I have noticed with computer generated analysis is it lacks built in bias when looking at an enormous amount of data and trying to find relationships. It does not start with "what makes sense" or focus on an expected relationship, it simply starts with a set of data values and then tries to develop predictive probabilities that suggest correlations and possibly causational relationships independent of "what we know" (which often is wrong).

Humans on the other hand, do not have the ability to shut off what they have already learned and so each incremental built of data is interpretated based on the prior biases.