Collaborative Decision Making at Scale: Bridging Theory and Practice
Abstract
Statement of Work: We will consider two kinds of decision making problems - structured and unstructured. The first covers situations where the underlying decisions are highly structured, such as making budgets, assigning water resources, and setting tax rates; in these situations, the underlying decision is quantitative, and any mechanisms and platforms heavily need to use this structure. We will focus on knapsack voting (i.e. budgeting), and develop algorithms and mechanisms that are strategy-proof, provide good cardinal guarantees, study fairness, and perform large-scale experiments. In unstructured decisions, we will perform experiments with consensus markets, develop initial algorithmic approaches, and perform at least one mid-scale deployment of our algorithms. Objective: While social media pervades many aspects of our lives, it has not yet proved to be an effective tool for large scale decision making: crowds of hundreds, perhaps millions, of individuals collaborating together to come to consensus on difficult societal issues. The objective of this research is to develop a unified algorithmic and empirical understanding of large scale decision making, and experiment with real-life deployments of our algorithms. Approach: For knapsack voting, our approach will be to assume special properties of the data. For instance, our initial data suggests that we often have recursive Condorcet winners. Accordingly, we will study Maximum Likelihood Estimators and optimum Ordinal guarantees under this assumption, as well as under natural noise models. We will also extend our previous work on approximate majorization to study the fairness of various social choice rules. For unstructured decisions, while the decisions themselves might not come from a well-understood decision space, we will structure the interactions so that users are collaborating in small groups or focused on small tasks (eg. rate or tank two alternatives, suggest an alternative that is midway between alternatives, etc). Overall Merit and ONR Mission/Relevance: The US Defense forces often find themselves embroiled in international situations where promoting democracy is a key objective. This is made harder when there are no pre-existing reliable democratic institutions. We believe that crowdsourced decision making platforms and mechanisms will play a large role in this transition in the coming decades. So we believe this is a core area for ONR to invest in. In addition, the Navy itself has many complex problems where collaborative decision making would be an asset.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 12, 2016
- Source ID
- N000141512786
Entities
People
- Ashish Goel
Organizations
- Office of Naval Research
- Stanford University
- United States Navy