Abstract

The main purpose of this thesis is to explore de Finetti's ideas and contributions to decision theory. Such ideas are not as well-known as his work on probability. ;The first part of the work is placed in a unisubjective decision-making context. It starts by including a discussion on predictivism, an approach to statistics which de Finetti insisted on and which has only recently been rediscovered and advocated. ;The second part is placed in the context of group, or multisubjective, decision-making. This area is well-known for its negative results, starting with Arrow's impossibility theorem. While being aware of such difficulties, de Finetti made important contributions to this field, which he saw as the appropriate context to approach basic concepts such as "objectivity" and Wald's admissibility theorem. Such concepts, together with exchangeability, illuminate the Bayesian position by also making connection with the non-Bayesian approach. ;The last part of the work starts with a critical discussion of classical methods which are unacceptable for Bayesians. Finally, an important application problem--environmental stress screening--is solved from a Bayesian predictivist point of view