loader image

Where AI Ethics Should Go

Éric Pardoux, Louis Devillaine

Towards processual AI ethics

Jeudi 30 juin 2022, 10h15-11h05

En présentiel :

Hörsaal, Doblerstr. 33, Tübingen

En distanciel :

ID de réunion : 96822543209 

Code secret : 863312


Programme :

Présentation :

The von Weizsäcker Zentrum and the Archives Henri Poincaré and LORIA (University of Lorraine at  Nancy) are organizing an interdisciplinary event.
The  premise of the event is that the subfield of AI ethics has experienced an explosive expansion in the last couple years, and it is time to take a breather from this frantic rate of expansion to think about the consolidation of the subfield and the directions it should take.

Argument de la conférence :
    Discourses on AI ethics have broadly focused on the principles and values that should be fostered in the development and use of AI systems, largely drawing from principlism and biomedical ethics. However, the translation of such principles into practice is still puzzling the AI community. Through this work, we want to complement theoretical ethical frameworks by highlighting how ethics by design could be promoted. Emphasizing the plurality of moments in which ethical decisions have to be taken pushes us to defend an ethical practice that would be more processual than procedural. We use the distinction between procedural and processual ethics to suggest an approach to AI ethics that would take place all along the design process and not only at certain milestones. This processual approach to ethics implies several conditions to be undertaken successfully: this is what we intend to start exploring through the research we are presenting. By sketching the main ethical concerns associated with peculiar steps of the design process of AI systems we put up some suggestions regarding the ways to tackle these challenges. More generally we propose some elements of an ethical culture that could be transmitted to all the stakeholders of the AI design process in order to stimulate the conception of ethical AI systems. Our position is based on the hypothesis that technological design is able to transmit some values to the artefacts and devices it creates. Therefore, we build up on the notion of ethics by design to devise different ways that methodologies of design could help promote the ethical development of AI. This implies both methodologies of work within structures developing AI as well as individual recommendations to the designers themselves. To achieve such an ethics by design, we intend to explore the socio-cultural conditions in which this approach could take place. Hence, we hope that our contribution may help strengthen the link between theoretical ethics and actual practices, by offering some ways to improve the application of AI ethics.