132
MOORE
There is a somewhat artificial restriction that will be imposed
on the action of the experimenter. He is not allowed to open up the
machine and look at the parts to see what they are and how they are inter
connected. In this military situation, such a restriction might correspond
to the machine being booby trapped so as to destroy itself if tampered with.
It might also correspond to an instance where the components are so
unfamiliar that nothing can be gained by looking at them. At any rate, we
will always impose this somewhat artificial restriction that the machines
under consideration are always just what are sometimes called "black boxes",
described in terms of their inputs and outputs, but no internal construction
information can be gained.
Another application might occur during the course of the design
of actual automata. Suppose an engineer has gone far enough in the design
of some machine intended as a part of a digital computer, telephone central
office, automatic elevator control, etc., to have described his machine in
terms of the list of states and transitions between them, as used in this
paper. He may then wish to perform some gedanken-experiments on his
intended machine. If he can find, for instance, that there is no experi
mental way of distinguishing his design from some machine with fewer
states, he might as well build the simpler machine.
It should be remarked that from this engineering point of view
certain results closely paralleling parts of this paper (notably the
reduction described in Theorem k) have recently been independently found by
D. A. Huffman in his Ph.D. thesis in Electrical Engineering (M.I.T.). His
results are to appear in the Journal of the Franklin Institute.
Still another situation of which this theory is a mathematical
model occurs in the case of the psychiatrist, who experiments on a patient.
He gives the patient inputs (mainly verbal), and notes the outputs (again
mainly verbal), using them to learn what is wrong with the patient. The
black box restriction corresponds approximately to the distinction between
the psychiatrist and the brain surgeon.
Finally, another situation of which this might conceivably be a
mathematical model occurs when a scientist of any sort performs an experi
ment. In physics, chemistry, or almost any other science the inputs which
an experimenter puts into his experiment and the outputs he gets from it
do not correspond exactly to the things the experimenter wishes to learn
by performing the experiment. The experimenter is frequently forced to ask
his questions in indirect form, because of restrictions imposed by
intractable laws of nature. These restrictions are somewhat similar in
their effect on the organization of the experiment to the black box
restriction.