754
‘A STRUCTURAL LINGUISTICS APPROACH
TO METHOD IN SYSTEM DYNAMIC MODELING
by
Richard G. Fritz
Department of Economics
University of Central Florida
Prepared for the 1983 International System Dynamics Conference, Newton,
Massachusetts, July, 1983,
755
Introduction
The purpose of this paper is to describe some of the impact that
Linguistic structure has had on the method of modeling in system dynamics.
In the structuralist framework, language is viewed as a system of signs which
structure our patterns of thought and influence our behavior. Learned
languages are incorporated into the structure of the unconscious which then
contains and constrains the capacity for communication and discourse.
Linguistic systems are not isomorphic. Thus, when the language used in
communicating social, political and economic ideas changes, (i.e., from
verbal to static linear mathematics; or from verbal to dynamic nonlinear
mathematics) this affects the theoretical structure of the discipline. The
concepts found in the previous language systeii do not receive equal "value"
when translated into the new symbolic linguistic structure. What had been
posited and affirmed in the luminous space of understanding in the previous
system detaches itself from the squares they inhabit and are reformed under a
new set of "signs" (Foucault, 1970. p. 217).
Many writers have attempted to attend to the epistemological questions
concerning the role of mathematics and its affect on knowledge of social
systems. A sample of this literature would include: Boulding (1666),
Georgescu-Roegen (1979), Katorizian (1980), Fusfeld (1980), Hardy (1978),
Dennis (1982), Quine (1960), Samuclson (1952), Dorfman (1954), and Lilienfeld
(1978). These authors (and many others) have recognized the constraints
imposed by the variety of verbal and non-verbal language systems. Generally,
the analysis is focused on the use of words or natural language systems
rather than mathematical symbols. On a few occasions, attempts are made to
compare the relative versatility of natural languages versus symbolic
language systems in evaluating and analyzing social systems. In practice,
social system modelers seldom attend to the impact linguistic structure has
on their working analytic paradigms. Even less attention has been applied to’
the comparison of various nonverbal linguistic structures employed in social
system modeling. One notable exception is Meadows (1980), who without the
use of explicit linguistic vocabulary covers several of the critical issues.
The primary differences between languages is not the symbols they use or
the meaning expressed in the symbols but in their structures. For the
structure or syntax of two languages to be identical or isomorphic, one must
be able to place their elements in a one-to-one (and onto) correspondence.
The obvious differences between a natural language and mathematical language
is in the richness of vocabulary and complexity of syntax of the former and
the poverty of those in the latter, Barbut (1970) argues that this
opposition points up the enormous efficiency of mathematical models, a
simplicity rarely encountered among the human sciences, The language of
mathematics is employed at the expense of a reduction in phenomena to which
those models may be applied. When reality is complex, symbolic language
retains only certain characteristics of the mental model translated through
the natural language system; those characteristics which matter most.
This paper will address the relative utility of employing the linguistic
structure used by systen dynamics compared to translating the modeler's
perception of reality imto: other symbolic language systens. The first
section will review the relation of language to the method of scientific
inquiry. This will include a discussion of the debate over the problem of
translating natural languages into symbolic languages for the purpose of
evaluating policy alternatives of social eystens. The final section of the
paper will specifically identify some of the differences between the imposed
linguistic structure of system dynamic models and the symbolic language
systems often employed in orthodox economic analysis.
Language and Epistemology
European epistemoligists, and notably the French, have contributed much
to the understanding of the problem of scientific description. The
relationship between the "truths" of a science and the descriptive schenata
or "language" used to arrive at and to describe these "truths" has broad
implications for the focus and direction of that science.
Anglo-American social scientists, being in the tradition of the British
epistemologists Karl Popper and Thomas Kuhn, have largely ignored this
problem of language as they tend to view the direction of science as a result /
of the conscious choices of scientists, System dynamicist have recently
employed the refutationist approach developed by Popper in order to show that
the method of system dynamics offers a large number and variety of "points of
contact" between theories and reality which represent genuine possibilities
of exposing errors in the theory (Bell and Senge; 1980), while others have
employed the logic of Kuhn's paradigmism in an attempt to compare the’
756
problem-solving qualities of system dynamics and its leading competitive
alternatives to modeling social systems (Meadows, 1980).!
The French philosophers of epistemology, such as Gaston Bachelard,
Georges Canguilham and Michel Foucault, see the movement of science as
relatively autonomous, proceeding by reorganizations, ruptures, mutations and
inseparable from its cultural frame. Scientists are not the cause of
scientific practice, only agents, subject to the external determinations -
social, economic, ideological, and political. Bachelard maintains that the
scientist constantly comes upon epistemological obstacles which are
crystallized and systemized in philosophy and which produce braking effects
in scientific practice. The epistomelogical obstacle emerges every time a
pre-existing organization of thought is thréatened, that is it appears at a
point where rupture with the past threatens. Its effect is to patch up, to
displace the question before it is posed, to prevent the question from being
posed. These obstacles are the perceptions, representations, values and
attitudes of a given society that intervene in science through language. The
images, metaphors, or in Bachelard's terminology "traces" present in ordinary
language inhibit the progress of science as they embody a certain
representation of the real, a reality offered to investigation. Expressions
such as "the rising/setting stn", remnants from pre-Newtonian science,
permeate the unconscious and result in the substitution of imaginary
questions for the real questions by which a science progresses. Bachelard
states that the dangers of metaphors for the formation of the scientific mind
is that they are not always passing images; they press on towards autonomous
thought.
