Blayne Haggart argues that in order to use machine learning effectively, we must update the scientific worldview to be more aligned with what scholars call "dataism."