Compromise programming: Non-interactive calibration of utility-based metrics

Utility functions have been used widely to support multi-objective decision-making. Expansion of a general additive utility function around the ideal results in a composite linear-quadratic metric of a compromise programming problem. Determining the unknown parameters of the composite linear-quadratic metric requires substantial interaction with the decision maker who might not always be available or capable to participate in such a process. We propose a non-interactive method that uses information on observed attribute levels to obtain the unknown parameters of the composite linear-quadratic metric and enables forecasting and scenario analysis. The method is illustrated with a small scale numerical example.