Stanley Mulaik
2005-10-05 01:56:06 UTC
Hi, Stan, thanks for your comments about abduction and IBE. Just
like every theory, there are many perspectives to interpret the
same thing.
It's quite common for debates to turn on the use of certain termslike every theory, there are many perspectives to interpret the
same thing.
rather than on the concepts themselves. The aim is to achieve
agreement or consensus on how to use the terms. Often priority
gives rights to the use of a certain term with a certain concept,
but not always, since popularity of a later concept with the
same term may establish rights for its use with that concept when
the earlier usage is not well-known. But I think Peirce's use
of the term abductive is well established and that takes in
the use of the phrase "inference to the best explanation".
According to Regt, abduction and IBE are separate in
the context of realism and anti-realism.
I would say this is a different context than Peirce uses thethe context of realism and anti-realism.
term, and a narrower one as well.
Realism of course seeks a final and absolute truth about the
world. Even Peirce suggested such a possibility, although he
saw science as iterative. He thought the iterations would
converge to a final 'truth' in the far distant future.
I personally don't think there is a need to believe that,
nor could we tell whether it does or doesn't. We can still
have 'objective knowledge' that is not absolute but always
provisional. What to me makes it possibile for the iterations
to never end is the possibility that we will always gain either
better and finer grain measurements that will reveal many new
heretofore unobserved things or travel to discover many new
phenomena, or observe (with telescopes) many new phenomena, etc.
so that there will always be newer phenomena that we
will have to incorporate into our theories, and quite possibily
they may not be adequately accounted for by our current
theories, demanding revisions. Thus for me knowledge is always
provisional, but objective by satisfying criteria of objectivity
to achieve provisional objective status.
Obviously I am not an ontological realist, but I am a critical
realist in that I believe we know the world directly in terms
of objects bearing attributes and having causal connections
among them. Maybe this is 'formal realism' as opposed to ontological
realism that believes in a final and unique truth. Formal
realism is that we experience the world in the forms of objects
bearing attributes and having causal connections among them. But
our experiences are always limited and confined to specific points
of view, meaning they don't encompass everything and can't be final,
so I doubt the usefulness of 'ontological realism' and absolute truth.
Objects are invariants in experience independent of changes due
to our own actions, perspectives, points of view, frameworks of
thought. That contrasts with phenomenalism which is that we
gain knowledge beginning with subjective experience of the
senses described in terms of things like 'sense data',
'colored patches' etc..
IBE is a research
tradition advocated by Harman in the mid of 20th century.
Although abduction and IBE share certain characteristics, Regt
asserted that "Peirce had nothing to do with this wrong-headed
1. Abduction is at the first stage of inquiry; IBE is at the
final stage for selecting a theory/model.
2. Abduction suggests a list of plausible routes of inquiry; IBE
picks one model only and hence the best explanation becomes the
only explanation: IBE --> IOE.
Regt, H. (1994). representing the world by scientific theories.
Tilburg University.
Well, I could not be an ontological realist, and that means Itradition advocated by Harman in the mid of 20th century.
Although abduction and IBE share certain characteristics, Regt
asserted that "Peirce had nothing to do with this wrong-headed
1. Abduction is at the first stage of inquiry; IBE is at the
final stage for selecting a theory/model.
2. Abduction suggests a list of plausible routes of inquiry; IBE
picks one model only and hence the best explanation becomes the
only explanation: IBE --> IOE.
Regt, H. (1994). representing the world by scientific theories.
Tilburg University.
could put the 'best explanation' as the product of the abductive
phase, as well as consider that there may be abductive paths
that different scientists take, and these may lead to the same
new data, and thus the one that fits while the others don't
becomes the 'best explanation'. This is a different
'best explanation'. And I don't think it ends there either.
The cycle starts anew with new phenomena discovered.
Although his context is about realism and anti-realism, I found
it germane to other situations. IBE assume that we have
"qualified" candidates and we just pick the best. But there is a
well-known argument for that: Bad lot agrument. What if everyone
in the pool is not "qualified"? To me TETRAD in the current form
(Glymour said hypothesis generation and testing can be conducted
along) does not distinguish abduction from IBE, which is like a
kind of one-step approach or a strong version of IBE.
objective truth'. We don't discover objective truth. Too much
has been telescoped into this idea, so that you don't see the
Peircean three stages contained within them. You have to have
the induction or generalization to new or different data to
establish invariants required by objectivity. TETRAD is still
working with just one data set.
