Download Administrative Decision Making: Progress and New Insights and more Lecture notes Decision Making in PDF only on Docsity!
Administrative Decision
Making
By HERBERT A. SIMON
Carnegie Institute
of Technology
T
HERE is no need, at this late date, to
justify the study of organization and
administration in terms
of the decision-
making
process, for decision-making concepts
and language have becomehighly popular
in
writing about administration.
1 This paper
will describe some of the progress that has
been made over the past quarter
century,
employing this approach,
toward deepening
our scientific knowledge what new
facts
have been learned
about human behavior in
organizations,
what new scientific procedures
for
ascertaining facts, what new concepts for
describing them,
and what new generaliza-
tions for
explaining them. This progress
extends both to descriptive and normative
matters: tothepure
scienceofadministration,
and its application
to the practical business
of managing.
To satisfy limits on
this journal's space,
your patience
and my time, the account will
be highly selective. Only a few
notable and
significant advances have
beenselected; others
for which equally plausible claims might
be
made are ignored. Afrequent
practice in the
social
sciences is to bemoan our present
ignorance while making optimistic predictions
about future
knowledge. It is a pleasure to
survey an area of social science
where, by
contrast,wecan speakwithout blushing about
our present knowledge indeed,
where only
asmall sample of
the gainsin knowledgethat
have been achieved in the past
quarter
century can be presented.
1 The term "decision-making" occurred
three times
in the titles of articles in the
first fifteen volumes
ofthe Public Administration Review—that
is,through
1955; it occurred ten times in the next eight volumes,
or about six times as often
per annum as in the
earlier period.
Amidstthe general depreciation of progress in
the social sciences, the author surveys an area
where"we canspeak without blushing
about our
present knowledge" the
examination of adminis-
tration
through decision-making concepts and
language.
Operations Research and
Management Science
Oneobvious answer
to the question "What's
new?" is the spectacular
development in the
normative theory of decision making that
goes
underthe labels of"operations research"
and "management science." Through these
activities, many classes of administrative
decisions have been formalized,
mathematics
has been applied to determine
the charac-
teristics of
the "best" or "good" decisions,
and myriads of arithmetic calculations
are
carried out routinely in
many business and
governmental
organizations to reach the
actual decisions from day to day. A number
of sophisticated mathematical tools linear
programming, queuing theory, dynamic pro-
gramming,
combinatorial mathematics, and
others have been invented or developed
to
this end.
Like all scientific developments,
this one
has a long intellectual history,
and did not
spring, full-grown,
from the brow of Zeus.
Nevertheless, the state of the art today is
so
remarkably
advanced beyond its position
before World War II that the difference of
degree becomes one of kind.
The quantitative decision-making
tools of
operations research have perhaps had
more
extensive application
in business than in
governmental
organizations. It is worth
8 Some notion
of the state of proto-operations-
research just before World War II, as it applied
to municipal administration, can be obtained from
Ridley and
Simon, Measuring Municipal Activities,
(Chicago: International
City Managers' Association,
first edition, 1938).
31
Reprinted from Public Administration
Review, Vol. XXV, No. 1, March, 1965
quarterly journal of the American Society
for Public Administration
1320 Eighteenth Street, N.W.. Washington.
D.C. 20036
32 PUBLIC ADMINISTRATION
REVIEW
recalling, however, that many of these tools
underwent their early development in the
American and Britishmilitaryservicesduring
and just after the Second World War (where
the terms "operations research" and "opera-
tions analysis" were coined). Among the
inventorsoflinearprogramming,for example,
were Tjalling Koopmans, seeking, as statis-
tician with the Combined Shipping Adjust-
ment Board, a means for scheduling tanker
operations efficiently; and George B. Dantzig
and Marshall K. Wood, in the Office of the
Air Force Controller, who used as one of
their first (hypothetical) programming prob-
lems the schedulingof the Berlin Airlift.
Operations research,
particularly in its
governmental applications, has retained close
intellectualtieswithclassicaleconomic theory,
and has soughtto findeffectivewaysof apply-
ing that theory to public budgeting and
expenditure decisions. This has been a cen-
tral preoccupationof the RAND
Corporation
effort,asexemplifiedbysuchworksasCharles
J. Hitch and Roland N. McKean,
The Eco
nomics of Defense in the Nuclear Age.s In
the past several years, Hitch, asControllerof
the Department ofDefense, and a numberof
his former RAND associates have played
major rolesin bringing thenew tools tobear
on Defense Department budget decisions.
Thus, while the quarter century begins with
V. O. Key's plaint about "The Lack of a
Budgetary Theory," 4 it ends with a distinct
8 Cambridge:
Havard University Press, 1960.
