You are on page 1of 8

PROCEEDINGS OF THEIEEE,VOL. 63, NO.

3, MARCH 1975 397

An Introduction to the Application of interpretive


Structural Modeling
DAVID W. MALONE

Abrtmct-Interpretive structural modeling is an emagingmethod- DIGRAPHS, BINARYMATRICES, AND INTERPRETIVE


ology which appears to be v a y useful as an aid to individuals and d STRUCTURAL MODELING
groups indeveloping an understanding of complex situations This
paper presents an introduction to the fundamental ccmcepts and opar- The process of ISM is based upon the one-to-one correspon-
tions of the methoddogy and reports on the results of two exercises dence between a binary matrix and a graphical representation
conducted with a group of graduate students who had nrinimdnutha of a directednetwork.Thefundamentalconcepts of the
m a t i d training. The hrst exercise mv&ed the structuring of personal
values and was intended to acquaint the individuals with the method- process are an “element set”and a “contextual relation.”
ology. The second was a group exacise focusing on b d g s to mvest- The element set is identified within some situational context,
ment in thecentnl city, a abject of substantive interest to the and the contextual relation is selected as a possible statement
partiapants. The results of the exercises demonstrate the utility of the of relationship amongtheelementsin a mannerthat is
methodology f a capturing md communicatiog individual a d p up contextually significant forthe purposes of theenquiry.
perceptions regding complex imes
(Examples of element sets and relations are presented in a
later section.) Theelementscorrespond tothe nodes on
INTRODUCTION a network model,and the presence of the relation between
HE TERM “inteqketive structural modeling” (ISM) is any twoelements is denotedby adirected line(orlink)
used here to refer to the systematicapplication of some connectingthosetwoelements(nodes). In the equivalent
elementary notions of graph theory in such a way that binary matrixrepresentation,the elements are thecontents
theoretical, conceptual, and computational leverage is exploited of theindex set forthe rowsand columns of thematrix,
to efficiently construct adirectedgraph, ornetwork repre- and the presence of the relationdirected from element i to
sentation, of the complex pattern of a contextual relationship element j is indicatedby placing a1in the corresponding
among a set of elements. intersection of raw i and columnj .
The mathematical foundations of the methodology can be Some basic terminology and mathematical operations will be
found in various reference works, e.g., [ 11. The philosophical introduced now by illustrating the process of developing a
basis for the development of the ISM approach has been pre- hierarchical restructuring of a given directed graph, or
sented in [ 21, and the conceptual and analytical details of the “digraph.” Consider asystemcomposed of fourelements
ISM process are outlined in [3]. The methodology has been S = { 1, 2, 3, 4 ) , and one relation, R = -, for which the
implemented ina man/machine interactiveenvironment in following digraph has been constructed:
such a way that human users are responsible for making sub
jective judgments while thecomputer is employed in an
unobtrusive manner for bookkeeping and for performing and
displaying the results of simple logical operations. Preliminary
versions of the computer programs were available in the Fall of
1973, when the firstcomputer-aided experiments in the use of
the methodology were begun. A binary matrix representation of this digraph is
This paper reports on two exercises which were conducted
in a classroom setting with a group of eight graduate students
in the Department of City and Regional Planning at The Ohio
State University, Columbus. The undergraduate degrees of the
students involved were scattered among economics, geography,
political science, andnatural resources. Thefirst exercise
involved the structuringof personal values and was intended to
1 2 3 4
1 1 0 1 1

3 1 0 0 1 .
4 0 0 1 0
j
acquaintthe individuals withthemethodology.The second This matrix is termed the adjacency matrix of D,and is con-
was a group exercise focusing on barriers toinvestment in the structed by setting aij = 1 wherever there is an arc in D
central city,a subject of substantive interestto the participants. directed from element si to element si, and by setting aij = 0
In the following, an introduction to the basic concepts and elsewhere.
operations involved in the ISM technique is followed by a Element sj is said to be reachable from element si if a path
description of each of the exercises, including the present* can be traced on D from s i to si. By convention, an element
tion and discussion of the results. si is said to be reachable from itself by a path of length 0. The
reachability matrix M of a digraph is defined as a binary matrix
Manuscript received June 14, 1974; revised October 8, 1974. This in which the entries mij are 1 if element S j is reachable from
work was supported in part bythe Batteile Memorial Institute,the element si; otherwise mij = 0. It can be shown that thereach-
Academy for contemporary Roblems, and the Kettering Foundation.
Theauthor is withthe ColumbusLaboratories,BattelleMemorial ability matrix can be obtainedoperationallyfromthe adja-
Institute, Columbus, Ohio 43201. cency matrix by adding the identity matrix and then raising
398 MARCH IEEE,PROCEEDINGS OF THE 1975

