1 paper
Mengya Hu, Susie Park, Suzana Ilic +5
Content-moderation classifiers are usually evaluated in isolation, but deployment requires choosing where to intervene and what follows a flag. We evaluate these choices using two…