Re: documentation... more

[Marc Stiegler <marcs>]

Rick's discussion of top-down-ness and bottom-up-ness bring up 
another interesting possibility with Xanadu: the ability to have 
it both ways. We could build 2 documents that cover the same 
material, one bottom up and one top down, using vcopies and resequencing 
in the inclusion list, and probably some version-specific material. 
By traversing a link (or a shared-content connection, or a reference, 
depending on whether we do this with links or vcopies or inclusion 
lists), you could switch contexts (bottom up or top down) for 
whichever particular item you were reading about--thus allowing 
the reader to read "middle-out" from any point in the whole hyperstructure. 
This might be useful, or it might be a nightmare--my head hurts 
when I think about writing it, and such a weaving would involve 
more cycles than you could shake a stick at. But it could be 
interesting to experiment with.

On a different tangent in discussing top-down and bottom-up educational 
tools, I do have one moderately interesting datum. I know of 
one semi-formal experiment investigating the relative success 
of top-down and bottom-up teaching methods. This experiment was 
conducted by one of my advisors while I was in grad school:

All freshman Computer Science majors had to take the same Intro 
to Computing course. There were 4 general sections for this course. 
Dr. Lee taught 2 sections bottom up, starting out with a simple 
simulated computer with an adder and memory and such pieces, 
and taught them how to program in machine language, then in assembler, 
then using a compiler. He taught the other 2 sections top down, 
starting with a compiler, working down to assembler, then discussing 
the hardware.

Both groups covered the same material. However, the groups that 
learned bottom up consistently averaged higher scores throughout 
the course, including the final exam which was identical for 
all 4 sections.

It would be wrong to draw too much of a conclusion from this 
one imperfectly controlled experiment; I myself could construct 
several single-cause explanations for it that have nothing to 
do with top-down and bottom-up teaching. But it was quite striking 
in its conclusions, and I know of no other similar studies. If 
anyone knows of other experiments along these lines I'd be very 
interested in hearing about them.


--marcs