Zilhão, José TeigaFernandes, Diogo Andrade Ribeiro2020-02-072020-02-072019-12-172019-07-31http://hdl.handle.net/10451/41714O objectivo deste trabalho consiste em avaliar os méritos explicativos e normativos da teoria matemática da acção racional conhecida como teoria bayesiana, tal como foi apresentada por Leonard Savage, em The Foundations of Statistics (1954), e mais tarde reformulada por Richard Jeffrey em The Logic of Decision (1965). O ponto de partida consistirá na apresentação analítica do Problema de Newcomb, um dos puzzles decisórios mais conhecidos e discutidos, o qual põe em causa esses referidos méritos. A tentativa de solucionar este puzzle oferecerá a oportunidade para colocar em confronto duas interpretações distintas do princípio da maximização da utilidade subjectiva esperada: a interpretação evidencialista e a interpretação causalista. Divergindo estas interpretações quanto à identificação daqueles que são os factores relevantes para a determinação de uma fórmula da utilidade esperada, a discussão em redor dessa divergência será o objecto central da investigação. Ao defender-se a necessidade de incorporar o conhecimento causal na construção de uma teoria da decisão adequada, defender-se-á a superioridade da teoria causal sobre a teoria evidencial. Essa superioridade é fortalecida pela constatação de que as estratégias revisionistas da teoria evidencial, a tickle defense e o ratificacionismo, não são capazes de lidar com certos problemas decisórios relevantes. Defender-se-á ainda que, num contexto decisório, a aplicação de um processo de revisão de crenças conhecido por visualização (Lewis 1976), alternativo à condicionalização bayesiana, permite neutralizar os contra-exemplos à teoria causal. Dada a opção pela versão contrafactual da teoria causal, constatou-se que, para efeitos de aplicação da teoria, é necessário adoptar-se uma resolução para a vagueza inerente à análise semântica de contrafactuais, resolução essa que dependerá de considerações pragmáticas. Defender-se-á, portanto, que a solução para o Problema de Newcomb dependerá do modo de resolução dessa vagueza, e que o confronto entre soluções distintas passará a dar-se no contexto da própria teoria causal.The aim of this dissertation is to evaluate the explanatory and normative merits of the mathematical theory of rational decision-making, known as Bayesian theory, as presented by Leonard Savage in The Foundations of Statistics (1954), and later reformulated by Richard Jeffrey in The Logic of Decision (1965). The starting point will be the analytical presentation of Newcomb's Problem, one of the most well-known and discussed decision-theoretical puzzles, which calls into question these merits. The attempt to solve this puzzle will offer the opportunity to confront two distinct interpretations of the principle of expected subjective utility maximization: the evidentialist interpretation and the causalist interpretation. As these interpretations diverge on the identification of those factors that are relevant to the determination of an expected utility formula, the discussion around this divergence will be the central object of the investigation. In defending the need to incorporate causal knowledge into the construction of an adequate decision theory, I will defend the superiority of causal theory over evidential theory. This superiority is confirmed by the realization that revisionist strategies of the evidential theory, the tickle defense and ratificationism, are not capable of dealing with certain relevant decision-making problems. It will also be argued that, in a decision-making context, the application of a belief revision procedure known as imaging (Lewis 1976), an alternative to Bayesian conditionalization, defuses the counterexamples to the causal theory. Given the choice of the counterfactual version of the causal theory, it was found that, for the purposes of theory application, a resolution for the vagueness inherent in the semantic analysis of counterfactuals is required, a resolution which will depend on pragmatic considerations. It will be argued, therefore, that the solution to Newcomb's problem will depend on the way in which this vagueness is resolved, and that the confrontation between different solutions will take place in the context of the causal theory itself.porTomada de decisãoEscolha (Filosofia)CausalidadeContrafactual (Lógica)Estatística BayesianaTeoria do conhecimentoTeses de doutoramento - 2019Problema de Newcomb : solução evidencialista ou causalista?doctoral thesis101622821