On 5/14/02 12:00 PM, "Hal E. Fulton" <hal9000 / hypermetrics.com> wrote:

> ----- Original Message -----
> From: "Chris Gehlker" <gehlker / fastq.com>
> To: "ruby-talk ML" <ruby-talk / ruby-lang.org>
> Sent: Monday, May 13, 2002 11:10 PM
> Subject: Re: OT:is software eng an art?
> 
> Now we're WOT (Way Off Topic)...  :)

Yeah! But this is kind of fun.

[big snip]

> 
> That's an interesting tally, but it doesn't address the issue of whether the
> actual experiements are valid. I'd suggest that even though you try to
> follow
> the scientific method, the more variables you ignore, the softer science you
> are. In some fields, there is an infinity (well, not literally) of
> circumstances and
> data that are ignored or are unmeasurable/uncollectable.
> 
> Constructing an experiment and performing measurements become harder
> and harder as the science gets softer; and the results become less and less
> meaningful.
> 
> Of course, as the psychologist pointed out in Robinson's _Green Mars_, all
> sciences were softer at one time. We do what we can. Come back in three
> hundred years and let's talk...

I see what you are getting at. If the nature of what you are studying such
that epsilon on all your models overwhelms the predictive variables, the
science is soft.

By this metric, I think a lot of the sciences that study groups of people
are a lot "harder" than you suspect. I had a chance to work with a PhD in
Social Psychology for McDonald's who fund a lot of academic work in that
area because they apply the results in their market research. When they
introduce a specialty product, they *know* within a few percent how many
units that they are going to sell at each store. And yes they do have a
formal, mathematical model.

I find that somewhat disturbing. I don't want that science to get better.

I think that I heard neither the Wharton nor the Bookings models have
misforecast GDP by as much a a percent since the Korean war. (But GDP has a
lot on inertia, so maybe that's not such a feat).

Then there are all those very accurate production cost models that the
micro-economists build (but they are predicting the behavior of machines,
more than people)

And the fact that traffic obeys the "laws" of fluid dynamics.

Anyway, thanks for responding.
-- 
C++: The power, elegance and simplicity of a hand grenade.