Games, Economic Behavior, and the Internet

Information Science 6230
Cornell University, Fall 2013

Tues-Thu 10:10-11:25am

Hollister Hall 320

Arpita Ghosh

There are many settings on the Internet where agents make decisions that involve responding to incentives, trading off costs and benefits, or strategizing about the behavior of other agents. Examples include online auctions (on Ebay, or advertising auctions); voting--- either explicitly, such as ratings on Reddit or implicitly, as in search engine rankings; crowdsourcing contests as in TopCoder or Innocentive; labor markets as in Mechanical Turk or ODesk; peer-to-peer file sharing; social lending platforms, and a vast range of social computing systems.

This course will examine models and theoretical foundations--- such as information elicitation, contest design, auctions, matching markets, public goods games--- that are relevant to economic behavior on the Internet, as well as discuss the design and behavior of actual systems.

The course will consist of a combination of lectures by the instructor, and seminar-style presentation of research papers by students. Assignments of papers to students for presentation will be made in the first few weeks of class.

Course Information

Instructor: Arpita Ghosh

Office: Room 110, 301 College Avenue

Email: arpitaghosh@cornell.edu

Office hours: Thursdays 1-2pm, Rhodes 490.

Outline of Topics

(Caveat: This is a tentative outline, and may evolve through the duration of the semester.)

Auctions on the Internet

Introduction to auction theory
- First, second price sealed bid, and Dutch and English auctions
- Private and interdependent valuations; the winner's curse
EBay auctions and bidder behavior
- Timing of bids in eBay auctions: Empirical evidence
- Understanding bid timing: Model and equilibrium bidding
Online ad auctions: The Generalized Second Price (GSP) auction for sponsored search advertising
- A theoretical analysis of GSP: Comparison with the VCG auction; equilibrium revenue and efficiency

Online Effort Elicitation: Incentives in Crowdsourcing

Introduction to contest design
- A basic model for effort elicitation via contests
- Equilibrium analysis for a winner-take-all contest; comparative statics.
How should a contest be structured?
- Equilibrium analysis for rank-order mechanisms
- Contest design: Optimal rank-order mechanisms
Other models and problems in contest design and effort elicitation; innovation races

Online crowdsourcing contests (Innocentive, TopCoder, ...)

Incentives in information elicitation

Eliciting information from an expert
- Proper scoring rules
The peer-prediction method
Extending the peer-prediction model; improved mechanisms
- Papers: Witkowski-Parkes'12; Dasgupta-Ghosh'13

Incentives in social search

Information seeking on a social network: Incentives and network structure
- Papers: Kleinberg-Raghavan'05
Incentive design in the DARPA challenge
- Papers: Pickard et al'11, Cebrian et al'12
Sybil issues in social search
- Papers: Babaioff et al.'12

Privacy and economic issues

Introduction to privacy issues online
- Famous privacy breaches on the Web
Notions of privacy
- Impossibility of absolute privacy. Differential privacy
Privacy and economic models

Acquisti'10

Pai and Roth'13

Egelman et al'09

Brandimarte et al (Working paper)

Acquisti et al'10

Carrascal et al'13

Coursework

The course will be evaluated on the basis of the following:

Class project: 35%
Research papers: In-class presentations; written summaries: 30%. (20% credit for your own presentation, 10% for written summaries in other presentations. Attendance is mandatory for student presentations and absence will carry a penalty.)
2 assignments: 15%
2 short summaries (2 pages) of any 2 relevant seminars in IS/CS/Econ over the semester. 10%
2 notes/`blog posts' about economic issues on the Internet. 10%

Class attendance is a critical component of the course since most learning will occur in class. While not explicitly assigned a percentage of total course credit, poor attendance will result in a penalty in the final grade.