- American Political Science Review, December
1940, p. 1142. Labels have an unfortunate tendency
to compartmentalize knowledge. Thus, the literature
of "budgeting" has been only partly informed by
the literature on "decision making," and vice versa,
and both of these have sometimes been isolated
from the economics literature on resources
alloca-
tion and public expenditure theory. Variants on the
same basic sets of ideas are rediscovered each gen-
eration: "measurement of public services," "program
budgeting," "performance budgeting," "engineering
economy," "cost-benefit analysis,""operationsanalysis."
What is genuinely new in this area in the past
decadeis the power and sophistication ofthe analytic
and computational tools. Some impression of these
tools may be gained from the Hitch and McKean
book previously mentioned; from Roland N. McKean,
Efficiency in Government Through Systems Analysis
(Wiley, 1958); Arthur Maass, et. al., Design of Water
Resource Systems (Harvard U. Press, 1962); or Alien
V. Kneese, TheEconomics of Regional Water Quality
Management (Johns HopkinsU. Press, 1964),
and the
references cited therein.
revitalization of the whole field of public
expenditure theory, and with a burgeoning
of new analytic tools to assist in allocating
public resources.
Optimality and All That
In many ways the contributions of opera-
tions research and management science to
decision-making theory have been very prag-
matic in flavor. The goal, after all, is to
devise tools that will help management make
betterdecisions. One exampleof apragmatic
technique that has proved itself very useful,
andhasbeenrapidly andwidelyadoptedover
the past fiveyears, isthe scheduling procedure
variously called PERT, or critical path
scheduling.
This technique does not use any
verydeeporsophisticated mathematics (which
may account
partly forthe speed of its adop-
tion), but is mainly an improvement of the
common sense underlying the traditional
GanttChart.
Contrasting with this pragmatic flavor,
advances in operations research have been
paralleledbydevelopmentsin thepuretheory
of rational choice a theory that has reached
a very high level ofmathematical
and logical
elegance and rigor. Among these develop-
ments perhaps the most important are: (1)
rigorous, formal axiom systems for defining
the concept of utility in operational terms,
(2) extension ofthe theory ofrationalchoice
to encompass the maximization of
expected
utility
under conditions of uncertainty, (3)
extension of the theory to repeated choices
over time dynamic optimization, and (4)
extension of the theory to competitive "gam-
ing" situations. These formal advances have
had animportantinfluence, in turn,on direc-
tions of work
in theoretical statistics (statis-
tical decision theory, Bayesian statistics), and
on the kinds of models that are preferred by
operations researchers or at least by the
theorists among their number. 5
An evaluation of thesecontributions on the
pure theory of rational
choice would return
a mixed verdict. On the positive side, they
have provided enormous conceptual clarifica-
tion for discussions of "rationality." For
8 Since Ihave discoursed at length on thesematters
elsewhere, I shall be brief here. See "Theories of
Decision Making inEconomicsand Behavioral Science,"
49 American Economic Review 253-283. (June 1959),
and Part IV of Models of Man (Wiley, 1957).
34 PUBLIC
ADMINISTRATION
REVIEW
processes
of social workers,
7 but similar
experiments
havebeenexceedingly
rare inthe
succeeding
twenty-five years.
One of the few
other examples to which
I can refer is the
study
done in the Prudential
Life Insurance
Company by the Survey
Research Center of
the University of Michigan.
8 Eitherresearch-
ersonorganizations decided
thatthe informa-
tion attainable from
field experiments was
not worth the trouble and
cost of carrying
out such experiments,
or they found it
diffi-
cult tosecure
the cooperationof business
and
governmental
organizations
in arrangingsuch
experiments or both. Whatever
the reason,
field
experiments havenotbeen
animportant
procedure for learning
about organizational
decision making.
In a few cases
researchers have tried
to
import relatively
sizeable organizations
into
the
laboratory hence, their
studies lie on
the
boundary line between
field and labora-
tory experiments. The
Systems Research
Laboratory of the RAND
Corporation, for
example,studied decision
makingby simulat-
ing,under
controlled conditions, an
entireair,
defense
control center and associated
early
warningstations,manned
onafull-timebasis
over a period of several months
by a staffof
some thirty subjects.
While the studies con-
ducted by the
Systems Research Laboratory
had as their
direct outgrowth a major
Air
Force
trainingprogram, the
laboratory proved
less tractable as
a setting for obtaining data
for testing theories
of the decision-making
process, and
there has been no subsequent
rashof studies
ofthis kind. 9
In contrast to the dearth
of field experi-
ments
andlarge-scalelaboratory
experiments,
laboratory experimentation
with relatively
small groups
has been a thriving enterprise.
Several examples of
methodological
advances
in
theartof small-groupexperimentation
can
be mentioned. FredBales,
with hisinteraction
process analysis,
developed
a scheme of data
processing useful for
studying the interaction
»HerbertA.Simon
and William R. Divine,"Human
Factors in an Administrative Experiment,"
1 Public
Administration
Review 485-492. (Autumn
1941).
