Better Collective Decisions via Uncertainty Reduction

TL;DR

This paper explores methods to improve collective binary issue decisions by reducing uncertainty through education, delegation, or appointing experts, analyzing the computational complexity of these approaches.

Contribution

It introduces a formal framework for decision enhancement tools and provides complexity results, including NP-hardness and fixed-parameter tractability, for their implementation.

Findings

Educating agents can significantly improve decision accuracy.

Delegation strategies can be optimized for better group outcomes.

Computational complexity varies across different decision-support tools.

Abstract

We consider an agent community wishing to decide on several binary issues by means of issue-by-issue majority voting. For each issue and each agent, one of the two options is better than the other. However, some of the agents may be confused about some of the issues, in which case they may vote for the option that is objectively worse for them. A benevolent external party wants to help the agents to make better decisions, i.e., select the majority-preferred option for as many issues as possible. This party may have one of the following tools at its disposal: (1) educating some of the agents, so as to enable them to vote correctly on all issues, (2) appointing a subset of highly competent agents to make decisions on behalf of the entire group, or (3) guiding the agents on how to delegate their votes to other agents, in a way that is consistent with the agents' opinions. For each of these tools, we study the complexity of the decision problem faced by this external party, obtaining both NP-hardness results and fixed-parameter tractability results.