The selection of test videos (the 5%) could address that -- just pick videos that a fair reviewer would ignore, and an at least slightly biased reviewer would r

The selection of test videos (the 5%) could address that -- just pick videos that a fair reviewer would ignore, and an at least slightly biased reviewer would reject. If a reviewer flags more than X of these videos, one might assume that the reviewer is biased. (X should probably be large enough to account for false positives from fair reviewers).