Fig. 1. A symbolic digraph of the process of extracting a digraph from


a situational context and then producing a hierarchical display.
Symbols are defined in the text. Arrows denote the presence of the
relation: ". is used t o determine * * * ."
1

the resulting matrix to successive powers until no new entries


are obtained. That is:
M = (A +I)" Fig. 2. A symbolic digraph of the principal operations involved in the
interpretive structural modeling process. (Symbols and operations are
where n is determined suchthat elaborated in text.)
( A +I)"-' < ( A + I ) " = ( A + I ) " + ' .
Here all mathematicaloperationsareBoolean;henceeach byinspection,itcanbeseen thatthesecond level of the
successivepowering operation preserves theentries of the example system consists of all the remaining elements: 1, 3,
previouspower, andmatrixequalityorinequalitycan be and 4.
determined on the basis of an entry by entry comparison. For Thisprocesshasrearrangedand partitionedthe original
the preceding example n = 2 and the reachability matrix is: element set S = { 1, 2, 3, 4) into hiernrchicnl components:
SH = (2; 1, 3, 4) -. . (Note that the hierarchical reordering

]
1 2 3 4 process could have proceededfrom the top downby identifying
1 1 0 1 1 the elements for which 6 l i = (ii n 6$.) The hierarchically r e
ordered digraph DH then becomes:
3 1 0 1 1 '
Level 2
4 1 0 1 1 =
I &re is a correspondence between the notionof reachability Lcvel 1
in graph theory and the notion of transitivity in binary rela-
tions. If the relation R used to establish the connectivity in What has been gained from this process is a partitioning and
the digrapb 4 is transitive-that is, aRb and bRc imply that rearrangement of the originaldigraph insucha way that
aRc-then the m a t e M represents the transitive closure D, of additional information about the system-that is, the presence
the digraph D. ( T W interpretation of the reachability matrix and content of levels-is transmitted through the format of the
is an essential aspect of the ISM process,which will be display.Although the value of thisadditionalinformation
elaborated shortly.) may appearto benegligible inthe caseof the preceding
The utility of the reachability matrix is that it can be used example,situations will appearlaterinwhichhierarchical
t o develop a hierarchical restructuring of the digraph. To do rearrangementgreatlyenhancesthecommunicability of the
this, a reachability set 6li is defined for each element si E S,as digraph.
all of those elements which are reachable from si, that is, all Asymbolicdigraphsummarizingtheprecedingprocess is
of those elements whose columns have an entry of 1 in row i. shownin Fig. 1.Itshouldbeevidentfroma review ofthe
This definition is stated mathematically as: @ = {si 3 mij = previous discussion that it is possible to bypass the construc-
1} , where mij is the entry in the ith row and jthcolumn of M. tion of theinitial digraph D and construct its equivalent
Further an nntecedent set cii is defined for each element si as adjacency matrix directly, continuing then through theprocess
all of those elements which can reach si, that is, all of those t o obtain DH. Whether to carry out such an exercise begin-
elements whose rows have an entry of 1 in column j . Stated ningwith theconstruction of D or A dependsuponthe
mathematically: ai = {si 3 mii = 1} . The intersection of cii preferencesanddegreeofmathematicalpreparationofthe
and 6$., denoted by ai fl&., consists of all of those elements modeler. In either case, for a large element set and/or a com-
which are common to both (ii and 61i. Those elements si for plex pattern ofrelationships,extensiveandlaboriousopera-
which ai = (ii n 6. are not reachable from any of the remain- tionscould be involved. ISM was conceived as aprocedure
ing elements of s, and hence can be denoted as bottom level for avoiding as much as possible the necessity for knowing-
elements. and theburden ofperforming-therelatedcalculations.A
Continuingwiththeexample,the ( i i , I&, andtheirinter- further advantage of removing this computationalburden is
sections are given in Table I. thattheconceptualutility of structural modelingcan then
Here it can be seen that the single element 2 is in the bottom be madeavailable to mathematicallyukophisticated users.
level. By strikingfromthematrix M therowandcolumn The ISM methodology employs the digital computer for book-
corresponding to element 2 andrepeatingtheprocess,it is keepingandperforming logical operations in the process of
possible t o determine the bottom-level elementsof the reduced directly synthesizing a digraph fr6m a specified set of elements
system,whichareactually the second-levelelementsof the andrelation.
a Someadditionalassumptionsandrelated
originalsystem. This operation is equivalent to simply r e limitationsarerequired,however,andthese will now be
moving the element 2 wherever it appears in the Table I ; thus elaborated.
MALONE:INTERPRETIVE STRUCTURAL MODELING 399

