Weight of argument and moral risk
Even if goodness and probability were measurable, the mathematical expectations of two alternative actions cannot be considered as the true measures of preferences because this method does not consider the weight of argument and the moral risk.
The epistemological concept of weight measures the reliability of probability. For an extensive analysis of this concept we direct the reader to the contributions by Levi (Chapter 3) and Vercelli (Chapter 7) in this volume. We mention here only that the weight of argument is the amount of relevant information available upon which a probability is established: if one argument is based on a greater amount of relevant evidence, it has more weight than another. Therefore, Keynes
(1973a , p. 347) doubts that 'a good whose probability can only be determined on a slight basis of evidence can be compared by means merely of the magnitude of this probability with another good whose likelihood is based on completer knowledge'.
As regards moral risk, in contrast, the mathematical expectation procedure does not answer the following question: is it certain that a larger good, that is extremely improbable, is precisely equivalent ethically to a smaller good that is proportionately more probable? Therefore, Keynes (1973a , p. 347) doubts that the moral value of a speculative action and the moral value of a cautious action 'can be weighed against one another in a simple arithmetical way'.