Professional Documents
Culture Documents
OF THOUGHT
AND PROCEDURE
NEW METHODS
OF THOUGHT
AND PROCEDURE
Edited by
F. Zwicky
and A. G. Wilson
Contributions to the Symposium on
METHODOLOGIES
Sponsored by the
Office for Industrial Associates
of the California Institute of Technology
and the Society for Morphological Research
Pasadena, California, May 22-24,1967
ISBN-13: 978-3-642-87619-6 e-ISBN-13: 978-3-642-87617-2
DOl: 10.lO07/978-3-642-87617-2
The use of general descriptive names, trade names, trade marks, etc. in this pub-
lication, even if the former are not especially identified, is not to be taken as a
sign that such names, as understood by the Trade Marks and Merchandise Marks
Act, may accordingly be used freely by anyone.
Title No. 1479
LIST OF CONTRIBUTORS
v
SHUBIK, MARTIN, Economic Growth Center, Yale
University, New Haven, Connecticut
VI
Table of Contents
LIST OF CONTRIBUTORS V
PROLOGUE 1
VII
SECTION IV INFORMATION THEORY 133
Chapter 1 A Survey of Information Theory 135
John R. Pierce
Chapter 2 Information Theory and Modern 163
Digital Communication
R. W. Lucky
EPILOGUE 333
The format followed at the conference and in this book is that the first paper
of each section presents a general review of the subject area and subsequent
papers within the section provide examples of specific applications of the
methodology.
VIII
PROLOGUE
1
thus the initiator of general education, so:me of whose ideas
were so profound and far reaching that even today they have
not yet been sufficiently taken advantage of inas:much as we
have not yet achieved the training of the whole of :man which
he visualized.
2
experts from all fields were brought together by the ingenious
founder of the Pestalozzi Foundation of America. Mr. H. C.
Honegger. to establish war orphan villages on all continents
and to deal effectively with the problems of destitute children
all over the world. In way of large scale constructive actions
for adequate housing the efforts of Profes sor Constantinos
Doxiades in Athens are outstanding, partly because of the
yearly Delos Conference which he has organized and in which
experts in absolutely every field of human endeavor partici-
pate. The establishment of the Cit~ de G~n~ralisation du
Canisy near Deauville, which is being promoted by the
French. is intended for occupation by experts and men of
universal outlook who will deal with large scale problems,
one after another. The Conf~rence des Sommets (Cultural
Top Conference) in Brus sels in 1961 should also be men-
tioned. At the invitation of the King and the Belgian Govern-
ment, outstanding representatives of all sciences. tech-
nologies and the arts were invited to attempt an integration
of all essentials of present-day knowledge. The organizer
of this conference, Francois Le Lionnais. pres ident of the
French association o~ scientific writers. had previously
edited a book, LA METHODE DANS LES SCIENCES
MODERNES. which may be regarded as a sequel to
Descartes' "Discours de la M~thode" and which contains
articles by s orne forty authors.
3
Associates at the California Institute of Technology.
After Dr. Warner retired, the project had to be post-
poned but is now being realized in this sym.posium.
through the cooperation of Richard P. Schuster, the
present Director of the Office for Industrial Associates.
F. Zwicky
4
SECTION I
OPERATIONS RESEARCH
CHAPTER 1
THORNTON PAGE
]V' esleyan Universzty
ABSTRACT
7
A. HISTORICAL INTRODUCTION
10 Definitions
8
distinguishing it frOITl the various branches of engineering.
or froITl other bounded disciplines. For this reason, 0 - R
studies are often undertaken by teaITlS of specialists froITl
various disciplines (statistician. physicist. cheITlist, air-
craft des igner. ITledical doctor. and psychologist. for
instance). Thus "teaITl res earch" and synthesis are often
cited as characteristic of operations research.
2. Pre-1930
9
Specific designs had received a good deal of attention, both
in military systems (such as fortifications. small arms,
artillery. tanks, submarines, and bombers) and in indus-
trial systems (such as steel mills, oil refineries, and rail-
roads). But a number of innovations such as radar. sonar.
aircraft-dropped depth charges ~ and torpedoes were being
developed. and there was no clear assessment of how these
new devices would affect military-system effectiveness.
10
bomber and submarine kills proved him correct. At this
point he coined the term "operational research" for his
systematic review of repeated military operations. The
changes in procedure or strategy were later called "deci-
sions" by the military high command, and operations re-
search is sometimes called the "science of decision. "
4. U. S. Wartime Studies
11
search patterns maximizing the probability of locating a
submarine after one position report. Another group in
London studied German submarine tactics from the many
torpedoings of Allied convoys, from daily radio fixes, and
sightings. Statistics gave the probability of submarines
sighting convoys (as a function of speed and size of convoy)
and the probability of penetrating destroyer-escort screens
(as a function of number of destroyers and number of subma-
rines}. An optimized strategy of convoying was derived,
based on shipping requirements and the number of destroyer-
escorts and search aircraft available. All this (and other
studies of a possible aircraft blockade of the German sub-
marine bases} involved a complex model of the submarine-
ship battle in the Atlantic. The "pay-off" {measure of
merit} was clear: maximize ship deliveries to Britain
during 1943 and 1944.
12
Operations Analysis (0. O. A.} and RAND Corp. and the
Joint Chiefs' Weapons Systems Evaluation Group (W.S.E.G.).
They conducted many studies in the ensuing decade. and
developed the techniques of quantitative operations research.
particularly "gaming" and mathematical simulation. The
diversified-team attack led to much broader studies. such
as the logistics requirements for overseas wars, the use and
suppression of guerrilla forces. the best use of manpower
in the U. S. Army ("integration" of Negroes started there).
and the economic effects of warfare.
-dM/dt = Ai2M2
13
stances not being the best overall. Such reasoning led to
broader studies; one of the broadest I know concerned
nuclear warfare. a "game" carried out jointly by O. R. O.
and RAND from 1954 to 1957.
6. Industrial Developments
14
involve several interrelated products, and decisions must
be made on the proportional output of each. One well-known
example is a study of the optimum outputs of a chemical
plant (11) where these proportions can be controlled to match
market demand and price.
15
there were 15 separate nations with 0 - R societies. and a
committee was formed (including 7 British and 10 Americans.
under the joint chairmanship of Sir Charles Goodeve of
London and myself) to arrange an international conference
(14). This took place at Oxford in September. 1957, and was
attended by 300 people from 17 countries. It led to the Inter-
national Federation of 0 - R Societies, which has since held
three more conferences (Aixen Provence in 1960, Oslo in
1963, Cambridge, Mass. in 1966). Papers presented at
these meetings covered topics ranging from the operation of
coal mines to the flow of intelligence messages in military
maneuvers in Germany. They have established some unity
of the new discipline ~ and stimulated broader and broader
applications to problems of transportation, urban planning,
medical services, and national economy.
16
Table 1
Universities Offering Advanced 0 - R Training, 1966
17
Table 2
Government Agencies with 0 - R Studies
18
Food and Drug Administration
General Accounting Office (Labor Statistics)
Health, Education and Welfare Dept. (Science Communica-
tion, Ed. Statistics, Program Planning)
Housing and Horne Finance Agency
Housing and Urban Development
Interior Dept.
Internal Revenue Service
Interstate Commerce Commis sion
Labor Dept. {Automation Manpower, Policy Planning,
Statistic s}
Library of Congress (Science Policy. Legislative Refer-
ence Service)
Maritime Administration
National Academy of Sciences
National Aeronautics and Space Administration (Resource
Anal. )
National Institutes of Health
National Institute of Mental Health
National Science Foundation
National Security Agency
Navy Dept. (O.E.G., Office of Nav. Res •• Nav. Res. Lab ••
Model Basin, Bureau of Weapons. Manpower.
Operations, Weather, Ships)
Naval Academy
Naval Medical Center
Patent Offic e
Peace Corps
Post Office Dept. (Transportation Research)
19
President's Council of Economic Advisors
Small Business Administration
Smithsonian Institution
Bureau of Standards {Technical Analysis Div.}
State Dept. (Arms Control, Overseas Program, Intelligence,
Technical Cooperation)
Office of Transportation
Treasury Dept. {Balance of Payments}
Veterans Administration
Weather Bureau
20
alternative programs. providing a basis for executive de-
cision in standard 0 - R manner (15). In fact. the Bureau
of the Budget now requires such studies. known as planned
programming or PPB.
B. TIME STUDIES
1. Logistics
21
reserve supplies cut the mobility of land forces - -a serious
limitation in modern warfare. Therefore, army logistics
are organized in a heirarchy of supply bases fed by a "pipe-
line" from home industry. Changes are frequently neces-
sary when new equipment is introduced, all delayed by the
pipe-line flow time. Heavy military action increases the
requirement for most spare parts and replacements; ad-
vances by military units increase the delivery delay. and
the logistics planner must take such possibilities into ac-
count, as well as living within his budget. A fairly accurate
mathematical model is pos sible, if failure data and usage
data are available.
2. Scheduling
22
3. Queueing
4. Long-terITl DevelopITlent
23
electronic computer. Until the last minute, there is little
chance of predicting such "break-throughs." but it is well
known that the average rate has been increasing for several
decades in the U. S •• and that it peaks just after a war. An
0- R study made by the National Science Foundation (18)
correlated the invention rate with money spent on research
and development. a conclusion so widely accepted that the
go,:ernment and most large industrial concerns now budget
an appreciable fraction of their income for Rand D. expect-
ing a long-term pay-off of several hundred per cent.
24
C. INTERDISCIPLINARY STUDIES
1. Human Engineering
2. Psychological Warfare
3. Manpower
25
1950 study of the Army's use of Negroes (22). The con-
clusion was that better use could be made if Negroes were
integrated rather than segregated. Quantitative data were
obtained on the proportion of Negro recruits, their perfor-
mance on intelligence tests, and their schooling. Opinion
polls showed that white officers who had served with
Negroes found them individually courageous, although
segregated regiments had poor morale and poor battle
records.
4. Public Interest
26
D. INFORMATION FLOW
1. Intelligence Messages
27
are plotted). Of course, a large proportion of the infor-
mation was irrelevant to anyone decision, but the effective
use of artillery, aircraft, and missiles required accurate.
up-to-date info rmation.
2. System Control
28
feasible and, in liIT1ited forIT1" are already in use. (Many
IT1ajor airlines use a single cOIT1puter IT1eIT1ory to keep track
of reservations, and IT10st banks are installing cOIT1puterized
accounting systeIT1s.) In the years since this early work
was done, autoIT1atic data proces sing (A. D. P.) has developed
rapidly. I recently witnessed the trial of a rapid-access de-
vice at the United Aircraft Corp. Research Labs. where a
year's worth of tiIT1e-sheet records, pay checks. publica-
tions. and assignIT1ents for hundreds of scientists and
engineers have been entered on tape. PrograIT1IT1ed for
tiIT1e-sharing on a large cOIT1puter. the alpha-nuIT1eric.
office control panel allows the Lab Director (after he punches
a confidential control input) to see on a video screen the
sUIT1IT1ary of all overtiIT1e work during the past week (or
IT1onth, or year), all expenditures on anyone of several
scores of projects. the list of projects overdrawn on their
planned budgets. or any other sUIT1IT1ary of the input data.
He has found such rapid access of great value in IT1anaging
research activity--allowing hiIT1 to shift personnel to lagging
projects. prevent overexpenditures. and evaluate profes-
sionals for pay raises. The hardware design and cOIT1puter
prograIT1IT1ing are fairly siIT1ple. but the selection of infor-
IT1ation stored in the rapid-access IT1eIT1ory, and the types
of sUIT1IT1ary possible. resulted froIT1 an 0- R study siIT1ilar
to the IT1ilitary studies a decade earlier.
******
This sketchy history of operations research is by no
IT1eans cOIT1plete. but serves to illustrate the rapid develop-
IT1ent and wide applicability. The next two chapters cover
two recent studies in greater detail.
29
REFERENCES
30
13. P. M. Morse and Thornton Page, O.R.S.A. Journal
.!' 1, 18, 1953.
140 Proceedings of the First International Conference in
0- R, John Wiley, N. Y., 1958.
TEXTBOOKS
32
CHAPTER 2
A DECADE
OF OPERATIONS RESEARCH IN HEALTH
CHARLES FLAGLE
The Johns Hopkim University
INT RODUCTION
33
produced the necessity for invention of the techniques of
mathematical programming. Perhaps it has been the
absence of competition in the health services, or perhaps
it is that optimization techniques are refinements of
existing rational and developed systems. Health service
systems in the large lack these properties.
34
BEGINNINGS OF OPERATIONS RESEARCH IN HEALTH
35
systems.
36
design and administrative processes. The motivation for
a new synthesis was not to achieve some optimum, but to
design an operating system compatible with the nature of
the load placed upon it. Progressive patient care has
already been mentioned as one alternative form of hospital
organization. Here it was recognized that random vari-
ation in census of patients in the various classifications
was inevitable. and that through built-in flexibility "gray
areas" could be designed to care for several classes of
patients. The magnitude of the gray areas is shown in
Figure 3.
