Date of Award

Spring 5-15-2017

Author's School

School of Engineering & Applied Science

Author's Department

Computer Science & Engineering

Degree Name

Doctor of Philosophy (PhD)

Degree Type

Dissertation

Abstract

When we have a population of individuals or artificially intelligent agents possessing diverse subjective inputs (e.g. predictions, opinions, etc.) about a common topic, how should we collect and combine them into a single judgment or estimate? This has long been a fundamental question across disciplines that concern themselves with forecasting and decision-making, and has attracted the attention of computer scientists particularly on account of the proliferation of online platforms for electronic commerce and the harnessing of collective intelligence. In this dissertation, I study this problem through the lens of computational social science in three main parts: (1) Incentives in information aggregation: In this segment, I analyze mechanisms for the elicitation and combination of private information from strategic participants, particularly crowdsourced forecasting tools called prediction markets. I show that (a) when a prediction market implemented with a widely used family of algorithms called market scoring rules (MSRs) interacts with myopic risk-averse traders, the price process behaves like an opinion pool, a classical family of belief combination rules, and (b) in an MSR-based game-theoretic model of prediction markets where participants can influence the predicted outcome but some of them have a non-zero probability of being non-strategic, the equilibrium is one of two types, depending on this probability -- either collusive and uninformative or partially revealing; (2) Aggregation with non-strategic agents: In this part, I am agnostic to incentive issues, and focus on algorithms that uncover the ground truth from a sequence of noisy versions. In particular, I present the design and analysis of an approximately Bayesian algorithm for learning a real-valued target given access only to censored Gaussian signals, that performs asymptotically almost as well as if we had uncensored signals; (3) Market

making in practice: This component, although tied to the two previous themes, deals more directly with practical aspects of aggregation mechanisms. Here, I develop an adaptation of an MSR to a nancial market setting called a continuous double auction, and document its

experimental evaluation in a simulated market ecosystem.

Language

English (en)

Chair

Sanmay Das

Committee Members

Roman Garnett, Brendan Juba, Roch Guerin, Yiling Chen,

Comments

Permanent URL: https://doi.org/10.7936/K7HH6HGT

Share

COinS