You have been given the training data and real-valued outcomes of a model trained to predict loan repayment. An outcome of 1 means that the person successfully paid back their loan, where an outcome of O means that the person defaulted. The affiliation feature is a protected attribute. Given these data and results, which of the following is true? Note: Please read the fairness primer for a detailed discussion on these topics.
a) The thresholds {red=0.62, green=0.55, blue=0.67} satisfy equal opportunity
b) The thresholds {red=0.67, green=0.51, blue=0.58} satisfy demographic parity
c) There exists a single threshold solution which also achieves maximum accuracy
d) The thresholds {red=0.59, green=0.50, blue=0.57} achieve maximum accuracy