oughts on explanation

CSGnet Archive CSG1993
1

[From Bill Powers (930110.1900)]

Quick note to Gary Cziko and Avery Andrews (930110 or some such)

Explaining how a music box works could go like this:

1. Turn the key until it won't turn any more.

2. Open the lid; this causes the music to play.

3. To stop the music, close the lid.

4. If the music stops by itself, return to step 1.

5. The above explains how the music box makes music.

Or like this:

The key on the music box winds a spring. When the lid is opened,
a catch is released and the spring turns a drum with little pins
sticking out of it. The pins bend and release flat springs, each
spring making a different sound. This explains what causes the
music to play.

Explaining why some one hates policemen could go like this:

1. When John was a teenager, he was arrested for a crime he
didn't commit.

2. The arresting officer was brutal to John.

3. John was found guilty on the testimony of the officer and had
to pay a large fine.

4. This explains why John hates policemen.

Or like this:

1. John believes that policemen arrest people for things they
didn't do.

2. John expects policemen to be brutal.

3. This explains why John hates policemen.

In each case, the second set of explanations is couched in terms
of the way a system is currently organized, while the first is
couched in terms of antecedent events. The first type of
explanation is descriptive; the second generative.

ยทยทยท

---------------------------------------------------------------
Best,

Bill P.

2

[Martin Taylor 930112 12:00]
(Bill Powers 930110.1900)

Quick note to Gary Cziko and Avery Andrews (930110 or some such)
Explaining how a music box works could go like this:
1. Turn the key until it won't turn any more.
2. Open the lid; this causes the music to play.
3. To stop the music, close the lid.
4. If the music stops by itself, return to step 1.
5. The above explains how the music box makes music.

Or like this:

The key on the music box winds a spring. When the lid is opened,
a catch is released and the spring turns a drum with little pins
sticking out of it. The pins bend and release flat springs, each
spring making a different sound. This explains what causes the
music to play.
....
In each case, the second set of explanations is couched in terms
of the way a system is currently organized, while the first is
couched in terms of antecedent events. The first type of
explanation is descriptive; the second generative.

There's a more general way of looking at the two descriptions. The first
one relates only to music boxes, whereas the second deals with objects
whose properties are known to the student/listener/trainee (SLT) in other
circumstances. The SLT is presumed to know what a "spring" is and does,
so the only new information brought to bear is that there is one in the
music box, and it can be wound (also a known concept) by the key. Springs
are known to unwind unless held by something, so the new information is
only what the detainer is. And so forth.

The information in the numbered set of instructions cannot be used outside
the world of music boxes. It would be effective with a mechanical moron,
if the intention were to allow the moron to play the music box. The second
would be useless with the moron, since it depends on the SLT having a good
idea as to the mechanical and acoustic functions of many different things.
To someone with that knowledge, the second set of descriptions provides
rather less information about the music box as such than does the first.
But it integrates the music box into a wide range of other tools and
instruments (such as clocks and tuning forks. It doesn't "explain" why
or how the box produces music. It only says that if you can explain why
a clock works or why a tuning fork works, you can explain why the music
box works. But can you "explain" those things? I think not. You can
only relate them to other things.

What you call "explanation" I would call description in terms of concepts
with a range of applicability beyond the immediate problem, and particulary
in terms of concepts already understood in some way by the SLT. The main
point about "explanation" is that it uses a great deal less information
or encompasses a much wider range of applicability than does "description."
I see no qualitative difference between the two concepts, only a quantitative
difference.

The same is true of "descriptive" vs. "generative" models. Generative models
provide descriptions over a much wider presumed range of applicability,
and often use fewer independent pieces of information (using the term
colloquially) to do it.

Martin