COMPUTER

Partitioning Embedding

Fig. 3. Functional representation of interpretive structural modeling, showing potential for man/
machine symbiosis.

Theprocedureoutlined here and summarized in Fig. 1 required by the user.) The corresponding minimum-edge
requires the initial specification of a digraph or its equivalent digraph DZ provides an efficient,hierarchically ordered dis-
binary matrix. Thenoteworthycontribution of the ISM play of both the direct responses and the indirect transitive
procedure,illustrated in Fig. 2, is that it operateswithout inferences
resulting from
the
embedding operation.
The
such an a priori knowledge of the structure. The process is superscript asterisks are added to A and D H in Fig. 2 to denote
initiated by specifying an element set and a transitive relational that they have been constructed using the ISM process, and
statement. An “embedding” operation is thenperformed,in specifically that all redundant links have been removed from
whichadvantage is taken of the ability to make transitive the digraph. Depending uponthecontext of the exercise,
inferences within the reachability matrix to guide the system- however, redundant links may be essential to convey the full
aticinterrogation of auserwith regard tothe\presenceor meaning and pattern of the relation. For this reason, the ISM
absence of the relation between pairs of elements. The user is process provides for a“comparison operation in which the
requested to respond to a query of the form: “(Is) (element user examines the result of the mathematical operations and
s i ) (relatedto)(element sj)?” where theverbdconstructs heuristics of the process and introduces modifications or
appearing inthe parenthesesare thoseappropriate tothe corrections to the digraph. Finally, a “substitution” operation
particular context, e.g., “does activity 23 precede activity 14?” consists of the introduction of elaborative text, interpretive
“doesthenumber of electric cars onthe road affectthe symbols, oradditional graphicalembellishmentswhich will
number of power plants in an urbanarea?” A digital computer make the fmal “interpretive structural model” comprehensible
is employed to keep track of the responses suppliedby the to a wider audience.
user (which are explicit entries to the reachability matrix), to ISM is intended for use when it is desired to utilize system-
provideimplicittransitiveinferences based upon previous atic and logical thinking t o approach a complex issue and then
responses, and to generate an efficient ordering of subsequent to communicatethe results of thatthinkingto others. The
queries. objective is to expedite theprocess of creating a digraph, which
Following theembeddingoperation,the resultingreach- can be converted to a structural model,and then inspected
ability matrix is used t o partition the element setaccording and revised tocapturethe user’s bestperceptions of the
to the procedure outlined in the preceding. Rearranging and situation. Theentire process has been implementedfor use
partitioningthe reachability matrix t o correspond to the in a manlmachine interactive environment in such a manner
hierarchical reordering of the element set, it is then possible thatthe user can concentrateon substantive concerns in
(asexplained in detail in [ 31) to systematically examine order to make subjective judgements regarding the presence or
“diagonalexpansions” of M inorder to discover the least absence of the relation between pairs of elements, andthe
number of connections which are required to form an adja- computer is assigned the task of bookkeeping androutine
cency matrix A * , whose transitive closure would be equal to calculations. Theentire ISM process can be represented
the constructed reachability matrix. This operation is referred graphically as shown in Fig. 3. Thisfigure is an interpretive
to as “extracting.” (Bothpartitioningandextracting opera- structural model of the process of building interpretive struc-
tions are performed in thecomputer,with no intervention tural models; it has been generated from Fig. 2 by substituting
PROCEEDINGS OF THE EEE, MARCH 1975

