Multi-Objective Decision Making

Many real-world tasks require making decisions that involve multiple possibly conflicting objectives. To succeed in such tasks, intelligent systems need planning or learning algorithms that can e ciently find di↵erent ways of balancing the trade-o↵s that such objectives present. In this tutorial, we provide an introduction to decision-theoretic approaches to coping with multiple objectives. We first present an overview of multi-objective decision problems, with real-world examples. Then, we show that di↵erent assumptions about these problems lead to di↵erent solution concepts such as the convex hull and the Pareto front. Next, we provide an overview of state-of-the-art algorithms for tackling them, such as multi-objective variants of dynamic programming. Finally, we highlight some applications of multiobjective methods and discuss some of the most important open questions.