Michel Foucault, in the tradition of Bachelard, relates science and the
discursive practice of a society at a given point in time. Science is
defined by the perceptual field of a given era - what is visible and
invisible, thinkable and unthinkable, stateable and unstateable. The objects’,
that will or will not acquire scientific status are dependent upon an
ensemble of interlocked and hierarchially structured discursive practices.
As the discourse of a given science or discipline acquires power and status,
it affects the perceptual field of other disciplines. It begins to function
as the norm, governing attitudes with respect to real objects and problems.
T yor a good survey of system dynamics and scientific method see: Bell and
Bell, 1980.
According to Foucault it is the structure and hierarchial relations of
discursive practices, which he calls the discursive formation of an era,
which assign the forms and limits to theory (Lecourt).
The existence of a pure intellectual space in which the concepts of
science are worked out by a body of scientists is pure fiction to these
French epistemologists. The ideological values of the social formation in
which the science is inscribed and the language through which these values
are passed permeate the consciousness of the body of scientists and render
their choices to a limited set of predetermined paradigns.
Language Thought and Reality
To understand how language influences scientific investigation it is
necessary to consider the relationship between language and perception,
language and thought, and language and society. ‘There is no pure act of
perception without thought. The flux of experience must pass through the
interpretative schemata of the mind. Language does not mirrow the mind but
rather it is language which gives structure and form to our thoughts. What
we see and think tends to be limited to what we can say. Perceptions and
thoughts are also socialized because the language that gives them form is
iumersed in the on-going life of a society and reflects the consciousness of
that society.
Languages are systems of categories and rules based on fundamental
principals and assumptions about the world. These principals and
assumptions are not related to thought: they are thought. Benjamin
Whorf called these fundamental organizing assumptions a ‘science’ and a
"metaphysic', that is a systematic account of reality and the apriori
assumptions on which that account rests. Such assumptions are embodied
in language, learned through language and reinforced in language use . .
Institutional science and metaphysics require a professional class of
scientists and metaphysicians to articulate their basic assumptions, but
even professional thinkers use a language and through their prior use of
its categories and processes these communal assumptions filter into .
their thinking. The interrelations between the two kinds of science and
the two kinds of metaphysics makes an important subject of study. (Kress
and Hodge, 1980, p. 5)
In the 1830's Wilhelm von Humbold postulated that one thinks in forms
limited and determined by the forms of one's native language. In the 1930's
Benjamin Whorf linked the structure of language'with a particular world view
and sought to reveal the metaphysics implicit in the structure of *
Indo-European languages (Coetzee, 1977).
The Indo-European languages and many others give great prominence to a
type of sentence having two parts, each built around a class of words —~
substantives and verbs . . . The Greeks, especially since Aristotle,
built up this contrast and made it into the law of reason. . .
Undoubtedly modern science, strongly reflecting western Indo-European
tongues, often does as well all do, sees actions and forms where
sometimes it might be better to see states. (Coetzee citing Whorf, p. 4)
The tendericy to perceive the universe in terms of objects and actions rather
than states is imposed by the use of a language which breaks down reality in
terms of subjects and verbs. The standard order so prevalent in English
(Subject-Verb-Object) imposes the reading of causality and temporality into
experience whereas these meanings would not necessarily be transmitted in
other families of languages. Coetzee affirms that the Subject-Verb order is
metaphorical because it imposes a temporal-causal order over the syntactic
order.
Modern science as we know it in the Western World mirrors the structure
and processes of Indo-European languages in that it seeks to give a
systematic account of reality by linking events to a network of causal
relations and to structures of objects and forces. Language provides a
theory of reality which is superimposed on the scientific theory it helps to
articulate.
There is evidence that Newton struggled with language while attempting
to explain the law of gravitation. The controversy that ensued over this law
(and which became a cause celebre in the history of science) was due to
language.
By using the standard syntactic order in English of Subject-Verb-Object,
Newton was obliged to assign temporal-causal relationships to heavenly
bodies.
In his general law of gravitation, Newton states that every two
particles of the universe attract each other with a force proportional
to their respective masses and inversely proportional to the square of
their distance apart. The key word is ‘attract’. Apart from occasions ‘
where the law is expressed in mathematical symbolism, there is no
statement of the law in Newton that does not include the word ‘attract’
or @ synonym equally metaphoric. The controversy that broke out over
the concept of gravity soon after the publication of the Principia was
published in 1686 centered on this metaphor. (Coetzee, p. 5)
Newton was obliged to withdraw this first version and replace it with a
more austerely mathematical treatment, "to prevent the disputes which might
be raised" (Coetzee, p. 7). He used mathematics to circumvent the added
mednings inherent in stating in natural language the relations between the
elements of the universe, and to make his theory more acceptable to the
professional thinkers of his day.
