You are on page 1of 25

Science& SocietyVol. 62, No.

3, Fall 1998, 375-399

Dialecticsand SystemsTheory

ABSTRACT: SystemsTheoryis bestunderstoodin itsdual na-
tureas an episodein thegenericdevelopment ofhumanunder-
standingoftheworld,and as thespecificproductofitssocial
history. On theone handitis a "moment" in theinvestigation
of complexsystems, the place betweenthe formulation of a
problemandtheinterpretation ofitssolutionwheremathemati-
cal modelingcan maketheobscureobvious.On theotherhand
it is theattemptof a reductionist scientific to come
to termswithcomplexity, and
non-linearity change through
sophisticated mathematical and computational techniques,a
gropingtowarda moredialecticalunderstanding thatis held
backbothbyitsphilosophical and


(MaynardSmith,1986),and in personalconversations,
JohnMaynard Smitharguedthatthedevelopment ofa rigorous,
quantitative mathematical
systemstheory makes obsolete.
Engels'awkward "interchangeofcause and can
effect" be replaced
by "feedback,"themysterious"transformationofquantityintoqual-
ity"isnow thefamiliar
phase or
transition threshold while
effect, "even
inmymostconvinced Marxistphase,I couldnevermakemuchsense
ofthenegationofthenegationortheinterpénétration ofopposites."
He couldhaveadded thathierarchy theorygraspssome ofthein-
sights "integrated or
levels" "overdetermination."
On theotherhand,MaryBoger,a leaderoftheNewYorkMarx-
subsumedundersystems Despitesystems
theory. theory's concern

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

withcomplexity, interconnectionand processshe has argued thatit

is stillfundamentally reductionistand static,and despite the power
ofitsmathematicalapparatusitdoes not deal at all withtherichness
of dialectical contingency,contradictionor historicity. Finally,she
added thatsystems-theoretic "interconnection"does not grasp the
subtletiesof dialectical"mediation."
This essayis a firstattemptto systematize myown viewsas they
have evolvedin discussionswithMaryBoger,RosarioMorales,Rich-
ard Lewontinand othercomrades.
As I entered thisexplorationI became aware of twoopposing
temptations.On the one hand I wantedto emphasize the distinct-
nessofdialecticsfromcontemporary systemstheory,to proclaimthat
our theoreticalfoundationsare not obsolete and continue to have
somethingimportanttosayto theworldofsciencethatsystems theory
has not already adopted. On the other hand, along withEngels I
found it gratifying to see science, grudginglyand haltinglyand in-
consistently but nevertheless inexorably,becomingmoredialectical.
Both affirmations are true,but theiremotionalappeal can also lead
to errorsof one-sidedness.I attemptedto use thisawarenessto ques-
tion myconclusionsas I made one or anotherclaim.
Anydescriptionof systemstheoryand of dialectical material-
ismis subjectto twokindsofproblems:in bothareas thereare many
practitionerswithquite divergentviews.I willnot attemptanykind
ofcomprehensivesurveyofsystemstheoryor "a systemsapproach,"
but limitmyselfto systemstheoryin the narrowsense as a math-
ematicalapproach to "systems" ofmanyparts.And second, systems
theory and dialecticsare not mutuallyexclusive.Some systemstheo-
ristsare also Marxistsor have been influencedbyMarxismin their
research contributionsto the development of the theory.Other
Marxistshave had at least a passingcontactwithsystemstheoryand
have used some ofitsnotionsin theirMarxistresearch.For example,
Goran Therborn,a Swedish Marxistsocial scientistinfluencedby
systemstheory,approached the nature of the state fromtwo per-
spectives:the traditionalMarxistviewof the role of the stateas an
expression of class rule, and the systemstheoreticexaminationof
its dynamicsas a systemwithinputsand outputs.The publisher's
blurbforhis book WhatDoestheRulingClassDo WhenitRules? (Ther-
born, 1978) summarizesthework:"Therbornuses the formalcate-
gories of systemsanalysis- inputmechanisms,processes of trans-

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

formation, outputflows- to advancea substantive Marxistanaly-

sisof state powerand stateapparatuses..."
Nonetheless, thetwoare quitedifferent in theirorigins,objec-
tivesand theoretical underpinnings. In what followsI willdiscuss
severalgeneralthemesthatuniteand differentiate them:wholeness
and interconnection, selectionofvariablesorparts,purposefulness,
and theoutcomesofprocesses.Materialist dialectics1is notoffered
as a completephilosophy ofnature,a System in theclassicalsense.
Dialecticians aretooawareofthehistorical contingency ofourthink-
ing expect thatthere willever be a final world view. Ratherit is
firstofall polemical,a critiqueoftheprevailing failings ofboththe
mechanistic reductionistapproach and itsopposite, the holisticide-
alistfocus.TogetherthesehavedominatedEuro-North American
naturaland socialsciencesinceitsemergencein 17thcentury Brit-
ain as a partnerin thebourgeoisrevolution. They have also domi-
natedpoliticsas thebroadliberal-conservative consensusthathas
definedthe"mainstream" politicsofdemocratic capitalism.
Thereforedialecticalmaterialism has focusedmostly on some
selectedaspectsofreality whileignoringothers.At timeswe have
emphasized thematerialityoflifeagainstvitalism,aswhenEngelssaid
thatlifewasthemodeofmotionof"albuminous bodies"(i.e.,pro-
teins;nowwe mightsaymacromolecules). Thisseemsto be in con-
tradiction withourrejectionofmolecularreductionism, butsimply
reflects differentmomentsin an ongoingdebatewherethemain
adversaries werefirstthevitalist emphasison thediscontinuity be-
tweentheinorganic and thelivingrealms,and thenthereductionist
erasureoftherealleapsoflevels.Attimeswe havesupportedDar-
winin emphasizing thecontinuity ofhumanevolution withtherest
ofanimallife,atothertimestheuniquenessofsocially drivenhuman
evolution.We could classify our speciesas omnivores, alongwith
togetitsenergy and substance byeatingotherlivingthings, and are
notlimitedtoonlyone kindoffood.Or wecouldunderlineourspe-
cial statusas "productivores"whodo notmerelyfindourfoodand
1 The term"dialecticalmaterialism"is oftenassociatedwiththe particularrigidexposition
of it by Stalin and its dogmaticapplicationsin Soviet apologetics,while "dialectical"by
itselfis a respectable academic term.At a time when the retreatfrommaterialismhas
reached epidemic proportionsit is worthwhileto insiston the unityof materialismand
dialectics,and to recapturethe fullvibrancyof thisapproach to understandingand act-
ing on theworld.Here I use materialistdialecticsand dialecticalmaterialismas synonyms.

