Profiling is an example of how pervasively AI is used today and of the great ethical challenges it poses. Both public and private institutions rely on profiling for very different purposes, which is why profiling has become a major source of income for mega-companies such as Meta. Profiling has consequences on all aspects of individual and collective life including highly sensitive ones such as health, economics, welfare and politics. This is why it is very important to take the ethical dimension into account when developing and adopting AI-based technology for profiling.
How can logic help to face this challenge? Profiling is an exercise in categorisation, but importantly, it is a very particular form of categorisation since it is driven by cost-efficiency considerations and is performed by actors who need to make a certain decision. So, while there are many ways to categorise, profiling is highly dependent on the agenda and aims of the actors performing the categorisation and the goal of profiling is directly connected to the actors’ need of making the optimal decision, whichever it might be.
Logic allows us to address and analyse profiling tasks as types of categorisations by systematically and explicitly relating them with the goals, interests and agendas of actors. I will discuss an approach which is not just novel but is also suitable to be used in the design of new generation learning algorithms which combine explainability and accountability.
Dora Achourioti