Elements (The EightLasswell Value Categories):

1. Wealth

2. skill

3. Lnlightelrrnt

4. Parer
5. Affection

6. Respect

7. Well-being

8. Rectitude

I h e Relation:

”(value I) is mre important to me than (value j).”

Fii 4. The elements and relations employed in the individual exercise.

r Exorcisela: A PrioriConstruction
Restructuring,Figure1
of Digraph andManual
x e r c i s el b : Cmputer
Construction of Digraph
Aided

- Hierarchically Using 1% Process, Figure 2


Student I n i t i a l Digraph Restructured Digraph

A
T
-

B
s
Fig. 5. Illustrative examples of studentresponses to value structuringexercise.Element numbers
correspond to the values listed in Fig. 4. Arrowdenotestherelation ‘‘.. . is moreimportant
than. .”

simpleverbalstatementsforthemathematicalsymbolsand INDIVIDUALEXERCISE:ORDERINGPERSONAL VALUES


byintroducingtheadditionaloperations of “substituting”
and “comparing.” Notethat Fig. 3 can be used to describeThefirstexercise w a s designed to introduce the students to
theprocesstoindividualswho havehadlittleor nomathe- theconceptsandoperationsunderlyixgtheprocess.The
training.
matical conceptual
context was chosen t o have a high degree of
MALONE: INTERPRETIVE STRUCTURALMODELING 40 1

personal relevance to the individual students and to involve tudent Initial Digraph ISH Digraph
nontrivial issues, while not being so extensive as t o preclude
hand calculations. Theelementset chosen wasLasswell’s
eight value categories [ 4 ] , and therelation was taken as 0
. is more important to me than -
6 6 . . as listedin Fig. 4.
a’’,

(This may or may not be a transitive relation, a point which cb


enhancestheconceptual
discussed later.)Thestudents
utility of this exercise,as will be
were asked to individually
C b
create an a priori digraph using these conditions and then to b
determinethe correspondinghierarchicalordering usigg the
operations summarized in Fig. 1. They were then given access b do
to a computerized version of the ISM procedure and asked to
repeatthe exercisewithin an environment characterizedby
Fig. 3. The results of these activities are discussed here.
- b b
Results for two of the students are presented in Fig. 5 . The 0
initial digraphs shown in column 1 of the figure clearly show
one construction strategy, in which the elements were laid out b
in a geometric pattern to aid the manual process of system-
atically making the pairwise comparisons. These digraphs
D
b
appear very complex, but when they are converted to binary
matrices and subjected to a hierarchical reordering calculation,
following theprocedure outlinedin Fig. 1, considerable
simplificationresults, as shownin column 2 of Fig. 5 . In
these figures two elements neither of which is “more impor-
tant’’ than the other,symbolically shown as

