Bureaucracies run through a cycle:
Problem is recognized as new.
Problem is solved quickly in a makeshift way.
Solution is optimized.
Solution is given to a new generation who only understand the solution procedures.
Time passes. The problem changes. The current solution becomes less successful. This is denied, in part to avoid changeover costs.
Solution fails badly enough that it is abandoned. No solution exists.
Problem is recognized as new.
And this continues. Historical example: British mathematics after Newton was paralyzed (according to C.P. Snow, Two Cultures) for perhaps a century after Newton because success in mathematics consisted of passing a standardized test in the Mathematics of Newton's time, and very few were the British contributions to mathematics. Eventually it was revised, and British contributions resumed. So says Snow.
That said, just turning a young person loose on Mathematica hasn't worked for me yet. The language is so, ah, persnickety, and requires so much knowledge of basic mathematics, that the young person can't understand it. ("There are different kinds of numbers? Why should I learn that? What use will this ever be to me?). Any successful approach will have to teach basic mathematics and present very specific Mathematica problems if it is to succeed. This should be a good thing, since an understanding of basic mathematics (at least of where numbers came from, and the problems that show why you need different kinds of numbers) makes things easier. It won't be easy to execute, however.
Remember how there were once tutorial languages such as Pascal? And note that Carnegie Mellon is now teaching only parallel algorithms to freshmen, with the single thread case taken as a degenerate case of the parallel case. Mathematica is comparably different from paper and pencil, hand calculators, and Excel, and will need just as careful a tutorial approach. If that approach exists, I don't know about it.
