O-232 New principles for the use of artificial intelligence in human reproduction: leave everything as it is?

Abstract

Abstract Study question Which fundamental principles in bioethics ought to be reinforced in the face of the novel introduction of artificial intelligence (AI) into in vitro fertilization (IVF)? Summary answer Contrary to recent arguments, there is no need for new ethical principles to address moral worries regarding AI-based IVF. What is known already Several AI-based technologies are currently introduced into various stages of the IVF process. These include, for instance, the prediction of success rates, automation of sperm selection and cryopreservation, and optimization of embryo evaluation. The anticipated benefits of such automation are couched with high hopes of improving efficiency, accessibility and consistency in ART. However, it is an open question which ethical principles are fit to address introduction of AI in the IVF lab. In the field of assisted reproduction, like in other medical contexts, arguments abound that the current set of ethical principles cannot adequately deal with anticipated moral issues. Study design, size, duration A literature study was performed to map principles formulated by international organizations and professional organizations. These are mostly formulated echoing the four leading principles of medical ethics–respect for autonomy, beneficence, non-maleficence, and justice—yet, with the addition of a new principle: explicability. The latter can be understood in terms of intelligibility (viz. ‘how does it work?’) and in terms of accountability (viz. who is responsible for the way it works?). Participants/materials, setting, methods Literature study, conceptual analysis, and ethical analysis. Main results and the role of chance One can expect that AI-driven reproductive medicine will bring new ethical challenges and potential harms that compound the anticipated benefits. The black-box character of many possible AI applications in IVF is a special challenge for informed consent as algorithms are becoming exponentially more complex in contrast to the relatively linear technological improvements of machines traditionally used in the profession. This opacity problem could undermine trust in the output of the systems. With such challenges at stake, how can we understand the justification for explicability as a separate principle? Based on the methodological criterion of simplicity, there are good reasons to avoid expanding the number of principles. Explicability can be subsumed under the four existing principles, which is morally relevant if bioethics is to retain a hands-on element. The contrary would have an unnecessary watering down effect on the separate, well-established principles, paradoxically making it more complex to follow their interactions. Furthermore, setting a high threshold of explicability could be perplexing for IVF patients (morally relevant in terms of autonomy) and doctors (morally relevant in terms of accountability and contextual factors of for example workload), possibly reopening discussion about paternalism in reproductive medicine. Limitations, reasons for caution This normative analysis does not presume to offer a complete bioethical theory. Rather, we take a pragmatic and principle-based approach which is crucial given the embryonic stage of AI technologies and to better understand the novel digital setting before us. Wider implications of the findings These findings instigate reflection about (bio)ethical analysis of AI by accentuating the importance of specification and contextualization of principles on a topic that will gradually mature. This analysis also connects with questions on how technological developments in the reproductive realm can contribute to moral progress. Trial registration number Not Applicable