NetNews Usenet Archive 1992 #27/NN_1992_27.iso/spool/sci/space/16223

Text File | 1992-11-21 | 3.5 KB | 68 lines

Newsgroups: sci.space
Path: sparky!uunet!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!cis.ohio-state.edu!news.sei.cmu.edu!fs7.ece.cmu.edu!crabapple.srv.cs.cmu.edu![e-mail]
From: [e-mail] (John Roberts)
Subject: Re: Scientific method
Message-ID: <[e-mail]>
X-Added: Forwarded by Space Digest
Sender: [e-mail]
Organization: National Institute of Standards and Technology formerly National Bureau of Standards
Original-Sender: [e-mail]
Distribution: sci
Date: Sat, 21 Nov 1992 18:43:30 GMT
Approved: bboard-news_gateway
Lines: 53


-From: [e-mail] (Henry Spencer)
-Subject: Re: Scientific method
-Date: 20 Nov 92 18:38:27 GMT

-In article <[e-mail]> [e-mail] (John Roberts) writes:
->... for instance, for instance, the Earth-impact model of the formation
->of the moon has risen from obscurity to the "most favored model", with
->(as far as I know) little or no new input of information - it's based on
->mathematical models and old Apollo and Voyager data...

-There is no problem testing a new theory quite rigorously using old data,
-if you do it carefully.  The trick is simply to get some testable predictions
-out of the theory before you look (closely) at the data, and then see if it
-checks.  This does get more difficult if the new theory has to be calibrated
-using the same data, but sometimes it can still be done.  It is more
-satisfying to have prediction precede experiment, because that *guarantees*
-that the theory was not custom-cooked to match the results,

I agree it can be done, but there are some potential difficulties that
investigators have to watch out for. Most obvious is "cheating" - theorists
making use of information that they pretend not to have when constructing
their models. To prevent that, somebody might get the idea of keeping
significant portions of the observational data "secret" - to be released
only after theories have been formulated. This of course would inhibit
the distribution of information (slowing the development of better theories),
and bring about the risk of "insider trading" of information ("so *that's*
why Fred Smith keeps coming up with the best theories"). I'm also not sure
how modifications of theories should be regarded in this context - if you
come up with a theory based on partial information, and analysis of more of
the data shows that some aspect of the theory needs to be modified, does
that mean that the theory is now considered questionable until it can be
checked against still more data? If so, there has to be some established
mechanism of releasing the data in small portions, so that a theory can be
taken through several iterations of evolution. But how can we be sure that
the decision on what data to release each time doesn't introduce biases into
the direction of development of the theory?

-but having a
-theory derived from general principles precisely explain measured phenomena
-in detail is a valid test, and often a fairly good one.

That seems to be the most straightforward approach in this context. I would
guess that in *most* cases, doing it this way rather than trying to force
in into the traditional mold would give the best results. (Spinoffs from
other research may provide additional confirmation anyway.) Of course, as you
say, the reason for the traditional method is to guard against researchers
inadvertently convincing themselves that events *had* to take place a certain
way because that's what was observed.

John Roberts
[e-mail]