37
Another ITIodel has eITIerged tiITIe and again in
health services research, a descriptive one dealing with
the flow of patients and personnel £rOITI one category to
another in the systeITI. The ITIodel appeared first in the
study of progressive patient care, where patients flow
froITI the outside world to and aITIong the three care cate-
gories within the hospital. The saITIe pressures that pro-
duced the interest in progressive patient care reflect a
broader probleITI, that of linking the ITIany subsysteITIs of
the health services. In addition to the categories of hos-
pital patients, such naITIes as "hoITIe care patients, "
"nursing hOITIe patients," "rehabilitation" and a variety of
other extended care resources ITIust be considered to
picture definitively the process of patient care in a COITI-
ITIunity. To be ITIeaningful the broadened picture, now
known as "coITIprehensive patient care," ITIust be concerned
with the decision rules, the adITIinistration processes for
transfer of patients froITI one facility to another. Res earch
in this area now involves probleITIs of the definition cate-
gories, and probleITIs of analysis of ITIultistate systeITIs
where flow is in part stochastic and in part deterITIinistic.
Figure 6 shows the cOITIprehensive care systeITI scheITIati-
cally. Note that the first state is the healthy population
period. The randoITI events producing flow £rOITI that popu-
lation into the health service systeITIs are the phenoITIena
of illness and accident. In several respects these phenoITI-
ena are age-dependent. The incidence, or rate of occur-
rence, of transition £rOITI health to one of the ill states is
dependent upon age. So also is the duration of stay in the
state. Both tend to increase with age. Since age changes
deterITIinistically the ITIodel ITIust sOITIehow take into account
the changing population in age cohorts and hence changing
incidence and prevalence in the various care states.
DECISION PROCESSES
One of the exaITIples of the cybernetic systeITI is
the physician-patient relationship, particularly the processes
of screening, diagnosis, and therapy. In association with
physicians the probleITIs posed to operations analysts have
largely concerned these ITIatters. Essentially the physician
is engaged in a gaITIe against nature. He knows sOITIething
38
of the prevalence of various diseases, that is to say, the
a priori probability that by chance the next patient he con-
fronts will have a particular disease. He has a battery
of screening and diagnostic tests, virtually all of which
have some degree of imprecision. One speaks of sensi-
tivity of tests, the power to detect the disease when it
exists, and the specificity of the same tests, their power
to detect the absence of the same disease. These numbers
are expressed as conditional probabilities and are almost
always less than one. The most difficult aspect of ex-
pressing diagnostic process in formal decision theory terms
is quantifying the loss function. The nature of the costs
of errors is well known. Scheff (6) has cited the high costs
usually placed by physicians on failure to detect a disease
when it is present.
39
WORK STUDY--TECHNOLOGY AND MANPOWER
40
time and a half for overtime have all become watchwords
within the past few months. A s a trans ient phenomenen
hospitals have responded to the increased demand placed
upon them by the new legislation by increasing personnel.
They have been called upon to explain the paradoxical
behavior of becoming increas ingly labor- intens ive with
rising costs. a pattern exactly opposite to other industries.
The explanation lies in the notion of the underdeveloped
system. There is no well developed labor saving technology
to draw upon. There has not been access to capital for
investments in labor saving equipment if it existed. The
task for administrators. researchers and participants
in the health services system is to direct the field into one
of a developed character where there is a balance between
man and machine. The challenge is to avoid the contro-
versies and antagonisms that have characterized industrial
technological advances.
41
BIBLIOGRAPHY
42
DISTRIBUTION OF POPULATION OF INTENSIVE CARE PATIENTS
(Johns Hopkins Hospital Osler Medical Clinic-6th Floor 29 Beds)
o 0 2 4 6 8 10 12 14 16
NUMBER OF INTENSIVE CARE PATIENTS
Fig. 1
50
>-
~
U 40
<t:
!L-
~ 30
lL.
0
~ 20
z
lJ.J
U
a:: 10
w
!L-
0
OCT
1957 1958
Fig. 2
43
EXAMPLE OF BED ALLOCATION FOR PROGRESSIVE CARE
Interchangeab.
Fixed Intensive a Fixed Interchangeable Fixed
Intensive Care Intermediate Intermediate Intermediate a Self Self Core
a:X
W
IIl Z
~<I
::>:1: 1.00
zl-
(f)
u.(f) 0.80
OW
..J
>- 0.60
1-1-
~~ 0.40
1Il-
<II-
0.20
o1Il~
a:u.
0..0 O.OO~~~~~~--~~--~~--~~--~~--~~--~~
CYBERNETIC MODEL
Fig. 4
44
HIERARCHICAL MODEL
STATE 0
EMIGRATION - DEATH
STATE
WELL
--
.,,-
".
----.--
ECOLOGICAL MODEL
(Unconstrained State Populations)
nj=f(Aijk' fLijk)
Fig. 6
45
CHAPTER 3
OPERATIONS RESEARCH
APPLIED TO CITY PLANNING
MARTIN L. ERNST
Arthur D. Little, Inc.
ABSTRACT
46
to improve the hous ing of its inhabitants.
47
Operations research--in the modern formal sense
of the term--got its start under government auspices in the
military organizations. It should be no surprise, then,
that on return to civilian life many of the early practitioners
made an attempt to apply their knowledge and experience to
non-military problems of local and national government.
While a wide variety of possibilities were recognized, or-
ganization and support for such work was soon found to be
lacking. Military acceptance of OR had been born in an
atmosphere of emergency, a great leveler of fears of inno-
vation. No corresponding sense of crisis existed in our
urban centers during the 1940's and most of the 1950's, so
little willingness arose to try something new.
48
so there was little immediate impact on city inhabitants or
government.
49
Purchas ing and Maintenance
Service Activities
50
complex} parallels in the operation of the terminal phase of
many physical distribution systems. Fire station locations
and numbers could be subject to analysis on a cost/benefit
basis, first to determine the relative desirability of many
smaller units versus fewer large units, and then to select
optimum locations; in most respects, the work would paral-
lel procedures used in studying highly responsive physical
distribution systems. The importance of rapid response in
the availability of police personnel on the scene of crimes
is currently being studied; this can lead to a variety of
improvements in law enforcement, derived from better
communications and perhaps better allocation of resources
such as police cars and patrolmen.
Transportation
Health
51
evidence that sociological problems. such as drug abuse
and addiction, have epidemiological aspects which can be
studied effectively by OR techniques.
S2
C. LIMITATIONS IN THE ROLE OF OPERATIONS
RESEARCH IN PLANNING
1. Creative
2. Analytical
53
ate specialists- -such as statisticians, com.puter program.-
m.e rs, architects and contractors ~ etc •• - - organized for a
team. effort. It should be em.phas ized ~ however, that the
purpose of the OR analysis is not to m.ake the final decision,
but to determ.ine the im.plications of following each of the
alternative paths of action. Although operations research
is vitally concerned with the decision-m.aking processes,
its goal should be to assist the decision m.aker. rather than
seek to m.ake the final determ.ination directly. sim.ply be-
cause not all relevant factors can be m.ade part of the
analytic proces s.
3. Evaluation
54
end far-term implications to be expected from following
each possibility, and the sensitivity of these results to
critical parameters and policies. The final evaluation
will incorporate these factors, plus the many which our
analytical techniques are completely inadequate to deal
with at the present time.
4. Commitment
55
as our techniques improve and our understanding grows,
it may be possible to expand the role of OR; but for the
moment, it appears desirable to accept limitations which
match our capabilities.
1. Background
56
porating a multitude of models to describe the operations
and interactions of key parameters characterizing an urban
area.
57
The city of San Francisco is a cOITlpact residential
and cOITlITlercial city, with alITlost all of its available land
already in active use. The ITlodel developed for its study
took cognizance of these characteristics in that it was con-
fined to residential housing and was concerned with estab-
lishing the "ITlarket" operations which influence the quantity,
quality and costs of housing available. The object was to
study the influence of feasible local governITlent actions on
this ITlarket. Accessibility, industrial growth and other
factors of iITlportance in the other ITlodels were considered
only indirectly.
58
described as a si:mulation in which the interaction between
population and housing characteristics is studied in an
effort to deter:mine the factors that lead to increasing and/or
i:mproving hous ing provided and to rais ing or lowering the
rentals charged. The concept of e:mploy:ment of the :model
involves co:mparison of a base run. in which a city takes no
actions other than those already underway. with results
fro:m a series of runs in which specific plans for future
action are s i:mulated. To develop the :model. the following
steps were taken.
a. Land Division
59
The fracts formed the "quanta" in terms of which all
changes in type or condition of phys ical hous ing space in
the city could be made.
b. Land Use
c. User Classification
60
code enforcement rules were constructed, capable of
forcing the renovation of a specified fraction of the sub-
standard housing in particular neighborhoods. These and
a host of other elements led to an eventual requirement
for some 15,000 data input items.
e. Simulation Structure
61
direct physical changes. such as public construction; or
they can represent regulatory changes. such as those
effecting zoning laws. code enforcement. etc. Finally.
various forms of subsidization. changes in taxes. etc ••
can also be scheduled in and out of the model at the start
of any given time period.
62
recomputes yields, etc. Successive recycles characteris-
tically result in smaller numbers of transitions being made.
and the total transitions can be limited to represent a
variety of facto rs. such as: unavailability of mortgage
funds; unavailability of labor without use of overtime and
resulting in high costs; lack of information by owners.
resulting in slower changes and construction than would
arise in a perfect market.
63
res ults. Many of the s e will match the expectations of com-
mon sense and experience, but a few surprises will arise.
When these surprises occur, it is necessary to trace them
back through the model interactions to establish why they
arose. In some situations (particularly in the early stages,
when programming and input data errors were common) it
could be determined that the surprises were due to faults
in the model. In other situations, however, it turned out
that the surprises represented the results of a series of
interactions that had not been anticipated, but which once
demonstrated turned out to make very good sense. In this
manner, the validity of the model has been of less impor-
tance than its educational function.
64
corne practical in recent years It must be recognized,
0
65
their near-term use in this manner. As a result, plans
which are developed for a suitable political unit- -the city--
tend to be greatly weakened by their limited treatment of
the hinterlands; while plans developed for a more natural
unit--such as a metropolitan area--simply cannot be imple-
mented because of lack of political cohesiveness.
66
controversial measures related to the subsidization of
housing and education which inspire conflict of opinion in
urban areas. To at least some extent rational analyses of
feasibility and costs can assist in resolving some of this
conflict. and making more clear what true choices are
available to urban dwellers and their political authorities.
The process will be a slow. educational one, but an impor-
tant one for the future.
3. Decentralization of Authority
67
participation of civilians of any sort in some of these de-
partments. much les s analysts who are studying the details
and efficiency of current operations is not apt to be welcome.
The city's services are developed by a host of specialists.
and each of these has been faced with practical problems.
limited budgets. and lack of recognition and response for
many years. Except in a period of crisis ~ it will be difficult
to achieve change. As suggested earlier. the barriers are
beginning to break down, but progress will not be rapid.
4. Continuity of Effort
5. Integration of Specialists
68
capabilities are not widely understood, while his publicity
in recent years has been almost too effective. Most large
scale efforts, therefore. have tended to develop consider-
able internal friction.
69
SIMULATION OPERATION
l
Insert New Population Data;
Execute Planned Public Actions
~
Allocate Available Users to Space;
, Compute "Pressures" of Unfilled
Demands and Resulting Rent Changes
J
List Potential Trans itions of Space Types
and Check for Zoning Restrictions, etc o ,
l
Execute Acceptable Transitions up
to Specific Limits
~
Summarize Results and Move to
New Time Steps
Fig. 1
70
SECTION II
SYSTEMS ENGINEERING
CHAPTER 1
SYSTEMS ENGINEERING-PLANNING
DEAN GILLETTE
Bell Telephone Laboratories, Holmdel, N. J.
INTRODUCTION
73
idea of how to use materials and methods to meet the ob-
jectives, and can be only as realistic as is the creator's
understanding of the underlying technology. Broadly,
Systems Engineering is the art in turning such a concept
into a reality. There are many activities in realizing a
concept, including mathematical modeling. preparation of
manufacturing info rmation. development of amain tenanc e
plan. tooling. installation and testing. The project systems
engineer will help to plan the details in many of these areas.
but, most importantly, he is uniquely responsible for under-
standing the relationships among them and for making sure
that the process is orderly and complete.
This description is recognized to be only a discussion
of the place of Systems Engineering in the innovative process,
and not to provide a definition that could satisfy the mathe-
matician or pedagog. However, just as the systems engineer
must often be satisfied, in early design stages, with describ-
ing the operation of the system in terms of the functions, or
outputs, of subsystems, so will the authors in these two
papers attempt to describe Systems Engineering in terms of
the actions in the output of various aspects of the work.
