Newsgroups: comp.ai,comp.ai.philosophy
Path: cantaloupe.srv.cs.cmu.edu!das-news.harvard.edu!news2.near.net!MathWorks.Com!europa.eng.gtefsd.com!howland.reston.ans.net!EU.net!uknet!festival!castle.ed.ac.uk!cam
From: cam@castle.ed.ac.uk (Chris Malcolm)
Subject: Re: Is Common Sense Explicit or Implicit?
References: <1994Aug19.113304.22524@unix.brighton.ac.uk> <332jo9$8cr@ux.cs.niu.edu> <1994Aug22.042736.25458@news.media.mit.edu>
Message-ID: <CvFMAI.3qG@festival.ed.ac.uk>
Sender: news@festival.ed.ac.uk (remote news read deamon)
Organization: University of Edinburgh
Date: Thu, 1 Sep 1994 03:42:17 GMT
Lines: 91
Xref: glinda.oz.cs.cmu.edu comp.ai:24028 comp.ai.philosophy:20253

In article <1994Aug22.042736.25458@news.media.mit.edu> minsky@media.mit.edu (Marvin Minsky) writes:

>What I'd like to see is an attempt to discuss
>the relations between **three or more** categories of knowledge types
>because I consider the explicit-vs.-not" to be evidently unproductive.
>A first step would be serious proposals about such triads.  Any
>offers?

EXPLICIT, IMPLICIT, and TACIT.

(I will use "creature" here to mean created agents both natural and
artificial: robots, beetles, men, and martians.)

Explicit knowledge: you could (in principle) find the representation
of the knowledge in the creature, and it is used by the creature in
generating the observed knowledgeable behaviour, i.e., knowledge as in
Brian Smith's Knowledge Representation hypothesis.

Implicit knowledge: not explicitly represented, but capable of being
made explicit by the available reasoning machinery operating on the
explicit knowledge. It is a design choice (space/time trade) whether
deduced implicit knowledge is left explicit once deduced, or deduced
afresh whenever needed.

Tacit knowledge: designed into the structure of the creature
(algorithmically, physically, etc.), and so correctly governing its
behaviour, but not available as meaningful knowledge to the creature
(unless it conducts a scientific investigation of its own nature of
course). By "meaningful knowledge" I mean cognitively penetrable, able
to be used by the creature's reasoning machinery, of indefinite
relevance. Tacit knowledge is therefore ad hoc, dedicated to specific
purposes, encapsulated, of definite relevance. 

For example, the longer legs on one side of the haggis tacitly know
(express the knowledge in tacit form) that the haggis runs round
hills. As far as the haggis is concerned, this is tacit knowledge,
since the haggis doesn't (explicit) and can't (implicit) know it, yet
this fact governs its behaviour. And should a scientifically inclined
haggis genius discover this fact, thus making it (for that haggis)
explicit knowledge, the explicit knowledge will serve to *explain* to
the haggis how it is fitted to its environmental niche, but will not
*govern* its hill-circumnavigating behaviour. That will still be the
role of the still tacit knowledge expressed in its bilateral
asymmetry. That knowledge is still tacit, because its now-explicit
form plays no part in governing the hill-running behaviour (it governs
its explanatory behaviour).

Note that whenever we (i.e. people who discuss not just human
knowledge, but robot knowledge, beetle knowledge, and the knowledge
known by component parts of knowledgeable creatures) use the term
"knowledge" we should make it clear who (or what) is the knower, since
what kind of knowledge it is, or whether it is knowledge at all, may
depend on the knower in question. Knowledge categorisation is knower
relative. This is a very important point, all too easily forgotten
since our epistemological history has been for milennia concerned only
with human knowledge.

Let us suppose for the sake of argument that there is a human language
grammar organ which contains the rules of our natural language grammar
in an explicitly represented form, i.e., the grammar organ learns
these rules explicitly, encodes them in some (in principle)
discoverable symbolism, and uses this explicitly represented knowledge
to parse incoming sentences. Thus, from the point of view of the
grammar organ as knower, this is explicit knowledge. But we also know
that the rules by which the grammar organ operates are not known to us
consciously, hence the science of linguistics, which conducts
experiments on the language organ in the attempt to reverse engineer
it. Therefore the grammar rules, which are explicit from the point of
view of my grammar organ as knower, are tacit from the point of view
of my own consciousness as knower. So whether some part of my
knowledge is tacit or explicit depends on whether I or some component
of me is considered to be the knower: knowledge categorisation is
knower relative.

These three categories concern knowledge within a single creature.
There are more categories once you consider communicating creatures
sharing knowledge, Popper's third world, etc., but given the current
state of AI we can leave that until later.  How do we know we can
*safely* leave it till later, that this omission will not distort our
view of a single agent? Because evolution demonstrates that social
knowledge is constructed incrementally on the basis of individual
knowledge.




-- 
Chris Malcolm    cam@uk.ac.ed.aifh          +44 (0)31 650 3085
Department of Artificial Intelligence,    Edinburgh University
5 Forrest Hill, Edinburgh, EH1 2QL, UK                DoD #205
"The mind reigns, but does not govern" -- Paul Valery
