Solution Pluralism and Metaheuristics
Steven Orla Kimbrough, Ann Kuo, Hoong Chuin Lau, Frederic H. Murphy, David Harlan Wood · Singapore Management University Institutional Knowledge (InK) (Singapore Management University) · 2011
Solution pluralism is an approach to problem solving and deliberation. It employs a plurality of distinct solutions for a decision problem for aiding decision making. The concept is well established in existing practice, although perhaps not recognized as such. This paper: (1) presents the concept as a generalization of established practice, (2) briefly describes successful uses of the concept in practice, and (3) presents several areas that appear would benefit from application of the concept. Throughout, the role of metaheuristics in finding the pluralities of solutions is emphasized. Solution pluralism is an approach or even a bit of philosophy for problem solving and deliberation. It emphasizes and explores the idea that obtaining a plurality of distinct solutions for a decision problem may often be both practicable for, and of great value in, aiding decision making. There are three principal steps in employing a solution pluralism approach to a given problem: (1) Define SoIs. Define or characterize the solutions of interest (SoIs) for the problem to hand. What, if we could get them, would constitute the plurality of solutions we would like to use for our decision problem? In our work, and in what we discuss below, we are mainly interested in optimization models, broadly construed. This should not be taken to indicate that solution pluralism is confined to mathematical models for decision making. (2) Obtain SoIs. Once the SoIs are defined, the problem is to get them. This is in general a non-trivial problem. If it is hard to find an optimal or even heuristically good solution, how are we to obtain a plurality of good or interesting solutions? Our work in this regard begins with the observation that metaheuristics generally, especially population-based metaheuristics and particularly evolutionary computation, are widely employed in practice and produce multiple, typically very many, solutions in the course of their searching. Can the solutions produced as a by-product of heuristic search be employed usefully for purposes of solution pluralism? We think so and will indicate why, although the present format requires extreme brevity. How best to exploit metaheuristics in order to obtain SoIs is, of course, a matter for research. One of our aims in this paper is to draw attention