The two talks are titled. "Planning" and "Implemen-
tation." These phases of Systems Engineering overlap sig-
nificantly in time; further, it is usually impossible. in a
productive organization, to identify one group of people as
the planners and another as the implementers. There may
be organizational or administrative advantage in charging one
group with the planning function and another with implemen-
tation, but if the planners lack a sound knowledge of the basic
technology or of the vicissitudes of implementation, or if the
implementers are not adept at task organization or in attaining
a balance between performance and cost. the system may be
a disaster. either functionally or economically. We can agree
on an operational division. however: planning results in in-
formation, in the broad, not technical, sense, and implemen-
tation in operating equipment. The bulk of the remainder of
this pape r is devoted to the products of the planning function:
information for management, for the implementation function,
and for additional planning. Since several of the papers to be
presented deal with mathematical modeling, the author will
74
take the opportunity to display some of his prejudices on
the value and use of such models in planning. Finally.
because this symposium is being held within academic
chambers, and is attended by educators. the author cannot
res ist making a few concluding remarks about the education
of a systems engineer.
75
of the system with the technical environment within which
it is to operate. When the system is to be a replacement
for an existing one, for example to achieve greater economy,
a plan for introduction must be developed. When the system
is an addition, for example to provide growth or greater
capability, it must be de signed to op erate with exis ting
facilities and equipment. Management must be informed of
any administrative problems. as well as costs, that can
arise with installation of the system.
76
is the first example of a reason why planning involves indi-
viduals with a sound understanding of technology and of the
design and manufacturing process.) The task is complicated
by the need to explore alternatives of system function. The
objectives of the system include a quantitative description of
the capability, capacity, or quality of the system. Variations
in these, with consequent modifications of system design,
have an influence on the cost and risk. Finally, there is a
need to investigate alternative system concepts that may
meet the same general system objectives, perhaps with
greater excursions from the initial quantitative objectives
than are possible simply with alternate designs. If the new
system is proposed as a replacement or an addition to
existing facilities for increased capability, an important
alternative in this last category is simply more of the old
system or a modification of it. A tested system in manu-
facture is often difficult to beat on economic grounds simply
because of the initial costs of a new system including those
of tooling, stockpiling, maintenance instructions, and
operational training.
77
should the planners propose iTIlpleTIlentation. The answer
is siTIlpler to state than to provide in practice: when a
useful service or function can be attained econoTIlically.
SOTIletiTIles technology suggests the functions. It is argued
that Bell discovered telephony rather than discovering a
way to provide audio cOTIlTIlunication at a distance. Edison,
however. actively sought and found the TIlechanisTIl of the
incandescent light. SOTIletiTIles technology lags. In 1945,
A. C. Clarke in a science-fiction story, installed a COTIl-
TIlunication relay on an earth satellite. At that tiTIle. arti-
ficial satellites were still only described in advance planning
docuTIlents; the technology to iTIlpleTIlent theTIl was, at best,
iTIlperfectly understood. However, the idea of a cOTIlTIluni-
cation satellite was attractive, if it could be built econoTIli-
cally. Soon. however, technology had caught up with the
concept, and, in 1954, J. R. Pierce described the COTIlpO-
nents of a cOTIlTIlunication satellite, and cOTIlputed t~e per-
forTIlance required of it. The Echo and TELSTAR experi-
TIlents followed.
78
engineer- -one who designs while conteTIlplating his feet on
his desk- -will stop with such a qualitative description.
The job of the planner that truly contributes to the creative
proces s only starts with identification of subsysteTIls. He
TIlust continue treating each of the subsysteTIls as a cOTIlplete
systeTIl in itself by describing its functions, and by quanti-
fying its objectives, now called "perforTIlance requireTIlents. II
79
requirements for them, but he must stop short of detailed
design. Perhaps the stopping point is at a subsystem
level at which all details are comprehensible to a single
man--he who has the responsibility for implementing
that particular subsystem. The complexity of today's
systems, and the scope of today's technology are so great
that the observation is not vacuous. The single man that
can con.ceive, plan and implement a modern, complex
system is only the hero of a science-fiction novel.
80
anticipation and pre-design for compatibility; or to identify
a service or function that cannot be fulfilled economically
with current technology. but which can with potential ad-
vances.
81
twenty years from now, when applicable technology may be
based on laboratory discoveries not yet made, but when it
is likely that economic pressures in evolution will be such
that the customer's display tube must have, for compati-
bility. the same bas ic parameters as will the first.
82
action is insensitive to a parameter value change. The
purist may say that such points and regions are associ-
ated with extrema. Such extrema may fas cinate the
mathematician, but to the systems engineer, they repre-
sent regions in which alternatives of action, perhaps not
quantified in the model, are most restricted or most free.
83
logical fashion.
84
CHAPTER 2
SYSTEMS ENGINEERING-IMPLEMENTATION
JOSEPH F. SHEA
Office of Manned Space Flight, NASA Headquarters
85
was required.
B. The body of knowledge was sufficient--in other words.
the fundaITlentals were there--but the understanding
of that available inforITlation was not fully utilized.
In other words. either the systeITl concept or the
detailed design was not as good as it ITlight, in retro-
spect' have been.
C. The iITlpleITlentation of the design- -in other words.
the translation into hardware--did not faithfully
reproduce the design intent.
D. The checks and balances which should be applied in a
reasonable ITlanageITlent structure to guard against
any of the above three were either inadequate or
iITlperfectlyapplied.
86
type of pitfalls outlined above and to provide an atmosphere
in which specific decisions can be considered in the total
context of the goals of the project.
87
country, must be set in motion. Inevitably it will involve
almost every element of our society- -from Capital Hill to
the least supplier of raw materials or piece parts. The
program will be so large that government funding is manda-
tory, and the complex process by which such vast enter-
prises get started must work itself out. The program will
probably follow the awkward, organizationally difficult pat-
tern of management by a government technical agency and
implementation by several tiers of industrial contractors.
88
Quite the contrary. the most promising is probably
a return to oldfashioned common sense in both the planning
and implementation of the program; above all, a recognition
that. all fads notwithstanding. the best management system
is competent people motivated to consider what can be
referred to as the "total management II problem. Again 9
another definition of systems engineering.
89
management task.
90
B. Design a test program which assures that all elements
of the system meet the requirements specified--and
monitor the results of this program to validate com-
pliance or note deviations which must be corrected
or compensated for elsewhere in the system.
C. Maintain the system specification. both performance
and interface, to pass on and control all recommended
changes so that overall system integrity and compati-
bility are assured. This area ultimately involves
everything from overall performance to minute details
of electrical or structural design which can affect
compatibility between individual systems.
D. Plan and observe the ultimate use of the system to
certify adequacy or spotlight areas where improve-
ments in design or operational concept must be
achieved.
91
to attempt to solve some of the fundamental problems which
plague large numbers of people attempting to work toward
a complex end by inventing management systems which are
intended to compensate for individual competence by forcing
stereotyped decision processes and standard, fool-proof
data gathering methods.
92
before it is set. Such a process would extend the definition
phase unnecessarily. Fortunately, the number of problems
which have to be cons ide red in this area is usually small
enough to be handled efficiently at the time of qualification
if sufficient visibility and understanding exist in the program
structure.
93
B. Assure that good practices are used in processes.
C. Assure that deviations in configuration or perfor-
mance are noted and either corrected or deemed
acceptable by responsible personnel.
D. Provide a record of the deviations for future review.
94
The above philosophy can be summed up in a phrase
which appears. at least in Voltaire. and can probably be
traced back much further in the evolution of human thought:
"The better is the enemy of the good. "
95
each problem they are faced with solving, but who would be
surprised to have so formal a name applied to their habits
of thought. The situation might be compared to that of a
gentleman who, at a somewhat advanced age, took a course
in English literature and was amazed to find he had been
"speaking prose" all his life.
96
SECTION III
DYNAMIC PROGRAMMING
CHAPTER 1
DYNAMIC PROGRAMMING:
A RELUCTANT THEORY*
RICHARD BELLMAN
Departments of Mathematics,
Engineering and Medicine, USC
99
There is a large group of people, of course, who
call themselves "philosophers" without any stated occu-
pation. My feeling is that no one should be allowed to
practice under the title of philosopher unless he is a
professional in some field. To engage in philosophy with-
out actually being a physicist, mathematician, artist.
engineer, businessman or politician is very much like
being the smile of the Cheshire cat. A major need in the
modern world is to have scientific philosophy discussed
by those who have actively engaged in science, and pre-
ferably are still engaged in scientific research.
100
In view of the foregoing. I thought that it Inight be
Inore interesting and rewarding to chscribe the develop-
Inent of dynaInic prograInIning as it really occurred~ rather
than to speak apparently sensibly, logically and profoundly
about the concepts of dynaInic prograInIning. The rational,
carefully built-up discussion would be alInost entirely
false. Those of us who teach have a guilty conscience
about this Inost of the tiIne. when we cOInpare the SInooth
flow of a polished lecture given after ten or twenty years
of work in a field with the agonizing effort and the Inany
false steps of our own early work. It is a bit unfair to
students who have a tendency to feel that their abilities
are inferior if they do not grasp the Inaterial the first tiIne
as effortlessly as we present it.
101
was rn.uch easier to do a quick respectable thesis in stability
theory than in analytic nurn.ber theory.
102
great deal of attention being paid to the theory, and possible
application, of an intriguing new theory. the mathematical
theory of games created by Borel and Von Neumann-
Morgenstern. Closely connected with this was the area of
"strategic bombing." This was of major interest at the
time since it was widely felt. (and correctly so in my
opinion), that only SAC, (the Strategic Air Command).
stood between Western Europe and Stalin's troops. My
interest in strategic bombing was natural in view of my
wartime work in the field of electronics with reference to
radar. and my participation in the development of the
atomic bomb at Los Alamos.
103
lation was that of a single-stage process; one bet, one
decision, and the game was over.
104
of action. Although. as indicated the ongms of the problem
were accidential. I was quite well prepared to study them
because of my general interest in functional equations.
differential-difference equations and the Abel-Schroder
equation from stability theory. theta functions from analytic
number theory. renewal equations from the theory of
branching processes. Thus. it was very natural to envisage
the use of functional equations to handle multistage decision
processes. It was not difficult to obtain an analytic formu-
lation by means of functional equations of unconventional
type~
105
implies the possibility of a new approach to many types of
equations.
106
the proper point of view of the mathematician. I was no
longer primarily interested in merely being useful, despite
my gratitude. I sensed the possibility of an entirely
different approach to all kinds of interesting processes
with the reward of being able to treat new types of functional
equations. I was not asking the usual question: "Do you
have equations for me to work on and possibly solve?"
I was asking a new que s tion, "Is it pos sible that many of
your fundamental problems involve multi-stage decision
processes that I can formulate in a new. and possibly.
more effective fashion?"
107
the fateful decision to leave the safe confines of academe
for the more exciting halls of the Rand Corporation in
Santa Monica.
108
forced to accept the fact that apparently simple decision
proces ses lead to intractable analytic equations. I should,
of course, have expected this On the basis of my experience
in the theory of differential equations and in the field of
mathematical physics in general. Simple physical processes
lead to nonlinear differential equations of simple form which
cannot be readily handled analytically.
109
subsequently. This is a significant part of the story of
scientific methodology. It is usually, if not always, im-
possible to predict where a theoretical investigation will
end once started. But what one can be certain of is that the
investigation of a meaningful scientific area will lead to
meaningful mathematics. Inevitably, as soon as one pur-
sues the basic theme of obtaining numerical answers to
numerical questions, one will be lead to all kinds of inter-
esting and significant problems in pure mathematics.
110
oneself solely to a descriptive theory of celestial mechanics.
However, in connection with investigations in economics,
engineering and operations research it was equally natural
to ask. "How can I make the proces s operate in a better
fashion?"
111
The mathematical structure of these perplexing
S
analytic problems was quite open. The problems read:
T
(2) Max g(u,v} dt ,
v o
where
du
{3} {a} dt = h(u,v), u{O} =c
(b) 0 .s v :s; k{u} •
What was remarkable was the level of analytic intricacy of
solution introduced by simple constraints. These con-
straints were an essential part of the background» intro-
duced by immediate economic, engineering and military
cons iderations.
112
restrictions none pointing in any particular direction. This
is pertinent to a comment made by Felix Klein, the great
German mathematician, concerning a certain type of mathe-
matician. When this individual discovers that he can jump
across a stream, he returns to the other side, ties a chair
to his leg~ and sees if he can still jump across the stream.
Some may enjoy this sport; others, like myself, may feel
that it is more fun to see if you can jump across bigger
streams, or at least different ones.
113
made at all.
114
This is basically a very difficult way to carry out research.
It is very easy to change the forIn of an equation in a large
nUInber of ways. The great Inajority of the new equations
are not Ineaningful. and. in cons equenc e. lead to both diffi-
cult and uniInportant probleIns. SiInilarly. there are Inany
questions that are difficult to answer, but hardly worth
asking. The well-trained InatheInatician does not Ineasure
the value of a probleIn solely by its intractability. The
challenge is there, but even very sInall boys do not accept
all dares.