Could it be that there are two 'best explanations' here? Peirce's
abduction to the best explanation begins with a new phenomenon,
which we then seek to understand, perhaps using prior experience,
but also using (metaphoric) schemas taken from perceptual experience,
with which we may create new concepts and theories to account for
the new phenomenon. It will be possibile to generate many of these,
and we will go through a process of trying to reproduce the data from
them, discarding those that don't fit, until finally we arrive at one
that fits (maybe perfectly). That would be the first 'best explanation'.
Peirce's theory of science involves three stages that are repeated
over and over again:
Abduction (just described), Deduction, and Induction.
Deduction is deriving some consequence (but not necessarily in a
formal manner using formal logic, but perhaps using the consequences
of certain embodied metaphoric schemas taken from perceptual
experience that underlie our original 'best explanation'). The
consequence may be some new thing we may predict will be observed
or found. This leads then to the next phase:
Induction. Induction is generalizing from particulars to other
particulars. Here the particulars from the initial phenomena
are joined with new particulars predicted in the deductive phase.
And so one goes out and gets the new kind of data that is predicted
and sees whether the prediction is actually upheld. That is a test
of the 'best explanation'. It is inductive because we get the actual
new data and see if we can generalize to it in the specified way.
If our prediction is upheld that supports the original 'best
explanation'. If not, we may go back to the drawing board and see
why the prediction didn't work. We enter a new abductive phase.
We may use everything we've gained in the way of observations up
to this point in this new abductive phase for a new cycle. But
we will have to make 'deductions' or, more weakly, 'derivations'
that lead to getting new data against which to test them.
Peirce, I believe, didn't say that science is about invariants,
but it is implicit in the inductive phase of his process. Making
an inductive generalization is asserting an invariant--in the
medium or process that links the observations to one another.
It also implies parsimony, because we have more observed entities
than entities used to account for them in having new data added
to the old being accounted for by the same theoretical entities
developed originally with the old data.
Often when people formulate models that are not especially conscious
of prior experiences or knowledge that they have used to formulate
them, or of numerous models discarded on the way, so they aren't fully
aware of the fact that they have gone through an abductive and deductive
phase to generate their model. But by testing their model they are
indeed testing an invariant that unites both the current data and
all prior data on which the model was based.
Back to the point that I left off last time about Goodman. Ian
Hacking framed the discussion of "construct" in terms of
Goodman's New Riddle in his book "Social construction of what."
Different ways of conceptualizing constructs at different time
(Bleen and Grue) may lead to opposite conclusions about the same
phenonmenon. A good structural model does not gaurantee capturing
the causal structure while the constructs are problematic. That's
why I think it is important to go back and forth across
multi-steps and always employ an abductive mind in all levels of
inquiry. In this sense, abductive reasoning is not only
applicable to the first stage, but also to the final stage!
equivalent models based on different structural frameworks.
I'm a real fan of the idea of the 'a priori', i.e. schemas that
organize experience in certain ways, providing frameworks that
are neither true nor false by themselves, but only in terms of
additional constraints imposed on them relative to specific
forms of data. Mathematics provides numerous different ways
of modeling the same data, and many of these will do so
to the same degree of fit. Rotations in factor analysis will
yield different solutions that all reproduce the data equivalently.
In SEM having models with free parameters in them gives rise
to the possibility of many equivalent models for the same data.
Philosophers in the 1980's became very much aware of these
same possibilities. To me SEM encompasses most of the issues
of philosophy of science in a concrete form that can be readily
grasped. Philosophers of science need to look at SEM.
I'm not sure whether science simply begins with one formal
framework that happens to best fit the data and continues to
do so, but could have begun with some other framework. Or
is it possibile that eventually some frameworks get cancelled
out as they get extended to new forms of data, so that there is
a process of survival of the fittest that reduces the number
of competing alternative frameworks? That will take more
thought to resolve.
Stan Mulaik
--------------------------------------------------------------
To unsubscribe from SEMNET, send email to ***@bama.ua.edu
with the body of the message as: SIGNOFF SEMNET
Search the archives at http://bama.ua.edu/archives/semnet.html