Identifying the problem, linking it to a requirement, selecting measures, and resolving conflicts: Alexandra Prégent and Alexei Grinbaum set out exercises and guides for trainers.
The main principles of AI ethics are well known; translating them into technical decisions is much less so. Alexandra Prégent and Alexei Grinbaum, philosopher and physicist at the CEA, offer a training framework for AI researchers and engineers, members of research ethics committees, public decision-makers and students.
Four skills to work on
The method is structured around four interrelated tasks: identifying the ethical issues of a project, determining what baseline requirements they bring into play, selecting and prioritizing technical and organizational measures to address them, and finally recognizing tensions that cannot be fully resolved in order to reason about trade-offs. It is based on an ethics checklist and a portfolio of measures that have already been published.
The article adds three types of practical exercises and guides for trainers, and draws lessons from pilot sessions, with recommendations for facilitating this type of workshop.
Read the full article on the original source
Go to the source →