I would argue that understanding the underlying distribution is useless when looking at a single datum (a man or a woman) but critical for understanding things

I would argue that understanding the underlying distribution is useless when looking at a single datum (a man or a woman) but critical for understanding things en masse. The author's point is that we can't be so quick to judge seemingly unfair trends or groupings which might actually be the product of statistics. He admits readily that there are exceptions to be found on all fronts.