115
Upon first gazing upon the complexities of the real
world t there is a certain temptation to return to number
theory. Number theory t however. is not rewarding enough
for continual effort. The problems are too difficult. and
the victories too few. Taking up the challenge of complexity.
I felt that the appropriate thing to do was to start with deter-
ministic control processes and to modify them stage by stage
to obtain theories which could be used to deal with basic un-
certainties in a more sophisticated fashion.
116
suffering from a serious disease. You are given two new
wonder drugs which have not been extensively used before,
which means that you possess only fragmentary knowledge
of their efficacy. How do you administer these drugs to the
patients! What procedures do you employ, keeping in mind
the administration of the inferior drug can result in the death
of a human being?
117
with speculation about "giant brains." Stuart Dreyfus and I
who had been trying to use computers to solve simple deter-
ministic control processes, as mentioned above were rather
miffed by these extravagant claims. There were only two
possibilities we could think of: Either these individuals
were making wild statements, or they were very much
smarter than we were. We couldn't even use computers
as effectively as we desired to solve these relatively simple
decision processes arising in control theory. Dimensionality
was, and is, a formidable barrier.
118
One gets quickly into the general area of decision processes
because when you have to use a cOlllputer, you can't wave
your hands. You can't say, "Behave in a reasonable fashion.
You know what I lllean." You have to be precise, "What's
the input? What are the data, what are the state variables.
what processing does one do with the state variables. how
doe s one evaluate, and so forth." C on seq uentl y, I feel as
far as scientific lllethodology is concerned, the biggest single
advance has been due to the developlllent of the digital COlll-
puter. For the first tillle we have a good chance of getting
rid of a great deal of hot air. This doesn't lllean that other
evils won't spring up in their place--but at least they'll be
different ones. We can put up with different ones because
as I said before, we can stand allllost anything rather than
be bored. There will be different difficulties and there will
be lllore sophisticated difficulties. It's going to be a great
deal harder to know what to do with the charlatan of five or
ten years frolll now. He's going to be a quite different
charlatan frolll the Morgan character in "THE WIZARD of
OZ." He is going to be a person who has his PhD, all of
the appropriate acadelllic titles, knows how to use a COlll-
puter, and has just a little kink loose SOllle place. But it's
not going to be so easy to find out where that kink is. It's
going to be a llluch lllore difficult task, and the general public
doesn't even realize the existence of the high level charlatan
yet. I could nallle SOllle nallles of people I'lll thinking of,
but you probably have SOllle of your favorites in lllind.
119
begin at the top. The question is that of distilling and sim-
plifying so that it can be made available at the elementary
levels, in high school and in college. This requires an
entirely different point of view and an entirely different
ability from the person who may have been able to develop
the theory or use the theory. This is a skill all in itself.
Since we don't know how people learn and we don't know how
we teach them what they do learn, it remains a very diffi-
cult problem.
120
way, even by computers we can think of ten years from
now. which will be a billion times more powerful. What
we now need is the same type of sustained effort in dynam-
ic programming that is required to do something meaning-
ful let's say, with the many-body problem in quantum
mechanics, with problems in celestial mechanics. and so
forth. We need very sophisticated approximation tech-
niques based upon intimate knowledge of the actual physical
processes. This means that progress will not be dramatic.
121
This leads to the question: What is the explanation of this?
There must be some reason, there must be some way of
processing the information we use for decision-making.
What techniques do we use? Is it an entirely different kind
of mathematics, is it something based upon gestalt rather
than just single bits of numbers? What are the state vari-
ables? Nobody knows. This is one of the outstanding
challenges in the explanation of human behavior. Again the
interesting feature of this study is it widens the horizons
of mathematics. It gets into the areas of biology and medi-
cine, it gets into the areas of psychology and psychiatry,
and it introduces entirely different kinds of problems.
122
CHAPTER 2
ROBERT KALABA
The RAND Corporation
I. INTRODUCTION
123
II. DYNAMIC PROGRAMMING AND OPTICS
(1) :y [ n ( x , y 0 ? J - :x {:y' [ n ( x , y ) p ] } = 0,
124
Next we expand in a Taylor series to obtain
(4) w(x,y) = :min {n(x,y).6 + w(x,y) + w (x~y).6 cos a
a x
+ w y {x,y).6 sina + O{.6) }
By subtracting w{x,y} fro:m both sides of the equation
above, dividing by .6, and letting .6 tend to zero. we
obtain the basic relation
x
This equation is actually a novel for:m of the eikonal
equation{3 ,4) 0To see this note that the :mini:mizing value
of e is obtained by setting the derivative of the su:m in
brackets with respect to a equal to zero. This yields the
opti:mal control law, in auto:matic control theory parlance,
(7) o= - w s in a + w cos a
x y
or
(8) tan e = w y (x.y)/w x (x,y) 0
( 12) o = dw = w x dx + w dy
y
125
or
(13 ) dy/dx= - w /w •
x y
The fact that the product of these slopes is -1 shows that
the wave-fronts and trajectories are orthogonal.
126
(22)
where the functions get) and k(t,y) are given, and the
function u is to be determined. A direct approach by
dynamic programming is given in reference (6). Here we
shall sketch an approach in which no optimization is in-
vol ved. but we do imbed the original problem, involving
the interval (0 ,c), within the class of problems for all
intervals (0 ,x). The equation (1) is rewritten
o
Furthermore, for ease of exposition. we shall limit our-
selves to the case
127
which is of prime importance in the theories of radiative
transfer(7) and multiple scattering. It is assumed that the
kernel k can be written in exponential form and is even.
rb -rz
(4) k(r) = J a e w(z) dz, r > 0,
=f
b
(9) ([>(t,x) J(t,x,z)w(z) dz •
a
In deriving equation (9) use has also been made of equation (3)
for the function J and the evenness of the kernel k. Thus
([>(t,x) is conveniently expressed in terms of J(t,x,z).
128
Finally, we explain how to determine J(x,x,z). According
to equation (3) for t = x, we have
=1 + J
b
(13) J(x,x,z) R(v.z,x)w(v) dv •
a
We now obtain a differential equation for the function R.
Differentiate equation (12) with respect to x to obtain
130
REFERENCES
INFORMATION THEORY
CHAPTER 1
J. R. PIERCE
Bell Telephone Laboratories, Murray Hill, New Jersey
ABSTRACT
135
"The Mathematical Theory of Communication,"
which Claude E. Shannon published in the Bell System
Technical Journal of July and October, 1948, was a re-
markably prescient paper. It gave us information theory
in the form in which we know it today. It proved the funda-
mental theorems, and it pointed out the areas in which
future problems lie.
136
For exaITIple, the sender ITIay hand in a longhand ITIessage
at a telegraph office; the recipient will receive a ITIessage
ITIechanically printed in block letters.
137
ter:ms. it is co:mpletely redundant.
138
the presc ribed level.
139
transmitter and receiver or between a directive microwave
transmitter and a directive microwave receiver on a hilltop
tens of miles distant.
140
the nature of the errors in the discrete channel. Shannon's
work tells us that errors can be corrected without trans-
mission of information from the message destination back
to the message source. However. if the communication
channel is merely interrupted infrequently for short periods
of time, a very effective means of avoiding error is to send
a message back to the transmitter. saying that a certain
stretch of the received characters had been lost or badly
garbled, and asking for a retransmission of this portion of
the message. Infrequent but powerful bursts of noise do
indeed garble transmission in just this way in some trans-
mission systems, such as local telephone lines. In such
cases, retransmis sion is often used in correcting errors.
141
require a huge dictionary, in which we would look up each
received block of digits in order to find the shorter trans-
mitted block. This is impractical. Other codes use parity
checks of the sort I have described in connection with Fig. 3.
Some of the best of these codes are called Bose-Chandhuri-
Hocquenghem codes.
142
times a second. It can be shown that this many samples
completely and perfectly describe the signal. Knowing the
sample amplitudes, we can reconstruct the signal exactly.
143
as great.
144
Now consider the same transmission circuit of band-
width B, but assume that we wish to transmit symbols from
an alphabet of 37 symbols at a rate of B symbols per sec-
ond. We can represent each message symbol by a point, as
shown in Fig. 5, and encode it by means of the two signal
samples which are given by its abscissa and ordinate. We
decode the noisy received signal as the point nearest that
given by a pair of received signal samples. As in the ana-
log case, the decoding fails disastrously if the noise added
in transmission is too large.
145
"quantum noise" appears in the output of the amplifier.
146
Table I. Binary codes for English text
Li p.
1
Code 1 Code 2 Code 3
147
In order to encode text with greater efficiency we
must know much more about English text than the proba-
bility of occurrence of the individual letters. We could
achieve a greater efficiency by encoding text word by word,
using Huffman's procedure to obtain codes for the various
words. This is not the limit. By an ingenious procedure,
Shannon estimated that in principle only about one bit per
letter should be required to encode English text.
148
specifying whether each tiny area of the mes sage is black
or white. Indeed. one system was devised as early as
1957 by W. S. Michel, W. O. Fleckenstein and E. R.
Kretzmer. This would reduce the number of digits re-
quired for transmission to around a tenth the number
required in conventional systems. Such systems have as
yet found no commercial acceptance.
149
noisiness of television pictures is set by the sharpness of
the eye.
150
Are we being misled? Suppose, for instance, that
we consider the use of television in conversing with another
person. Suppose that we don't try to transmit a continuous
picture at all. Before the conversation, we will transmit
a description of the talkers, a description detailed enough
so that we can make little models of the talkers' heads and
shoulders, and clothe and light these models properly. A
model will be so constructed that the head can move, the
eye swink, and the lips and jaws move, with the s arne con-
straints as those of a human being. In order to transmit
a satisfactory picture we need merely find out how a person
who is talking moves the various parts of his face and head,
and have the model at the receiving end reproduce these
motions. This would seem to reduce the channel capacity
needed to transmit all the info rrnation about the person
talking to around telegraph speeds, rather than television
speeds, which are perhaps 10,000 times as great.
151
of the talker's voice. The important thing is that the
fidelity criterion of the message destination is satisfac-
torily met by this synthetic speech. Unfortunately. the
vocoder tends to have an electrical accent which. along
with its great cost. militates against its general use.
152
cryptography.
153
equal probabilities. As we see in Fig. 7, the groups of
letters are larger than English words, and they do not
resemble English words.
154
they may have tried to include in grammar something which
would be easier to include in the general idea of an orderly
world and sanity and sense in speaking about it. Informa-
tion theory and linguistics may yet corne meaningfully
together in discussing the probabilistic constraints of
grammar and a meaningful environment on the utterances
of man.
155
REFERENCES
156
10. C. E. Shannon, "Communication Theory of Secrecy
Systems, II Bell System Technical Journal, Vol. 28,
pp. 656-715, October, 1949.
157
INFORMATION TRANSMITTER RECEIVER DESTINATION
SOURCE (CODER) (DECODER)
SIGNAL RECEIVED
SIGNAL
MESSAGE MESSAGE
NOISE
SOURCE
Fig. 2.
0 0
0 0
0 0 0
Fig. 3.
OUI
01lJ z.J
z.J oD..
on. ()~
u:::t UI<I
IIJ« mm
"'''' • • •
• • •
• • 0
• • 0
FIRST
SAMPLE
0 0
• • •
• •
0 0 0 0
Fig. 5.
Fig. 4.
158
5
~
t(
,
4
\\
Io-UPPER
\\ ~
'~ BOUND
3 \
\ I. ~
f!!
I-' CD ~
l ,
lJ1
~ 2
~
"" ....
~
,
LOWER .........
.. ...,A ) l
BOUND
""" r----. - ~
N """'" . . r---- ) l
(
-
I (
Fig. 6. Upper and lower experimental bounds for the entropy of 27-letter English.
10 Zero-order approxiInation (symbols independent and
equi-probable) •
XFOML RXKHRJFFJUJ ZLPWCFWKCYJ FFJEYV-
KCQSGHYD QPAAMKBZAACIBZLHJQD.
Fig. 7
160
5. First-order word approximation. Rather than continue
with tetragram ••••• n-gram structure is easier and
better to jump at this point to word units. Here words
are chosen independently but with their appropriate
frequenc ies.
REPRESENTING AND SPEEDIL Y IS AN GOOD APT
OR COME CAN DIRRECT NATURAL HERE HE THE
A IN CAME THE TO OF TO EXPERT GRAY COME
TO FURNISHES THE LINE MESSAGE HAD BE
THESE.
Fig. 8
161
ABSTRACT
Fig. 9
162
CHAPTER 2
R. W.LUCKY
Bell Telephone Laboratories, Holmdel, New Jersey
ABSTRACT
163
devoted to an exposition of current coding techniques. In
these techniques extra, redundant, digits are added to the
message before transmission. At the receiver this redun-
dancy is used to correct errors which may have occurred
during transmission. The use of coding represents a
practical approach toward the ideals of information theory.
The rapidly decreasing cost of physical implementation and
the increasing need for digital communication promise new
uses of infQrmation theory in communications technology in
the near future.
