CAREER: Making Better Decisions: A Proposal for Human-Centered Computational Social Choice using Artificial Intelligence and Data

Tulane University, New Orleans LA

Investigators

Abstract

For centuries groups of agents have come together to make social choices: using formal processes to select alternatives and allocate resources while taking into account the diverse constraints and preferences of the members of society. These group decisions range from high stakes, e.g., who should receive a kidney; to low stakes, e.g., where should we go to dinner? In all these settings and applications, each agent arrives with their own preferences, and a mechanism (algorithm) makes a decision that is then shared by that group. This research focuses on improving mechanisms and processes for two classic settings: selection problems (e.g. voting), where a single or small set of winners must be shared by the group, and allocation problems (e.g., matching), where each agent receives a personal allocation or bundle. This research will leverage the fact that since computers mediate much of the communication between agents, large amounts of data are available to analyze, understand, and improve the decision-making process. Hence, there is a unique opportunity to study group decision making in the complex, realistic contexts in which it takes place by leveraging human-centered artificial intelligence (HCAI) principles that prioritize human needs and values. The formal study and application of models, algorithms, and axioms for collective decision-making are fundamental to many areas, including computational social choice, electronic commerce, and recommender systems. This award will advance techniques from human-centered artificial intelligence and computational social choice (HCAI-COMSOC) to close the gap between the abstractions for modelling and reasoning about mechanisms and the use-inspired features of human behavior. The research is organized around a Grand Challenge of building systems and developing the underlying algorithms necessary to facilitate better decisions in online, deliberative, and interactive environments in ways that align with human values and under human control. To tackle this grand challenge, the research will focus on (1) developing new data, models, and metrics of agent preferences for both outcomes and mechanisms, and (2) fundamental research to create novel algorithms and mechanisms for human-centered computational social choice, to create a use-inspired feedback loop with the broader impacts of the award. All code and data collected and developed as part of this project will be released in the unified PrefLib framework and the project will support students in service learning classes to work with local non-profit stakeholders. This project is jointly funded by the Division Of Information & Intelligent Systems / Robust Intelligence (IIS/RI) and the Established Program to Stimulate Competitive Research (EPSCoR). This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.