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

our habitatbutproducethem.Bothare true;therelationofconti-

nuityand discontinuity in processis an aspectofdialecticsthatsys-
temstheory does notdeal withat all.
Butcritiqueis notjustcriticism, and dialecticsgoesbeyondthe
rejection of reductionist or idealistthinking to offera coherental-
ternative,moreforthewayin whichitposesquestionsthanforthe
specificanswers itsadvocateshaveproposedat anyparticular time.
Itsfocusisonwholeness andinterpénétration, thestructureofprocess
morethanofthings, integrated and contradiction.
Allofthisis appliedto theobjectsofthestudy, to thedevelopment
ofthought aboutthoseobjects,andself-reflexively tothedialecticians
ourselves so as nottolosesightofthecontingency of
and historicity
ourowngrappling withtheproblemswe study.
Dialecticalmaterialism isuniqueamongthecritiques ofscience
in thatitsrootsare outsidetheacademyin politicalstruggle as well
as within,thatitdirectscriticism bothat reductionism andidealism,
thatitisconsciously andthatitrejectsthegoalofa final
"system."Butitisunlikepostmodernist ofsciencewhichuses
thecontingency ofscientific claimsto denythehistorically bounded
but no lessrealvalidity ofsome claimsoverothers,in favorofan
Systems theory hasa dualorigin, inengineering andinthephilo-
sophical criticism ofreductionism. On the one hand itcomesoutof
as the of
engineering cybernetics, study self-regulating mechanisms
withoftenrathercomplexcircuitry. Norbert Weinerintroduced the
termcybernetics in hisbook ofthatname ( Cybernetics, orControlin
theAnimalandMachine, 1961). The termbecamepartofcommon
usagein theSovietUnion,butwasmostly replacedin theUSA by
controltheory, thetheory ofservomechanisms, orsystems theory.In
thisformitis themathematics offeedback, thestudyofmathemati-
cal models.The prefaceto TheTheory (James,etai,
1947),one oftheearlyclassicaltextsin thisfield,states:

The workon servomechanismsin the [Livermore]Radiation Laboratory

grewout of itsneed forautomaticradarsystems.It was thereforenecessary
to develop the theoryof servomechanismsin a new direction,and to con-
sider the servomechanismas a device intended to deal withan input of
of knownstatis-
knownstatisticalcharacterin the presence of interference

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

A servomechanisminvolvesthe controlof powerbysome means or other

involvinga comparisonof theoutputof thecontrolledpowerand theactu-
atingdevice. The comparisonis sometimesreferredto as feedback.(2.)

Thisformofsystems theory is highly mathematical and formal.

Itsearlierversionsassumedsystems thatweregiven,theequations
known, andmeasurement precise. soonsystems
But analysis wastaken
up bymilitary designers,with the idea ofa weaponssystem replacing
thedevelopment ofparticular weaponsas thetheoretical problem,
andbymanagement systems as the scientific of
aspects directing large
enterprises. Here themeasurements are fuzzier,theequationsnot
known,and therefore othertechniquesbecomenecessary. Herbert
Simonat CarnegieMellonUniversity, Mesarovicat Case Western
Reserve,theInternational Institute forAppliedSystems Analysis in
Austria as wellas mathematicians and engineers in theSovietUnion
andothercenters workedtoadvancetheconceptual frameworks and
mathematics ofmanyvariablesinteracting at once and thecomput-
ing routines forfollowing what happens. More recently, theSanta
Fe Institutehasmadethestudyofcomplexity itselfthecoreintellec-
The majorrole of engineeringand managementsystems in
developingsystems theory is reflected in the assumption of goal-
seeking.Thus Meadows,etai (1992) definea system as "an inter-
connectedsetofelements thatiscoherently organized around some
purpose.A system is morethanthesumofitsparts.It can exhibit
dynamic, adaptive, goal-seeking, self-preserving and evolutionary
Butthe"system" ofsystems theory isnotreality itselfbuta model
ofreality, an intellectualconstruct thatgraspssomeaspectsof the
reality want to study butalso differs fromthatreality inbeingmore
manageable and easierto study and alter.Therefore models arenot
"true"or"false." are
They designed to meet a number of criteriathat
are in partcontradictory, suchas realism,generality and precision
(Levins,1966).It is thehope ofsystems analysts that the departures
fromreality thatmakethemeasiertostudydo notlead tofalsecon-
The wholeness, interconnectedness ofpartsand thepurposeful-
nessofsystems are emphasized. The first twoqualitiesare inherent
in whatwe meanbya system.