@=a
or equivalently as
E
b
00 b
areinterpreted as being “equally important.” (This conven- do
tion was establishedprior to the exercise.) The hierarchical
digraphs are arranged such that all equally important elements
appearonthe same level; elements in the higher levels are
more important than elements in the lower levels.
Q
Sequential elimination was anotherstrategyemployedby Fig. 6. Further examples of student responses t o value structuring
some studentsforconstructingtheinitial digraph. In this exercise. Elements as b t e d in Fig. 4. Upper elements are “more
strategy the list of elements was scanned for the most impor- important” than l o w e r elements.
tantelementorgroup of elements,which would then be
placed in the highest level and eliminated from the set; the legitimacy of the concept of a “heriarchy of values,” and the
reduced set would then be scanned for the next most impor- implications of differencesbetweenindividualand collective
tant elements, and so on. Three graphs which were constructed value hierarchies. The general concept of values hasbeen
in this manner are shown in column 1 of Fig, 6 ; column 2 of discussed inmanyrecent works, including the previously
the same figure presents the digraphs which were constructed referenced work by Lasswell [ 4 ] , so no further discussion
by the same students by using the computerized ISM process. will be presented here. Thequestion of the legitimacy of a
Notethatforthestudents using this constructionstrategy hierarchy of values appears t o be closely related to the notion
there was relatively little difference between the initial and of “transitivity”, which is central to the ISM process, and so
the ISM structures, in contrast with the previous cases illus- is worth considering more closely.
trated in Fig. 5. a -
The relation ‘‘- is more important than * -” is logically
This exercise provided an experiential context within which transitive. It is perhaps an open question, however, whether
to discuss several considerations related to the process of individual perceptions of thisrelationaretransitive; that is,
interpretive structuralmodeling. First, the comparisonbetween althoughthe relation is transitive, individuals may not be!
thehandcalculationsandthe computer-aided operations Consider, for example the following construct, taken from the
demonstrated theconsiderable computational leverage provided initial digraph of Student A (Fig. 5 ) :
by the ISM process, both in terms of time-required to process
thestructure-two to fourhoursforthe manual approach,
five to tenminutesforthe ISM approach-and in reduced
requirementsfor methodologicalknowledge-the computer
performs all the necessary computations.
In addition, the exercise provided “objective” data for use This student presumably inserted double arrows wherever he
in general
a discussion of theconcept of “values,” the perceived the pair of values to be equally important. Why
402 PROCEEDINGS OF THE IEEE, MARCH 1975

mnt, Ri Element Score. Vi


5. 7.5 Affection 44 ( n o a t Important)
7. 1.5 Well-Being 44
6 8. Rectitude 41
5 6. Reapect 39.5
4 3. Enlightenment 37
3 2. skill 19.5
2 4. Parer 14.5
1 1. Weelth 12.5 (Leaat Important]

Fig. 7. Combined rankand scoresforthe Lasswellvalue categories


bnsed upon responses of seven graduate planning students.

then did he leave out the arrow (relation) directed from value
6 to value 3? There is clearly atransitivelyinferred link by
virtue of the path from6 to 5 t o 3. If this student deliberately
left out the link from 6 to 3, thusintendingthat value 3
(enlightenment) was strictly more important t o him than value
6 (respect), then he was being intransitive. In such a case, the
operation of taking a transitive closure will insert a link from
6 to 3, and the resulting hierarchical digraph will not portray
the original intent of thestudent.(Notethatthere is no
intrinsic reason for human beings to be transitive, but it may
be of value to know when they are and when they are not.)
The ISM processmakes use of theproperty of transitive
inference to reduce thenumber of explicit pairwise com-
parisons which must be made among the elements, and thus
never gives the user the opportunity to enter an intransitive
response. Althoughthe process forces the user initially to
produce a logically consistent,transitivedigraph,intransitive
substructures may be introducedthroughthesubsequent
“correction” process, which allows the user to explicitly insert
or delete elementsand/or links. Thus it is importantthat
n p r r d hy R w l Errore R-mh Corpmrra
J
users of the ISM process understand both the value and the Fig. 8. Relativesignificance of specificobstaclesin hinderingprivate
implications of the transitivity assumption. investment in different parts of central cities.
Having collected a set of individual rankings of a particular
element set,as containedin Figs. 5 and 6 , it is of some
involved in the exercise seriously questioned the conceptual
interest to consider the possibility of aggregating these ranking
legitimacy or wisdom of structuring personal values. In fact,
to produce a collective ranking (or hierarchy). One simple
the possibility that one value may be instrumental to another
way to do that in this case is to assign weights wij for the ith
leads to some interesting questions regarding the utility of a
elementaccording to its relativepositioninindividual fs
simplepreference ordering. Furthermore, since the ranking
hierarchy. By summing over the individuals, collective a
and aggregating approach used liere ignores relative intensity of
score W j can be assigned to each element:
preferences, there aresome possible theoretical flaws that
“i shouldbeconsidered. A full discussion of these points is
wi = wij. beyond the scope of this paper. More sophisticated measuring
j-1 and aggregating techniques are available, however, andthe
interested reader is referred to tlie psychometric literature,
This canthen be used to establish relative positions (or e.g., [SI.
“ranks”) R i in a collective hierarchy. For the seven students
who participated in thisexercise, the collective hierarchy of
GROUPEXERCISE:BARRIERSTO INVESTMENT
values is displayed in Fig. 7. A measure of individual consis-
IN THE CENTRAL CITY
tencywith thisrankingcan be obtainedbycomputingan
individualconsistencyscore Z j as the average difference b e The second exercise was conducted the week after the first
tween theindividual‘s ranking and the collective ranking: and involved theparticipation of the full group of eight
students inusing the computerizedversion of the ISM method-
ology to collectively create a network representation of the
pattern of interaction among a set of obstacles to investment
inthe central city.Theset of obstacles was takenfrom a
For the seven participating individuals, the worst consistency comprehensivepaperby Downs [ 6 ] , which served as the
score was an average difference of 1; in the relative positions substantivebackground for the exercise. The primaryobjec-
along the hierarchy, the best consistency score was and the g, tives of the following discussion are to describe the conditions
average was 1 A. This is aremarkableconsistency, and sug- under which the group exercise was conducted,to present
gests that these students form a fairly homogenous group. some statistics regarding the operation of the process, and to
Raising the issue of the significance and legitimacy of such a illustrate the typeof “model” that is produced.
strategy for generating a collective perception led to a spirited The element set used is described in detail in [ 6 ] . Fig. 8 is
discussion among the participants. Only one of the students takenfromthat paper and summarizes Downs’ perceptions
MALONE: INTERPRETIVE STRUCTURAL MODELING 403

