Bo Waggoner

Maxwell-Dworkin 115

I am a Ph.D. candidate at Harvard University in Computer Science (what is Computer Science?), studying theoretical CS, artificial intelligence, and mechanism design. I am a part of Harvard's EconCS group (what is EconCS?), advised by Yiling Chen. I graduated from Duke University in 2011 and am an avid runner.
Page last modified: 2016-01-27.
Contents: Research Interests, Teaching, Papers, Talks, Writeups

See also: Running page, Misc page

Research Interests

Mostly in theoretical CS and microeconomics, especially mechanism design (from an algorithmic perspective).

Specific topics: algorithmic auction design, information elicitation (particularly in crowdsourcing settings, from both game-theoretic and machine-learning perspectives), online algorithms (particularly online bipartite matching problems), differential privacy, computational social choice (e.g. fair division, voting), computational learning theory (e.g. testing of discrete distributions).


Spring 2016:
CS 236r Prediction, Learning, and Games. with Professor Yiling Chen.
Teaching Fellow
Office hours: Mondays 2:30 - 4:00, or email me to find a time. Maxwell-Dworkin 115.

Fall 2013: CS 284r Incentives and Information in Networks. with Professor Yaron Singer.
Teaching Fellow.

Fall 2012: CS 121 Intro to Theory of Computation. with Professor Salil Vadhan.
Teaching Fellow.


Informational Substitutes for Prediction and Play (working paper - draft coming soon)
Yiling Chen and Bo Waggoner.
    Introduces a definition for when two pieces of information, "signals", are substitutes or complements. The definitions are natural from several perspectives; the main result is that they turn out to capture (respectively) "good" and "bad" equilibria of prediction markets. Also has some algorithmic applications to information acquisition under constraints.

A Market Framework for Eliciting Private Data (NIPS 2015 - upload coming soon)
Bo Waggoner, Rafael Frongillo, and Jacob Abernethy.
    Designs a prediction-market like mechanism for "purchasing" information or data from the crowd. Gives a contest-like structure where participants' contributions collaboratively update a single collective hypothesis. Preserves differential privacy of contributions.

Low-Cost Learning via Active Data Procurement (EC 2015)
Jacob Abernethy, Yiling Chen, Chien-Ju Ho, and Bo Waggoner.
    Designs mechanisms/learning algorithms to purchase data from agents who arrive one by one. The key idea is to actively change the prices we offer based on the current state of the learning algorithm. Proves learning guarantees as a function of the budget constraint.
    Supplementary: simulation code.

Fair Information Sharing for Treasure Hunting. (AAAI 2015)
Yiling Chen, Kobbi Nissim, and Bo Waggoner.
    Designs a mechanism for the following problem: A group of agents (e.g. pirates) each hold private information about the solution to a search problem (e.g. location of buried treasure); we want them to truthfully report to the mechanism, which then assigns search tasks fairly.

lp Testing and Learning of Discrete Distributions (ITCS 2015)
Bo Waggoner.
    Examines uniformity testing and learning of discrete distributions, given access to independent samples, under ℓp metrics. One result is that a 6-sided die is easier to test for fairness than a 2-sided coin, and a 52-card shuffler easier than the die, if the measure of fairness is ℓp with p > 4/3. In general, gives upper and lower bounds on number of samples needed (tight everywhere for learning and in many cases for uniformity testing).

Online Stochastic Matching with Unequal Probabilities (SODA 2015)
Aranyak Mehta, Bo Waggoner, and Morteza Zadimoghaddam.
    Designs an algorithm for online bipartite matching when edges are labeled with a probability of success and the goal is to maximize the expected number of matches. Prior work considered the case where all edge probabilities are equal; we consider the general case with unqual (but vanishing) edge probabilities. The approximation ratio is 0.534.

Output Agreement Mechanisms and Common Knowledge (HCOMP 2014)
Bo Waggoner and Yiling Chen.
    Considers equilibria of "output agreement" games, where two people get the same input and are asked to answer a question about it, rewarded based on how closely their answers agree. Rather than fully truthful, such games elicit "common knowledge" (though some subtleties arise).
    Preceded by the following workshop version, containing essentially the same results with different presentation:
    Information Elicitation Sans Verification. Bo Waggoner and Yiling Chen. The 3rd Workshop on Social Computing and User-Generated Content (SCUGC-13), at EC-13.

Designing Markets for Daily Deals. (WINE 2013)
Yang Cai, Mohammad Mahdian, Aranyak Mehta, and Bo Waggoner.
    Designs auctions for maximizing welfare when the bidders, like marketers on a Daily Deals site (e.g. Groupon), have private information about how valuable their deal is to consumers in the form of beliefs or predictions. We elicit this information truthfully, select an outcome, and assign payments to maximize a combination of bidder, auctioneer, and consumer welfare.

Evaluating Resistance to False-Name Manipulations in Elections. (AAAI 2012)
Bo Waggoner, Lirong Xia, and Vincent Conitzer.
    Considers voting when voters might cast multiple votes ("false-name manipulation"). We consider how best to design a deterrent against such manipulation in order to maximize the probability that the outcome of the election matches the true population preference, and how to statistically test whether this occurred.
    Supplementary: simulation code.


2015-10-14. Low-Cost Learning via Active Data Procurement.
Microsoft Research NYC.
slides (Google presentation), slides (pdf).
    Almost identical to the version below.

2015-09-09. Low-Cost Learning via Active Data Procurement.
Duke CSEcon Group.
    slides (Google presentation), slides (pdf).

2015-06-19. Low-Cost Learning via Active Data Procurement.
EC 2015, Portland, Oregon. 18min.
    slides (Google presentation), slides (pdf).

2015-02-06. Fair Information Sharing for Treasure Hunting.
Harvard EconCS group. 60min.
    slides and notes (Google presentation), slides (pdf).
    12min version given at AAAI 2015, 29 January 2015, Austin, Texas.

2015-01-13. lp Testing and Learning of Discrete Distributions.
ITCS 2015, Rehovot, Israel. 20min.
    slides (pdf), slides with notes (pdf).

2015-01-06. Online Stochastic Matching with Unequal Probabilities.
SODA 2015, San Diego, CA. 20min.
    slides (Google presentation), slides (pdf).

2014-01-10. Toward Buying Labels from the Crowd.
Indo-US Lectures Week in Machine Learning, Game Theory, and Optimization, Bangalore. 20min.
    slides (pdf), slides and notes (pdf).

2013-12-13. Designing Markets for Daily Deals.
WINE 2013, Cambridge, MA. 20min.
    slides (Google presentation, with notes), slides (pdf), slides and notes (pdf).

2013-06-16. Information Elicitation Sans Verification.
Workshop on Social Computing and User-Generated Content (SCUGC), at EC 2013, Philadelphia, PA. 20min.
    slides (pdf).
    Different version given at HCOMP 2014, 4 November 2014, Pittsburgh, PA.

2012-03-26. Evaluating Resistance to False-Name Manipulations in Elections.
Harvard EconCS Group. 45min.
    slides (pptx), slides (pdf), slides and notes (pdf).


Useful Tips, Tricks, and Techniques. A list-in-progress I am compiling of some of the most common and useful approximations, inequalities, bounds, proof techniques, etc. for theoretical computer science.

Jumpstart: Introductions by examples to some useful tools.

Possible misspellings of my name include "Bo Wagner" and "Bo Wagoner".