Article in Internet Citizen. Good piece. I remember reading the classic 'How to Lie with Statistics', by Darrell Huff. It was not a 'how to', but rather a how not to get fooled by simple statistics and numerical arguments. Though simplistic, it still reads well. This article makes a similar case, but takes it into the realm of big data and deep learning. In a world with much more available data. So biases are much easier to support today. And they can be supported not only at the basic statistical level, but also at a much deeper algorithmic level, that is harder to explain or validate than a bar chart.