23. High borrowing


costs and non-
availability of
financing

t
16. Superior
land suburban
ownership competition

public property and


insecurity
schools
1. Highcrime
rates
-
c 7. Poor
maintenance
-9.

-.
neessy
neighborhood
appearance

Deleted Obstacles:
19. High
property 2. Vandalism
taxes
come shift -
low demand

14 quality of
other publi
services
supply of
housing

Fig. 9. Interdependence of obstacles to investment in the Columbus, Ohio, central business district:
as perceived by a group of Ohio State University planning students. (Lower level obstacles intensify
or aggravate higher level obstacles.) The list of obstacles and their interpretation were taken from
161.

of the relative importance of the various obstacles, indicated perspectivewith regard totheconceptualcontext.For this
by brief descriptive phrases, in hinderingprivate investment exercise, the perspective was established partly by having all
invarious parts of central cities. Note that for any column participants read and study [ 6 ] . Furthermore,the exercise
in the figure the symbols denoting degree of importance was begun by conducting a group discussion to establish con-
disaggregate the element set into a three level hierarchy of the sensus statements concerning the objectives of the exercise.
sort constructed in the previous exercise. The purposes of the This discussion led to the following set of statements.
exercise conducted by thestudents,then, was to extend
Downs’ exercisebyinvestigating the pattern of causal inter- Theme: To investigate obstaclesto capitalinvestment in
action among these elements. For this purpose, the relation downtown areas.
-
was taken as ‘* * * aggravates or intensifies * *.” Focus:Tostudythe interdependence among obstacles to
In conducting a group ISM exercise, it is necessary to investment in commercial development in downtown C e
expendsomeeffort in establishinga commonor shared lumbus, Ohio.
404 PROCEEDINGS OF THE IEEE, MARCH 1975