8 N. C.Morse
andE.Reimer, "Experimental
Change
of a Major Organizational Variable,"
52 Journal of
Abnormal and
Social Psychology 120-129. (1955).
Robert L.Chapman, etal, "The
System Research
Laboratory's Air
Defense Experiments," 5 Manage
ment Science 250-269. (April
1959).
of task-oriented and
social-system oriented
behavior in small
problem-solving groups.
Alex Bavelas devised
a small-group task that
permitted the experimenter
to alter the
decision-making
process byopeningorclosing
particular channels
of communication
be-
tween members
of the group. In succeeding
years,theBalescoding
schemeandtheBavelas
small-group
task have both been used
in a
substantial
number of studies, manipulating
agreat
many different independent
variables.
Both have
proved exceedingly
valuable in
permitting the cumulation
of comparable
knowledge fromawhole
seriesofexperiments
carried out by different
investigators in dif-
ferent laboratories.
It is impossible
to summarize here,
oreven
to reference,
the numerous contributions
to
the substantiveknowledge
of decisionmaking
that have been contributed
by the small-
group experiments.
A single example will
convey the flavor
of such work. Cyert
and
March
were able to produce bias
in the esti-
mates
of members of asimulated
organization
by creating partial conflict
of interest among
them, but showed
that under certaincircum-
stances
this biasdid not affect organizational
performance.
10
New knowledge about organizational
deci-
sion making can be obtained
from appro-
priately planned
experiment on individuals
as well as
from small-group experiments.
Andrew
Stedry,for example, has tested
in this
way theoriesabouthow
budgetcontrolsaffect
behavior in organizations.
11 The series
of
studies of influence
processes carried
out at
Yale by
the late Carl Hovland and
his asso-
ciates
belonginthesame category.
12
Persuasion
and Evocation
Mention of
the Yale research on influence
processes
marks a good point
in our discus-
sion to
turn to several substantive
develop-
ments
in the theory of decision
making. The
notion
that a decision is like
aconclusion de-
rived from aset of premises has
beenauseful
10 Richard M. Cyert
and James G. March, The
Behavioral Theory of the
Firm (Prentice-Hall, 1963),
pp.67-77.
11 Budget Control and
Cost Behavior (Prentice-
Hall,
1960), Chapter 4.
12 See theYale
Studies in Attitude andCommunica-
tion, edited by Hovland
and Rosenberg, and pub-
lished by the Yale
University Press.
ADMINISTRATIVE DECISION
MAKING
35
methaphor for
analyzing the decision-making
process. Following
the metaphor a step
further, we can
view each member of an or-
ganization as "inputting"
certain premises,
and "outputting"
certain conclusions, or
decisions. But each
member's conclusions
become,in turn, the inputs,
thatis tosay,the
premises, for other
members. For one person
to influence another
involvesinducing himto
use appropriate
premises in his decision mak-
ing.
What happens
in an organization, or inany
kind of social system, when there
are con-
flicting premises pushingaparticular
decision
in different directions? Much of
the research
on influence processes has
been aimed at
answering this question.
In much of this re-
search, influence has beenconceived
asakind
of "force," so thatwhen
several influences are
brought to
bear simultaneously, the outcome
is interpreted
as a "resultant" of the imping-
ing
forces. Persuasion is then a process
of
exerting
suchaforce.
Animportantadvancein understanding
de-
cision making has been to complement the
notionofpersuasion
with the notionofevoca-
tion. When we
want someone to carry out a
particular action,
we may think of our task
as one
of inducing him to accept latentdeci-
sion premises favorable to the
action that he
already possesses. Thus, writing about
food
will often make a reader
hungry, but we
would hardly say
that we had "persuaded"
himthathewas
hungry;itwould bebetterto
saythatwe
had "reminded" him.
Processes of persuasion play
their largest
role in decisionmaking in conflict situations
where the issue is already posed,
and the
alternatives present.
This is the framework
withinwhich most of
theYale studies on atti-
tude change
were carried out. It is also the
framework for theimportant and
well-known
study of Voting by Berelson, Lazarsfeld,
and
McPhee. 13
On the other hand, in
studies of decision
making where the
focus of attention of the
participants
is one of the main independent
variables,
the evokingprocesses take on larger
importance.
The recent study of the
Trade
Agreements Act renewal, by
Raymond Bauer,
Ithiel Pool, and Lewis
Dexter indicates that
these processes playedamajor
role in deciding
the issue. 14 The authors describe the setting
oftheir study thus (p. 5): "We are
interested
in thesourcesofinformationforeach
of these
populations,
the bases of its attitudes on
the
tradeissue, and thecircumstances
which lead
some individuals to take active
roles in the
makingof
policy." (Emphasissupplied.) They
demonstrate
convincingly that the behavior
of particular
Congressmen on the trade issue
depended as much on the alternative
claims
on their time and attention as on the
distri-
bution of interests of their constituents.
