Information Technology Reference
In-Depth Information
approaches to this phenomenon. On the naive side, “learning” is inter-
preted as a change of ratios of the occurrence of an organism's actions
which are predetermined by an experimenter's ability to discriminate such
actions and his value system, which classifies these actions into “hits” and
“misses.” Changes are induced by manipulating the organism through elec-
tric shocks, presentations of food, etc., or more drastically by mutilating, or
even removing, some of the organism's organs. “Teaching” in this frame of
mind is the administration of such “reinforcements” which induce the
changes observed on other occasions.
On the sophisticated side, learning is seen as a process of evolving algo-
rithms for solving categories of problems of ever-increasing complexity
(Pask, 1968), or of evolving domains of relations between the organism and
the outside world, of relations between these domains, etc. (Maturana,
1969). Teaching in this frame of mind is the facilitation of these evolution-
ary processes.
Almost directly related to the level of conceptual sophistication of these
approaches is their mathematical naiveté, with the conceptually primitive
theories obscuring their simplicity by a smoke screen of mathematical pro-
ficiency, and the sophisticated ones failing to communicate their depth by
the lack of a rigorous formalism. Among the many causes for this unhappy
state of affairs one seems to be most prominent, namely, the extraordinary
difficulties that are quickly encountered when attempts are made to
develop mathematical models that are commensurate with our epistemo-
logical insight. It may require the universal mind of a John von Neumann
to give us the appropriate tools. In their absence, however, we may just
browse around in the mathematical tool shop, and see what is available and
what fits best for a particular purpose.
In this paper the theory of “finite state machines” has been chosen as a
vehicle for demonstrating potentialities and limitations of some concepts in
theories of memory, learning, and behavior mainly for two reasons. One is
that it provides the most direct approach to linking a system's external vari-
ables as, e.g., stimulus, response, input, output, cause, effect, etc., to states
and operations that are internal to the system. Since the central issue of a
topic on “molecular mechanisms in memory and learning” must be the
development of a link which connects these internal mechanisms with their
manifestations in overt behavior, the “finite state machine” appears to be
a useful model for this task.
The other reason for this choice is that the interpretations of its formal-
ism are left completely open, and may as well be applied to the animal
as a whole; to cell assemblies within the animal; to single cells and their
operational modalities, for instance, to the single neuron; to subcellular
constituents;
and,
finally,
to the molecular building blocks of these
constituents.
With due apologies to the reader who is used to a more extensive and
rigorous treatment of this topic, the essential features of this theory will be
Search WWH ::
Custom Search