14.0642 black-box vs glass-box methods

From: by way of Willard McCarty (willard@lists.village.Virginia.EDU)
Date: Mon Feb 05 2001 - 15:24:44 EST

  • Next message: by way of Willard McCarty: "14.0643 Corpus Linguistics 2001"

                   Humanist Discussion Group, Vol. 14, No. 642.
           Centre for Computing in the Humanities, King's College London

       [1] From: lachance@chass.utoronto.ca (Francois Lachance) (40)
             Subject: cross-dressing or the container contained

       [2] From: "Mary Dee Harris" <mdharris@acm.org> (29)
             Subject: Re: 14.0638 black-box vs glass-box methods?

       [3] From: Willard McCarty <willard.mccarty@kcl.ac.uk> (51)
             Subject: black-box vs glass-box methods

             Date: Mon, 05 Feb 2001 20:21:53 +0000
             From: lachance@chass.utoronto.ca (Francois Lachance)
             Subject: cross-dressing or the container contained


    "The temptation/ To take the precious things we have apart/ To see how
    they work" sings Billy Bragg on _Workers Playtime_ on the track entitled
    "Must I Paint You a Picture" (P) 1988 Go! Discs (C) 1988 Billy Bragg [A
    musicologist or a librarian may be able to explain the (P) -- I'm new at
    the practice of citing music lyrics.]

    Yes indeed the text is a black box. A two black box problem. Like a string
    of pearls when neither object of study nor methodology is transparent.

    The metaphorics of transparency are sometimes refractory. Consider text as
    glass box -- a dazzling strand of glittering gems. How is it built? The
    quality of light as it passes through... the prismatic moment of wonder.

    There is the transmission of a Christian text which has always puzzled me
    and I've not done the philological work to explicate "through a glass
    darkly". I have however pondered the importance of scale: obsidian thick
    cut to act as a mirror; thinner, a tool to view the eclipse and spare the
    eye. And what of the cyberpunk mirror shades?

    Without due recourse to the source of the lighting, the metaphorics of
    transparency privilege knowledge in the synoptic mode. If neither glass
    box nor not-glass box receives no illumination, the difference is moot.

    History is built of a mix of senses and a braid of time. Billy Bragg
    opens his song with this complexity simply stated "It's bad timing and me/
    We find out a lot of things out this way/ And there's you/ A little black
    cloud in a dress" -- not a black box, not a dark continent, but...

    The other questions follow: who congregates around the black boxes and who
    surrounds the glass boxes and who runs the course between them? when?

    David L. Hoover who gave us pause to reflect upon the black box nature of
    text may perhaps provide a further gloss over a glass of a dark brew given
    that the Institute for the History of the Production of Knowledge is
    associcated with the English department at NYU and the Association for
    Computers and the Humanities & Association for Literary and Linguistic
    Computing is holding its Conference at NYU, June 13-17, 2001. I'm sure
    many glasses will be half full and half empty and emptied and refilled
    before the question of the nature of containers and contained is settled.

    I just hope we here who will not be there then got to hear by here reading


    Francois Lachance, Scholar-at-large
    Member of the Evelyn Letters Project

             Date: Mon, 05 Feb 2001 20:22:20 +0000
             From: "Mary Dee Harris" <mdharris@acm.org>
             Subject: Re: 14.0638 black-box vs glass-box methods?

    Until fairly recently, I was fairly skeptical about a lot of the statistical
    methods being used in Natural Language Processing because they do seem to be
    'black box' methods. But in the last couple of years, I have changed my
    mind, after investigating the results and learning more about the methods.
    It is an interesting phenomenon in that the researchers themselves often
    understand the statistical methods well, but not why they show the results
    that they do. In other words, they know the mathematics of the procedure,
    but can't explain why these features revealed by the statistics are being

    For example, several of the methods employ the 'bag of words' method of
    taking all the words in a text and treating them as if they had been thrown
    into a bag. In that method, word order and cooccurrence are no longer
    considered. But with even a cursory knowledge of linguistics, we KNOW that
    word order and cooccurrence influence the meaning of language, so how is it
    that ignoring those features doesn't change the results of the analysis?