Newton attempted to eliminate metaphorical content in his scientific
discourse by using two linguistic techniques. These techniques have been
adopted by the modern scientific community and contribute to the scientific
discoursive style as we know it: the predominate use of passive constructions
and nominalisations.
Passivisation is a transformation of the basic transactive model
(Subject-Verb-Object) in which there is a source, a verbal process, and an
affected entity. The transactive model indicates clearly the causal process
as all agents in the process are specified. When this model is transformed
into a passive construction the source or agent of the process may be
omitted. This information is lost or obscured. The passive construction
enabled Newton to avoid philosophical questions about causality by omitting
the syntactic agent. Nominalisations reduce both agents, source and affected
entity, and the process to a state, thereby eliminating all temporality and
causality.
When causal and temporal relationships are blurred, discourse is vague
or ambiguous as the source and consequence of phenomena. “The science that
proceeds through non-transactive models will tend to be a large collection of
particular facts about self-caused events which co-exist" (Kress and Hodge,
p. 39). Kress and Hodge maintain that such a style is functional for the
community of scientists in that it allows one to avoid making distinctions
when accounting for data beyond the scope of theory.
Labov adds that groups create kinds of languages which serve to
reinforce a sense of identity and exclude others. The distribution of power
is reflected in and sustained by differences in language. Scientific
language also sets up a barrier around the priviledged knowledge of its
community of specialists. The repetitive use of the expression “given . ."
is a case in point. What is given? Who determines the goals? How is the
theory defined? What is the status of the investigator? This common
linguistic device can serve to remove important questions from public
consideration (Kress and Hodge).
758
Linguistic Style, Theory, Ideology
Linguistic transformations are used to effect theoretical
transformations and are not free of ideological determinations, according to
Fowler, et.al., (1980, p. 63).
Anomalies constantly face scientific theories and are resolved either by
changing the theory or by reinterpreting the event. Awkward facts may
be successfully denied, suppressed or reinterpreted (through linguistic
devices). Anomalies are a challenge not simply to the ideology but to
the legitimacy of the order.
Kress and Hodge maintain that linguistic transformations serve two
functions: economy and distortion. Linguistic forms allow significance to be
conveyed and to be distorted. In this way the hearer/reader may be
manipulated and informed (p. 6). What is significant is the disappearance
of deleted material and its non-recoverability in the text.
“Presenting anything in or through language involves selection — how
the speaker/writer chooses to present reality" (Kress and Hodge, p. 15). A
profession is not self-contained. It has links with institutions, groups,
and movements. Its credibility depends on which forces it gives expression
to and to which institutions or segments of society it identifies with,
supports and respects. Language serves to confirm and to consolidate the
organizations which shape it, being used to manipulate people, to establish
and maintain them in economically convenient roles and statuses, to maintain
the power of state agencies, corporations, and other institutions" (Fowler,
et.al., p. 190). i
Linguistic styles are socially determined patterns of language.
Preferred syntactical arrangements can encode a world-view without the
conscious choice of the speaker/writer. World-view comes from relations to
institutions and socio-economic structure of society but is facilitated and
confirmed by language use. "The systematic use of certain linguistic
structures is connected to the texts’ place in the socio-economic system and ‘,
exist prior to the production of the text and our reception of it" (Fowler,
et.al., p. 185), We are socialized into holding theories and judgments
because of the social meaning, reinforced in the lexical and syntactic
structures we use.
It is unnecessary to assume that groups deliberately construct a "syntax
of mystification". Once a style comes into existence it becomes appropriate
for expressing a given content. Groups do not consciously recognize the
pufposes they encode in language and that the aims which they mediate in
their professional capacities may not coincide with their beliefs or
sympathies.
Natural Language Versus Mathematics
Adopting mathematics as an instrument of investigation and communication
of scientific research does not negate the problem. Maher asserts that
mathematics and logic are only "parasitic systems", outgrowths of the
Processes of natural language. He sees natural languages as palimpsets as
they bear the imprint of different eras. Language surface, its forms and
structures, reflects not the present but the past. The grammatical system
tends to persist indefinitely and will in time cease to symbolize the
cultural forms which motivated its existence. "Surface structures are handed
down from one generation to the next, while the underlying values are subject
to revolutionary change" (Maher, 1977, p. 5). It is the decalage between the
surface structure and the shifting values that motivated it that creates the
metaphor.
Mathematics is also metaphorical in that it grows out of abstraction.
There is no pure abstraction as there is no pure perception.
The equation, the syllogism, all their complex superstructures . . are
intrinsically nothing but metaphors. “The source of those metaphors is
figure-ground differentiation of configuration, with abstraction of
certain salient features, preceding from other features of the bundle.
(Maher, p. 8)
No matter what linguistic medium is adopted, scientific description will
remain problematic. In reflecting on this problem one must consider two
questions. What is the relationship between the linguistic medium and the
material and what does this medium impose on the material.
