Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site fortune.UUCP
Path: utzoo!linus!security!genrad!mit-eddie!mit-vax!eagle!mhuxl!ihnp4!fortune!rpw3
From: rpw3@fortune.UUCP
Newsgroups: net.ai
Subject: Re: Information sciences vs. physical sc - (nf)
Message-ID: <2073@fortune.UUCP>
Date: Wed, 21-Dec-83 04:02:24 EST
Article-I.D.: fortune.2073
Posted: Wed Dec 21 04:02:24 1983
Date-Received: Thu, 22-Dec-83 04:29:22 EST
Sender: notes@fortune.UUCP
Organization: Fortune Systems, Redwood City, CA
Lines: 55

#R:sdcsvax:-8400:fortune:21500003:000:2367
fortune!rpw3    Dec 20 20:15:00 1983

I don't think you can limit the "observational sciences" by saying they
cannot do experiments. What they cannot do is experiments where the
experimentor has a high degree of perceived control over the total
environment of the experiment. (In fact, because of that, sometimes
they can do better work because they are not fooling themselves with
the illusion of a "controlled" experiment.)

The key to an "experiment" is that there is a theory relating a quantity
of previously observed data which PREDICTS that if certain as-yet-unobserved
data were taken, the result would be so-and-so. The experiment consists
of taking such NEW data (not correlated with previous data) and,
by analysis, confirming/rejecting the hypothesis. Notice that nowhere
is there the necessity that the experimentor be in control of any of
the data.

While it is very, very true that it is easier to make advances in a
field of study if the subject permits the experimentor to tightly
control one or more independent/orthogonal(izable) observables
which can be shown to account for the vast majority of the standard
error of the observed data, it is not "necessary" for science.

If one's universe supplies a sufficient diversity of sample populations
(stars, cultures, etc.), good science can be done without experimentor
"control". Conversely, terribly sloppy "science" can be (and has been)
done when the experimentor assumes that he/she knows what the "independent"
variables are and that the experiment "controls" them.

The classic example (which was fortunately rescued by the experimentor's
insistence to the funding organization that more data was needed) is
Elton Mayo's studies of production at the Western Electric plant at
Hawthorne. It's so famous that it's now known as the "Hawthorne effect"
(why not Mayo? who knows?).

	Turn the lights UP, production goes up...
	Turn the music on, production goes up...
	Introduce coffee breaks, production goes up...

	then...

	Turn the lights DOWN, production goes UP...
	Turn the music OFF, production goes UP...
	Take away coffee breaks, production goes UP...

Proper experiment design is FAR more critical than picky catagorizations
of what is and is not science.

Rob Warnock

UUCP:	{sri-unix,amd70,hpda,harpo,ihnp4,allegra}!fortune!rpw3
DDD:	(415)595-8444
USPS:	Fortune Systems Corp, 101 Twin Dolphins Drive, Redwood City, CA 94065