Lectures, hwks


       UW CS
       Theory Group

Course Description

Modern technology has enabled communication, and therefore social and economic interaction between people, at unprecedented scales. One consequence of this mass interaction is the growth of new large-scale economic systems such as the Internet, online advertising, Ebay, and Mechanical Turk. While some of these electronic markets are just large-scale versions of traditional economic systems, their scale necessitates a computational approach to designing and analyzing them. Other systems such as the Internet present new challenges that have not been studied before. The emerging field of algorithmic game theory studies such challenges.

In this course we will present a computational and algorithmic approach to designing and analyzing economic systems. The course will be divided into the following three parts:

• Algorithmic Mechanism Design: Optimization in economic scenarios where the input to the optimization problem is unknown and must be elicited from self-interested participants.
• Price of anarchy type analyses: How well do economic systems behave at equilibrium?
• Efficient computation of equilibria: What kind of games and markets are tractable? When do simple uncoordinated dynamics over participants converge to equilibria?

For further information about these topics, see this survey paper by Tim Roughgarden.

Prerequisites: No knowledge of economics is necessary. We will assume some basic knowledge of computational efficiency. In particular, any one basic course in algorithms (577 or 787) or complexity (520, 810, or 830) or equivalent is required. If you are unsure of whether or not you have sufficient background for this course, please contact the instructor within the first week.

Textbooks: There is no textbook for this course. The following two books are recommended references. In addition, we will distribute lecture notes compiled from various sources.

• Algorithmic Game Theory by Nisan, Roughgarden, Tardos, and Vazirani. (A non-printable version can be accessed here for free).
• Multiagent Systems by Shoham and Leyton-Brown.

Evaluation: We will have three to four homeworks over the semester and no exams. The homeworks are expected to be challenging. Collaboration will be allowed and encouraged, but solutions must be submitted individually. Students taking the course for credit will also have a project requirement. The project will involve self-studying a recent work related to course topics and preparing a survey report. Suggested topics and instructions for preparing your report will be announced at the beginning of the semester.

Staff: Shuchi Chawla (instructor) and David Malec (TA).

Office hours: By appointment for both Shuchi and David.