From Trapped in the Virtual Classroom by David Bromwich, in the July 9, 2015 issue of the New York Review of Books [my boldface]:
When H.L. Mencken, an avowed atheist, was asked if he believed in baptism, he replied “Believe in it? I’ve seen it done!” For thirty-eight years, I’ve been a teacher of a discipline of interpretation that is fostered in university departments of literature, philosophy, the history of ideas, and to some extent psychology and political science, a discipline that might best be described as an art that can be taught; and if someone asks, “Do you believe in it?,” I can’t do better than “Believe in it? I’ve seen it done.” The discipline I have in mind is not a religious ritual but an educational practice that can seem to the uninitiated as monkish as some aspects of an alien religion.
This discipline can’t be called a science as a natural scientist would understand the term. It is not progressive. It has few formulas worth memorizing. It doesn’t show its glory in “research programs,” airtight systems of classification, or pathbreaking discoveries that refute a previous century’s scientific orthodoxy. Yet the discipline does have perceptible boundaries and teachable methods of confirmation or falsification, based in large part on the relationship of evidence to assertion, and on the common sense and insight of qualified judges. It imparts a fund of practical wisdom that grows over time and alters its emphasis with the disposition and character of the practitioner.
Is mathematics more like “science as a natural scientist would understand the term” or is it a “discipline of interpretation?” Bromwich’s article is about mechanization of teaching, but his comments, for example
Does this movement cooperate with a pressure to make human life ever more machinelike? In the process of approval and acceptance, are we being asked to conceive of knowledge itself as mainly constituted by information? And does knowledge come to be seen therefore as a social good that can be disseminated and assimilated in a uniform and mechanical way, so that finally the amount of good accomplished is to be judged by criteria of efficient satisfaction