Automated design of social choice rules by optimization and rule extraction

点击翻译按钮获取中文摘要

The goal of this research project is to develop a method for automatically generating social choice rules that can be applied to large-scale problem instances by combining existing automated mechanism design techniques (which can be applied only for small-scare problem instances) and rule discovery techniques. In this research, we first develop an important core technology for achieving this goal, i.e., a method for extracting social choice rules that satisfy desirable properties based on the results obtained by automated mechanism design. This result was presented at Forum on Information Technology (FIT) and obtained the best paper award (Funai best paper award). Furthermore, the existing automated mechanism design techniques use integer programming to obtain optimal rules for a given criterion. As a result, the types of participants must be represented as a set of discrete values. We develop a new method for directly obtaining social choice rules based on another optimization technique called quantifier elimination, which can handle non-linear, continuous variables.