From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!usc!cs.utexas.edu!bcm!aio!gothamcity.jsc.nasa.gov!dlyndes Tue Jan 28 12:15:03 EST 1992
Article 2960 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!usc!cs.utexas.edu!bcm!aio!gothamcity.jsc.nasa.gov!dlyndes
>From: dlyndes@gothamcity.jsc.nasa.gov (David Lyndes)
Newsgroups: comp.ai.philosophy
Subject: Re: Cargo Cult Science
Message-ID: <1992Jan21.182930.8953@aio.jsc.nasa.gov>
Date: 21 Jan 1992 18:29:30 GMT
References: <92Jan15.081805est.14473@neat.cs.toronto.edu> <6030@skye.ed.ac.uk>
Sender: usenet@aio.jsc.nasa.gov (USENET News Client)
Reply-To: dlyndes@deltahp.jsc.nasa.gov
Organization: Barrios Technology @ NASA/JSC; Houston
Lines: 158

I apologize if I've missed anyone's response to my claim that empirical
theories are not refutable.  Our news server went down and we lost quite
a few postings.

Gene Miller writes:
> Newton's theory, like all scientific theories, is refutable, and,
> in fact, was constantly refuted as your analysis makes clear.

When was it refuted?  Was it in the early 1700's when the Grenwich Observatory
observations failed to agree with the predicted orbit?  This forced the
"post-observational fudge" to calculate the orbit taking into account
the orbit of venus.  Later "post-observational fudges" were to include the
orbits of other planets.  More "post-observational fudges" were to perform
more iterations for a more precise approximation of the n-body problem.  Even
more "post-observational fudges" took into account the spins of the planets.
The magnetic fields of the sun and planets were "fudged" in to get a yet better
agreement with observed orbits.  The "force" of the radiation from the
sun was "fudged" in in some calculations.

       [Note that all of these "fudges" are retained in calculations
       using general relativity.  Does that mean that general relativity
       is refuted?]

Does a change to our value of the speed of light due to new measurements
count as a "post-observational fudge?"

How about a more rigourous calculation of the perfect gas law constant
due to high pressure measurements?

If I calculate using ONLY general relativity (not, as Eddington did,
newtonian/euclidean far from the sun but general relativity close by)
in order to get a better match to observation of mercury, is that a fudge?

Was it a "post observational fudge" to posit the existence of nutrinos
to account for the missing mass in certain sub-atomic interactions?  Maybe
we should have scrapped E=mc^2 instead.

Was Heisenberg's exclusion principle a "fudge" in the face of observations
that electrons do not collapse into the nucleus of the atom?

How about neptune's orbit and the search for pluto; was it a "post-
observational fudge" to say "hey, maybe there's another planet out
there tugging on neptune."  That strategy worked for neptune and pluto.
It seemed reasonable to try the same tactic on mercury.

"Fudging" in drag goes a long way toward explaining why boats behave as they do
when they move through water.  It seemed reasonable to "fudge" in aether
drag to help explain mercury's orbit.

      [Please note that the above "questions" are rhetorical.]

> A theory is refuted when post-observational fudging is needed.
> Refutation is ever-present in science.

While I would agree that scientific theories are changed, that the scientists
change their theories due to observations, the changes are NOT necessitated
by REFUTATION.

1) Post obseervational "fudging" goes on all the time.  As the above
   examples should make clear, "fudging" is an essential part
   of the scientific process.  The fudging allows assumptions to
   be made more precise, other factors to be included, etc.  Which factors
   need to be "fudged in" is not (in general) something which can be known
   a-priori.  We can only calculate, check, then "fudge".

2) When post-observational fudging counts against a theory, it is not
   a refutation!  It is effective ONLY when there is an alternative theory
   available AND it at best indicates that one theory MAY be a more fruitful
   path of research than another.

3) The charge of "irrefutability" against pseudo-science is badly misnamed
   and often misapplied.  As Minsky pointed out this weekend, the fact that
   there are no (and can be no) contrary observations should not count
   against an empirical theory which is true.  In addition, "irrefutability"
   should not count against math.  Nor should "irrefutability" automatically
   count against a general methodological statement of a research program
   such as Strong or Weak AI.

   I believe that (and I am prepared to argue for this) a legitimate
   charge of "irrefutability" has more to do with whether the "theory"
   has any explanatory content.

4) I repeat my argument: Nobody (at least nobody in the news which did
   get through to me) said anything about the auxiliary hypotheses, which
   is where my argument gets its force.  So I should probably assume that
   I stated my argument poorly.  I hope this version is more clear.

   A) A theory such as F = k m1 m2 / r^2 makes NO predictions whatsoever.
   B) However, with the addition of auxiliary hypotheses predictions can
      be made.
   C) Auxiliary hypotheses (a.k.a. Boundary Conditions) come in two sorts:

      - Simplifying Assumptions: mostly unstated assumptions about what
        is assumed to have no significant effect such as:  there are no
        other planets, the planets move through a hard vacuum, there are
        no other forces, the planets behave as point-masses, the graduate
        students aren't playing tricks on my equipment, God (or the devil)
        isn't interfering, etc.

        There are ways in which some simplifying assumptions can be
        "pinned down" and I would be happy to discuss these if anyone
        is interested.  But there is no way or set of ways to pin them
        ALL down.

      - Initial Conditions: these include a list of the planets, their
        positions, velocities, masses, etc.

   D) Thus the logical form of a theoretical prediction is

      (T and S and I) implies expected observation O.

      where:
             T is the theory
             S is the set of simplifying assumptions
             I is the set of initial conditions

   E) Failure to find O does NOT imply that T is false.  Any 1st quarter
      logic text will tell you that given:

      1   (1) if (T and S and I) then O      calculated, observed, whatever
      2   (2) not O                          calculated, observed, whatever

      we can conclude:

      1,2 (3) not (T and S and I)            follows from 1 and 2 by modus tollens
      1,2 (4) (not T) or (not S) or (not I)  follows from 3 by demorgan's rule

This argument has nothing to do with experimental error, subjectivity, or
any other "Phil Donohue" style arguments.  Neither does it show that science
is irrational or non-rational.  It is an argument concerning the logical
form of theories-predicting-observations.  It merely shows that there
is a limit to what can be done with observations in an otherwise purely
logico-deductive science.

Note that "not T" does NOT follow.  In fact, it seems to me (given all
the examples at the top of this page) that under normal circumstances
using a successful theory (like newtonian mechanics and gravitation), it
makes MORE sense to "fudge".  That is, it makes more sense to try to
adjust the auxiliary hypotheses than to scrap the theory.
--------------------------------------------------------------------------------
A separate but related argument was suggested by Daryl McCullough.  He
pointed out that typically, T is not a "single theory", but several.  In
which case, we can replace "T" by "T1 and T2 and ... and Tn" in the above
theorem, and the result is:

       1,2 (4) (not T1) or (not T2) or [...] or (not Tn) or (not S) or (not I).

I suppose that we could count "theories" by counting equations, at least for 
the purposes of Daryl's argument).
--------------------------------------------------------------------------------
-- 
+-------------------------------------+---------------------------------------+
| David K. Lyndes                     | "In 50 years, if we're nice to them,  |
| Barrios Technology                  |  computers will keep us around as     |
| email: dlyndes@deltahp.jsc.nasa.gov |  pets." --anon.                       |
+-------------------------------------+---------------------------------------+
| The opinions expressed are not necessarily those of my employer nor of God. |
+-----------------------------------------------------------------------------+


