Fostering Ethical Sensitivity in AI Practitioners Through Ethics-based Assessment

Adrian Gavornik, Juraj Podroužek

The field of AI ethics is said to be dominated by principilism approaches. Still, bridging the gap between theoretical and often ambiguous principles and their practical operationalization remains a challenge. Simultaneously, there are growing calls from within the field to include a more virtue ethics approach to AI ethics, focusing on cultivating practical wisdom in AI practitioners, without abandoning principles completely. Several approaches have been proposed, but empirical evaluations regarding their implementation and effects in practice are still lacking. In this short position paper, we would like to reflect back on our empirical experience with conducting ethics-based assessment of various AI systems with their developers. Done in a workshop format, they include activities such as stakeholder identification, data sources mapping, risk identification and countermeasures prescription, with the aim to support the process of moral sensitisation of participants. We highlight the potential of such assessments to incorporate aspects of virtue ethics, fostering practical wisdom and ethical sensitivity of the participants. We first summarize the main debates in AI ethics relating to the dominance of principilism and its criticism. We then turn to some of the proposed ways out, such as attempts to operationalize AI ethics principles through various (mostly technical) tools and more importantly to the growing body of literature calling for a virtue ethics approach to AI ethics. We aim to discuss how an ethics-based assessment and its facilitated process of moral self-reflection and deliberation  can be compatible with virtue ethics, while also making use of existing AI ethics principles and guidelines. We build on existing understanding of ethical sensitivity as a construct that operationalizes various stages of moral process in the context of technology design teams. It consists of the recognition of ethical issues, its particularization, and judgment. However, we contend that in order to make such assessment meaningful, we should strive to find consistent methods to trace and measure the impact of ethics-based assessments on AI practitioners’ ethical sensitivity. Our aim is to initiate a dialogue on ethical sensitivity in AI and the role of trustworthy AI tools in fostering this quality in practitioners.

Cite: Adrian Gavornik, Juraj Podroužek. Fostering Ethical Sensitivity in AI Practitioners Through Ethics-based Assessment. ESDiT & 4TU Ethics Conference Rethinking Ethics – Reimagining Technology. University of Twente, 2024.