Regnier states that this relationship is always characterized by
transformation and deformation.
Entre la representation et le represente, le rapport n' est pas
simplement d' abstraction et d'approximation mais de transformation et de
deformation. (Regnier, 1974, p. 23)
A model is the interpretation of a theory and must furnish a description
which is not contradictory with the theory. All aspects of phenomena are not
represented in a’ model. One choses properties which present a certain
coherence and one negates the accessory. What the model retains and what it
ignores poses the problem of what is pertinent and what is negligable.
759
10
Judgment and interpretation are closely linked with perception, linguistic
conditioning and ideology.
A descriptive medium, whether natural language, logic or fields of
mathematics, has certain semantic limits that orients research with the
boundaries of its own particular representation of the world. Scientific
output may be viewed as a compromise between the necessities of the
descriptive medium and those of the real. Much of the debate over the
mathematical modeling methods used to represent social systems stems from
the modeler's world view (Meadows, 1980). System dynamicists assume that the
systems are primarily closed, interacting with the environment which
influences it, and are more interested in the dynamic path of a response than
the end state. Orthodox economics, through econometric models assume that
the world is dualistic and open. This means that the environment (markets,
government action, foreign influences, or institutional settings) delivers
inputs (exogeneous) to which the system provides specific responses. System
dynamicists believe the problems are predominately addressed as long run
issues, while the microscopic view of econometrics confines itself to the
short run. Meadows (p. 237 ,1980) summarizes several characteristics useful
for comparing modeling paradigms. However, linguistic structure is not
included in the comparative categories.
In order for the linguistic structures to matter between modeling
methods, we must first show that they are different. That is that the mental
model in our head is originally formed in a natural language system, then
translated into a syubolic language structure consistent with the
quantitative modeling tool. Further that the translation of the mental model
differs depending upon which symbolic language system is receiving the model.
Natural Language Systems, Mental Models and Mathematics
By the end of the 18th century the Newtonian scheme was decisive in
convincing the world that nature 1s mathematically designed and that the true
laws of nature are mathematical. Newton's amazing contributions were made
possible by his reliance on mathematical description even where physical
understanding was completely lacking. Newton placed mathematical description
and deduction at the forefront of all scientific accounts and prediction.
While this position was attacked: by David Hume and others, Immanuel Kant
affirmed that all axioms and theorems of mathematics were "truths" (1781).
ul :
However, Kant argued that science was a world of sense impressions arranged
and controlled by the mind in accordance with innate categories such as
space, time, cause and effect and substance. "The mind contains furniture
into which the guests must fit" (Kline, 1980, p. 77).
The development of non-Euclidean geometry finally led to the recognition
that mathematics was not a body of truths. The debate over the “anticipatory
function" of the language of math continues (Kuyk, 1977, pp. 141-170). That
is, does the axiomatic language "run ahead" of verbal language such that the
manipulation of a formula leads to a result that could not be thought to be
true before the manipulation. Whether mathematics offers a more useful
(powerful) linguistic structure for the social sciences is not the issue of
this essay. However, those readers interested in the application of this
problem in the language structure used in system dynamics are referred to the
work of Forrester on the counterintuitive behavior of social systems
(Forrester, 1971).
The salient point is that orthodox economic analysis adopted the
Linguistic structure of differential calculus with the "marginaliet"
revolution in the late nineteenth century. The structure of the adopted
calculus resulted in the dominant theoretical role played by a single
economic agent. Ideologically, this shifted the focus away from the
“political economy" of society to the "economics" of utility maximization by
the individual. The new linguistic structure introduced new "words";
derivative and infinitesimal. The marginalists used these to isolate
relationships by the necessary linguistic constraint of assuming "other
things remain equal" so that the changes in the economic variable on which
they focused was not to be systematically related to the variation in the
variables they were ignoring. The need for. simplification in the new
symbolic language structure was invoked to support the position that the
assumptions of a theory should be removed from the ambit of criticism
(Friedman, p. 14). Abstraction assumptions are not an element of the
“axiomatic structures" of the theory and therefore may be ignored by the
formalistic language of the model, Some believe that the deterministic
language of differential calculus and the “representative” individual
economic agent were merely expository devices chosen because of their
pedagogical utility (Gill, p. 76).
12
The translation of classical economic theory into the calculus of
exchange employed a new-linguistic structure. In the new language system of
“pure exchange", production relations, which had been a dominant feature of
the mental models of the classical economists, became unnecessary elements.
Menger, Jevons and Walras stressed the notion of exchange as expressing the
essence of the structural system: "production to ‘some extent appeared merely
as an indirect way of exchanging intitial holdings" (Arrow and Starrett,
1973, p. 133). Although there have been numerous attempts to define the
compliment of classical production within the new conceptual plane adopted by
the marginalist, the theory of exchange exists in metaphysical space, cut off
from reality by the absence of a theory based on real inputs and discrete
capital goods (Fusfeld, 1980).
‘Thus, the natural language system of the classical economists
represented a different set of meta-assumptions (or methodological priors)
than the symbolic language system of the neoclassical economists.
Translating mental models into the working models of the discipline led to
differences in policy conclusions under the conceptual plane of calculus.
System dynamists have recognized that differences in analytic paradigms can
lead to differences in policy conclusions (Anderson, 1981), although they
have not related the meta~assumptions. back to linguistic structure. Having
established that linguistic structures matter between modeling methods, the
next issue is the extent that the translation of the mental models differ
when translating from the natural language system into competing symbolic
language systems of the receiving quantitative models.
The Translation Problem and Mental Models
In his attack on systems theory, Robert Lilienfeld (1978) argues that
the road to reality is traversed through everyday language. The major
unresolved problem of epistemology is how we come to know our own mental life ’,
and that of others. Philosophers continue to debate how we come to be aware
of physical objects and how far subjective elements enter into our experience
of them. Systems theory to Lilienfeld represents outmoded thought patterns
because one of the basic findings of "modern science" is that:
development and analysis of modern physics is the experience that the
concepts of natural language, vaguely defined as they are, seem to be
more stable in the expansion of knowledge than the precise terms of
scientific language, derived as an idealization from only limited groups
of phenomena. This is in fact not surprising since the concepts of
13
natural language are formed by the immediate connection with reality;
they represent reality . . . On the other hand, the scientific concepts
are idealizations; they are derived from experience obtained by refined
experimental tools, and are precisely defined through axioms and
definitions. Only through these precise definitions is it possible to
connect the concepts with a mathematical scheme and to derive
mathematically the infinite variety of possible phenomena in this field.
But through this process of idealization and precise definition the
immediate connection with reality is lost. The concepts still
correspond very closely to reality in that part of nature which had been
the object of the research. But the correspondence may be lost in other
parts containing other groups of phenomena (Lilienfeld, p. 251, quoting
Heisenberg, p. 200).
Lilienfeld argues that systems theory as technique - in its computer-based
simulation models, in the mathematical foundations of cybernetics ~ is based
on deterministic categories (p. 256). Accordingly, to Lilienfeld systems
theory is an ideology which offers nothing new except a new vocabulary.
Idlienfeld's conclusion that systems theory is neither a philosophy nor a
science and thus on pragmatic grounds it appears to make no difference,
ignores the contribution of the new linguistic structure. Even if his
conclusions are correct, and there is much controversy over the issue,
systems theory would "make a difference” if there is value in developing an
alternative symbolic language system from which to translate the mental
models social scientists have conceptualized in their natural language
systems.
‘The new vocabulary of systems analysis fits the semiofficial doctrine of
translation developed by Dennis (1982). This follows the development from
the nineteenth century to present that the concepts and propositions of
economics could be translated into the symbols and formulas of mathematics.
The doctrine has some credibility. Mathematical symbols, formulas, and
methods do enter into economic theorizing but not in a way as to prove
behavioral propositions about human beings and their economic actions.
Mathematics, traditionally developed is the logic of numbers and number ‘
relations, It is not a logic about events and the conditionality of the
gccurrence of events (Dennis, p. 107). Even though number systems and
measurement systems have been shown to be homomorphic relational structures,
numerical functional formulas do not express behavioral propositions about
events and the contingency of their occurrence (Krantz, et.al., 1971).
Moreover, contrary to Samuelson the syncopation of homomorphic identity
found in the translation of natural languages to mathematics cannot yield the
761
14
"Jogical identity of words and symbols" (Samuelson, 1952, pp. 59-60).
Translation of a sentence conveys the same information as the original, in
that it expresses the same proposition (Rescher, 1969, p. 322). The mere
correspondence of synonymous words does not meet the necessary conditions of
adequate translation. Thus translation may take on different relative
qualities that may range from "naive" to "fad
706-710). Employing special symbolism in scientific work is primarily to
achieve notational conciseness as an aid to logical manipulation.
to "strict" (Dennis, pp.
Abbreviation through linguistic symbolism does not afford greater degrees of
precision than ordinary language, only clarity gained by the use of
abbreviated symbolisms. Therefore translation of language systems may be of
different quality with the highest quality resulting in symbolic notation
that yields the most clarity in expressing the original propositions found in
the natural language. When the propositions being translated are scientific
argumentation, the translational adequacy becomes a vital aspect of ensuring
the logical rigor of the argument.
Translation, Transformation and Syntactic Order
Recall the previous example of syntactic transformation when Newton
adopted the passive construction and nominalisations rather than the basic
transactive model (Subject-Verb-Object). The transactive structure of
Indo-European languages indicates the causal and temporal processes as all
agents in the process are specified. The passive construction enabled Newton
to avoid causality and temporality by reducing the source, affected entity
and the process to a state. Nominalisations and passivisation may be
achieved in linguistic structure by either transforming the existing language
system to a non-transactive model or by translating one linguistic system
into another while at the same time altering the syntactic order. Both
result in obscuring the causal and temporal relationships so that discourse
is rendered vague or ambiguous as to the source and consequence of phenomena.
