Mayank Mahajan

… problem. There are excellent algorithms available for learning linear classifiers (e.g., SVM), but no efficient algorithm is known for learning an arbitrary combination of two linear classifiers.

The lesson is that statistical patterns that apply to the majority might be invalid within a minority group. In fact, a variable that’s positively correlated with the target in the general population, might b…

I say “favor” because majority groups will enjoy higher rates of accuracy in decision-making. In this case, accuracy as a proxy for fairness seems appropriate. A classifier that performs no better than a coin toss when assessing minorities while accurately s…