1 / 21

CS522: Algorithmic and Economic Aspects of the Internet

CS522: Algorithmic and Economic Aspects of the Internet. Instructors: Nicole Immorlica (nickle@microsoft.com) Mohammad Mahdian (mahdian@microsoft.com). Reputation Mechanisms. Mechanisms that collect, aggregate, and distribute information about the “reputation” of participants.

wilton
Download Presentation

CS522: Algorithmic and Economic Aspects of the Internet

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. CS522: Algorithmic and Economic Aspects of the Internet Instructors: Nicole Immorlica (nickle@microsoft.com) Mohammad Mahdian (mahdian@microsoft.com)

  2. Reputation Mechanisms • Mechanisms that collect, aggregate, and distribute information about the “reputation” of participants. • PageRank and HITS can be considered reputation mechanisms. • Today, we talk about feedback-based reputation mechanisms used in electronic marketplaces.

  3. Why is reputation important? • Trust in a long-term relationship • Classical example: repeated prisoner’s dilemma • If both players follow a “tit for tat” strategy, they will always cooperate (if  is close to 1). B A

  4. Why is reputation important? • The expectation of reciprocity or retaliation in future interactions creates an incentive for good behavior. • What about repeated interaction with different people? • Society enforces good behavior by keeping track of reputation. • Also, organizations like Better Business Bureau, Zagat, Consumer Reports.

  5. Reputation on the Internet • Connection between individuals is significantly weaker. • People can share their opinions through reputation mechanisms. • Information: Lower quality, higher quantity

  6. Without reputation… • Buyers will not pay the full price of a high-quality item. • Sellers with high-quality items will not accept discounted prices. • Eventually, only low-quality sellers remain. • “market for lemons” [George Akerlof]

  7. Impact of reputation on prices • Experiment by Resnick et al. • ~200 postcards; half sold by an ebay account with strong reputation, the other half sold by a new account operated by the same person • The price was 8.1% higher for the seller with strong reputation. • Many other experimental studies…

  8. Challenges • Eliciting information • People don’t bother to give feedbacks Solution proposed: payment for feedback • Hard to elicit negative feedback (fear of retaliation) • Ensuring honest reports

  9. Challenges, cont’d • Distributing information • Name change New comers must be distrusted until they pay their dues. alternatively, can try to prevent name change. • Portability from system to system (e.g., between amazon and ebay)

  10. Challenges, cont’d • Aggregating information • It’s not obvious what’s the best way to aggregate the information. • Current methods: number/percentage of +/- feedback • Perhaps should be weighted by the value of the transaction and/or reputation of the feedback provider.

  11. A game-theoretic model • Chris Dellarocas (2003) • One seller, many buyers • Each buyer buys once. • In each round, a number of buyers bid for an item, which can be of high quality, or low quality. Auction is 2nd price. • The quality depends on the level of effort the seller exerts.

  12. The model • Seller offers a unit of good of high quality • Buyers bid their expected valuation based on seller’s feedback profile. Let w1, w2 denote the highest and second highest valuation for a high-quality good. • Seller decides whether to exert high effort at cost c, or low effort at cost 0; corresponding probability that the quality is low is  and  (<). • Buyer receives the good and leaves feedback. The feedback profile of the seller is updated.

  13. Binary feedback mechanism • Buyer leaves either a positive, or a negative feedback. • The feedback profile of the seller is a collection of N recent feedbacks. • Every time a new buyer gives feedback, it replaces a random feedback in the profile. • Feedback profile can be summarized by the number x of negative feedbacks.

  14. Equilibrium play • If s(x) is the probability that the seller exerts high effort, then the expected valuation of the buyer i is [s(x)(1-)+(1-s(x))(1-)]wi, (wi is i’s valuation for a high quality good.) • )Revenue of the auction = G(x) = [s(x)(-)+(1-)]w2.

  15. Equilibrium play • V(x) = seller’s expected payoff starting from reputation x. • V(x) = G(x) + s(x).Ucoop(x) + (1-s(x)).Ucheat(x). • Ucoop(x) = -c +  [ (1-x/N) V(x+1) + ( x/N + (1-)(1-x/N)) V(x) + (1-)(x/N) V(x-1) ]. • Similar expression for Ucheat(x).

  16. Results • Theorem. If w2/c is large enough, then the seller’s optimal strategy is:

  17. Distribution at the equilibrium

  18. Efficiency as a function of w2/c

  19. Conclusion • A reasonably high (but not perfect) degree of efficiency can be achieved. • Model can be generalized to cases where buyers sometimes don’t leave feedback. • The model predicts that it is optimal to treat missing feedbacks as positive. • Open question: more than one seller?

More Related