Navigation auf uzh.ch
Navigation auf uzh.ch
Date and time: 25.04.2023; 15:15-17:45h
Location: RAA-E-30
Registration needed (see below)
This workshop for early-career researchers focuses on the possibilities and implications of using machine learning systems in clinical contexts of fertility treatment and assisted reproduction. The practice and ethics of present as well as possible future applications of artificial intelligence in assisted reproduction will be presented. We are then going to discuss what the role and significance of AI and machine learning is for our current research, and how we can best account for its potential and challenges within our projects.
Vasilija Rolfes is a philosopher and currently works at the Department of History, Philosophy and Ethics of Medicine, Heinrich-Heine-Universität Düsseldorf. Vasilija Rolfes does research in stem cells ethics, ethical implications of experimental technologies of reproductive medicine, chances and risks of artificial intelligence in clinical practice and stigmatization and discrimination of patients in the medical context. She was also an expert consultant on behalf of the German Bundestag on ethical aspects of germline intervention via CRISPR.
Atoosa Kasirzadeh is a tenure-track assistant professor (Chancellor’s Fellow) in the philosophy department and the Director of Research at the Centre for Technomoral Futures in the Futures Institute at the University of Edinburgh. Prior to this, she was a visiting research scientist at Google DeepMind in London and a postdoctoral research fellow at the Australian National University. She holds a Ph.D. in philosophy (2021, specialized in philosophy of science and technology) from the University of Toronto and a Ph.D. in applied mathematics (2015, specialized in large-scale algorithmic decision-making) from the Ecole Polytechnique of Montreal. Her research is on philosophy and ethics of data and AI (opacity, justice, generative AI, recommender systems), philosophy of science (explanation, representation, prediction, complex systems, automating science), and increasingly their intersection with socio-political philosophy and philosophy of language. She also collaborates with public and private institutions as an advisor.Website
Abstract Vasilija Rolfes: Predicting and selecting - an ethical perspective on the application of artificial intelligence in reproductive medicine
In reproductive medicine, artificial intelligence methods can be used to improve sperm, oocyte and embryo selection and to generate better predictive models for in vitro fertilization. However, complex normative issues arise with potential clinical applications. In the context of research ethics with view to evidence and efficacy of such supportive application, the challenge of informed consent, risk-benefit ratio for future children. Furthermore, because of the many actors involved in making the diagnosis, questions of responsibility may not be fully resolved. Likewise, questions arise about the equitable reimbursement of resource allocation. If involuntary childlessness is considered to be an affliction, the increase in the chance of having a child can be positively evaluated from an ethical perspective. By using artificial intelligence, the time gained by the physician for the patient can be seen as a gain for the benefit for the patient in the medical context. Until the clinical application of artificial intelligence, the quantity and quality of data must be critically examined and questions of transparency must be solved. In the medium and long term, it would be necessary to deal with undesirable effects and social dynamics that could accompany the use of artificial intelligence in reproductive medicine.
Vasilija and co-authors’ recent 2023 paper on the matter titled "Artificial Intelligence in Reproductive Medicine – An Ethical Perspective" is available open access at Link
Abstract Atoosa Kasirzadeh: Explainable AI and values
The use of ineliminably opaque machine learning algorithms in various areas of inquiry, from science to high-stakes decision making, is expanding quickly. This expansion has resulted in a surge of interest in two seemingly separated problems. (1) The explanation problem asks why and how the outputs of opaque algorithms are obtained. (2) The value problem asks whether and how the outputs of opaque algorithms are in accordance with value-neutral ideals. In this paper, I argue that the explanation problem and the value problem are two sides of the same coin. I show, in particular, how values are inherent to the demands of machine learning explanations and how explanations reveal value judgments essential to the machine learning practice. The recognition of the relation between these two problems highlights how reactions to the value problem help characterize our demands from a vigorous research program — Explainable Artificial Intelligence — for resolving the explanation problem.
