Weighting Evaluations

Conrad Heilmann (Rotterdam) gives a talk at the MCMP Colloquium (4 July, 2012) titled "Weighting Evaluations". Abstract: This paper discusses two approaches to weighting numerical evaluations. We start by assuming that we have obtained numerical evaluations of some type of object, such as a utility function that represents preferences over acts. We then ask how we can weight these evaluations and identify two approaches. One approach, called the probability approach, derives a separate measure on the domain of objects on which the value operation was defined. The other approach, called the discounting approach, enriches the description of the domain so as to amend the value operation directly, which yields an implied measure. We analyse and compare the two approaches along four lines, asking how they (i) can be motivated conceptually, (ii) derive and elicit the measures, (iii) can apply the respective measures, and (iv) update the measures. The upshot of this discussion is that the probability approach should be favoured over the discounting approach on all four accounts.