As an imposed value judgment, linguistic translation which results in
non-transactive models where transactive mental models were the goal, are
categorized as “naive” translations. Natural language translated into
symbolic systems that retain the original transactive model meet the
necessary condition of "fair" translation. (Here, Dennis’ categories are
adopted to fit this essay's theme.)
15
. It has been argued that orthodox economics, relying on the logical
empiricism of statistical (econometric) modeling, achieves the passivisation
of theoretical discourse by "nouning" (Neale, 1982). Although Neale's
terminology is combersome, he succeeds in identifying the linguistic
transformation problem.
Nouning has contributed to confusion about meaning, and to confusing
word order with cause. It is important to distinguish between value as
a noun (thing), which it is not, and to value as a verb. The former
generates a bunch of non-questions, such as... , "Why do diamonds
have more value than bread?" Diamonds do not have value. They have
mass, density, .. . Instead . . . one should be asking, "How do people
value diamonds?" And answering by saying, "They value them by stealing
them, buying them, insuring them, killing for them," and so on.
Stealing, buying, . . . are how people value diamonds. (Neal, pp.
362-63)
In orthodox economics, nouns are explanations: utility, preference, tastes.
These are not things but verb processes. Neale believes a clearer
understanding of social systems is achieved by rejecting "nouning” and
arguing for "processual verbing". Verbing influences our ideas about cause
while nouns exist separately from their being or doing.
Broadly described, econometric models translate natural language systems
into symbolic language systems, transforming the transactive structure into
passive syntax. At first glance it appears that the causation analysis in
the robust literature of probability theory is transactive. However, the
question rests not with the power of statistical models but with what is the
sentence (represented by an equation) saying. Are econometric models
paraphrasing orthodox economics by avoiding the verb processes of the
transactive structure?
Following the characteristics used by Meadows (1981, pp. 174-200),
econometric models are detail decision making based, product oriented, and
structured as open systems requiring many exogeneous variables to drive the
model. When two-way causation does appear as in simultaneous-equation
formulation, equilibrium is achieved without temporal analysis. The primary
focus of econometric models is on the noun of the equation sentence. In the
basic linear open model there is no feedback describing the temporal and
causal path that represents the Indo-European syntax of natural language
structure, The result is similar to that of Newton adopting the mathematical
linguistic structure in order to avoid the debate over causal source. The
reduced form estimation process means that econometric models tend to
762
16
represent surface phenomena only, with much causal structure implicit”
(Meadows, p. 229).
The econometric translation has benefited from the recent developments
in the theory of measurement. In the modern literature, attention has
focused on "the construction of homomorphisms (scales) from empirical
relational structures of interest into numerical relational structures that
are useful" (Krantz, et.al, 1971, p. 9). The relational concepts of
structural identity (isomorphism) and structural similarity (homomorphism)
are used to justify "the direct application of computational methods to the
results of measurement" (Luce and Suppes, 1968, p. 72). The symbolic
sentence: x + y = z, can serve the dual purpose of expressing 'z' as the
empirical result of measurement and the result of numerical computation
whereby the abstract operation of addition is performed upon the numbers x
and y. ‘The consequence of isomorphism between empirical and numerical
relational structures is that the same symbolism may be adopted for both
systems. Thus the rationale for the use of numerical algebras to espouse and
describe certain properties of empirical relation systems.
Krantz, et.al., argue that the problem or representation is the heart of
the measurement development: “When measuring some attribute of a class of
objects or events, we associate numbers (or other familiar mathematical
entities, such as vectors) with the objects in such a way that the properties
of the attribute are faithfully represented as numerical properties" (Krantz,
et.al., 1971, p. 1). The problem of representation, while important, is not
the source of concern in most social science theories. Most social policy
analysis concerns causal connections (or patterns of connectedness) that are
open to empirical inspection (corroboration or refutation), "The exploration
of systems of causal connections relating distinct events or conditions seems
not to have been a part of the agenda for theorists of measurement, even in
chapters devoted to the measurement of probability involving an "algebra of ,
events" (Dennis, p. 1058).
While there are some examples of fair translations employing logical
granmers in simultaneous equation econometric models, the dominant linguistic
structure of econometric models transform the transactive structure of natur-
al language into a passive structure. Causality becomes the problem of the
measurement of probability considered in an open structure (Granger, 1969;
Sims, 1972). However as Bell and Senge (1980) have pointed out, if a model
W
includes multiple exogenous time series inputs, disentangling internally
generated behavior from externally generated behavior may be difficuit «©
impossible. From the viewpoint of linguistic structure, their conclusion is
a logical outcome of avoiding the transactive model of natural language.
The system dynamicists' model based on endogenous explanations
parrallels the refutationists insistence on causal explanations. According
to Bell and Senge, the power of simulation testing to reveal flaws or
corroborate model assumptions is enhanced by employing endogenous
explanations of behavior. Without exogenous time series inputs, a system
dynamics model should generate the empirical behavior of interest. An array
of simulation tests conducted without time series inputs guarantees that
model behavior arises from feedback loops. The interactions necessary for
understanding the causes of behavior are found within the wodel structure.
