A Framework for Testing Incentive Compatible Reputation Mechanisms

Student

Aurélien Frossard

Period:

Winter semester 2004-2005
Ecole Polytechnique Fédérale de Lausanne
Computer Science Department
Artificial Intelligence Lab

We consider a context in which users can buy goods through an online market (e.g. Amazon). Besides the complete description, buyers can also access textual reviews written by other buyers (or independent reviewers) about the goods. Every time a buyer buys a good, she has the right to submit feedback about the reviews she has read before buying. Feedback is binary (0 if the buyer does not agree with the reviewer, 1 otherwise). The goal of the reputation mechanism is to assess the trustworthiness of reviews by considering the feedback of previous buyers.

Reputation information about reviews cannot be trusted unless the mechanism makes it in the best interest of all buyers to submit honest feedback. The goal of this project was to develop a framework in which different incentive compatible reputation mechanisms can be tested and evaluated.

Download: