"All data comes with bias"
An essential moment in this MOOC: The last video by Gaël about ethics in Machine Learning. From plain statistical blunder to insidious reproduction of social prejudice, MLearners have a responsibility to “watch out” what they are doing (thank you for the link to fairlearn.org).
Hence my suggestion:
A couple times in the MOOC, maybe already in chapter 1, replace a few Penguins or Californian homes with a dataset whose processing neatly exemplifies such caveats: poor choice of data - great model - absurd conclusions.