Eliciting Honest Feedback Empirical Results for Quadratic and Logarithmic Scoring Rules

Electronic markets face the challenge of motivating honest feedback from strangers who have limited interaction with each other. Miller, Resnick and Zeckhauser [3] propose a mechanism for eliciting truthful evaluations by providing monetary incentives. However, their paper does not present valuable empirical data that would help one better understand their system. For example, knowing how noise and deception affect a buyer’s feedback payment gives insight into the mechanism’s robustness to poor evaluators and prevaricators. Payment distribution data also allow us to assess the strengths and weaknesses of various scoring rules. In this experimental study, we simulate auctions with sequential interaction involving 500 buyers, 5 discrete signal types, and 5 discrete seller types. We generate payment distribution data across 10,000 auctions using the quadratic and logarithmic scoring rules. Additionally we explore the relationship between payment distributions and the degree of buyer noise and dishonesty.