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions


The othersourceof"systems" theoryhas been in criticalattempts

to counterthe prevailingreductionismin science since the lastcen-
tury.Here itsboundariesare not well definedbut shade offgradu-
allyinto variousholisms.
Holism is not new.The historyofscience is not the historyofits
mainstream,the successionof dominantparadigmspopularizedby
Thomas Kuhn. There has alwaysbeen dissidencein science,dissatis-
factionwiththe dominantideas, alternativeapproaches withinthe
variousdisciplines,and quite divergent"mainstreams" among disci-
plines. "Holistic"criticismhas alwayscoexisted with the dominant
reductionism.It was expressedin such currentsas vitalismin devel-
opmentalbiology,Bergson's"emergence,"in psychology(Bronfen-
brenner,Perl, Piaget), ecology (Vernadsky'sbiosphere,the Soviet
"geo-biocoenosis,"Clements'and laterOdum's ecosystems) , anthro-
pology(Kroeber's"superorganic")and other fieldsas a graspingfor
wholenessand interconnection.In thisaspect it is usuallyreferred
to in theUnitedStatesas a "systemsapproach"or "systems thinking."
Some authorsengage in systemstheoryin both the narrowand the
broad meanings.Especiallyambitiousand centralwas L. von Berta-
lanffy'sGeneralSystems Theorystartingin the 1930s (vonBertalanffy,
1950). Biologicalcomplexity wasusuallya centralchallenge.W. Ross
Ashby'sDesignfora Brainposes the problem as one of reconciling
mechanisticstructureand seeminglypurposefulbehavior:

We take as basic the assumptionsthatthe organismis mechanisticin na-

ture,thatit is composed of parts,thatthe behaviorof thewhole is the out-
come of the compounded actionsof theparts,thatorganismschange their
behaviorbylearning,and thattheychange it so thatthe laterbehavioris
betteradapted to theirenvironmentthantheearlier.Our problemis,first,
to identifythenatureofthechangewhichshowsas learning,and secondly tofind
why such adaptationforthewholeorganism. (Em-
changesshouldtendtocause better
phasis in original.)

Ecologyalso has broughtto public consciousnessthe richinter-

connectednessof theworld.Examplesare regularlyput forthof the
unexpected,oftencounterproductive effectsofinterventionsdirected
at solvinga particularproblem.Pesticidesincrease pest problems,
draininga wetlandcan increasepollution,antibioticsprovokeanti-

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

bioticresistance, clearingforeststoincreasefoodproductionmaylead
to hunger.And BarryCommoner'sdictathateverything is connected
to everything else and thateverything goes somewherehave become
part of the common sense of at least a partof the public.
The powerfulimpactoftherealizationthatthingsare connected
sometimesleads to claimsthat"youcannotseparate"bodyfrommind,
economicsfromculture,the physicalfromthe biologicalor thebio-
logical fromthe social. Much verycreativeresearchhas gone into
showingthe connectednessof phenomena thatare usuallytreated
as separate.It is even said thatbecause of theirinterconnectedness
theyare all "One," an importantelementof mysticalsensibility that
assertsour "Oneness"withthe Universe.
Of courseyoucawseparatetheintellectualconstructs "body"from
do it all the time,as soon as we label them.We have to in order to
recognizeand investigate them.Thatanalyticalstepis a necessarymo-
mentin understandingtheworld.But it is not sufficient. Aftersepa-
rating, we have tojoin them again, show theirinterpénétration, their
mutualdetermination,theirentwinedevolutionand yetalso their
distinctness.They are not "One." The pairs of mutualistspecies or
predatorand preyare certainlylinkedin theirpopulationdynamics.
Sometimesthe linkageis loose, as when each affectsthe lifeof the
otherbut theeffectis notnecessary.Sometimesverytightly, as in the
symbiosis algae and fungi in lichens. Snowy owls and Arctic hares
drive each other's population cycles in a defining feedback loop.
Mutualistsmayevolve to become "one," as Lynn Margulishas pio-
neered in arguingfortheoriginsofcellularstructures. But predator
and preyare not"One" untilthelaststagesofdigestion.Psychothera-
pistsworkboth withassertingconnection in examiningfamilysys-
temsand withcriticizing"codependence," the pathologicalloss of
boundaries and autonomy.There is a one-sidednessin the holism
thatstressesthe connectednessof the worldbut ignoresthe relative
autonomyof parts.
As againsttheatomisticand absolutizedseparationsofreduction-
ism,holistscounterposethe unityof the world.That is, theyalign
themselves at the"oneness"end ofa spectrumfromisolatedto "One."
Theylook forsome organizingprinciplebehind thewholeness,some
"harmony" or "balance"or purposewhichgivesthewholestheirunity
and persistence.In technologicalsystems,there is a goal designed

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

by the engineersthatis the criterionforevaluatingthe behaviorof

the systemand formodifying the design.To the extentthatthe de-
velopmentofsystems theoryhas been dominatedbydesignedsystems,
goal-seeking behavior appears as an obvious propertyof systemsas
such, and therefore it is soughtalso in the studyof naturalsystems.
In the studyof society,thismaylead to a functionalism which
assumesa commoninterestdrivingthe society.But a societyis not a
servomechanism; itscomponentclassespursuedifferent, bothshared
and conflictinggoals. Thereforeit is not a "goal-oriented"system,
even when manyof itscomponentsare separatelygoal-seeking.
Withintheframework to accommo-
ofstaticholismitis difficult
date change as otherthan destructive, so thatconservationbiology
oftenemphasizespreservationof a particularspecies or ecological
formation,ratherthan conditionsthatpermitcontinuedevolution.
Dialecticiansvalue the holisticcritiqueof reductionism.But we
rejectthesharpdichotomyofseparation/connectionor autonomy/
wholenessand an absolutesubordinationofone to theother.This is
not a complaintabout being "extreme.""Extreme"is a favoritere-
proach byliberals,forwhomthe desiredconditionis moderation,a
middleground"somewherein between,"mainstream,compromise.
Their favoritecolors are "not black or whitebut shades of gray."In
contrastthe dialecticalcriticismis "onesidedness,"the seizingupon
one side of a dichotomouspair or a contradictionas ifit were the
whole thing. Our spectrum all thegrays
is nota gradientfromblackthrough
to white,but a fractalrainbow.
Of course,despite Hegel's dictumthat"the truthis the whole"
we cannot studyThe Whole. The practicalvalue of Hegel's affirma-
tion is twofold:
First,thatproblemsare largerthanwe have imaginedso thatwe
shouldextendtheboundariesofa questionbeyonditsoriginallimits.
Evensystems theoryconstruesproblemstoo small,eitherbecause the
domain is assignedto theanalystas a given"system"or because addi-
tionalvariablesknownto interactwiththeinitialsystemare notmea-
sureable or do not have knownequations,or because of traditional
boundaries of disciplines.Thus a systemsanalysisof the regulation
ofblood sugarmayinclude the interactionsamong sugaritself,insu-
lin,adrenalin,cortisoland othermoleculesbutis unlikelyto include
anxiety,or the conditionsthatproduce the anxietysuch as the in-
tensity oflabor and the rateofusingup ofsugarreserves,whetheror

