Howard Zinn pointed out that History is always biased. In my own work, I've noticed that Machine Learning is always biased. Basically, any historian picks out which facts to discuss, and is not immune from social pressures when they interpret those facts. Similarly, after running enough machine learning algorithms, I realized that the features measured, the way they are measured, and their relationship to other features is considered "factual raw data" - yet it is the very essence of subjective while often ignoring the greater and more important reality.