F1.12: Multicriteria Decision Making and Evolutionary Computation

Applying evolutionary computation (EC) to multicriteria decision making addresses two diicult problems: (1) searching intractably large and complex spaces and (2) deciding among multiple objectives. Both of these problems are open areas of research, but relatively little work has been done on the COMBINED problem of searching large spaces to meet multiple objectives. While multicriteria decision analysis usually assumes a small number of alternative solutions to choose from, or an \easy" (e.g., linear) space to search, research on robust search methods generally assumes some way of aggregating multiple objectives into a single gure of merit. This traditional separation of search and multicriteria decisions allows for two straightforward hybrid strategies: (1) make multicriteria decisions FIRST, to aggregate objectives, then apply EC search to optimize the resulting gure of merit, or (2) conduct multiple EC searches FIRST using diierent aggregations of the objectives in order to obtain a range of alternative solutions, then make a multicriteria decision to choose among the reduced set of solutions. Over the years a number of studies have successfully used one or the other of these two simple hybrid approaches. Recently, however, many studies have implemented Pareto-based EC search to sample the entire Pareto-optimal set of non-dominated solutions. And a few researchers have suggested ways of INTEGRATING multicriteria decision making and EC search, by iteratively using EC search to sample the tradeoo surface while using multicriteria decision making to successively narrow the search. Although all these approaches have received only limited testing and analysis, there are few comparable alternatives to multicriteria EC search (for searching intractably large spaces to meet multiple criteria).