COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	metaep[f82,jmc]		A proposal for meta-epistemology
C00008 ENDMK
C⊗;
metaep[f82,jmc]		A proposal for meta-epistemology

	In their everyday lives and work, physical scientists, the man in
the street, and the artificial intelligence researcher share a common
naive realist epistemological viewpoint.  There is a physical world, and
part of it consists of people seeking knowledge of it, and now part of it
consists of machines programmed to seek knowledge.  Philosophers find this
naive, and, indeed, the same scientists, men in the street, and
artificial intelligence researchers that use it in their normal activity
also find it naive when they think about the foundations of their
knowledge.

	Nevertheless, we propose to base the study of the nature of
knowledge and the processes for obtaining it precisely according to this
naive realist viewpoint.  Namely, consider a dynamic system evolving
in time according to suitable laws.  An example of such a system is
a collection of interacting finite automata such that the state of
each automaton at time  t+1  is a function of its state at time
t  and the inputs it receives from the other automata.  However,
other models of causal systems such as system of differential equations
for positions of particles or for quantum mechanical wave functions
have similar properties for the questions that interest us here.

	We are interested in systems in which some of its components
can be regarded as knowledge seekers.  This means that we are
interested in rules that interpret some functions of the states of the
subsystems as assertions, i.e. beliefs, about the system as a whole.  We are
interested in what is asserted about the world and the relation
of the component to it and how these beliefs change with time.
We propose to study this with suitable mathematical and mathematical
logical theories of epistemological systems.

	An epistemological system is then a tuplet with the following
components: (1) A system of interacting automata.
(2) A language for expressing assertions about the system including
both assertions about particular states  and about the temporal
behavior of the system.  (3) Functions mapping states of certain
subsystems into sentences of the language.  The following questions
can be studied mathematically.

	(1) When do such functions exist in reasonably describable
form.

	This will be true only for certain subsystems of certain
systems and certain classes of mappings.  This is a key point, and
I call it the argument from cryptography.  Consider a lump of granite
with specks on it of many colors.  One might suppose that in a suitable
language, the pattern of specks might encode a textbook on chemistry.
If the number of bits in the pattern is sufficent, this is true, but
extremely probably, only in a language with an extremely lengthy
description.  This is supported by the Kolmogorov-Chaitin theory of
computational complexity but actually elementary cryptography supplies
convincing experience.  Even 30 character simple substitution cipher
cryptograms have unique solutions, and if 30 characters are written
at random, it is extremely unlikely that the text will have an
interpretation as a simple substitution cipher in any human language.
Therefore, the remark of xxx that a novel in one language might
be a chemistry text in another is mistaken.

	For this reason, if we have one expressible rule that translates
states of a subsystem into assertions about the world, then it is
extremely unlikely that there is another.

	(2) Suppose we imagine varying a component of the system in
ways that correspond to varying the way a robot represents knowledge
and seeks it and pursues other goals.