[PREV - MASTER_META_POP_FALSE]    [TOP]

MASTER_CYLLINDER

                                             March 15, 2024

                                                            MASTER_PARADIGM
     Margaret Masterman, "The Nature of a Paradigm",
     _Criticism and the Growth of Knowledge_ (1970)

In the last four pages of her paper on Kuhn's "paradigm"
concept, Margaret Masterman goes off in an interesting
direction, essentially speculating on the possibility that
a computer program might be capable of identifying new
paradigms... or perhaps, determine the range of extent
of existing ones?


   "Now there are two forms of formal thinking which are relevant to
   the to the analysis of main-feature replication: both of these have
   emerged from the computer sciences.  The first of these, on which
   there is now quite a literature, [p85-2 Parker-Rhodes] is the
   mathematics of classification, or of 'clumps'; i.e. the
   formalization of the process of finding Wittgensteinian families.
   The second of these, on which there is almost no literature, apart
   from the general literature on mechanized pattern recognition,
   [p85-3 Barus] is the set of procedures for making a digital computer
   make an 'inexact match' between two formulae which are highly
   similar to one another, but not quite the same."



   "In both of these methods, the conglomerates of data in question
   have to be characterized by reference to a set of properties with
   regard to which an answer can always be given to the question, 'Has
   this conglomerated this property or not?'"


    "... a strong *prima facie* case could be made for saying that
    this 'inexact matching', if and when it can be achieved, is the
    'replication- relation' which we are looking for.  It is not
    certain in which sense it is a relation: it is reflexive and
    symmetric, for instance, but not transitive (from the fact that A
    has it's main features similar to those of B, and B to those of C,
    it by no means follows that A has its main features similar to
    those of C, unless each replication has an identical P). Thus
    replication-relation logic, in its crude state is a
    one-step-at-a-time logic which never gets off the ground; a logic
    in which the whole putative effort is to see under what
    conditions, and with what weighting, and with what feedback of
    information to change the weighting, and at what cost to the
    richness and completeness of the characterization scheme, a
    limited amount of recursiveness within some particular sequential
    pattern of replications, can be established."


    "It is not even certain that replication is, strictly speaking, a
    form of inference.  I do not see, for instance, how any
    inference-theorem could be proved of it.  In fact, when contrasted
    with normal simple deduction, replication, and controlling
    replications, is logically horrible.  It is however what of all
    things, the human brain in its unconscious recognition-processes
    seems most easily to do; the artificial intelligence men have now
    thrown new light on it [p87-1, Good]; and it is (I think) how
    Kuhn's paradigm extends itself.  Quite a few very simply
    replicating-systems have actually been made; with the field of
    information-retrieval, for instance, any retrieval algorithm which
    has a scale-of-relevance-procedure attached to it counts as a
    replication-system within the description which I have given, as
    does any search-procedure which distinguishes main features ..."


                                           By the way, what do you
                                           do when you do quotations
                                           of text with footnotes?
                                           Me, I'm taking a few liberties.

                     [p85-2 Parker-Rhodes]
                      MASTERMAN_NOTES_PARKER-RHODES

                      [p85-3 Barus]
                      MASTERMAN_NOTES_BARUS


                                           [p87-1, Good]:
                                           MASTERMAN_NOTES_GOOD


--------
[NEXT - MASTERMAN_NOTES_PARKER-RHODES]