164
Information theory started about two decades ago as
a technical tool for the communication engineer. In the
early years it received a great deal of publicity as scientists
attempted the application of information theoretic concepts
to many seemingly-alien fields. Although some of these
applications have withstood the test of time, many more
have proved to be only superficial. In the late 1950's a
journal devoted to information theory, decrying the flood of
worthIes s interdisciplinary paper s submitted to it. editori-
alized about the gene ric paper entitled "Information Theory,
Photosynthesis and Religion." In this type of paper the
author describes the close relationship between the concep-
tual framework of information theory and. e. g •• psychology.
"Having placed the discipline of psychology for the first time
on a scientific base. the author modestly leaves the filling
in of the outline to the psychologists. "
165
practical aspects of digital communication have undergone
an enormous expansion. Yet a large gap exists between
the theoretically possible and the practically achievable.
In fact. it is still difficult to point to substantial contribu-
tions of information theory to the practice of digital com-
munication.
166
The Theory
167
upper bound to the achievable error probability.
168
decision theory.
169
capacity.
170
Quantized Signal Samples - Modulation and Coding
171
of error for the vector itself is small.
172
Time Series (5) Although this report was affectionately
0
173
telephone channel. Similar gros s inadequacies of the
Gaussian model are found on high frequency radio and
tropospheric channels. On the other hand the deep space
and, to a lesser extent, the satellite channel are fairly
well characterized by the Gaussian model. In the case of
the telephone and h. f. radio media the real channels do
something other than just add Gaussian noise to the
transmitted signal. Probably their most striking charac-
teristic is that they also distort the signal in a linear (but
possibly time varying) manner. This distortion is a form
of time dispersion created by the arrival at the receiver of
various frequency components of the signal at different
times. In the radio channel the signal may also arrive via
a number of diverse electrical paths - -the same phenomenon
which produces ghosts in television pictures.
174
structures. (8) In these systems the receiver attempts to
learn the channel characteristics as closely as possible at
any given time and to redesign itself for optimum reception
on the particular channel it has identified. Adaptive methods
have considerably improved practical system performance
on h. f. radio and telephone channels. In the telephone
example which we mentioned previously data rates of 10 ,000
bits per second are obtained with the aid of adaptive receiver
circuitry.
Block Codes
175
so that code words can be calculated from the input infor-
mation digits. For this reason coding theorists have turned
to the theorem'S of modern algebra dealing with groups,
rings, and fields. Recently, codes have also been formed
based on projective geometry, perfect difference sets and
graph theory. In certain situations the best existing codes
have even been found by computer-aided trial and error or
search procedures. Although many portions of coding
theory possess a beautiful mathematical framework, the
field as a whole seems to have developed in patchwork
fashion when compared with the original foundations of
information theory.
P1 = d 1 + d Z + d 3
Pz =d Z + d 3 + d4
P3 = d 1 +d Z +d 4
Since we are talking about binary digits {bits}, the arith-
metic used for the three parity check equations is modulo-Z,
=
i. e. , 0 + 0 =
0, 0 + 1 1, 1 + 1 =o. The check digits are
called parity bits because they check on the evenness or
oddness of the sum of the number of 1's in a certain subset
of the information digits. For example, in the (7,4) code
the first parity bit p 1 is 0 if the sum of l' s in data bits
d 1 , d Z ' and d 3 is even and is 1 when this sum is odd.
176
0000 000 0100 111 1000 101 1100 010
0001 011 0101 100 1001 110 1101 001
0010 110 1001 110 1010 011 1110 100
0011 101 0111 010 1011 000 1111 111
We can demonstrate many of the properties of more com-
plicated codes with this simple example. We have already
seen that the process of encoding consists of calculating
the three parity check bits from the four input data bits.
Practically this can be accomplished with an extremely
simple 3-stage feedback shift register circuit. This is a
circuit with three binary storage elements and means of
shifting the contents of one storage unit to the next at
regular bit intervals. Two feedback connections enable
the circuit to calculate the required parity check bits on
successive shifts of the contents after the data bits have
been loaded into the stores.
177
error. Each nonzero word in the code is at least distance
three from the all-zero code word since every othe r code
word has at least three l' s. This holds true not only for
the all-zero code word, but between any two code words.
The code is said to have minimum distance 3 and can clearly
be used to correct all single errors. Indeed the rationale
for the code construction rules was to create a set of 16
code words which were as far apart as the given amount of
redundancy would allow- -in this case a minimum distance
of three between any two code words.
178
another code word by the channel--and this requires at
least three errors in the (7,4) code.
179
Once again the spectre of exponential growth in
complexity with block length afpears, for it seems neces-
sary to use a table listing Zn- different syndromes and
their associated error patterns. Actually for some short
codes a similar procedure is used in which all correctable
error patterns are added in turn to the received sequence.
The syndrome of the resultant is calculated in each case
=
if it is a code word. In the (7,4) code there are only Z3 8
correctable error patterns, each of which could easily be
tried and tested by the decoder. However, for longer and
more powerful codes so many patterns would have to be
tried that it would be physically impossible for the decoder
to complete all the calculations before the arrival of the
next code block. For these longer codes it is necessary
to have some way of calculating the error pattern directly
from the syndrome. Just how this is done depends upon the
structure of the particular code. In some cases it is re-
markably simple; in others no constructive procedure is
known.
180
words of the {7,4} example. The code words form a group
under the operation of addition. (The sum of any two code
words is another code word. the addition is associative»
the all zero code word is the identity element, and each
word has an inverse within the code.) Also, the code is
cyclic. Any cyclic shift of a code word results in another
code word. For example. a one unit right shift of the code
word 0100 111 gives 1010 011, which is als 0 a code word.
181
of simultaneous nonlinear equations. Late r a method us ing
Newton's identities was discovered which required the in-
version of a matrix and a relatively simple substitution
process. A number of modifications of the decoding pro-
cedure have been proposed in recent years. Still, however,
decoding BCH codes of even modest lengths and error cor-
recting abilities is not easy from an equipment complexity
viewpoint. While many coding systems have been con-
structed which use BCH codes to detect errors, only a very
few systems have been designed for forward-acting error
correction.
182
some will not. A "vote" is taken to see if the majority are
satisfied. If so. the bit is accepted as correct; otherwise,
it is changed. Unfortunately. the families of codes which
have been found to be majority-logic decodable are sparsely
populated and generally less efficient than comparable
BCH codes.
183
techniques designed for random error correction are used.
184
systeITl was iITlpleITlented with 210 logic transistors and a
2800-bit store. In tests it reduced the error rate of a 2000-
bit per second data systeITl froITl 1.2 X 10- 5 to 8.5 X 10 -7 •
The latter figure is sOITlewhat deceptive since decoding
errors. when they do occur, tend to occur in large bursts.
Actually the ITlean tiITle between these error bursts at the
decoder output was about 12 hours.
185
our example code) sliding window. Each sub block of two
bits is corrected based on the information within the de-
coding window. With these bits corrected the window moves
two more bits to the right and corrects the next subblock,
etc. This method of decoding ignores information which
is outside the window at any given time. Also, because
previously corrected bits are used in the decoding of sub-
sequent bits. any errors made in the decoding process tend
to propagate, i. e. , to make subsequent decoding errors
more probable.
186
nonzero code rates. On the other hand. the next method of
decoding convolutional codes which we shall describe,
sequential decoding, enables the attainment of arbitrarily
small error rates at significant fractions of channel
capacity. (17)
187
theory the decoder never really makes a decision at all.
since later it may have to change its mind about an earlier
decision. In practice of course the decoder must turn over
decisions to the user after it has proceeded well enough
past the given bit to feel reasonably sure about its decis ion.
While the bit is in limbo waiting to be released it must be
placed in a storage unit in the decoder known as a buffer.
188
lations to decode a given bit decreases only as N-f3.
Therefore the buffer ITlUst be extremely long (or the arith-
metic unit very fast) to keep the probability of buffer over-
flow small. The sequential decoding method solves one
problem of exponential growth, but introduces another
event (buffer overflow) which cannot be made exponentially
smaller by additional complexity.
189
Conclusion
190
ignored by theo rists and syste:ms designers. This factor
:may beco:me i:mportant in future co:mputer-to-co:mputer
co:m:munication.
191
REFERENCES
192
Tufts, D. W., "Nyquist's ProbleIl1--The Joint
OptiIl1ization of TransIl1itter and Receiver in Pulse
AIl1plitude Modulation," Proc. of the IEEE~
March, 1965.
193
13. Burton, H. O. and Weldon, E. J., Jr •• "An Error
Control System for use with a High Speed Voiceband
Data Set." National Communication Conference.
Boulder, Colorado, June, 1965.
194
a. Message
101
b. Corresponding waveform
formed from 2 samples
c. Transmitted vector
d. Received vector
t Noise
I ,
I
I ""
I
Figure 2
10
co
rl
ill
6
~
u
'""
C
-rl
rl
4
'"
ill
'd -----------
H
ill
>
0
"'<>: 2
~
<Il
25 50 75 100
M
Figure 3
196
101
Coded Message
o 1 ----> 10 11
1 ________ _
,,-£1)-
01-------------
00-
Quantized Samples
QUANTIZED SYSTEM
Figure 4
10 20 30
Measured
Pr~b. of Error
Theoretical
Figure 5
197
I~
\\
~
"" \
P(m, 31)
Einom al
f\.
\
o 4 8 12 16 20 24 28
m
Figure 6
d1 d2 PI P2 P3
6 21 26 31
7 22 27 32
8 23 28 33
4 9 24 29 34
5 10 25 30 35
INTERLEAVING
Figure 7
198
Infol"Jlation bIt
C Check bjt;
l' \'
I C I C
CONV;:-r;~J'I'TONAl
l' il
,
I
;~:'JCUDJNG
c
r!Jod-2
SUM
Figure 8
214
10
11
10
11
OJ
r
r----- . . Example Path,
Message 0111
0 r ---
11
10
I
L ___
00
00
11
CODE TREE
Figure 9
199
SECTION V
GAME THEORY
CHAPTER 1
GAME THEORY:
A NEW PARADIGM OF SOCIAL SCIENCE
OSKAR MORGENSTERN
Princeton University
ABSTRACT
203
I. Bas ic Concepts *
One of the principal problems in science is the
development of proper concepts. This task assumes
particular significance in the social sciences where there
is a great wealth of phenomena that are difficult to describe,
to order and to analyze. Only proper, sharply defined con-
cepts can give exactness; data, no matter how good, never
can do this.
204
typical for the social sciences as the great and powerful
tool of analysis.
205
on the arbitrary. seemingly indifferent behavior of nature.
Consequently no statistical assumptions can be made re-
garding lithe others II as one can in the case of nature. nor
can such assumptions guide the participant who is bent on
obtaining a maximum advantage. One cannot play chess
successfully by attributing statistical ideas to the other
player.
206
standing: Game Theory does not assume rational behavior
but wishes to explain it. to give it precise mathematical
meaning. How could it assume something that has neither
intuitive meaning nor is otherwise established! Game
Theory merely assumes that each individual strives for
his greatest advantage and will follow- -within the unalterable
rules of the game- -every lead that he believes will secure
it for him, no matter how small the advantage may be. We
can safely claim that outside of Game Theory general talk
about some behavior as being "rational" or not in situations
that can be and must be modeled by games of strategy has
no meaning within the constraints set by scientific standards.
207
the addition of another player to a given number possibly
transforming the game in a substantial manner. E. g. ,
when n> 2, coalitions among players are possible and
will happen if their members derive advantages from com-
bining. There are games where the sum of gains and
los ses add up to zero; and there are those where that sum
may be positive or negative. constant or variable. Non-
zero-sum games are models for social and economic
relations since they may show productivity or loss. These
games are an extension of zero-sum games to which they
can be reduced by introducing a (fictitious) player who
makes contributions to. or withdrawals from. the game.
as the case may be. a fact that unifies the theory for all
pos sible cas es.
208
one which is both the maximum of the minima which the first
player can ever obtain and also the minimum of the maxima
the second player may ever have to pay. Not strictly deter-
mined games have no saddle point in pure strategies and
there is doubt what optimal play should consist of in such
cases. Von Neumann proved in his famous minimax theorem
that if, however, each player, instead of choosing a pure
strategy, uses his strategies with certain specifiable proba-
bilities, then even these games have saddle points. Thus
there is always a safe way of playing and the correct proba-
bilities can be computed. This is the advice to be given to
the player.
209
chance device will have to be used.
210
an immediate, intuitive appeal, A maximum problem is
deterministic; even when probabilities are introduced, they
are fixed, But the uncertainties associated with the actions
of others who control vital variables in their own interest
are of a different order, and as said before, they cannot be
treated in a conventionally probabilistic manner Yet we
0
211
IV. Randomization and Reality
212
If Game Theory is descriptive then we would expect
to find randomization in actual, individual economic behavior
since it seems reasonable to assume that individuals have
learned to adapt themselves to the underlying game situations
and have discovered that mixed strategies are to be used to
their advantage. They may not have arrived at the precisely
computed probabilities which the complete theory yields, but
they would at least have dis covered the principle. In answer
we note that matching pennies is played according to the
theory and it is a prototype for large classes of games.
