A method to assess trustworthiness of machine coding at scale

I would be interested to hear others opinion of this paper.