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

not thejob allowsa tiredworkerto restor take a snack. Models of

heartdisease are likelyto include cholesteroland the fatsthatare
turnedinto cholesterolbut not the social classes of the people in
whom the cholesterolis formedand breaksdown. Systemsanalysis
would not knowhow to deal withthe pancreas under capitalismor
theadrenalsin a racistworkplace.Models ofepidemicsmayinclude
ratesof reproductionof virusesand theirtransmission but not the
social creationof a sense of agency thatmayallow people to take
chargeof theirexposure and treatment.
The second applicationoftheunderstandingthatthetruthis the
wholeis thatafterwe have defineda systemin thebroadesttermswe
can at the time,thereis alwayssomethingmore out therethatmight
intrudeto change our conclusions.
Dialecticsappreciatesthepre-reductionist kindofholism,butnot
itsstaticquality,itshierarchicalstructurewitha place foreverything
and everything in itsplace, nor the a prioriimpositionof a purpose-
fulnessthatmayor maynot be there.Thus it "negates"materialist
reductionism'snegation of the earlier holism,an example of the
negationof the negationthatMaynardSmithfoundso opaque but
could have recognizedas the non-linearity of change.


Wholesare thoughtofas made out ofparts.Systemstheorylikes

to take as its elementsunitaryvariablesthatare the "atoms"of the
system, unchangingas theyebb and flow.
priorto it,and qualitatively
Their relationsare then "interactions"as a resultof whichthe vari-
ablesincreaseor decrease,emit"outputs"and thusproducetheprop-
ertiesofthewholes.But thewholesare not allowedto transform the
parts,exceptquantitatively. The long distanceconversationdoes not
transform the telephone,the marketdoes not change the buyeror
seller,and powerdoes notaffectthepowerfulnor love thelover.It is
the priority of the elementsand along withit the separationof the
structureof a systemfromitsbehavior- rationalassumptionsfor
- theorystill
designedand manufacturedsystems thatkeepssystems
vulnerableto the reproach being large-scale reductionism.
The partsof dialecticalwholesare not chosen to be as indepen-
dent as possible of thewholesbut ratheras pointswhereproperties
ofthewhole are concentrated.Their relationis not mere "intercon-

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

nection"or "interaction" buta deeperinterpénétration thattrans-

formsthemso thatthe"same"variablemayhavea verydifferent sig-
nificanceindifferent contexts andthebehavior ofthesystem canalter
Forinstance temperature isimportant inthelivesofmost
species. temperature many has different meanings. Itactson the
rateofdevelopment oforganisms andtherefore theirgeneration time
and also on thesizeofindividuals; itlimitsthesuitablelocationsfor
nestingor reproduction; itmaydetermine theboundariesofforag-
ing or the time available for searching food.It influences
for the
availablearrayofpotentialfoodspeciesand thesynchrony between
theappearanceofparasites and theirhosts.Itmodifies theoutcomes
Buttemperature isnotsimply giventotheorganisms. The orga-
nismschangethetemperature aroundthem:thereisa layerofwarmer
airatthesurfaces ofmammals; theshadeoftreesmakesforests cooler
thanthesurrounding grassland; the construction of tunnels in the
soil regulatesthetemperatures at whichgroundnestingantsraise
theirbrood;thecolorofleaflitter and humusdetermines thereflec-
tionand absorption ofsolarradiation. the
Through physiology and
demography oí" theorganism, effective its
temperature, range and its
predictability quite different from the weather box temperature
ofa place.On anothertimescale,temperature actsthrough various
pathways pressures of natural selection, the
changing species, which
againchanges itseffective temperature. "temperature"a bio-
Thus as
logicalvariable within an ecosystem isquitedifferent fromthemore
easilymeasuredphysical temperature thatcanbe seenintheweather
box as priorto theorganisms.
Although systems theory is comfortable withtheidea thata cer-
tainequationisvalidonlywithin somelimits, itdoesnotdealexplicitly
withtheinterpénétrations ofvariables initsmodels,theirtransforma-
tionsofeach other.In a sense,Marx'sCapital wasthefirst attempt to
treata wholesystem ratherthanmerely tocriticize thefailings ofre-
ductionism. Hisinitialobjectsofinvestigation inVolumeI, commodi-
ties,are notautonomous buildingblocksor atomsofeconomiclife
thataretheninserted intocapitalism, butrather arestudiedas "cells"
ofcapitalism chosenforstudyprecisely becausetheyrevealthework-
ingsofthewhole.Theycan be separatedoutforinspection onlyas
aspectsofthewholethatcalledthemforth. To Marx,thiswasan ad-
vantage because the whole is reflected in theworkings ofall theparts.