Perspective: To attempt to describe conditions as they exist tool for those who wish to exercise careful, logical thinking in
today, according to the shared perceptions of a group of approachingcomplex issues, and then to communicatethe
graduate studentsin city and regional planning. results of thatthinkingtoothers.The exercises described
During this discussion it was further decided to delete three inthis
paper were designed to present and illustrate the
elements from the list as not being pertinent to the focus on utility and limitations of fundamental concepts inherentin
downtown Columbus. the methodology. It is hoped that experiences such as this
A computer-assisted ISM exercise was then conducted on the will encourage the participants to experiment with the applica-
remaining 22 elements.Queriesappeared onan interactive tion of the methodology within more specific substantive or
computer terminal in the following form: process frameworks. Applications which have been completed
at the time of this writing include: the modularization of a
“High crime rates and insecurity large econometric model [ 71, program planning and modeling
aggravate or intensify for public and societal systems [ 81, and a systematic approach
Vandalism ?” tohumansettlement planning [9]. Documentedcomputer
programs implementingthemethodology became generally
Group discussion was thenconducted in order to seek con-
available late in 1974,and further experiments andapplications
sensus on whether to respond“yes” (an entry of a1in the
are underway.
reachability matrix)or“no”(anentry of 0). If it was not
possible to reach consensus by discussion, a vote was taken. ACKNOWLEDGMENT
Inordertocompletetheentirestructure,ittook approxi-
Theauthor has benefitted greatly fromthe writings and
mately 2.5 h to respond to 192 such queries out of a possible
counsel of Dr. J. N. Warfield, who originated the approach to
total of 462 (22 X 21) if all possible pairwise comparisons had
structural modeling discussed in this paper. G. Clendening of
been made. The logic of the ISM process thus produced 270
Battelle’s Columbus Laboratories, has provided programming
entries by transitive’inference. Thestructureproduced by
assistance and invaluable advice and guidance in the selection
this initial pass through the ISM process is shown in Fig. 9.
andimplementation of varioushardware configurationsfor
It is beyond the scope of this paper to presentadetailed
running the demonstrations.
discussion of the substantive concerns of this exercise or to
make adetailedevaluation of thestructuralmodel which REFERENCES
resulted.Although such discussion and evaluation were an
F. Harary,R. 2. Norman, and D. Cartwight, Structural Models:
essential part of the original exercise, timeand procedural An Introduction to the Theory of Directed Graphs. New York:
limitations precluded extensive modifications and corrections Wiley, 1965.
to themodel of Fig. 9. J. N. Warfield, “An assault oncomplexity,” Battelle Memorial
Inst., Columbus, Ohio, Battelle Monogr. 3, Apr. 1973.
In evaluating their experience in participatingin this.exercise, -, “Structuring complexsystems,” Battelle Memorial Inst,
thestudents’ views weremixed. I n general, they were not Columbus, Ohio, Battelle Monogr. 4, Apr. 1974.
satisfied that the model of Fig. 9 was either a true representa- H. D. Lasswell, A Preview of Poh’cy Sciences. New York: Ameri-
can Elsevier, 1971.
tion of reality, or of the perceptions of the group as a collec- L. L. Thurstone, The Measurement of VuJues. Chicago, Ill.: Univ.
tive. They were in agreement, however, that the exercise had Chicago Res,1959.
been a worthwhile learning experience, for both the individual A. Downs, “Stimulating capital investment in central city down-
town areas and inner-city neighborhoods,” prepared for a con-
participants and the group. Had the exercise been conducted ference sponsored by the National Urban Coalition and the
within the contextof an ongoing project, the shared experience Johnson Foundation, February, 1973.
A. M. El-Mokadem, J. N. Wafileld, and D. M. Pollick, “Modulariza-
would have been invaluable in supporting future interactions tionof large econometric models: An application of structural
among the participants. modeling,” Battelle Memorial Inst., Columbus, Ohio, Mar. 1974.
k P. Sage, and R. W. Hawthorne, “On the application of systems
engineering methodologyto program planning and interpretive
CONCLUSIONS structural modeling of public and societal systems,” Information
and Control Sciences Cent., Inst. Technology, Southern Methodist
Interpretive structural modeling is a context free technique. Univ., Dallas, Tex. 1974.
Like othersuch mathematicalmethodologies itcannot be A. N. Christakis, and D. W. Malone, “A systematic approach to
human settlement planning,” Science-Based Planning Project, The
evaluated independently,butratheronly with regard to its Academy for Contemporary Problems and Battelle Columbus
utility in specific application contexts. ISM is seen as a useful Laboratories, Columbus, Ohio, Rogr. Rep. I, Jan. 1974.

You might also like