[ontolog] Re: Knowledge Representation/Access
Great post, Chris - a most wonderful synthesis! Can't help myself forwarding this to other pals.
Tx. -ppy (01)
=== (02)
Subject: Re: Knowledge Representation/Access
Date: Thu, 27 Jun 2002 23:03:23 -0500 (EST)
From: Chris Dent <cdent@burningchrome.com>
To: <ba-unrev-talk@bootstrap.org> (03)
Elaborating on why (I think) knowledge access structures are far more important that knowledge
representation formalisms. (04)
There's a lot of context for this for which there isn't really space. If I seem to leave something
out, feel free to ask me about it. These ideas are fairly rough, as I'm trying to synthesize pretty
much everything I've learned in the last year or so into a simple lesson. Much or most of this is
opinion. As far as I can tell none of this is particularly revolutionary, it has all been said in
similar ways by other people in other places. (05)
It is common in the information science field (that's the course of study I'm on at the moment) to
categorize aspects of the human comprehension and communication of signal as data, information,
knowledge and wisdom. What these things are is an open debate, and rests in the land of religion,
but a model I like goes like this: the four are arranged as a pyramid with data on the bottom,
information next, then knowledge, then wisdom. (06)
Data is essentially perceived phenomena. (07)
Information is data which has been placed in context or identified as having pattern. Facts. (08)
Knowledge is the discovery of new perspective through the synthesis of information. It is
considered by some to be contained in individuals only, but the notion of institutional memory or
knowledge throws this into doubt. It is the multi-faceted lens through which we see the world. (09)
Wisdom is an ethical perspective, a distillation of knowledge that has value for a community. (010)
Learning can be thought of as a process of gathering data, using learned perspectives to make it
information, synthesizing the information into knowledge. (011)
An adjustment in the view of this model reveals it to be a dynamic model. One person's knowledge
can be someone else's information, from which they may create knowledge. (012)
Okay, that's the stage. The conclusion I make is that knowledge generation is a synthetic, creative
and perception altering process using information, all kinds of information, as a source of fuel. (013)
I conceive of knowledge generation as a process of accretion. We start with some small concepts and
gather new learning about them to form greater capacity to perceive. Sometimes concepts collide and
we have the paradigm shift that educators love. (014)
I do _not_ think of knowledge generation as part of the scientific method. The scientific method is
for proving or disproving recently generated knowledge. Knowledge creation might be considered
hypothesis generation, except frequently we accept our knowledge without question, trusting the
analogous process which has brought us to new knowledge. (015)
Formal knowledge representations, by their nature, are closed systems and thus not knowledge at
all. Some claim to be able to represent all things, but if they are in fact complete and precise
languages they are, to me, by definition, resistant to synthesis. Interpretations of those people
who dig them suggests they like them for precision in the processing of facts. Knowledge is not
facts. Knowledge is interpretation (of all sorts of stuff, including facts). (016)
That's problem one: formal knowledge representations are not what they think they are. Problem two
is a simple question of usability and distribution of power, and leads directly into the
conclusion. (017)
Knowledge generation is a process of gathering and comparing. To effectively gather and compare one
needs access to a lot of information. Human discourse is not a strict formal representation
(although we could argue that language is in some ways a formal representation because it exists in
social space, but that's a different discussion) so in order for it to be used in a formal
representative system it needs to be translated to the new representation. There's no way to
correctly automate this (with computers) and I don't think there ever will be. (018)
This is because computers can _only_ work with formal representations and translating human
discourse is an intepretive or comparative or (if I'm getting the word right) analogous process. (019)
Getting an army of people to translate discourse to a formal representation would be a huge
process. Some might think it worthwhile. I don't. Here's why: (020)
It is presumably possible to declare, within a subset of the population, that henceforth we will
attempt to communicate with formal representations so we might engage in an orgy of precision.
Unfortunately this won't work very well: (021)
- formal representations must be learned
- formal representations are hard to use
- formal representations are not very expressive or persuasive
- formal representations exist for inference not communication (022)
In addition to being hard to use, attempting to use formal representations for communication raises
questions about elitism. (023)
But most importantly: while a computer can easily be instructed to make inferences from a formal
representation a human can't easily read them and thus can't do much in the way of knowledge
generation. Yes, they can establish fact, but that's not the same thing (although it is an
important thing). (024)
Also, discourse exists to convince. Communication is an art. Facility with the language, especially
nuance, makes the art. Precision is antithetical to nuance. Nuance creates intepretive resonances
in the receiver which may cause them to adjust the perspective. (025)
Formal representations are important in the way they can augment a human trying to make decisions.
In a collaborative knowledge environment they are a crucial part of the picture but not the crucial
part. [believe he meant "... but not the picture." -ppy] (026)
As was made evident to me from the discussion on the portreview list, failed collaboration is the
result of an inability of people to reach a shared change in perspective. A change in perspective
is new knowledge. Here on the bootstrap lists the underlying motivation is solving complex and
urgent problems. The belief is that through collaboration and augmentation we can make progress on
problems that unaided we can not address. (027)
Addressing information (or knowledge if you choose to use that term) is the key to creating
knowledge. In order for me to learn a new thing I need to be able to compare some stuff. Likewise,
in order for me to convince you, or for us reach a consensual new perspective, we need to share
information, evaluate it, place it in context, and synthesize it. At the same time, we can be more
effective if we can easily discover that we are treading covered ground. (028)
So, whether the knowledge (actually information) is stored in a formal representation or in human
discourse being able to access it in effective ways is the crucial key to collaboration. (029)
--
Chris Dent <cdent@burningchrome.com> http://www.burningchrome.com/~cdent/
"Mediocrities everywhere--now and to come--I absolve you all! Amen!" -Salieri, in Peter Shaffer's
Amadeus (030)
=== (031)
On Wed, 26 Jun 2002, Jack Park wrote: (032)
> At 04:21 PM 6/26/2002 -0500, Chris Dent wrote:
> >Earlier today I decided that knowledge access structures are far more
> >important than knowledge representation formalisms. Access helps to
> >generate new theory while avoiding repetition.
> >
> >(Formal information representations is another cup of tea entirely.)
>
> Chris,
>
> Having nicely sidestepped the personal labels ascribed to you, you finally
> opened up a great area for discourse. So, I relabeled the thread.
>
> You said: "Earlier today I decided that knowledge access structures are far
> more important..."
>
> In the context of unrev, I don't see where that happened. I see hints of
> it on the PORT list. So, let me ask you to elaborate.
>
> Thanks,
> Jack (033)
--
An archive of the [ubl-ontolog] postings can be found
at http://ubl.cim3.org/lists/ubl-ontolog/ (034)