Anson Kahng

Computational Perspectives on Democracy Degree Type: Ph.D. in Computer Science
Advisor(s): Ariel Procaccia
Graduated: August 2021

Abstract:

Democracy is a natural approach to large-scale decision-making that allows people affected by a potential decision to provide input about the outcome. However, modern implementations of democracy are based on outdated information technology and must adapt to the changing technological landscape. This thesis explores the relationship between computer science and democracy, which is, crucially, a two-way street–just as principles from computer science can be used to analyze and design democratic paradigms, ideas from democracy can be used to solve hard problems in computer science.

   Question 1: What can computer science do for democracy?

To explore this first question, we examine the theoretical foundations of three democratic paradigms: liquid democracy, participatory budgeting, and multiwinner elections. Each of these paradigms broadly redistributes power from the few to the many: For instance, liquid democracy allows people to choose delegates more flexibly and participatory budgeting enables citizens to directly influence government spending toward public projects. However, bcause these paradigms are relatively new, their theoretical properties are relatively unexplored. We analyze each of these three settings from the point of view of computational social choice, which is a mathematical framework for collective decision-making. In particular, we focus on a combination of robustness, fairness, and efficiency with the end goal of providing actionable advice for future iterations of these paradigms.

   Question 2: What can democracy do for computer science?

Toward this end, we explore two settings in which democratic principles can be used to augment approaches to making difficult decision in our case, automating ethical decision-making and hiring in online labor markets. Both of these problems are difficult in the sense that there is no universally agree upon function to optimize, making them a poor fit for traditional approaches in computer science. Instead, we try to emulate a world in which we can get input from people in order to arrive at a "societal" decision. In each of these settings, we first propose and analyze a theoretical approach that leads a single decision, and then, in collaboration with HCI researchers, run experiments in the real world to test the efficacy and practicability of our approaches in the real world.

Thesis Committee:
Ariel Procaccia (Chair)
Chinmay Kulkarni
Nihar Shah
Vincent Conitzer (Duke University)
David Pennock (Rutgers University)

Srinivasan Seshan, Head, Computer Science Department
Martial Hebert, Dean, School of Computer Science

Keywords:
Computational Social Choice, Theoretical Computer Science, Artificial Intelligence