213
directors, in order to discover to what extent randomization
has taken place. This is not easy to do. The critical phases
may be submerged in the subconscious--especially if Borel
is right in seeing difficulties which he clearly takes to pre-
sent themselves in the conscious domain. This is part of a
wider task for descriptive work to be done: social reality
described so far is known to us by the use of the old concepts
discussed at the beginning of this paper. But they differ
radically from those of Game Theory. Consequently also
the account of what really goes on according to Game Theory
in social and economic life differs in significant part from
what is currently taken to be important.
214
v. Rules of the Game
215
since their basic assumptions are at any rate and for other
reasons in more profound conflict with reality.
216
think ••• " variety of argumentation still often encountered
but which leads nowhere. Instead t the minimax principle t
being abs olutel y convincing. is the only one to offe r the
way out. It gives advice to each player since it first assures
him his optimal result regardles s of the actions of the other
and lets him only profit should the latter deviate from his
optimal strategy, and second it defines the value of the
game, thereby imparting full information about the objective
possibilities residing in the situation.
2.17
one, taken singly, is dominated by an imputation not in the
solution set, but this dominating one is in turn dominated
by another one belonging to the solution set. The imputations
belonging to the solution set are each separately ringing for
acceptance by the players, but once one of them is chosen
the others protect it by their "virtual" existence. They
offer acceptable alternatives to the players. The solution
set represents a "standard of behavior." All this is des-
cribed by the so-called characteristic function of the game,
which is a numerical set function derived from the applica-
tion of the minimax theorem to the n-person case.
218
organizations. In particular the troublesome old problem
of complementarity of economic goods is now adequately
taken care of by the superadditivity property of the charac-
teristic function mentioned above.
VII.
219
actions and interactions of persons and their consequences.
If these cannot be discovered by a scientifically acceptable
procedure- - i. e. Game Theory- -then there are only un-
scientific methods and this will not do. There is no knowl-
edge outside scientific knowledge. There is "art," "intui-
tion," "faith" and other- -so far- -unanalyzed mental activity;
all may from time to time prove successful. The normative
element of the theory comes through, however: advice is
to be given how to behave optimally. The mere fact that
individuals, corporations and governments take action in
situations for which there is as yet no complete scientific
understanding and for which science can give no correct
advice, does not mean that these men and agencies some-
how, mysteriously, have discovered the secret. They have
learned from experience, of course, but men cannot com-
municate experience easily whether from one generation to
another or to an outside observer. We each live but are
unable to say how we do it, except in the grossest terms.
It takes biology and medical science to tell. And govern-
ments are very hard to describe; the fact that interactions
among individuals take place makes communication even
more difficult. So there is no way out: either we have
scientific knowledge or we don ft. If a theory carries us
only up to a certain point, that is precisely as far as our
knowledge will reach.
220
too is reached and more and more results fall into this
category. In the present paper I can, unfortunately, give
no discussion of these cases. Between these fall numerous
others, all of interest and some surprising in nature. The
following is a partial list, the order in which the items are
mentioned is of no consequence.
221
as disadvantageous for a player as being treated
an equal by the other players, but be ing pushed
by them into a completely losing position.
222
tions strong enough to exhaust the contributions
offered. In that case the social organization of
that game. as represented by alternative coali-
tions, will break down completely. (This may be
taken to be a mathematical description of the
corrupting influence certain forms or amounts
of foreign aid have on underdeveloped. primitive
societies. Similar events may be observed also
for parts of highly organized societies.)
223
scope of game theory beyond those commonly given. They
are but a small segment of what has been discovered. I
use this word intentionally: few of thes e res ults could have
been anticipated. Prior to Game Theory none could have
been proved mathematically or otherwise; most of them
are the consequence of investigating the mathematical
properties and implications of comparatively few and
intuitively transparent assumptions. Let us also recall
that without exception there is behind all of them one uni-
fying theorem, that of the minimax. This is a remarkable
property of the theory, probably of considerable interest
also to philosophers of science.
224
captured by precise concepts. Of course, the mere use of
a new terminology is not equivalent to application of the
theory. The frequent occurrence in many wr itings of the
word "strategy" is not by itself an indication of a proper
use of Game Theory.
VIII. Outlook
225
appear to work in all three directions.
226
process will develop. It is important to realize that the
raw social evidence does not automatically reveal the under-
lying mathematical structure and that propositions from
pure mathematics are only seldom directly useful and
applicable to concrete empirical problems. The "inter-
face." to use a word beloved by engineers. between the
two worlds is where the main difficulty lies. But if that
situation is recognized by both sides it will surely lead to
further interesting and. we hope. practically important
(agreeable!) consequences.
227
CHAPTER 2
M. SHUBIK
Economic Growth Center, Yale University
ABSTRACT
228
A. Economics and Game Theory
229
How should it distribute its goods and services? What is
meant by economic efficiency? What limits are there or
should there be upon the economic powers of the individual
or the group? These and many more are natural economic
or politico- or socio-economic questions. The simplified
model of man combined with game theoretic analysis will
help to give some insights into and partial answers to the
questions. In particular an interpretation is given of the
virtues and limitations of the price system in the freely
competitive market guided by the "invisible hand."
230
which any piece can be moved; they define checkmate and a
dra wand in the cas e of a tournament they as sign a point
value to win, lose, or draw. In an economy obviously the
rules are not so well defined. Nevertheless the technology,
laws, and institutions of society provide the bas is for the
construction of models sufficiently relevant to merit study.
231
choice between two bundles of commodities consisting (for
the sake of the example) of xi apples and Yi oranges or
X z apples and yZ oranges. The first alternative Ai or
(x1'Yi) will be preferred or indiffere:i,t to the second alter-
native A Z Ot' {xz,yZ}' or vice-versa. If we consider all
pairs and we as sume a condition of trans itivity (i. e. , if A 1
is preferred or indifferent to A Z and A Z to A3 this im-
plies that A1 is preferred or indifferent to A 3 ) we may
draw a map as is shown in Figure 1 , where each line is the
contour of alternatives to which the individual is indifferent.
ORANGES
APPLES
Figure i
232
ADMIRAL GENERAL
Figure 2
233
A
GAIN OF 2
GAIN OF I
8 c
Figure 3
234
if we dispense with interpersonal cOITlparisons and trans-
ferability.
Table 1
1. Pareto OptiITlality
235
Figure 4
2. The Core
v(9)=0 v(9)=0
v(1,2,3)=3 v(1,2,3)=3
236
The characteristic function specifies the worth that
a coalition can achieve if its members decide to act in
unison by themselves. For a three person game there are
eight coalitiions (count~the one with no members). In the
two examples above v(1.Z) =
zt means that the coalition of
players 1 and Z in the first game are able to obtain zt
by independent action. v( 1 .3) = 1 means that the coalition
of players 1 and 3 in the second game can obtain 1 by
indepe ndent action.
ai - Z-2.1
+ aZ:::::
a. az + a3 2: zt
a1 + a3 2: zt
b.
Z(a1 + a Z + a 3 ) = 7t
or a1 + a Z + a3 = 3-3/4
237
which is inconsistent with condition (b). This means that
the core is empty. In the second game, however, we have:
a1 + aZ 2: 1
c. aZ +a3 2: 1
a1 + aZ 2: 1
d. a1 + aZ + a3 =3 •
Adding the three inequalities together we obtain:
1
a1 + aZ + a 3 2: 12
238
strategies, one for each player, forms an equilibrium point
if each player, knowing the strategies of the others. is not
motivated to change.
5. The Value
239
solutions which hav~ been suggEjsted such as those of
Shapley. (6) Nash. (7) Harsanyi .l8) and Selten. (9) They are
all cooperative solutions. i. e •• all participants will eventu-
ally cooperate but use the solution to determine their
shares of the final proceeds). These solutions are based
upon the axiomatization of concepts of symmetry. fairness
or equity.
240
To begin with, we assume that the structure of pre-
ferences and technology cause no problems. We examine
the solutions to a simple market model of an economy by
applying all five solutions suggested in Section C. Suppose
there were three traders in an economy, Figure 5 shows
what the solutions might look like:
u,
Figure 5
241
the core.
242
3. The Wrong Structure
4. Conclusions
243
References
245
CHAPTER 3
ON COMMITTEES
L. S. SHAPLEY
The RAND Corporation
INTRODUCTION
246
a specific goal, a "unique factorization" theorem that de-
scribes how a simple game is decomposed into a hierarchical
arrangement of committees, subcommittees, and individual
agents. Since this is a new theorem, we give it a fully
rigorous treatment, and almost half of the following text is
taken up with the sometimes complicated (but entirely
elementary!) proofs that are required to "keep us honest. "
These proofs may be skipped on first reading without loss
of continuity.
247
1. SIMPLE GAMES
248
must also win. A sort of converse would also be a natural
condition:
(1d) if SnT =cf> and S E::W. then T tW.
i. e., any coalition completely disjoint from a winning coali-
tion must lose. But we do not impose this condition a priori.
Garnes that satisfy (1d) are called proper; all others, im-
proper. Tho\:lgh rarely found in application. improper
games play an important role in the structural theory,
somewl;tat analogous to that of imaginary numbers in algebra.
No. of "primes" • • • 1 o 1 6 1 01 1
249
in larger ga:rnes.
2. COMMITTEES
250
consisting of single individuals. By a proper committee we
shall mean one of intermediate size. Obviously. only proper
committees can lead to significant decomposition of the
game structure.
251
THEOREM 1. Let r(C, W C} be a
committee of r(N. W}. Then for every
S E: Wm(C) there exists B E: We such
that SnC =
B. Conversely, for every
B E: We there exists S E: Wm such
that S nC =
B, unle s s the committee is
inessential.
c_-----_ c_-----_
T shaded R shaded
(a) (b)
252
Theorem 1 reveals that except for the inessential
case the elements of We are precisely the intersections
of the elements of Wm{C) with C. Thus the following
corollary holds, justifying our double use of the term
"committee" :
253
therefore repeat the argument of (a), with primed
letters, and obtain T' E:: W. where T' = B' U(S'-C).
But T' S S; hence T' =
S by the minimality of S.
Thus we have
S'-C =T'-C =S-C = T-C.
Also, we have S'nC C We by Theorem 1. But
s' nc S; B; hence
S'nC = B = TnC,
by the minimality of B. The two displayed equations
establish that S' =
T, and hence that T E:: W m •
Finally. since T ~ B ~ cf>, we have T C Wm(e).
This completes the proof.
c c
~----- ......
,,--_~~O~""""'S'
T shaded T' shaded
(a) (b)
254
(4b) (sOC)U(S'-C) E: W m •
255
clumsiness involved in testing committeehood directly from
the definition.
5. SUBCOMMITTEES
c,-______________ ~
T nC shaded
256
(TnC)U(S'-C) = (BnD)U(B'-D)U(S'-C)
= [(SnC)nD] U[ (S'nC)-D] U(S'-C)
= (SnD)U(S'-D)
(same diagram illustrates). A final application of
Theorem 3 shows that D is a committee of r(N, W).
6. CONTRACTIONS
257
0-
c
o
c
258
If DeC, then of cours e D would dis appear in any
contraction on C, and we can infer nothing about D's
conunitteehood. There remains the case where D and C
"overlap," i. e., where the sets CnD, C-D. and D-C are
all three nonempty. This is a more complex situation.
since contraction on C now violates the integrity of D--
some of the players in D are eliminated, others remain.
It is not clear intuitively how the committeehood of D
is related, if at all, to the committeehood (in the contracted
game) of either D-C or (D-C) U{i C }' the two most likely
candidates for comparis on.
259
K[ Hi' ••• , Hm]
respresent the compound game, defined by taking as players
all the players of the component games Hi' and by taking
as winning coalitions all sets that include winning contingents
from enough of the components to spell out, by their indices,
a winning coalition of the quotient game K.
260
Subcommittees can be displayed by using compound
rep res entations as components. For example, the symbol
K[H 1 [G 1 ,G Z ,p,q] ,H Z ,H 3 ,r,s]
reveals that G 1 and G Z are committees of H 1[ G 1 ,GZ ,p,q] ,
which is in turn a committee of the full game.
261
games in much the same way as Boolean operations act on
sets. or as logical operations act on truth values. Two
extreme cases. among the possible quotients, playa special
part in the theory. and it will be convenient to represent
them as operations rather than functions. The first,
denoted by G:>, corresponds to quotients having the maxi-
mum possible number of winning coalitions; the second,
denoted by ®. corresponds to quotients having the least
pos sible number of winning coalitions.
262
winning coalitions are precisely the minimal winning coali-
tions of its components. No minimal winning coalition
meets more than one component. *
Conversely, any game
whose minimal winning coalitions "fit" within a partition
in this manner is decomposable as a sum. It is easy to
see that such a game has a unique representation as a sum
of games that are not themselves sums--we merely take
the finest partition that does not split any minimal winning
coalition.
