An approach recently developed by Tanguiane for constructing quadratic objective functions from a minimal set of answers to simple questions about indifference is shown to suffer from a number of deficiencies in practical work. Most importantly, objective functions turn out to be highly non-robust with respect to small errors in the decision maker's answers. For definite objective functions an alternative method is proposed which avoids these problems. In addition a modification is suggested for the case of an indefinite objective function which will at least increase the robustness of the function constructed.
Contact: