Paolo Baldi: Logical Consequences for Inferences with Data

Invitation.pdf

Paolo Baldi (University of Salento): Logical Consequences for Inferences with Data

Abstract
Data is often used in science as a guide to discard erroneous scientific hypotheses. The process is however far from trivial, since the rejection can be a matter of degrees, and possibly mistaken. In this talk, I will present a family of logical consequence relations, to capture patterns of inference from data to the rejection of hypotheses. I will introduce different definitions of 'rejection degrees' each yielding a distinct non-monotonic consequence relation. If time allows, I will hint at ongoing work, embedding our framework in a formal argumentation setting and extending it to machine learning predictions. The presentation is based on joint work with Hykel Hosni and Esther Corsi (University of Milan).