Abraham Othman
Contact Info
About Me
Call me Abe. I'm a second-year PhD Student in the Computer Science
Department at the School of Computer Science at Carnegie Mellon. My
advisor is Tuomas Sandholm, which makes me part of the AMEM research
group, a bunch of inveterate gamblers with a computing problem.
This is what I looked like on my first day of graduate school.
Things I Do
I run the Game Theory Discussion
Group. I currently serve as the malevolent dictator of Dec/5, SCS's
semi-secret social organization. I also developed the Gates Hillman Prediction
Market, which we're using to forecast when CMU's new CS building
will open.
If you're at CMU, you might be interested in how to make a door label, or maybe the rebuilding of the free food cam.
When I'm at work I like to listen to new music. You can check out my hype machine page to see what I'm currently digging.
Previously
I have a degree in
Applied
Math from Harvard, but I spent
more time hanging around the Computer Science,
Economics, and Comparative Literature
departments. My undergraduate advisor was David Parkes and I also
worked with Uli
Doraszelski. I ran Adams
House Intramurals and really enjoyed rowing House Crew.
I grew up in Ann Arbor, Michigan, which gave me an appreciation for college sports and an ability to find humor in the cruel vicissitudes of fate.
Publications
Othman, A. and Sandholm, T. 2008. The Cost and
Windfall of Manipulability , at the Second International Workshop on
Computational Social Choice (COMSOC). Download
pdf.
A mechanism is manipulable if it is in agents' best interest to
misrepresent their
private information (lie) to the center. We provide the first formal
treatment of the
windfall of manipulability, the seemingly paradoxical quality by which
the failure of
any agent to play their best manipulation yields a strictly better
result than an optimal truthful mechanism. We dub such mechanisms manipulation optimal. We
prove
that any manipulation-optimal mechanism can have at most one manipulable
type
per agent. We show the existence of manipulation-optimal multiagent
mechanisms
with the goal of social welfare maximization, but not in dominant
strategies when
agents are anonymous and the mechanism is symmetric, the most common
setting.
For this setting, we show the existence of manipulation-optimal
mechanisms when
the goal is affine welfare maximization.
Work related to this paper was presented at GAMES 2008.
Work related to this paper was presented at GAMES 2008.
Corwin, I. and Othman, A. 2008. Time
Inconsistency and Uncertainty Aversion in Prediction Markets, at the Third Workshop on Prediction
Markets, in conjunction with the ACM Conference on Electronic Commerce (EC). Download preprint.
Starting from first principles we derive a method for detecting price
biases in prediction
market data. Using this method on Tradesports contracts from the 2005-06
NBA season, we
demonstrate that trades executed in the last hour of trading have a
significant longshot price
bias, while trades occurring earlier do not. We present a new
theoretical model which uses
uncertainty aversion to explain our findings.
Othman, A. 2008. Zero-Intelligence Agents in
Prediction Markets, in Proceedings of the
7th International Conference on Autonomous Agents and Multiagent Systems
(AAMAS). Download pdf.
Conlee B., Othman, A., and Yetter, C. 2007. What to Feed a Gerrymander. The UMAP Journal 27(3): 261-280. Download preprint.
We construct a novel agent-based model of prediction markets in which putative human qualities like learning, reasoning, and profit-seeking are absent. We show that the prices which emerge from a market populated by a class of distinctly inhuman agents, Zero-Intelligence agents with diffuse beliefs, replicate the findings of empirical market studies. We use this result to argue against the prevailing descriptive theories of price formation in prediction markets, which have stressed the role of expert, rational participants.
Conlee B., Othman, A., and Yetter, C. 2007. What to Feed a Gerrymander. The UMAP Journal 27(3): 261-280. Download preprint.
This was Harvard's winning entry in the 2007 Mathematical Contest in Modeling (MCM). The prompt was to design an algorithm to simply and fairly redistrict states, and to demonstrate our method using the state of New York. Our solution interpreted the problem as an issue of large-scale combinatorial optimization. Our paper earned an "Outstanding" rating (top 1%) and won a prize from INFORMS, an MCM sponsor, as their selection for best paper.