[disgruntledphd2]

As long as you can get a reason from the model, it's not that bad.

Black box automatic decision making is much more problematic.