The cited Medium post says many things. Most of them seem to be background or diversions. I did my best to pick out the essential points and discuss them in my

The cited Medium post says many things. Most of them seem to be background or diversions. I did my best to pick out the essential points and discuss them in my comment above. To the extent I could understand, those essential points seemed questionable. For what it's worth, another commenter posted this and I found it much clearer, if still imperfect: https://fairlearn.org/main/user_guide/datasets/boston_housing_data.html https://fairlearn.org/main/user_guide/datasets/boston_housin... The FairLearn authors here explain, like I did above, that the ethical issues depend on use case, and are not inherent to the data or the dataset developers' choice of transformation for the B variable.