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

Butforlarge-scale reductioniststherelationship goesfromgiven, fixed

partstothe wholes thatare their The and
product. priority autonomy
ofthepartis essentialto systems analysis."Autonomy" does notof
coursemeantheyhaveno influence on eachother.The "variables" of
a system may increaseand decrease but remain what they are.
Partsofa system maythemselves be systems withtheirownstruc-
tureand dynamics. Thisapproachis takenbyhierarchy theoryin
whichnestedsystems eachcontribute as partstohigherlevelsystems
(O'Neill,etai9 1986).Thisallowsus to separatedomainsforanaly-
sis.However,thereverseprocess,thedefining and transforming of
thesubsystems the
by higher level,is rarelyexamined.
Muchstatistical forinstancein epidemiology,
analysis, separates
theindependent variableswhicharedetermined outsidethesystem
fromthedependentvariableswhichare determined bythem.The
independent variablesmight be rainfall or family income; thedepen-
dentvariablemightbe theprevalence ofmalariaor thesuiciderate.
In contrast,
systems approaches recognize thefeedbacks thatgivemu-
tualdetermination: predators eat theirprey,preyfeedtheirpreda-
tors;pricesincreaseproduction, production leads to surplusesthat
lowerprices;snowcoolstheearthbyreflecting awaymoresunlight,
and thena coolerearthhasmoresnow.In feedbackloops,changes
ineachvariableareina sensethecausesofthechangesintheothers.
Whatthenhappenstocausation? Whatmakesone "cause"morefun-
Wecanattempt toanswerthisquestionin twoways.First, wemay
askwherea particular pattern ofchangewasinitiated at a particular
time.Forinstance wemightaskofa predator/prey system, whydoes
theabundanceofbothpredator andpreyvary overa fivehundred mile
gradient?We can analyze thefeedback relationship toshow that ifthe
environmental differencesalongthegradient enterthesystem byway
oftheprey,saythrough temperature increases its
increasing growth
rate,thiswillincreasethepredator population so thatthetwovariables
are positively correlated.Butiftheenvironmental differences enter
bywayofthepredator, perhapsbecausethepredator is itselfhunted
moreinsomeplacesthanothers, thenincreases inhunting reducethe
predator andtherefore increasetheprey.Thisgivesus a negative cor-
relationbetweenthem.Therefore ifweobservea positive correlation
wecansaythatthevariation isdriven fromthepreyendandifa nega-
tivecorrelation thenthevariation isdriven fromthepredator end.The

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

preymediatestheactionoftheenvironment and is the "cause"ofthe

observedpatternin the one system,the predatorin the other.Simi-
larlyin a studyofthecapitalistworldeconomyI examinedproduction
and pricesduringthe1960sand 70s and foundthatthemajoragricul-
turalcommoditiesexhibiteda positivecorrelation betweenproduction
or yieldper acre and priceson the worldmarket.This supportsthe
viewthatpricefluctuations in thelargereconomyand affect
production decisions rather than appear as responsesto fluctuations
in production,and thisdespiteobviousand dramaticchangesofpro-
ductiondue to theweatheror pests.
Whetherthisis generallytrueor not is an empiricalquestion.In
a complex networkof variablesthe drivingforcesforchange may
originateanywhere.When we attemptto ask "does economics or
geopoliticsdetermineforeignpolicy?"or "isthecontentofTV driven
bysales or ideology?"the question is unanswerablein general.The
complex networkof mutualdeterminationsrequiresa complexan-
swerthatis hintedat in theawkwardterm"overdetermination" which
recognizescausal processesas operatingsimultaneously on different
levelsand throughdifferent pathways.Or itbringsus back to Hegel:
the truthis the whole.
Then whereis the locus of historicalmaterialism?Doesn't it re-
quire thatthe economydeterminesociety?
No! "The economy"as a set offactorsin social lifehas no inher-
ent priority overanyofthe othermyriadinterpenetrating processes.
Sometimesit is determinantof particularevents,sometimesnot.As
long as we remainwithinthe domain of a systemsnetworktracing
pathways,everything influenceseverything else bysome pathwayor
other. Changes in the productivetechnologychange economic or-
ganizationand classrelationsand beliefsabout theworld,butchanges
in thetechnology arisethroughtheimplementation ofideas,and exist
in thoughtbeforetheyare made flesh.Or as thefoundingdocument
of UNESCO stated,"Since warsare made in the mindsof men . . ."
Then is social lifea productofintellect?Or is intellectan expression
of class and gender?Approached in thisway,all is mediations,and
the assignmentof absolute priorityis dogmatism.
But thisis quite different fromidentifying themodeofproduction
and reproduction,which is presentnot as a "factor"in thenetworkbut
as the networkitself.It is the structureof thatnetwork,thatmode,
thatdefinesworkersand capitalistsas theactorsor "variables"in the

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

network,makesitpossibleforsexismtohavecommercial value,makes
a politicalactivity,
legislation or allowsmajoreventsto be initiated
by caprices of monarchs. It is thecontextwithin
ousmediations play themselves out and transform
thana factoramongfactors.

Goal Seeking

The thirdqualityof systems, purposefulness, also betraysthe

origin systems theory. The outcomes are evaluated fortheircorre-
spondence to the built-inpurpose, while deviationsfromthatpur-
pose are seen as non-adaptive, contradictory and self-destructive
haviors.Theseappearas system failures.The engineercan discard
ora managercan reorganize thestructures thatlead tothem.Butin
realityonly some are
systems purposeful even whentheyare con-
structed tosatisfysomepurpose.In others, whilethe"elements" are
actorseachwiththeirownpurposesand maybe said to seekgoals,
thesystem as a wholedoes not.
Dialectical"wholes" arenotdefinedbysomeorganizing principle
suchas harmony orbalanceormaximization ofefficiency.
In myview,
a systemischaracterized byitsstructured setofcontradictory processes
thatgivesmeaning toitselements,maintains thetemporary coherence
ofthewholeand also eventually transforms itintosomething else,
dissolvesitintoanothersystem, or leadstoitsdisintegration.


Once mathematical systemstheorydefinesa setofvariables

interrelations whatis
fromsuchand such
Fromthenon, all dependson themathematical
agility the orthecomputer
analyst program tocomeupwith"solu-
tions"oftheequations.A solutionis thepathofthevariables. The
desiredresultis prediction, betweenthetheo-
reticaland observedvaluesofthevariables.
Thereare onlya fewpossibleoutcomesofequations:
a) The variablesmayincreaseor decreaseoutofbounds.This
maymeana realexplosion, thesystem.
disrupting Butitcanalsomean
thatpasta certainpointtheequationsare notvalid.

