Let’s say you are seeking to maximize variable x which is inversely correlated with variable y. If there is some practical reason that variable y is more constrained than than variable x, then it makes sense to maximize variable x since your marginal decrease in variable y will be reduced.