This attends to the scientific goal of highly corroborative theories
requiring multiple “points of contact" with reality (Bell and Senge). The
refutationists view of scientific method defines objectivity in terms of the
degree the theory presents opportunities to test it against reality.
The linguistic structure of feedback loop analysis (mathematics of
integration used in control theory) provides increasing objectivity through
refutability. System dynamics has a strict syntax and structure. All models
must be have the property of closure containing at least one feedback loop.
The model closure test requires that "starting from any point in the
influence diagram it must be possible to return to that point by following
the influence lines, in the direction of causation, in such a way as not to
cross one's track (Coyle, 1978). "Closing the loop" can only be accomplished
in a temporal setting with an explicit delay intervening between initial
action and the resulting feedback (Roberts, 1981, p. 7). The syntactic
structure of the variables (sentence components) is determined uniquely by
the type of each variable.
Coyle outlines the relationships between the three variable types;
rates, levels and auxiliaries (Coyle, p. 523-28). Level variables (an
accumulation or integration over time) are stocks (nouns) that change as
flows come into and go out of it. Rate variables are the flow, decision,
action (verb) or behavior that changes over time as a function of the
influence processes. Auxiliary variables are combinations of information
763
18
inputs into concept (predicates) terms. Rates must be the preceding variable
of a level. Auxiliary or rate variables may succeed levels. Other
combinations are explained by Coyle, but the syntactic order must remain
transactive (Subject-Verb-Object) around the closed loop. ‘This focuses the
attention on the general system reaction to general disturbances and on the
dynamic path of a response rather than its end state (Meadows, pp. 227-28).
In Neale's teriinology, system dynamics employs the linguistic structure of
verb processes rather than "nouning" the hypothesis. The result is a "fair"
translation of the natural language system into a symbolic language structure
which is more concise and facilitates computer modeling of causal relations.
Conclusion
The translation of natural language systems into symbolic linguistic
systems cannot produce isomorphic structures, ‘hus the problem is to
minimize the loss of coherence that can result from the transformation of the
transactive structure found in Indo-European language grammars into the less
causally and temporally explicit form of the passivitive structure.
Differences in analytic paradigms can lead to differences in policy
conclusions (Anderson; Phillips; Meadows). Some of these differences can be
explained by the impact linguistic structure imposes on mental models and
quantitative models.
As languages are not isomorphic, what is imposed on science by their
models will vary. Natural language as an abstract system of classification
eubodies a theory of reality in its forms and syntax. Groups of languages
present a preferred model for interpreting and perceiving phenomena.
However, these systems are made actual by human agents in social interaction
and is renegotiated in response to forces outside the language system (Kress
and Hodge, p. 63).
Mathematics as an abstract system also imposes semantic limits which
orient research, It may, as natural language, be manipulated to present
certain points of view. The preference for mathematics over natural language
to investigate and explain social science forces may be said to be in part
ideological. As in the case of natural language this choice may or may not
be conscious.
System dynamics employs a linguistic structure which yields "fai
translations from natural language system. This translation provides
19
numerous points of contact with reality and thus offers an opportunity for
refutable hypothesis to be tested against reality. This essay has attempted
to enhance awareness of the placement of linguistic structure within the
methodological critique of the field of system dynamics. Evaluation of the
contribution by the discipline should not only include policy and theoretical
implications of the field but also methodology, included in which must be a
consideration of the linguistic structure.
764
20
Georgescu-Roegen, N-, "Methods in Economic Science," Journal of Economes
Issues, Vol. 13 (2), June 1979, pp. 317-28.
Gill, Flora: "Some Methodological Implications of the Marginal Revolution,"
Australian Economic Papers, June 1981, pp. 72-82.
Granger, C.W.J., “Investigating Causal Relations by Econometric Models and
Cross-Spectral Methods," Econometrics, Vol. 37, July 1969, pp. 424-38. °
Hardy, Williams, Language, Thought and Experience, Baltimore, MD: University
Park Press, 1978.
Hume, David: A Treatise of Human Nature (Oxford: Clarendon Press, 1739).
Jevons, W.S., Theory of Political Economy, second edition, 1879 (London:
MacMillan Publishers, 1871).
Kant, Immanuel: Critique of Pure Reason, Traduit en anglais pas F. Max
Muller, 2 editon, (New York: MacMillian, 1905).
Katouzian, Homa, Ideology and Method in Economics, (New York: New York
University Press, 1980).
Kline, Morris: Mathematics, The Loss of Certainty, (Oxford: Oxford University
Press, 1980).
Krantz, David H., et-al., Foundations of Measurement, Vol. 1, (New York:
Academic Press, 1971).
Kress, Gunther and Hodge, Robert, Language as Ideology, (London: Routledge
and Kegan Paul, 1979).
Kuyk, Willem: Complementarity in Mathematics, (Boston: D. Reidel Publishing
Company, 1977).
Labov, William, Sociolinguistic Patterns (Philadelphia: University of
Pennsylvania Press, 1973).