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

b) The variablesmayreach a stableequilibrium.It thenremains

thereunless perturbed,and returnstowardequilibriumaftera per-
turbation.Ifthe processesinclude randomness,thena solutionmay
be a stableprobabilitydistribution.
c) There maybe more thanone equilibrium,in whichcase not
all oftheequilibriaare stable.Each stableequilibriumis theend result
forthe variablesthatstartout "near" thatequilibrium,withinsome
range called itsbasin of attraction.The basins of attractionaround
the equilibriaare separatedbyboundarieswherethereare unstable
equilibria.The outcomethendepends on thestartingplace, and the
variablesmove towardthe equilibriumin whose basin of attraction
d) The variablesmayshowor approach cyclicbehavior,in which
case how quicklythe variablescycleand the magnitudeof the fluc-
tuationsdescribethe solution.A cyclicalpatternalso has itsbasin of
attraction,the range of initialconditionsfromwhichthe variables
approach thatcycle.
e) The trajectories mayremainbounded butinsteadofapproach-
ing an equilibrium or a regularperiodicityshow seeminglyerratic
pathways, sometimes lookingperiodicfora whileand thenabruptly
movingaway, and différent initialconditionsno matterhow similar
maygive quite different trajectories.This is referredto as chaos al-
thoughin factit has itsown regularities.
The behaviorof a systemwill depend on the equations them-
selves,the parameters,and the initialconditions.Much of the con-
tentofsystemstheoryis the descriptionof the relationsbetweenthe
assumptionsofthemodel and theoutcomesforthevariables,or iden-
tifying the proceduresforvalidatingthe models.
The outcomesare expressedas quantitativechangesin thevari-
ables. This is an extremelyusefulactivity formakingpredictionsor
decidingupon interventions in the systemor systemdesign.But it is
also limiting,and imposes constraintson the models. Most models
requirespecifying the equations and estimatingthe parametersand
variables.Thereforethosethatare notreadilymeasureableare likely
to be omitted.For instance,we can writecompartmentmodels for
epidemics thattake as variablesthe numbersof individualsin each
compartment,thosewho are susceptible,infectedbut not infective
yet,infective,or recoveredand immune.We make some plausible
assumptionsabout thedisease (ratesofcontagion,durationoflatent

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

and infectiveperiods,rateof loss of immunity)and turnthe crank,

watchingas numbersshiftfromone compartmentto another.Then
we can ask questionssuch as, willthe disease persist,how long willit
taketo pass thepeak,howmanypeople willdie beforeitis over,what
wouldbe theeffectofimmunizingx% ofthechildren?We could add
complicationsofdifferences due to age and evensubdividethepopu-
lation into classeswithdifferent parameters.
Contagionalso depends on people's behavior,the levelof panic
in thepopulation.Thischangesin thecourseoftheepidemicas people
observeacquaintancesgettingsickand dying,and maytakeprotective
action.Buthowmuchexperienceis needed to changebehavior?How
muchpanic beforetheywilllose theirjobs ratherthanfaceinfection?
Whatdegreesof freedomdo people have?How long willan altered
behaviorlast?Do people reallybelievethatwhattheydo willaffect what
happens to them? Will they remember fornext time? Since we have
neithertheequationsfordescribingtheseaspectsnor measurements
ofpanic or historicalhorizonor economicvulnerability, such consid-
erationswillnot usuallyappear in the models but at bestonlyin the
footnotes.In recentyears,modelinghas become a recognizedmajor
researchactivity. But thishas had the effectof reducingmodelingto
the quantitative modelsdescribedabove.
Most systemsmodelers take it forgrantedthatquantitativein-
formation("hard"data) is preferableto qualitative("soft")informa-
tionand preferpredictionor fitting ofdata to understanding.In their
viewofscience,progressgoes simplyfromthevague,intuitive, quali-
tativeto the precise,rigorousand quantitative.The highestachieve-
mentis the algorithm,the rule of procedurewhichcan be applied
automatically byanyoneto a whole class ofsituations,untouchedby
humanminds.That is therationalebehind MaynardSmith'ssugges-
tionthatsystems theoryreplacesdialectics.Marxistsarguefora more
complex and non-hierarchical relation between quantitativeand
qualitativeapproaches to the world.
A much smaller effortgoes into qualitativesystemsmodeling
whichwould allow us to deal withthese "soft"questions.Instead of
the goal of describinga systemfullyin order to predictits future
completelyor to "optimize"itsbehavior,we ask how much we can
get awaywithnot knowingand stillunderstandthe system?
Whereastheengineeringsystems presumerathercompletecon-
trolover the parametersso thatwe can talk about optimizingthe

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

parameters, thesystems we are mostconcernedwithin natureand

in societyare notunderourcontrol.We tryto understand themin
ordertoidentify thedirections inwhichtopushbutdo nottrust our
modelstobe morethanusefulinsights intothestructure orprocess.
Dialecticianstakeas theobjectsofourinterest theprocessesin
complexsystems. Our primary concernis understanding themin
ordertoknowwhattodo. We asktwofundamental questionsabout
thesystems: are the
why things waythey are instead of a littlebitdif-
and are
ferent, why things waythey the are instead of verydifferent,
and fromthesethepracticalquestionsofhowto intervene in these
complexprocessestomakethings betterforus.Thatis,weseekprac-
ticaland theoretical understanding ratherthana goodfit.Precision
and prediction or
may may not be useful in thisprocess,buttheyare
The Newtonian answertothefirst questionis,things remainthe
waythey are because nothing much is happening to them. Stasisis
thenormalstateofaffairs, andchangemustbe accountedfor.Order
is thedesiredstate,anddisruption istreatedas disaster. A dialectical
viewbeginsfromtheoppositeend:changeis universal and muchis
happening changeeverything. Therefore equilibrium and stasis
are specialsituations thathavetobe explained.All"things" (objects
or patterns ofobjectsor processes)are constantly subjecttooutside
influences thatwouldchangethem.Theyarealsoall heterogeneous
internally, theinternal
and dynamics isa continuing sourceofchange.
Yet"things" do retaintheiridentities longenoughtobe namedand
sometimes persistforverylong times indeed.Someofthem,much
too long.
The dynamic answertothefirst questionishomeostasis, theself-
regulation thatis observed in physiology, ecology,climatology, the
economy andindeedinallsystems thatshowanypersistence. Homeo-
stasistakesplace throughtheactionsofpositiveand negative feed-
backloops.Ifan initialimpactsetsprocessesin motionthatdimin-
ishthatinitialimpact, werefertoitas negative feedback, whileifthe
processesmagnify theoriginalchangethefeedbackispositive. Thus
positiveand negativeappliedto feedbackhavenothingto do with
whether welikethemornot.Whenpositive feedbacks haveundesir-
able resultsthatincrease out ofbounds, we referto them as vicious

