On Algorithmic Fairness and Stochastic Models for Combinatorial Optimization and Unsupervised Machine Learning