of AI-based tools with the capability to deliver both the analytical and communications capabilities needed by Phronimo. I call it VirtBERT. With a fully developed, classroom oriented, Phronimo in hand I dedicate Chapter 5 to demonstrating and defending its capabilities. This involves envisioning the inexperienced teacher (introduced in Chapter 2) seeking guidance from Phronimo when faced with the theft of her phone by one of her students. This is a mundane, but typically complex, example. It provides me with the opportunity to demonstrate the importance of Phronimo’s Socratic reasoning method, as well as the effectiveness of its grounding in a synthesis of Swanton’s PVE and VirtBERT. This completes my account of Phronimo. The remaining chapters are dedicated to addressing a range of potential objections likely to be raised against a Phronimo. The most damaging objections to the notion of an agent relying on the moral testimony of others are referred to as ‘the problems of moral deference’. They include concerns such as the erosion of moral autonomy, of moral understanding, and of virtue – in the deferring agent. These objections become more worrying when the moral guide is a machine. Chapter 6 is therefore dedicated to defending Phronimo against these objections. In the literature there is no consistent taxonomic framework by which the problems of deference are ordered, so I begin by proposing a methodology which allows for the consideration of machine-based moral advisors. I then offer a detailed defence against each in turn. I conclude that Phronimo not only avoids most of the objections but has the potential to enhance the moral reasoning capabilities of its users. In the final chapter Phronimo meets its competitors. Although this is a new field, a number of machine-based moral advisors have been implemented – or have been described in the literature. I undertake a critical comparative analysis of the best of these. I conclude that Phronimo shows great promise as a potential leader in the field. I complete the chapter by reverting to the question of which theory is best equipped to provide the moral grounding for Phronimo. I imagine a Phronimo grounded in both a consequentialist and deontological theories – as well as an atheoretical AI-based alternative. I conclude that a PVE-grounded Phronimo is best positioned to overcome the main objections to the idea of human deference – particularly if that moral guidance is provided by a machine. This dissertation is first and foremost a doctoral thesis in philosophy. However, its content and structure are such that it can also be read as a philosophically minded feasibility study by anyone interested in creating a machine-based ethical assistant.