Coherence Through Partial Information in an Additive Multiattribute Utility Analysis

摘要

This report addresses the specific problem of the resolution of incoherent weight assessments in an additive multiattribute utility analysis. The approach taken is that if incoherencies occur because actual numerical assessments are too precise, then it would be useful if the ramifications of less precise, but coherent, information were made clear. Two types of information are considered. First, it is supposed that the decision maker can order the attributes on the relative importance of the weights. The implications of any given ordering are shown to be very simply analyzed. Second, it is supposed that, in addition, inequality assessments can be made between certain pairs of weights. The analysis demonstrates the implications of these, and also suggests which inequality assessments are likely to be most useful.