Newsgroups: comp.ai.philosophy
From: lupton@luptonpj.demon.co.uk (Peter Lupton)
Path: cantaloupe.srv.cs.cmu.edu!das-news.harvard.edu!news2.near.net!news.umass.edu!news.mtholyoke.edu!news.byu.edu!news.kei.com!MathWorks.Com!europa.eng.gtefsd.com!howland.reston.ans.net!pipex!demon!luptonpj.demon.co.uk!lupton
Subject: Is common sense explicit or implicit?
Distribution: world
Organization: No Organisation
Reply-To: lupton@luptonpj.demon.co.uk
X-Newsreader: Newswin Alpha 0.4
Lines:  183
Date: Tue, 13 Sep 1994 22:09:22 +0000
Message-ID: <292384118wnr@luptonpj.demon.co.uk>
Sender: usenet@demon.co.uk

Common sense - Implicit or Explicit? Marvin Minsky 
asks us to suggest classifications other than 
explicit/implicit. I would like to suggest the notion of 
knowledge and understanding and common sense being 
"use-bound". This is not intended to be all or nothing - 
knowledge and understanding may be use-bound/non-use-bound 
more or less.

(BTW, I don't know whether this is the same thing as TACIT
 as described by Chris Malcolm).

In order to make sense of the distinction, I feel obliged to
argue for two positions:
 
  1. That use involves the ability to project
     regularity in an essential way

  2. That, insofar as sensory data supports use,
     this support is through our ability to detect
     and to project regularity.

Both these statements stand in need of clarification
and the reader should bear with me until I do so.

First, I should clarify the way I am using the word 
"use". According to the dictionary:
 
to use - "to cause to act or to serve for a purpose"

a definition I will adopt.

When we use the noun "use" it can mean "instance
of use" or, for example, "potential for use". I shall
try to restrict myself to instance of use and be 
more explicit if "potential for use" is meant.

The word "cause" might be considered a little 
troublesome, since cause can occur in more than 
one way - non-accidentally and accidentally. I think 
it should be clear that a accidental cause does not 
count in the meaning of the word "use".

For example, I may wish to destroy an ugly vase. I
may also accidentally break it with a hammer as I 
pass by. In doing so I have *not* used the hammer 
to break the vase. In contrast, I have accidentally 
broken the vase with the hammer.

So in giving an example of use, one should be able
to answer the questions:

  1. In what sense non-accidental?
  2. For what purpose?

Consider a thrush which uses hard surfaces to help
break the shells of snails. It is non-accidental in that
it is, we suppose, an evolved behaviour which has 
evolved *because* it satisfies the purpose of getting 
the thrush fed.

Use is also a success word. To use is to *succeed* in
causing something to act for one's purpose. In order to
subtract success from use, we must resort to locutions
such as "try to use" or "attempt to use".

The word "cause" might produce difficulties in another
way. We don't mean by the word "cause" that there
must be a specific action by the user. One can, for example, 
use piped music to relax with, even though one is being 
subjected to it and one has no choice in whether the music 
is played or not. One can, in this case, deliberately
choose to listen to the music with the purpose of
relaxation. This would be in contrast to being in a 
situation where piped music was being played and 
subsequently becoming relaxed. Such a case, on that
account, would not imply any *use* by the person who
became relaxed (the proprieter might be using the music,
but that is another matter).

With all this preamble out of the way, the relationship
between use and projection of regularity is now easy:
 
In order to say that there has been a use, one must
give an account of the way in which the use was
non-accidental and to what purpose. Such an explanation
must involve the projection of regularity from the
here-and-now to the putative events which will satisfy 
the purpose in order to discharge the requirement of not 
being accidental.

So much for the first argument.

The second assertion requires some clarification also.
Sensory data can enter into a use in more than one way.
One such way is in the satisfaction of the purpose.
For example, one may use a chair by getting up on it to 
see better. Clearly, sensory data enters into this use as 
part of the satisfaction of the purpose. The purpose is,
after all, to see better and this plainly involves sensory
data. Again, one may use a CD to listen to music 
for pleasure and once again sensory data is part of what 
satisfies the purpose.

Discounting that particular way in which sensory data enters 
into use, however, we can then ask: how else does sensory data
contribute to use? The answer is, I think, as stated above:

     That, insofar as sensory data supports use,
     *and that sensory data is not fulfilling the purpose*,
     this support is through our ability to detect
     and to project regularity.

The argument is this: sensory data tells you how things are,
but a use is extended over time. How can it help to know
how things are, if what you want to know is how things
will be and, indeed, how to "cause things to act or to 
serve for a purpose"? The answer is, of course, through 
our ability to use sensory data to detect and to project 
regularity. It is precisely because we can do these things 
that sensory data has that particular role to play in use.

Of course, detection and projection of regularity also
occur over time. The point here is that an organism may
*invest* in the processes by which it may detect and project
regularity without having a use for those regularities yet.
So, for example, a child might spend hours soaking up 
sensory input without any apparent use in sight. Such 
activity can, however, permit the child to invest in its 
ability to detect and project regularities. And such 
abilities may well prove their usefulness in the future.

The two arguments I have made indicate that, in a sense, detection
and projection of regularity are universal for use. Sensory
data can, of course, be *used* to detect and project regularity
and so detection and projection of regularity are special uses
that sensory data can be put to. What we are now saying is that 
these apparently rather special uses of sensory data are, in fact, 
an essential part of all uses and (except for purpose) are the 
way in which sensory data enters into use.

This is not to say that separation between detection and projection
of regularity and use *must* occur in practice - just that it always
*could* occur in practice. One would also claim that, very often for 
human beings, this separation *does* occur in practice.

Definition: 

use-bound - being intertwined with and not readily available 
                  except in the context of a specific use.

Clearly the above definition is a matter of degree.

An example where projection of regularity is use-bound might be 
a thermostat. A heating control system *could* be made with a
component which just predicts the future temperature of the house by 
detecting and projecting regularity. There *could* be a sharp 
separation between how such projections are made and how they
control the boiler. However, in the thermostat, we see a conflation 
of the two.

The reason the thermostat conflates the two is that, given the single
use such predictions will be put to, there is no need to split the two
functions apart. Given a single use it is more efficient (cheaper) to
smash the two functions together. However, if projections of the future 
temperature of the house could be put to *many* uses (for example, 
controlling the opening and closing of windows etc. etc.) then there 
could well be benefit in separating the function of prediction from the 
function of controlling such devices.

In short, if a projection can be used in more than one way, there may well 
be a benefit in separating out the processes of detecting and projecting
regularity from the other processes of making use of such projections.
One imagines that, for insects, there is little benefit to be had from such
separation because the lives of insects are so constrained. For humans, 
however, with our diverse lives, there will often be real gains to be made 
in creating a separation between the detection and projection of regularity 
on the one hand and the use that such projections could be put to on 
the other hand. *Who knows* what use sensory data could subsequently
be put to?

-------------------
Peter Lupton
