Re: Mathematica courses in College

*To*: MathGroup at yoda.physics.unc.edu*Subject*: Re: Mathematica courses in College*From*: fateman at peoplesparc.berkeley.edu (Richard Fateman)*Date*: Thu, 19 Aug 93 08:54:53 PDT

I think it is a mistake to give a course on Mathematica per se, for college credit, except perhaps for some nominal amount. I feel the same way about Fortran, COBOL, Lisp, Matlab, Maple, ... A course in "Problem solving with Mathematica" which is what Nancy Blachman teaches, may be justified if the emphasis is on Problem Solving. I teach a course "Structure and Interpretation of Computer Programs" which some people mistakenly view as "the Scheme course" because students read and write programs written in Scheme. But that is not the point of it. I think that the accreditation teams who periodically visit colleges have a similar view. I would hope that they would be quite unhappy with a CS degree assembled from 1 unit of "Programming in Assembler for 8086" 1 unit of "Programming for Sparc" ... 1 unit of "Programming in Fortran" 1 unit of "Programming in ... " ... 1 unit of "Programming in Lotus 123" ... Now if you are going to award 1 unit for learning (the first, presumably) programming language, should that language be Mathematica? For CS students, I think most CS faculty would answer, resoundingly, NO! For non-CS but engineering/physics students, there is perhaps more of a debate. These are the people who, until recently, and maybe now, are taught Fortran 77. I think they should be taught Matlab or similar languages, and I have also argued for introducing symbolic manipulation. Ideally, this would be in a course on Problem Solving, or Scientific Computation, or Numerical Computation (not usually "Numerical Analysis" which is more mathematical in orientation) or Programming per se. In spite of a recent book or two, I think that Mathematica's treatment of numbers is so at odds with conventional numerical error treatment, its algorithm speeds at odds with usual analysis, and its bag of tricks so undocumented, that a course using it for numerical analysis would have to ignore or explain away too many anomalies to be comfortable. Why should a professor have to spend time on this? Your comfort may vary. RJF