Erratum to: Prediction Mechanisms That Do Not Incentivize Undesirable Actions

No Abstract

Definition 10 of the paper entitled “Prediction Mechanisms That Do Not Incentivize Undesirable Actions” contains an error. Here is the correction:

Definition 10. We say a proper scoring rule provides incentive c if the agent cannot guarantee within c of the optimal expected payment by giving some constant dummy report $\vec{r}$. (More precisely, ∀ dummy report $\vec{r}$, ∃ distribution $\vec{p}$ under which reporting truthfully instead of reporting the dummy value $\vec{r}$ pays off by at least c: $\tilde{S}(\vec{p},\vec{p}) - \tilde{S}(\vec{r},\vec{p}) \ge c$.) A one-round prediction mechanism guarantees incentive c if for each agent j and each combination of others’ reports r _− j , the corresponding proper scoring rule provides incentive c.

This work is supported by NSF IIS-0812113, the Sloan Foundation, and a Yahoo! Faculty Research Grant. We thank the anonymous reviewers for helpful comments.

The original online version for this chapter can be found at http://dx.doi.org/10.1007/978-3-642-10841-9_10