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

Itisoftensaidthatnegative feedbackstabilizes andpositive feed-

backdestabilizes a system. But thisis notalways the case. If positive
feedback exceedsthenegative thenthesystem isunstableinthetech-
nicalsensethatitwillmoveawayfromequilibrium. In thatcase,an
increaseofnegative feedbackis stabilizing. Butiftheindirect nega-
tivefeedbacks bywayoflongloopsofcausationare toostrongcom-
paredto theshorter negativefeedbacksthesystem is also unstable
and willoscillate.Then positivefeedbackloops can havea stabiliz-
ingeffect byoffsetting theexcessivelongnegativefeedbacks. Long
loops behave likedelays in the system. The significance of a feedback
loop dependson itscontextin thewhole.The complexsystems of
concerntous usuallyhavebothnegativeand positive feedbacks.
Homeostasis does notimplybenevolence.A negativefeedback
loop shouldnotbe seenas theelementary unitofanalysis or ofde-
A the
sign. simpleequationmaygive appearance of "self regulation"
in thesensethatwhena variablegetstoobigitis reducedandwhen
itgetstoo smallitis increased.Butthereductionand theincrease
mayhavequitedifferent causes.An increasein wagesmaylead to
employers cutting thelabor force,increasing unemployment andthus
making it easier to reduce wages. A decrease in wagesmaylead to
labormilitancy thatrestores some of thecuts. The outcome(ifnoth-
ingelse happens)is a partialrestoration of theoriginalsituation.
Neither is
party seeking homeostasis, and the wage/employment feed-
backis notdesignedor pursuedbyanyoneto maintaineconomic
stability. one possiblemanifestation ofclassstruggle. Thus
homeostasis doesnotimplyfunctionalism, a viewwhichassignspur-
pose to the feedback loop as such.
Thisdistinction isimportant, especiallywhenweexamineappar-
entlyunsuccessful attempts toachievesocially recognized goals.Mead-
ows,Meadowsand Randers(1992) presenttheproblemas follows:
Thisbookisaboutovershoot. Humansocietyhasovershot forthe
samereasonthatotherovershoots occur.Changesare toofast.Signalsare
late,incomplete, ignoredor denied.Momentum
distorted, is great.Re-
sponses are slow. . . (2.)

error-correcting inadequate.Andifyouassumethat

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

social processesare aimed at sustainable,healthful,equitable rela-

tionsamong people and withthe restofnature,thenthe defectis in
thefeedbackloops, the mechanismsforachievingthesegoals. But if
agriculturefailsto eliminatehunger,ifresourceuse is notmodulated
to protectpeople's healthand long-termsurvival, itis notbecause of
the failingsof a mechanismaimed at these goals. Rather,most of
worldagricultureis aimed at producingmarketablecommodities,
resourcesare used to make profits,and the welfareeffectsare side
effects oftheeconomy.It is thecontradictions amongopposingforces
(and betweenthoseoftheecologyand theeconomy)ratherthanthe
failureofa good trybyinadequate information systems and deficient
homeostaticloops thatare responsibleformuch of the presentsuf-
feringand the threatof more.
When a change occursin a component(or variable)ofa system,
thatinitialchange percolatesthrougha networkof interactingvari-
ables. It is amplifiedalong some pathwaysand bufferedalong others.
In the end, some of the variables(not necessarilythe ones thatre-
ceived the initialchange or those nearestthe pointof impact) have
been altered,while othersremain prettymuch the waytheywere.
Thereforewe identify"sinks"in the system,variablesthatabsorb a
large partof the impactof the externalshock,and otheraspectsof
the systemthatremainunchanged, protectedby the sinks.We can
even havesituationswherethingschange in waysthatcontradictour
common sense, where forexample adding nitrogento a pond can
lowerthe nitrogenlevel or an inflatedmilitary budgetundermines
nationalsecurity.(This outcomedepends on thelocationofpositive
feedbackswithina system.)
But "unchanged"requiressome furtherexamination.The "vari-
able" is not a thingbut some aspectofa thing,perhapsthe numbers
of individualsin a population,not "the population."
One simplesystemconsistsof a predatorthatfeeds on a single
prey.All else is treatedas "external."It is sometimesthecase thatthe
predatoris regulatedonlybythe prey.Then a change in conditions
thatacts on the reproductionor developmentrate or mortality of
thepreydirectly, thatis not due to thepredator,willbe passed along
to thepredator.Increased preyleads to increasedpredatorsand this
reduces the preyback towarditsoriginalvalue. The "prey"variable
may remain unchanged while the predator population either in-
creases in response to increasedavailabilityof preyor diminishesif