    I also want to respond to part of David Hoover's message, "To require that
    the researcher become a theorist of humanities computing in order to be
    considered to be doing humanities computing seems to me (to analogize
    further) rather like requiring that all scientists be historians of
    science." I would disagree with his analogy in that theory and history are
    two very different types of knowledge. Theory implies we know how something
    works while history imples that we know how we got to the point of using
    that method. The two might be related but are definitely not the same

    I do agree with David Hoover though that one doesn't necessarily need to
    know the theory of humanities computing to be able to do it. While one
    would be able to do more innovative research if the theory were understood,
    there are plenty of mechanical methods one can follow to produce results.

    Mary Dee Harris
    UT Austin

             Date: Mon, 05 Feb 2001 20:22:38 +0000
             From: Willard McCarty <willard.mccarty@kcl.ac.uk>
             Subject: black-box vs glass-box methods

    I'm grateful to Chris McMahon for establishing what seems the philosophical
    end-point of the question I raised, and to David Hoover for laying out much
    of my own several-mindedness about it. I certainly agree with him that
    setting up stringent requirements for humanities computing is
    conterproductive. At that end of the question I'm interested in description
    rather than prescription, with the qualification that we should be able now
    to identify directions and tendencies that are more and others that are
    less productive.

    At the opposite end from Chris M's Black Box of Being is what seems to me
    at first blush to be an anti-intellectual attitude and practice, though as
    I've observed it, the anti-intellectualism falls only in our common
    bailiwick. Again, let me use a fictitious example, this time a contrast
    between two musicologists.

    Musicologist A applies some piece of software to a sequence of notes. Let's
    say the transformation thus effected stimulates him or her to think great
    thoughts about this music. Our musicologist pays no attention whatever to
    the transforming process itself, only to the result, in fact when asked
    treats the enquiry as an impertinence, as totally irrelevant -- that, A
    says, is a purely technical matter outside his or her domain. Not
    interested. A black box.

    Musicologist B uses the same software on the same music, also is greatly
    inspired to think new and great thoughts about the music. In contrast with
    A, however, B goes to X, a colleague in computer science. Together they sit
    down with the transformation algorithm to see how it does what it does. B
    and X then work together (or X teaches B enough about the algorithm that B
    can then work on the transformation process him- or herself); the result is
    -- let us be generous -- another transformation and a set of insights into
    the music that come from understanding the former process applied. A glass box.

    It seems to me descriptively that A, who is using the computer as "just a
    tool", is not doing humanities computing (or "digital scholarship" or
    whatever else we choose to call it), however good the resulting musicology
    may be. In contrast B is, and it seems so is X. Certainly it's fair that
    someone specialising in a discipline keep his or her nose to the
    disciplinary grindstone. But since A is using the music-transformation
    software no more profoundly than he or she uses a wordprocessor or e-mail
    program when he or she writes out great scholarly thoughts, I cannot see
    anything of interest for us here intellectually.

    What we do about the situation socio-politically, institutionally, is
    another matter. If we're to clarify our practice, however, it seems to me
    unavoidable, indeed desirable, that we be able to discriminate between what
    does and does not belong under the rubric of the interdiscipline. Or if it
    is better to be all-inclusive, that we be able to tell where the food is
    and where it isn't. What I think we need to avoid is the attitude that it
    simply does not matter, that anything goes.



    Dr Willard McCarty / Senior Lecturer /
    Centre for Computing in the Humanities / King's College London /
    Strand / London WC2R 2LS / U.K. /
    +44 (0)20 7848-2784 / ilex.cc.kcl.ac.uk/wlm/

    This archive was generated by hypermail 2b30 : Mon Feb 05 2001 - 16:14:38 EST