263
9. OVERLAPPING COMMITTEES
c E3
El
Ez
264
Proof. (a) Assume first that no element of Wm
meets more than one of E l • E Z' Er To show that E
is a committee we must show that {S n E} U(S ,_ E) e Wm
for all S, S' e Wm o Without loss of generality, S'
meets Co
265
Substituting S1 for this set in D, we obtain
(S1nE1)U(SZnEZ}U(S1nE3}U(Q-E) E: W m •
In the committee C z E1 UE Z ' we therefore have
(S1nE1}U(SZnEZ) E:W~ ,
as well as
and
266
COROLLARY. Let G be a simple
game without dummies. Then either
G is a sum or product. or the maximal
proper committees of G are disjoint.
267
decomposition of G. Thus, none of the Gi's are them-
sel ves sums.
268
individual player.
269
The i:mportant point is that the quotients that appear
in the representation are all pri:me. For when a quotient
has a proper co:m:mittee. then players in the corresponding
co:mponents of the co:mpound for:m a proper co:m:mittee of
that ga:me. which necessarily includes the players of at
least two co:mponents in the representation. In the present
case, however, the co:mponents are :maxi:mal proper co:m-
:mittees (or unaffiliated players). so that this cannot happen.
270
SECTION VI
MORPHOLOGICAL RESEARCH
CHAPTER 1
F. ZWICKY
California Institute of Technology
ABSTRACT
273
A. The Origin and Development of the
Morphological Approach
274
In order to achieve all of these goals we must con-
ceive of an approach which allows us to implement and to
integrate our knowledge of all of the essential interrelations
among the physical objects of the world and the physical,
chemical and biological phenomena which govern their
interaction. And further we must clearly understand the
nat ure and interplay of the ideas, concepts and actions of
men. With these goals in mind, the morphological approach
was cori.ceived which concerns itself with the development
and the practical application of basic methods which will
allow us to discover and analyze the structural or morpho-
logical interrelations among objects, phenomena and con-
cepts. and to explore the results gained for the construction
of a sound world.
275
use of which might completely alter the course of the
destinites of men and of nations. For instance if nuclear
fusion of any common materials, such as ordinary rocks
were achieved by a group of scientists in some specific
country, the rest of us might be forced to accept a treaty
about the future of inner and outer space as dictated by
those in power in that particular country. Obviously. the
members of the International Academy of Astronautics. as
well as all men of good will and sound mind will spare no
effort to keep themselves informed on all scientific and
technological progress. or, as we morphologists express
it, on all new pegs of knowledge. Such info rmation will be
indispensable if we wish to construct for our good use a
world image which is true in the sense that it will enable
us to fully comprehend and control the happenings around
us as well as those within us. The visualization of this
true world image can guide us on sure paths to discoveries
and to basic inventions and. last but not least it will enable
us to predict correctly the course of future events and to
adjust our actions accordingly.
276
B. Some Characteristics of the Morphological Approach
277
a. The Conversion of Chemical Energy into Kinetic
Energy of a Missile
278
instance. to get one gram of a payload off the Earth into
interplanetary space requires theoretically 6 X 10 11 ergs
(plus an additional amount of a few per cent to compensate
for the frictional losses of the large rockets in traversing
the atmosphere). m present day practice of launching
spacecraft as the final stages of multiple rockets. per gram
of payload 100 grams of propellant are needed which release
a useful energy of 100 X 1011 ergs. Over 90 per cent of
this available energy is therefore wasted in the conventional
types of rocket launchings of artificial satellites and space
capsules.
279
to be impossibleo As a consequence he will attempt to
visualize all of the possible ways of converting chemical
energy into kinetic energyo In parenthesis I like to point
out that comprehensive reviews of this type are seldom
given in the textbooks or in lecture courses. although they
would be highly instructive. The teaching profession in
science and engineering should seriously occupy itself with
this subject and explore it in all of its vast aspects
0 Here
I must restrict myself to a brief sketch of some typical
considerations which have led to rather startling results.
Diagram
a)
®® - - V ---+-
b)
© - - V ---+-
c)
@® --V--
280
a gun, the missile being in front. This method of launching
has the drawback tha t no muzzle velocity can ever be achieved
which is materially greater than the average thermal velo-
cities of the molecule s of the reaction products at the peak
temperatures reached after the explosion of the pr opellant
charge. For molecules such as N 2 , H 2 0, CO 2 as they
are relased by the common chemical explosives these
velocities are of the order of 2 km/sec. Once the pro-
jectile has reached this speed it will not be accelerated
any further since the molecules are too slow to follow and
cannot transfer any more forward momentum to the missile.
281
energy of the exhaust gases. Under the best conditions,
and using a high energy propellant capable of producing
the assumed exhaust velocity we therefore can achieve
escape velocity only with a thermo propulsive efficiency
of less than 20 per cent. For multi-stage rockets the
losses are still greater since the empty cases of the initial
boosting rockets will be jettisoned into space with con-
siderable kinetic ene rgies with respect to the frame .6,
relative to which the missile was launched. This energy
is entirely wasted, in addition to that carried away uselessly
by the exhaust gases of the various stages.
282
Wall
283
b. Nuclear Fusion Ignition
284
1. The Method of the Morphological Box
We proceed as follows.
285
phenomenological morphology and absolute morphology.
Also, the system of multi-language teaching which will be
discussed in another report to this symposium was in part
developed with the aid of the method of the morphological
box.
286
P4 = Type of thrust augmentation. P 41 = None, P 42 = inter-
nal, P 43 = exte rnal.
P5 = Type of ignition. P 51 = self igniting, P 52 = external
ignition.
= Sequence of operations. P 61 = continuous, P 62 = inter-
mittent.
287
using only first principles of mechanics and of thermody-
namics. I proposed to call the resulting expression the
universal thrust formula(3}
U 2 1/2
I :: U /g::-f3o[f3-1+(1-f3+2f3LH:jU)] (2)
sp e 0
288
power plants and propellants we started at the end of
World War II work on various vast projects of space re-
search. The first one among these was an attempt on
December 17. 1946 at the White Sands Proving Grounds
to fire artificial meteors from shaped charges mounted on
a V-2 rocket. (3) (7) The first man-made objects (slugs of
fused TiC and A1 2 0 3 ) were subsequently successfully fired
into interplanetary space and free of the gravitational pull
of the Earth from an Aerobee rocket which was launched at
Alamogordo. New Mexico,on the night of October 16. 1957.
These may now appear to be modest efforts when compared
with later spectacular achievements, but they nevertheless
remain the first practical attempts to start the SPACE AGEJ3)
289
there is a close connection between these two inasmuch
as during the formation of compact bodies energy must
be liberated in the form of electromagnetic and gravi-
tational waves, corpuscular radiation such as the cosmic
rays and through the ejection of various types of material
bodies.
y) Compact bodies may be formed both slowly or rapidly
(implosively) , the terms slow and fast being defined by
the statement in the following paragraph.
290
of all types, including the BCO, (which are called BSO or
blue stellar objects by some) and the more numerous red
and infrared RCa (RSO) and ICO (ISO), many groups of
compact galaxies, as well as the apparently rare large
clusters of compact galaxies. I could also show that inter-
galactic matter, the existence of which had been vehemently
denied by E. P. Hubble, W. Baade and other experts in the
field(3) is distributed throughout all of cosmic space between
the commpn galaxies and that this matter consists of stars,
dust, gases and plasmas of protons and electrons, in addi-
tion to the cosmic rays. New classes of galactic objects
such as the Humason-Zwicky stars, pygmy stars and certain
variables were also discovered.
291
to negate the absolute certainty expres sed by the uncertainty
principle, (3) followed by the formulation of a law which is in
better accord with the principle of the flexibility of scientific
truth will lead to new fundamental insights in theoretical
physics.
292
large cOITlpared with those for which we have tested it. As
a result of this inquiry I found that there are no clusters
of rich clusters of galaxies. (5) The ITlost direct interpre-
tation of this fact leads to the conclusion that the gravitational
attraction between bodies which are separated by ITlore than
fifty ITlillion parsecs decreases ITlore rapidly than Newton's
law would indicate. Likewise» Euclidian space ITlay be a
good approxiITlation for those liITlited regions within which
we usually operate but it probably does not represent real
space any ITlore if we deal with nuclear diITlensions or the
universe as a whole.
293
6. The Method of Generalization
294
realize his very particular genius can then choose from this
box those tasks to whose solutions his genius can contribute
most. If I were asked as to what I personally have in mind
I should say that I am most interested in the elimination of
all human aberrations. On the way to the realizati on of a
sound and harmonious human society. however. all men and
women will have to make their own unique contributions.
and nobody else will be able to take their place if each one
puts his very own genius to work.
295
REFERENCES
296
(12) Proceedings of the lAU Sym.posium. at Byurakan 1966
(to appear). Also Com.ptes Randus 261, 649 (1965)
and 262, 1566 (1966).
297
CHAPTER 2
A. G. WILSON
Douglas Advanced Research Laboratories
Huntington Beach, California
298
hypothesis generation through morphological analogy.
10 NON-QUANTIFIED MODELS
299
A concrete example of this approach attempted by
Alexander and Manheim may be found in the MIT report
entitled, "The Use of Diagrams in Highway Route Location."
Alexander and Manheim I s problem was to locate the route
for a freeway covering a 20 mile stretch in Massachusetts
starting at Springfield and ending somewhere near North-
hampton. They first morphologically derived all of the
individual abstract forces whose interaction would deter-
mine the path which the freeway should take. Shown on
Table 1 of the freeway design parameters ~ is the goal or
objective of the study. which was a freeway to meet major
current traffic desires. In this case the aggregate solution
was restricted to be a new freeway. rather than a morpho-
logical examination of all possible solutions to meet current
traffic requirements. This new freeway had to be considered
in the context of its interaction with existing freeway systems
and in support of the competition with other transportation
systems. Future transportation systems as visualized also
had to be given representation. However. the largest num-
ber of constituent forces fall into two classes; those which
determine the internal structure and behavior of the free-
way. and those reflecting the interaction of the freeway
with the environment. Table 2 of freeway design param-
eters shows the decomposition of the internal and environ-
mental parameters into their different values. Under
internal parameters. are first the construction parameters
including earthwork costs. bridge costs. pavement and sub-
grade costs. and construction interference. Secondly.
there are economic factors: land costs. public financial
losses. user costs. obsolescence; and thirdly, operational
factors: travel time, local accessability, safety, mainte-
nance, and self-induced congestion.
300
siderations such as eyesores and noise must be considered.
301
II. MORPHOLOGICAL ANALOGY
302
com.plex structures is in the analysis of how complexity
and bigness are treated in the natural order. We observe
throughout nature that the large and complex is constructed
in a hierarchical modular manner from the small and
simple. Direct confrontation of the large and small is
avoided, a hierarchical linkage is always interposed.
Bigness is avoided in the sense that the ratio between the
size of any structure and the modules out of which it is
built is functionally bounded. If there are demands for a
structure to continue to grow in size or complexity. then
a new level in the hierarchy and a new module are intro-
duced so that aggregate to module ratios may remain
bounded.
303
direction of changes in an urban complex. Ernst's approach
might be called a reductionist approach, decomposing the
city into components and sub- components, and looking at
the "portfolio of possibilities." This is an important part
of the analysis of any complex problem. However, the
morphologist wants to add something. There may exist
some parameters which place limits on the portfolio of
possibilities but which are not evident in the reductionist
approach. I would like to look at the city in this alternate
manner. For this purpose the important properties of
hierarchical modular structures to abstract are the bounds
or limits to which the modules and the aggregates may be
subject.
304
in Calcutta in 1756.) However, we :must bear in :mind that
in the :modern city for purposes of density li:mits, the real
inhabitants are :motor vehicles, not people. The :maxi:mu:m
density is deter:mined by the :mini:mu:m space needed for
:maneuvering, parking, and servicing auto:mobiles.
305
Since R::; c7' ::; C7 and a: < iT, where barred 2
quantities are m.ean values, we have N =
aO'R2 < ac 2a:'T <
s
ac 2 iT 7 2 :s ac 2 iT 2 T2. In a three dim.ensional m.odel we
m.ay introduce the m.ean height, n, of the city and use
three dim.ensional densities, p and p, giving
2-~ ~2
N<a'c hpT
If we designate the absolute lim.it p7 2 by l/H, then
HN < a' (1)
c2n
306
where p is the density and B is a fixed bound (w, 'ssume
that G and c are constants), we see that for a given
den s ity- - as for example, mean stellar dens ity- -the maxi-
mum possible radius of a star is determined. Such an
inequality not only defines a limit to stellar size but forbids
close packing of stars in space. Stars can be organized
together into a larger aggregate only if a lower value of p
obtains. If p assumes the mean value of galactic density
the argument may be repeated. The maximum size of a
galaxy is determined by the same bound but with a lower
value of p. The repeated application of a potential bound,
like in the Schwarzschild inequality, can account for the
levels in the hierarchical modular structure observed in
the universe. However, the inequality does not explain
the particular set of pIS which are observed in the universe
nor does it indicate at what level the hierarchical modular
structure may terminate. Potential bounds like the
Schwarzschild li.Init may also be interpreted as bounding
the maximum velocity a module may possess in a coordinate
system at rest with respect to the aggregate. With this
last interpretation, we see from Figure 1 that cosmic
bodies are either "density limited" or "velocity limited. "
The "slope 3" line represents the limiting density of matter
in a non-degenerate form. Solid cosmic bodies lie on or
to the right of this line. (On the logarithmic scales used
in the diagram, the planetary bodies appear to have
ess entially the same dens ities.) The" slope 1" line repre-
sents the observed location of the velocity limited bodies,
Le •• the star, galaxy, cluster, and derived super cluster
having the largest potentials or escape velocities. (This
is an observed potential bound and differs in numerical
value from the theoretical Schwarzschild bound. The
objects falling on the observed bound, like those on the
density bound, are non-degenerate.) The inequalities (1)
and (2) may be put in the respective forms,
PT 2 < B* and pR 2 < B
These inequalities have the same ingredients and we might
expect them to have the same significance even though the
values of the coupling constants are quite different.
