IDSAI Seminar: 'A Capability Approach to AI Ethics' with Emanuele Ratti (Lecturer, Department of Philosophy, University of Bristol)

Independently of the type of AI tool used (e.g. a simple decision tree or a complex deep neural net), their relative autonomy and scalability pose serious legal and ethical challenges, as AI tools may erode human selfdetermination and multiply wrongdoings. Over and above legal challenges, there are energetic debates on how the ethical impacts can be identified, and the right governance mechanisms that can address them. As a solution to these problems, it has been proposed that AI ethicists should take inspiration from auditing processes. This approach has been called ethics-based auditing (EBA), and both procedurally and content-wise is based on an underlying conception of ethics that significantly draw from the recent principled turn in AI ethics, which is notoriously fraught with difficulties. In this talk, I introduce an alternative framework for EBAs that is not based on the principled approach. In particular, I aim at conceptualizing EBA – and AI ethics altogether - on the basis of the capability approach. Rather than checking for compliance to vague principles, EBAs should investigate the impact of AI tools to capabilities. I formulate a preliminary characterization of capability-based EBA in the context of medical AI. This talk is part of a pilot project, which deliverable is the creation of a computational prototype that can be used for both internal and external auditing.

Delivery and Registration:

To be delivered hybrid. To manage the registrations, we ask participants to complete a simple form, which closes on the morning of the seminar, but please do not let that put you off. If you do miss the registration cut-off, then please email IDSAI.