This sounds like something new that Harvard has spotted or stumbled across, whereas the lack of model understanding (ie why does it predict, or cluster or group or features relate) is the primary reason such methods have not been adopted in safety-critical engineering for at least the 5 years I have been interested in Data Science.

Data Scientist and Chartered Aeronautical Engineer (MEng CEng EUR ING MRAeS) with over 15 years experience in the Aerospace, Defence and Rail Industry.