373e4c5d8edfa8b74fd4b6791d0cf6dc-AuthorFeedback.pdf

We believe that label bias is omitted frequently enough that it affects25 people'sthinkingonmattersoffairness. Thanks for pointing out our error about the Fish et al paper, which indeed does use simulated data to get access31 to the unbiased labels. Maybe this36 is obvious, but the idea was initially met with criticism because of this purported obstacle.