I study how an unbiased and uninformed decision maker should design a mechanism that provides enough incentives for possibly biased agents to share their information impartially. In the first part of the paper, I provide a general characterization of the optimal mechanisms under a strong independence assumption. In the second part of the paper, I study gender discrimination in recruitment as an application of the general model. I define bias-proof mechanisms as mechanisms such that the report of each one of the various recruiters of the firm does not change the expected number of women hired. I then use the results from the first part of the paper to show that the optimal bias-proof mechanism is a flexible quota mechanism. Unlike fixed gender-quotas, flexible quotas allow for the number of women hired to depend on the recruiters' report. I find that fixed quotas are only optimal in very special circumstances. Furthermore, while, by definition of bias-proofness, the recruiters' gender does not matter, the gender distribution of the set of applicants that each reviewer analyses does matter. In particular, I find that it is best to have gender diversity within each set of applicants, i.e., if possible, it is best that each reviewer - male or female - reviews the same number of female and male candidates rather than having some reviewers review only women while other reviewers review only men.