SummaryI contend on both theoretical and historical grounds that quantitative risk assessment is relevant for policy determination only as a cost estimate. In particular, it provides a method for estimating the costs of a hypothetical insurance policy against the potential liabilities associated with a given course of action. It is not relevant to the question of rational choice under risk.RésuméJe montre, en partant d'arguments aussi bien théoriques qu'historiques, que le calcul quanti‐tatif des risques n'aide à la détermination d'une politique à (...) suivre qu'en tant qu'estimation des coûts. En particulier, il fournit une méthode pour estimer les primes d'une police hypothétique d'assurance contre les responsabilites potentielles associées à une suite donnée d'actions. 11 ne permet pas un choix rationnel dans des situations comportant des risques.ZusammenfassungIch argumentiere, sowohl aus theoretischen als auch aus historischen Gründen, dass eine quantitative Risikoberechnung für die Wahl einer Handlungsweise nur als Kostenschätzung relevant ist. Namentlich bietet sie eine Methode, um die Prämie einer Versicherungspolice gegen pot‐entielle Haftbarkeit, welche mit den Vorgängen eines Aktionsverlaufs verbunden ist, zu bestim‐men. Bei der Frage einer rationalen Wahl in einer Risikosituation ist sie nicht relevant. (shrink)

Subjective probability considered as a logic of partial belief succumbs to three fundamental fallacies. These concern the representation of preference via expectation, the measurability of partial belief, and the normalization of belief.

In the theory of explanation it is imperative that self-explanation be excluded; we cannot allow an explanandum to explain itself. This applies to the explanation of events and to the explanation of laws. In Nagel's account the following three features of a valid explanation of a universal law serve as a prophylaxis against self-explanation: “... all the premises are universal statements, there is more than one premise, each of which is essential in the derivation, and the premises, taken singly or (...) conjointly, do not follow logically from the explicandum”). (shrink)

Inleiding wetenschapsfilosofie voor 3e jaarsstudenten wis- en natuurkunde aan de TH Delft en Universiteit van Amsterdam.

Los's probability semantics are used to identify the appropriate probability conditional for use in probabilistic explanations. This conditional is shown to have applications to probabilistic reasoning in expert systems. The reasoning scheme of the system MYCIN is shown to be probabilistically invalid; however, it is shown to be "close" to a probabilistically valid inference scheme.

Probabilistic reasoning is traditionally represented by inferences of the following form (also called probabilistic explanations):where A and B are one-place predicates in a first order language, P(A | B) is the conditional probability of observing A among individuals having property B, and q is close to one.This argument is not logically valid, as the premises may be true while the conclusion is false. Moreover, as it stands, the premises do not even make the conclusion plausible. It may be the case (...) that 90% of the population dies before the age of 85, and that individual j is a member of the population, but this in itself does not make it plausible that individual j will die before the age of 85. (shrink)

Gaisi Takeuti has recently proposed a new operation on orthomodular lattices L, ⫫: $\scr{P}(L)\rightarrow L$ . The properties of ⫫ suggest that the value of ⫫ $(A)(A\subseteq L)$ corresponds to the degree in which the elements of A behave classically. To make this idea precise, we investigate the connection between structural properties of orthomodular lattices L and the existence of two-valued homomorphisms on L.

Gaisi Takeuti has recently proposed a new operation on orthomodular lattices L, ⫫: $\scr{P}\rightarrow L$ . The properties of ⫫ suggest that the value of ⫫ $$ corresponds to the degree in which the elements of A behave classically. To make this idea precise, we investigate the connection between structural properties of orthomodular lattices L and the existence of two-valued homomorphisms on L.

Gaisi Takeuti has recently proposed a new operation on orthomodular latticesL, $\begin{array}{*{20}c} \parallel \\ \_ \\ \end{array} $ :P(L)»L. The properties of $\begin{array}{*{20}c} \parallel \\ \_ \\ \end{array} $ suggest that the value of $\begin{array}{*{20}c} \parallel \\ \_ \\ \end{array} $ (A) (A) $ \subseteq $ L) corresponds to the degree in which the elements ofA behave classically. To make this idea precise, we investigate the connection between structural properties of orthomodular latticesL and the existence of two-valued homomorphisms onL.