Just had to review a paper that used a black box model as a magic solution to a complex problem, with no effort to understand or interpret the results. having a model that "works" but can't be explained or improved is not actually working.