Dealing With Intransitivity, Non-Convexity, And Algorithmic Bias In Preference Learning