When using the Pearson r correlation coefficient (as well as virtually all other statistical procedures), it is significant to pay attention to the sign of a coefficient (plus or minus). Why, then do we ignore the sign when interpreting the point-biserial correlation coefficient?