Black Magic

Recently,  I was invited along with a few others to interview a data scientist. The candidate was going to present one of his past projects so we got a sense of his skills and experience. Generally, all good ideas except one thing. The candidate was more qualified and credentialed than all of us put together. If I was in his shoes, I may have felt insulted that such a motley crew was going to make a determination about my abilities. While its nice and magnanimous of this person to submit to such unqualified scrutiny, it made me wonder what if any assessment we could possibly make regarding his fitness for purpose. 

This incident also made me think about the state of data science and its disconnect from the lay people whose problems and questions it is supposed to answer. Once the mechanics of how the data is used to arrive at the answer becomes so complex that the person asking the question has no way of validating the approach from a commonsense perspective, the solution is some kind of black magic they need to trust, Furthermore, they need to assume their problem statement was correctly understood, interpreted and translated into the data science space from plain English. They have no assurances that any of that is true.