Multi-Objective Optimization: Practical Use and Policy Applications is an elective course at Pardee RAND Graduate School.

Multi-Objective Optimization: Practical Use and Policy Applications

Professor: Marler
Units: 0.5
Elective Course
Concentration: Quantitative Methods


Policy development and analysis often boils down to decision making, and most decisions involve multiple objectives or goals, with various constraints. In fact, many decisions can be posed as multi-objective optimization (MOO) problems. Yet, the struggle to pursue more than one objective is frequently inefficient, often depending on trial-and-error. And, this trend occurs in any discipline involving different people or institutions with different intentions, or involving a single decision-maker with more than one target. Although the idea of resolving interactive scenarios between multiple people has long been a topic of the social sciences, originally growing out of the field of economics, the process of systematically and computationally optimizing a collection of objective functions is less common. Furthermore, interpreting results can be especially challenging, as there may be no single “best” solution.

This five-week course focuses on concepts and methods for MOO, with policy applications ranging from human performance to national strategy. This topic bridges single-objective optimization, a mainstay of the logistics community, with game theory. It strives to help answer questions like, how should equipment be distributed to maximize effectiveness while minimizing load, or how should body armor be designed to maximize mobility and survivability simultaneously? Rather than depending too heavily on mathematical aspects, the intent is to help students identify real-world MOO problems, understand solution concepts, formulate and solve problems using existing tools (like Excel), and interpret results. A variety of common methods are reviewed, all with a focus on problem formulation and modes of preference articulation. We will then touch on methods for representing fuzzy preferences, applicable to scenarios where priorities may not be distinct. We will also touch on methods for incorporating uncertainty, which relates to robust decision making. Finally, game theory is introduced, simulating how information is exchanged between multiple policymakers.