Well, if you train a ML-based hiring system on the data you have - and all the top performers just happen to fall within a certain narrow group (for example: Ma

Well, if you train a ML-based hiring system on the data you have - and all the top performers just happen to fall within a certain narrow group (for example: Male, 30s, White or Asians, College educated, etc.), then that said system will most likely predict similar individuals to be good hires - while radically different will get random, or even negative predictions. Imagine it as a movie recommender system. If you only watch a certain type of movies, then that system is going to keep recommending similar movies to you.