Why do I like people like me?

In many dimensions the ability to assess knowledge depends critically on the observer's own knowledge of that dimension. Building on this feature, this paper offers both theoretical and empirical evidence showing that, in those tasks where multidisciplinary knowledge is required, evaluations exhibit a similar-to-me effect: candidates who excel in the same dimensions as the evaluator tend to be ranked relatively higher. It is also shown that, if races or genders differ in their distribution of ability, group discrimination will arise unless evaluators (i) are well informed about the extent of intergroup differences and (ii) they may condition their assessments on candidates' group belonging. JEL codes: J71, D82