Their model gets flipped inside out because it shifts the blame from self to someone else. The summary conclusion of the model is seen as the most important attribute in the cost function, thus the model is kept, wrongly. The lack of resonance around the graph nodes and edges is ignored as inconsequential, however they should be seen as evidence that your cost function is rewarding steps taken in the wrong direction.
Too much "ends justify the means", and not enough "truth for truths sake". It was better when you just hurt yourself for handouts.