Lecourt, Dominique. Marxism and Epistemology, Bacheland, Canguiihem, and
Foucault, (London: Humanities Press, 1975).
Lilienfeld, Robert. The Rise of Systems Theory (New York: John Wiley and
Sons, 1978).
Luce, R. Duncan and Patrick Suppes. "Mathematics", in International
Encyclopedia of the Social Sciences. Vol. 10, (New Yor!
Collier-MacMillian, 1968, pp. 65-76).
Maher, Peter, " 'Upside-down' Language and Mind: The Linguistic Basis of
Logical and Mathematical Formalism," Language Science, December 1977,
pp. 1-9.
ai
= BIBLIOGRAPHY
Anderson, David F. “How Differences in Analytic Paradigms Can Lead to
Differences in Policy Conclusions," in Elements of the System Dynamic
Method, Jorgen Randus, ed., Chapter 3, pp. 61-74 (Cambridge, MA: MIT
Press 1980).
Arrow, K, and Starrett, D.A. "Cost and Demand - Theoretic Approaches to the
Theory of Price’ Determination," in J.R. Hicks and W. Weber, eds. Carl
Mimger and the Austrian School of Economics, (Oxford: Oxford University
Press, 1973).
Barbut, Marc, "On the Meaning of the Word Structure in Mathematics,”
Introduction to Structuralism, Michael Lane (ed), (New York: Basic
Books, 1970).
Bell, James A. and James F. Bell, "System Dynamics and Scientific Method,”
Elements of the System Dynamic Method, J. Randers (ed.), (Cambridge, MA:
MIT Press, 1980), pp. 3-22.
Bell, James A. and Senge, Peter M., “Methods for Enhancing Refutability in
System Dynamic Modeling," Studies in the Management Sciences ~ System
Dynamics Vol. 14, 1980, pp. 61-73.
Boudling, Kenneth, "The Economics of Knowledge and the Knowledge of
Economics," American Economic Review, Vol. 61, No. 2, May 1966, pp.
1-13.
Coetzee, J.M. "Newton and the Ideal of a Transparent Scientific Language,"
Journal of Literary Semantics, Vol. 11, No. 1, April 1982, pp. 3-11.
Coyle, R. Geoffrey, "System Dynamics: An Approach to Policy Formulation,"
Journal of Business Policy, Vol. 3, No. 3, Spring 1978, pp. 40-48.
Dennis, Ken, "Economic Theory and the Problem of Translation," Journal of
Economic Issues, Vol. 16, No. 3, September 1982, pp. 691-712.
Dorfman, Robert, "A Catechism: Mathematics in Social Sciences," Review of
Economics and Statistics, Vol. 36, November 1954, pp. 374-77.
Forrester, Jay W., "Countaintuitive Behavior of Social Systems," Technology
Review, Vol. 73, No. 3, January 1971, pp. 1-16.
Foucault, M. (1970) The Order of Things, (New York: Vintage Books,
Translation of Les Mots et les Choses, 1973).
Fowler, R., Hodge, Robert, et.al.: Language and Control (London: Routledge
and Kegan Paul, 1979).
Friedman, Milton; "The Methology of Positive Economics,” dans Essays in
Positive Economics (Chicago: Univeristy of Chicago Press, 1953).
Fusfeld, Daniel: "The Conceptual Framework of Modern Economics," Journal of
Economic Issues, Vol. XIV, March 1980, pp. 1-52.
765
22
Meadows, Donella H. "The Unavoidable A Priori," in Elements of the System
‘Dynamics Method, edited by Jorgen Randus, Chapter 2, pp. 23-56.
(Cambridge, MA: MIT Press, 1980).
Neale, Walter C. "Language and Economics," Journal of Economic Issues, Vol.
14, No. 2, June 1982, pp. 355-69.
Newton, Isaac, Mathematical Principles of Natural Philosophy, traduit en
anglais par Andrew Motte, 1729, revise par F. Cajori, (Berkeley:
University of California Press, 1947).
Phillips, Bernard. "Paradigmatic Barriers to System Dynamics," Proceedings of
the International Conference on Cybernetics and Society (Cambridge, MA:
MIT Press), October 1980, pp. 682-688.
Quine, W.V. Word and Object. (New York: John Wiley and Sons: 1960).
Quine, W.V., Methods of Logic, 3rd ed. (London: Routledge and Kegan Paul,
1974).
Regnier, Andre, La Crise du language scientifique. (Paris: Editions
Anthropos, 1974).
Rescher, Nicholas, "Translation as a Tool of Philosophical Analysis," Journal
cf Philosophy, Vol. 52, 1969, pp. 152-55.
Roberts, Edward (ed.), Managerial Applications of System Dynamics (Cambridge,
MA: MIT Press, 1981).
Samuelson, Paul A. "Economic Theory and Mathematics, An Appraisal," American
Economic Review, Vol. 42, May 1952, pp. 56-66.
Sims, Christopher, "Money, Income, and Causality," American Economic Review,
Vol. 62, No. 4, September 1972, pp. 540-552.
