A lot of the advanced applications of algorithms are using deep neural networks. Unfortunately, it's notoriously difficult to figure out specifically why a neural network did what it did. You just end up with "because the algorithm said so, that's why."