Totheextent thatthe mechanism
ofevoca-
tion is important for decision making,
many
new ways arise in which
organizational ar-
rangementsmayaffect
behavior. Asexample,
one of the findings
of the study just men-
tioned (p.229)can be cited:
In summary, we would
suggest that most signi-
ficant of
all toan understandingofwhatcommunica-
tion went out
from business on foreign trade was
neither self-interest nor
ideology, but the institu-
tional
structure which facilitated or blocked
the
production of
messages. Whether a letter to a con-
gressman would get written depended
on whether
organization facilitated it, whether
the writer's round
of
daily conversations would lead up to it,
whether
a staff was
set up to produce it, and whether the
writer conceived writing
this letter to be part of
his job.
Evoking mechanisms
take on special promi-
nence wherever dynamic
change is occurring.
Studies
of the diffusion of innovations show
that the timingofadoption of
an innovation
depends critically on the
means for getting
people to attend to it.
15 From everypoint of
view, the new knowledge
gained about evok-
ing and attention-directing
processes is a
major substantive
advance in our understand-
ing
oforganizational decision making.
The Structure
off Decisions
A decision is
not a simple, unitary event,
but the product
of a complex social process
generally
extending over a considerable
period
of time. As noted, decision making
includes attention-directing or intelligence
processes that determine the occasions
of de-
cision, processes
for discovering and design-
1 University of Chicago
Press, 1954.
14 American
Business and Public Policy: The Poli
tics of Foreign Trade
(Atherton Press, 1963).
16 SeeJ. Coleman,E. Katz,and H.Menzel, "Diffusion
of an Innovation
Among Physicians," 20 Sociometry
253-270. (1957); also, H. A. Simon
and J. G. March,
Organizations (Wiley,
1957), Chapter 7.
ADMINISTRATIVE DECISION
MAKING 37
tion of a department store buyer anda simu-
lation of a bank trust investment officer. 19
I am not aware thatany single comparable
simulation of a decision-making
process in
the areaofpublic administration
has yetbeen
carried out, but it appears that several are
under way in current research. Perhaps the
most likely target for initial attempts is pub-
lic budgeting. If we examine the strategies
described in recent empirical studies, like
those of Wildavsky, 20 wewillseethat they can
be rather directly translated intocomponents
of computer programs.
Parallel with these simulations of adminis-
trativedecision making there has been a con-
siderable exploration of individual thinking
and problem solving processes, also using
computer simulation as the tool of theory
formulation and theory testing. 21 Today, we
have a considerable specific knowledge on
how human beings accomplish complex
cog-
nitive tasks. We have reasons for optimism,
too, that this body of knowledge will increase
rapidly, for in the digital computer language
we havean analytic tool anda means forac-
curate expression whose powers are commen-
surate with the complexity ofthe phenomena
we wish to describe and understand.
19 Descriptions of these two simulations may be
found in Chapters 7 and 10, respectively, of Cyert
andMarch, Behavioral Theory of theFirm, opcit.
"Aaron Wildavsky, The Politics of the Budgetary
Process (Little, Brown and Company, 1964).
21 For asurvey, and numerous examples,seeEdward
Feigenbaum and Julian Feldman, Computers and
Thought (McGraw-Hill, 1963).
Landmarks and New Roads
These, then, are some of the more promi-
nent landmarks along the road of decision-
making research over the past twenty-five
years. On the normative side, the analytic
tools of modern operations research have se-
cured an important place in the practical
work ofmanagement. Their role ineveryday
decision making promises to be much en-
larged as present techniques are supple-
mented by new heuristic approaches.
On the side of the pure science of admin-
istration, there have been equally fruitful
developments. The experimental method, in
the small-group laboratory, can now be used
to study a widerange ofdecision-making be-
haviors thatarerelevant to organizations. We
haveintroduced theconcept of evocationinto
our theories of influence,andhave used it to
gain new understanding of the decision-
making process in changing environments.
Finally, the modern digital computer, a
powerful new tool, has provided both a lan-
guage for expressingour theories ofdecision
making and an engine for calculating their
empirical implications. Theories can now be
compared with data of the real world of
organizations.
The attention-directing
mechanisms so im-
portant in decision making also have played
their part in determining the particular de-
velopments sampled in this paper. Another
scientist, with a different set of research con-
cerns, would chose a different sample. The
fact that even one such sample exists shows
howfarwehave comeduringthe past twenty-
five years toward understanding human be-
havior in organizations.
