COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002		Let m? be a little more precise about how this approach is
C00006 ENDMK
C⊗;
	Let m? be a little more precise about how this approach is
fundamentally non-positivist.  Positivism always starts its
conceptualization of a phenomenon by considering the observations
that can be made.  It does this to the extent of considering an
unknowable phenomena as some kind misuse of language.
On the other hand, we consider possible structures that might
exhibit the phenomenon, study their properties as much as
possible mathematically, and only later ask about how observation
can be used to tell which possibility is realized.  It is always
possible that the opportunities for observation are insufficient
now or forever to make the discriminations we would like to make.
(Of course, we should also be aware of any possibilities of
constructing homomorphic images A' and B' of two theories
A and B such that A' and B' are isomorphic and consistent with
observation.  Under such circumstances, we are inclined to regard
the distincion between A and B as unreal, but we may still
remember them in the offchance that improved observational methods
may resuscitate the distinction).

	In the automaton model the reasons for the non-positivistic
approach is particularly clear.  Imagine a system of interacting
automata some of which are regarded as persons trying to understand
their automaton world as a whole.  It is a highly contingent question
whether their observational possibilities are sufficient to permit
this, and small modifications in the automaton system may make
large changes in their ultimate ability to infer or even conjecture
the structure of their world.  It would be a difficult technical
problem to design the world so that it is always positivistically
modelable.  Indeed, the question can depend on the initial state
of the world and on the initial actions of the philosopher-automaton.
Some of the possibilities are discussed in (Moore 1956).

	As a further illustration, we have started our discussion
of mental qualities with known machines where the mental qualities
are to be defined in terms of the state of the machine.  When we
study the mental qualities of people, whose internal states we
cannot presently observe, we continue the same approach.  We regard
the mental qualities as functions of the unknown state.  Naturally,
it turns out that we must develop ways of relating the mental
qualities to observation, even though we are not able to infer
the state as a whole from observations that can be made.