SI 683 Reputation Systems

Instructor: Rahul Sami

1.5 Credit, 7-week course module
Second half of Winter 2008
Tuesday, Thursday 10:30am-12:00noon, 409WH
Office hours: Monday 3-4PM 3246ESI-N; Thursday 2-3PM 314WH

Course Goal:

In this course, you will learn about the design of reputation systems: design choices, benefits, threats, and limitations. At the end of this course, a student should be able to critically analyze a reputation system design to identify strengths and potential weaknesses, and to design a reputation system for a particular domain with a clear idea of the tradeoffs involved.

Overview:

The Internet enables interactions, commercial and non-commercial, between people across geographical and social boundaries. Ideally, these interactions are mutually beneficial, but they may also be exploitative or fraudulent. For example, buying an item on an online auction may get you a bargain, but you may also lose money if the other party ships you poor-quality goods, or does not send anything at all. Ideally, you would like to engage in interactions with reliable and trustworthy entities, but this is often difficult for two reasons: (1) You may not have information about the quality of the other party, or the value of the exchange; (2) The other party may have incentives that conflict with your own, and may thus seek to manipulate the interaction to your detriment.

Reputation systems are widely used to address these problems in online forums; for example, eBay.com, Amazon.com, and Yahoo.com all use reputation systems in their auctions. A user's reputation is an aggregate of feedback from all her past trading partners. Ideally, it will reveal information about the average quality of those transactions; further, the desire to maintain a high reputation gives her an incentive to honestly execute the current transaction. However, a reputation system needs to be designed carefully, and tailored to its application setting: poorly designed reputation systems may be uninformative or prone to manipulation.

In this course, we will study the design and critical analysis of reputation systems. We will discuss incentive issues involved in motivating users to behave honestly and to give honest feedback, as well as other practical aspects of designing a reputation system, such as the format of feedback input and retrieval. We will also study ways in which strategic parties may try to circumvent the system, and techniques to defend against these attacks.

Prerequisites

A basic introduction to game theory: old SI 625, SI652 or equivalent, or permission of instructor.

Course Schedule

Week 1:
Introduction to reputation systems
Measuring the value of reputations
Week 2:
Incentives for feedback contribution
Reputations as an incentive tool.
Week 3:
Economic models (continued).
Negative feedback: incidence and response
Week 4:
Identity attacks on reputation systems: Changing names
Identity attacks on reputation systems: Sybil attack
Week 5:
Midterm
Distributed reputation systems for peer-to-peer systems
Week 6:
Reputations based on textual feedback
Alternatives to reputation systems
Week 7:
Case study of a reputation system(ePinions))
Summary and review

Course Work and Assessment

4 Assignments 30%
Assignments will include exercise problems on the economic models studied, and short-answer questions on the papers and topics discussed in class.
Class Participation 10%
Final Project (due on 21st April) 60%
The final project will involve designing a reputation system for a particular domain. Students will submit a paper (about 6-8 pages) describing the design, outlining the strengths and weaknesses of the system (including potential attacks), and explaining the design choices they made.