307
On the basis of these similarities we might propose
a theorem of the form:
Given
1. The existence of an aggregating force tending to
bring modules into a condition of maximum com-
pactness, (gravity in the case of cosmic bodies.)
2. The existence of a maximum limiting density, (the
limit set by non-degenerate matter in the cosmic
example. )
3. The existence of a potential bound or its equivalent,
(such as the Schwarzschild Limit, in the gravita-
tional cas e. )
then hierarchical modular structures provide a way for
accommodating indefinite size while satisfying these
intrinsic limitations. ¥ecifically we are led to inequalities
of the p R2 < B or p'T < B* type. If we assume we may
apply such a theorem to a city. then frcim 0: R2 S ac 2 ;.;:2.
we see that for a given density, the size depends on a bound
set by the effective velocity of travel and the maximum
acceptable commuting time. The bound may be satisfied
as N increases by increasing c, or alternatively the
solution may be found in hierarchical structure.
308
the bound Illay be satisfied by selecting a density distri-
bution o-(r)- r-(Y+l}where '{ > 1. In this case, the city
Illay grow and still satisfy the bound if it is built in a ring
shape. Several suggestions of this sort have been Illade
including a city which is nothing but a series of linear
structures several stories high with freeways on top.
309
BIBLIOGRAPHY
310
FREEWAY DESIGN PARAMETERS
I
MAJOR CURRENT
TRAFFIC DESIRES
LV OTHER
l OTHER
I-' NEW
I-' TRANSPORTATION 4 ~ 4 ~ FREEWAY
FREEWAY
SYSTEMS SYSTEMS
STRUCTURE L ENVIRONMENT
I ~ FUTURE
TRANSPORTATION
SYSTEMS
--_._- -
FREEWAY DEffiGN PARAMETERS
II
Internal Environmental
Construc tional Physical
Earthwork costs Catchment areas
Bridge costs Drainage patterns
Pavement and subgrade costs Weather effects
Construction interference Air pollution
W Economic Economic
I-'
N Land costs Regional land development
Public financial los ses Local land development
User costs Non - recompensable public
Obsolescence and private loss
GALAXY-_~
+ 10
SEQUENCE
'"
'"
t.)
en JUPITER
:::> SATURN
c EARTH
c:(
a::: MARS
...J
c:(
-10 MOON
z
0
~
c:(
~
-20
>
c:(
a:::
c.!l
C>
<.::J
a -3D
DENSITY LIMITED
AND VELOCITY
..J
LIMITED BODIES
-40
-so
313
CHAPTER 3
MORPHOLOGY OF
MULTI LANGUAGE TEACHING
F. ZWICKY
California Institute of Technology
ABSTRACT
314
A. Problems of Communication
315
feet. Fortunately in the United States some commendable
efforts are now being made to remedy this situation.
316
from all prejudice, bias and prevaluations. I therefo re was
searching for a noun which expresses the opposite of bias,
a word which I had naturally used extensively in the German
edition and which I needed very badly for the translation.
I discovered to my dismay that no such single word exists
in English. a fact which probably has occurred to very few
individuals who speak only this one language. Under these
circuIpstances I therefore propose to create a new word,
that is the expression "unbias" which will be exceedingly
useful if the morphological approach and morphological
research will have the great future which I venture to pre-
dict for them.
317
2. The ParaITleters of the ProbleITl
First ParaITleter P 1 :
Second ParaITleter P 2 :
children. P 2q =
ordinary travellers. •
Fourth ParaITleter P 4:
318
Morphological Box of Multilanguage Teaching
Pi :: Language (Set 1)
Pi :: Language (Set 3)
P 2 :: Background
of pupil
P 3 :: Method of
teaching
type
319
therefore is to evaluate all of the possible chains and to
establish through successive approximation and back feeding
of experience gained what the most effective means of
language teaching are.
a, International Expressions
a. Technical Expressions
320
canon, chell1istry, Chri stian, COll1ll1une, conse rve, coopera-
tive' dell1on, despot, discuss. edict, effect, engineer.
finance, forll1, general, genius, gigolo, hell1isphere, hotel,
intill1ate, interpret, locoll1otive, ll1achine, ll10ral , ll1otor,
negro, observatory, pill, popular, quartz, register, spec-
trull1, taxi, technical, t ele scope, telephone, telegraph,
universal, university, ventilate, wine, xylophone, zone,
etc. , etc.
321
As to the introduction into any language of new tech-
nical expres sions they should be chosen so as to be usable
in all languages and designations such as ramjet or athodyd
should be avoided and the universal word aeroduct should be
used instead. In this connection I have proposed to the
International Academy of Astronautics that in our journal.
the Astronautica Acta. every author, in a brief appendix to
his article should suggest translations of new technical terms
he might have introduced. Reviewing my own activities in
science, engineering and philosophy, I was surpris ed to dis-
cover that I am responsible for having coined over a hundred
new special words, most of which are generally understand-
able all over the world while others require expert transla-
tions. To mention some of my creations, they are. in
astronomy: neutron stars. pygmy stars. whittled down
stars, nuclear goblins {Kernkobolde, lutins nucleaires.
duendes nucleares, folletti nUcleari, iadrenie tshortiki},
compact galaxies, neutron star studded compact galaxies.
cosmic balls of light, neutron star studded cosmic balls of
light, symbolic velocity of recession {relating to the cosmo-
logical redshift}, or in engineering: aeroduct, aeroresonator,
{aeroturbojet}, aeropulse. aeropistonjet, and the correspond-
ing propulsive power plants which travel under water or
traverse the solid Earth, such as hydropulse, the terrapulse
and the "fuels" they need, that is the hydrofuels and the
terrafuels which are the analogs to the aerofuels or ordinary
conventional fuels; and in Physics: cooperative phenomena,
secondary structure of crystals. mosaic structure of
crystals {now called dislocations}, gravitons, mosaic
gratings; in chemistry: fragment chemistry, metachemistry,
monopropellants, coruscatives {heat detonators} and finally
in methodological phil os ophy: the principle of the flexibility
of scientific truth, morphological research, the method of
morphological field coverage, the method of the morpho-
logical box and so on.
322
ments of the Pope we encounte r a multitude of Latin words
or phrases. Starting from such pegs of knowledge acquired
from everyday reading we cannot only learn to understand
quite a bit of Latin but, by tracing Latin words through the
Western languages we can retain parts of these as well.
323
but is used only in combinations with, for instance,
cura = care in German and English as Sinekure and sine-
cure; also sine die = indefinitely (without naming the day).
Tangere = touch survives as tangent (in Russ ian tangens),
tangible and the dance tango; canis = dog in canine and in
.-
canicular days, journees caniculaires, giornate caniculare,
and in Russian kanikuli = vacation, dogs days; vox = voice
and vocare = call, in voice, voix (FR), voce (IT), voz (SF)
and in vocation or advocate, while lawyer in French,
German, Italian, Spanish and Russian is 1 'avocat, Advokat,
avocato, abogado, and advokat. Avus--grandfather seem~
to live only in Spanish as abuelo, domus = house remains
as an ordinary house (dom) only in Russian while it has
suffered a substitution of meaning in all other languages
where it has become a cathedral (or a dome of an observa-
tory, etc.). The same is true of the Latin sinister = left,
which it still is in Italian (sinistra) while sinister in all
other languages means baleful or foreboding evil.
324
played such as alto, andante, allegro, basso, crescendo,
forte, finale, piano, sotto voce, vivace and soon. Mus icians
therefore have at their disposal SOTIle words with which they
can cOTIlTIlunicate inte rnationall y.
325
c. Regularities Within Certain Groups of Languages
The following words start out the same way in Spanish and
in French:
esperar esp€rer sperare
espanol espagnol spagnuolo spanish spanisch ispanski
326
j3. Relative Strength of the Words in the Romance
Languages
d. Grammar
327
masculine, feminine and neuter have practically disappeared,
except in references to objects by means of the pronouns he.
she or it. Declension is achieved through the use of prepos i-
tions such as with, for, of, from, by and so on. but without
any change in the endings of the nouns or adjectives, except
for the occasional addition of a terminating letter s in the
genitive. Most other languages, however, use far more com-
plicated patterns. In German for instance, the nouns. ad-
jectives and associated definite and indefinite articles are
all being declined. German on the other hand allows us to
combine nouns, adjectives and prepositions freely and to use
most verbs as nouns which enormously facilitates the ease
of expression and allows one to state in one word thoughts
which need a sentence in other languages.
13. Verbs
328
Z. Background of Pupils, Methods of Teaching, Psycho-
logical Types
329
pants each of which spends effort and tiITle as he chooses
and as he feels theITl to be adequate for the purposes which
he has in ITlind. In way of illustration one of the ITlany
observations which I have ITlade along these lines ITlay here
be presented.
330
Russia and Switzerland.
331
EPILOGUE
METHODOLOGY--A DISCIPLINE
333
for problem solving be concerned not only with the definition
and solution of specific problems but also with the totality
of that growing complex consisting of the set of problems
competing for our attention. The discipline of methodology
should investigate criteria by which to assign priorities t
the appropriate levels of resources - -funds and talent- -to
be thrown against a problem, the nature of the interrelated-
ness of problems, the consequences of solutions to problems
and the anticipation of derivative problems.
334
times delayed, the full extent of the damage created by the
pollution probably will not be evident for some years. As
expected, there was widespread comment on this disaster.
However, criticism did not focus on the navigational situ-
ation which was the immediate cause of the wreck, nor on
the structural feas ibility of large tankers (they are quite
feasible- -there is a tanker of 300.000 tons currently under
construction and one of 500,000 tons on the drawing boards),
rather comment focussed on the defects in a technology that
could blindly and blandly create the set up for this sort of
disaster. This isolated example made some of the blind
spots of technology visible to many for the first time. One
of our own cabinet officers commented, "The environmental
backlash we confront today cannot be eliminated just by
applying more of the ~ science and technology that put
us in our present predicament. "
335
habits of thought that we unconsciously carry to new situ-
ations in which they are no longer applicable. An example
of such a habit of thought that affects our application of
technology is the making of decisions primarily on the
basis of feasibility. One of the severe deficiencies in the
present use of technology is the failure to note that at some
level of the state of the art the answers to the two questions:
how big can we build a tanker. and how big should we build
a tanker, begin to diverge. For decades technology has
been primarily concerned with finding ways to do things
hitherto impossible. The emphasis has been on pushing
back the limitations of nature and ignorance in order to
make more products and activities feasibile and broaden
our spectrum of choice. In an increasing number of tech-
nological areas we have recently moved from the regime
of finding a way to the regime of choosing the best way.
The task is no longer to remove natural limitations but to
set up limitations of our own, to define the constraints
and restraints which are prerequisite to sensible choice.
In a regime of limited capability, choice is usually properly
made for the limit of feas ibility- - build a plow that will cut
as many furrows simultaneously as possible. However,
the habit of thinking developed in this regime tends to carry
over into the second regime; the difficult problems of choice
being ignored and option being made simply for the limit
of feasibility. For example, in typical past wars the level
of tolerance to destruction and ability to recover was higher
than the level of any enemy's capabilities to destroy. How-
ever, in the past two decades. this inequality has been
reversed. It is now possible to destroy beyond any nation's
tolerance to absorb. We have entered the regime of choice.
There is the neces sity for limited and restrained actions,
but some spokesmen still adhere to first regime thinking.
336
We had best rapidly acquire the techniques essential for
decisions in a choice regime. The new developments in
biology, for example, are leading us to a capability level
where we may shortly be able to determine the sex of our
offspring. extend our life spans indefinitely. and even
create new varieties of organisms. Clearly the responsi-
bilities of choice imposed by such developments are likely
to be as demanding as any ever faced by man. The tempta-
tion to be guided purely by feas ibility. say in producing
selective viruses. could put an end to the human experiment.
337
for new systems with the impossibility of paying off
development costs before obsolescence may place us in
the same situation as an organism whose life span drops
below its gestation period.
Albert Wilson
338