Convergence of Interactive Procedures of Multiobjective Optimization and Decision Support

The paper presents an overview of issues of convergence of inter-active procedures in multiobjective optimization and decision support. The issue of convergence itself depends on assumptions concerning the behavior of the decision maker — who, more specifically, is understood as a user of a decision support system. Known procedures with guaranteed convergence under classic assumptions are reviewed. Some effective procedures of accelerated practical convergence but without precise convergence proofs are recalled. An alternative approach to convergence based on an indifference threshold for increases of value functions or on outranking relations is proposed and illustrated by a new procedure called Outranking Trials.