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

fewerpreyare produced. The predatorvariableactsas a sinkin this

system.Tracingthe ups and downsofpredatorand preyfinishesthe
tasksof the systemsanalysis.
But what I referredto as "prey"is reallyonly the numbersof
prey.If preyreproductionhas increased withmore food but the
populationofpreyhas not changed, itis because the preyare being
produced fasterand consumed faster.That is, the preypopulation
is younger.Individualsmaybe smallerand thereforemore vulner-
able to heat stress.They maybe more mobile,migratingto findun-
occupied sites.If the preyare mosquitoes,a shorterlifespan may
mean thattheydo notspread as muchdisease even ifthereare more
of them.They mayspend more timein cool moistshelterswhere
theymeet additional predatorsand the model has to be changed.
Natural selection in a younger population mightfocus more on
thosequalitiesthataffectthe survivaland earlyreproductionof the
young.Thus the variable,"prey,"thatwas unchanged in the model
can be activelytransformed in manydirectionsnot dealt within the
The particularsof the dynamics,the relationsamong the posi-
tiveand negativefeedbacksin a system, sourcesand sinks,connectiv-
ityamongvariables,delaysalong pathwaysand theireffectsare all in
the domain of systemstheoryin the narrowsense. The partsof the
systembecome thevariablesofmodels,and equationsare proposed
for theirdynamics.Systemstheorystudies these equations. Math-
ematicalrules have been discoveredfordeterminingwhen the sys-
temwillapproach some equilibriumconditionor oscillate"perma-
nently,"thatis, as long as the assumptionsstillhold.
Modern computationalmethodsallow forthe numericalsolu-
tionsof large numbersof simultaneousequations. The parameters
are measured,the initialconditionsofthevariablesare estimatedor
assumed. (The distinctionbetweenparametersand variablesis that
theparametersare assumedto be determinedoutsidetheboundaries
of the "system"and are onlyinputswhilethe variableschange each
otherwithinthe "system.")The computerthencalculatessuccessive
stepsin theprocessand comes up withnumbers,thepredictedstates
ofthevariablesat different times.The numericalresultsare compared
to observations.Ifthecorrespondenceis good enough,itis assumed
thatthe model is valid,thatit "accountsfor"the behaviorof the sys-
tembeing studied,or 90% of the behavior,or whateverlevelwe de-

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

cide is acceptable. If not, more data maybe collected to get better

estimatesof parametersor the equations maybe modified.
However,systems theorystartswiththevariablesas givens.It deals
withthe problemsof selectingvariablesonlyin a verylimitedway.
When we approach anyreal systemof anycomplexity,the question
ofwhatthe rightvariablesare to include in the model is itselfquite
complex.Itis theclassicalMarxistproblemofabstraction(see Oilman,
1993, fora detailed examinationof dialecticalabstraction).Some
practicalsystemsmodelingcriteriaare: reciprocalinteraction,com-
mensuratetimescales, measureability, variablesthatbelong to the
same disciplineand can be representedbyequationsofchange.The
systemshould be large enough to include the major pathwaysof in-
teraction,withidentification ofwhereexternalinfluencesenterthe
network.Systemstheorymakes use of growingcomputingcapacity
to givenumericalsolutionsto thedifferential or difference equations
thatdescribe the dynamics.In order to have precise outcomesit is
necessaryto have good estimatesof the parameters,thingslike the
reproductiverateofa population,theintensity ofprédation,thehalf-
lifeof a molecule, or the cost/priceratioin an economic produc-
tion function.The gatheringof these measurementsis difficult, so
thatestimatesare oftentakenfromthe publishedliteraturerather
thanmade afresh.Parametersthatcannotbe measuredreadilycan-
not be used.
Once variables are selected, theyare then treatedas unitary
"things,"whose onlypropertyis quantity.The mathematicswilltell
us whichquantitiesincrease,whichdecrease,whichfluctuateor re-
main unchanging.The source ofchange is eitherin thedynamicsof
the variablesin interactionor in perturbationfromoutside the sys-
tem. ("Outside the system"means outsidethe model. In a model of
species interactionsa geneticchange withina species is regardedas
an externalevent,since it is externalto the demographicdynamics
althoughitis located insidethecellsofthebodies ofindividualmem-
bersof a population.) But all variablesare themselves"systems" with
internalheterogeneity and structure,withan internaldynamicsthat
is influencedby eventson the systemscale and also changingthe
behaviorofthevariables.Thus dialecticsemphasizestheprovisional
natureof the systemand the transitory natureof the systemsmodel.
The variablesof a systemchange at different rates,so thatsome
are indicatorsoflong-termhistorywhileothersare more responsive

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

tothemostrecentconditions. Thusinnutritional surveys weuse the

heightofchildren theirage as an indicator of long-term nutri-
tionalstatus, thegrowth overa lifetime, whileweight forheightindi-
catesfoodintakeoverrecentmonthsor weeksand therefore mea-
suresacutemalnutrition. Becauseeachvariablereflects itshistory on
itsowntimescale,theyare generally notin "balance"or harmony.
Ideology need not "correspond" to class position,politicalpowerto
economicpower,or foreststo climate.Rather,thelinksbetween
variablesin a system identify processes:ideologyresponding, not
corresponding, to classposition,economicpowerenhancingpoliti-
calpower,political powerbeingusedtoconsolidate economicpower,
colderclimatetreessuchas spruceand hemlockgradually displac-
ingtheoakandbeechofa warmer period.Butall oftheseprocesses
taketime,so thata system doesnotshowa passivecorrelation among
itspartsbuta network of processesconstantly transforming each
other.In Darwinianevolutionary theory both the adaptednessofa
speciestoitssurroundings anditsnon-adaptedness arerequired, the
former showing the outcomes of natural selectionand the iden-
tifyingitas a processthatis nevercompleteand showingthehistory
ofthespecies.Completeadaptedness wouldhavebeenan argument
forspecialcreation, notevolution, proclaiming a harmony thatmani-
feststhebenevolent wisdomoftheCreator.
The secondquestion,whythingsare thewaytheyareinsteadof
verydifferent, is a questionofhistory, evolution, development. It is
concernedwiththelong-term processesthatchangethecharacter
ofsystems. The variablesinvolvedin long-term changemayoverlap
withtheshort-range ones,butare notin generalthesame.Manyof
theshort-term processesarereversible, accordingtocon-
ditionswithout accumulating to contribute to the longrun.
Atanyone momenttheshort-term eventsare strongprocesses,
temporarily overwhelming some of the long-term directional changes
thatare imperceptible in theshortrun.Yet thetwoscalesare not
independent. The reversible short-term oscillationsthrough which
a system confronts changing circumstances have themselves evolved
and continueto evolveas a resultof theirfunctioning in thelong
run.Andtheyleavelong-term residues:thebreathing in and breath-
ingoutofordinary respiration mayalso resultin theaccumulation
oftoxicor abrasivematerials in thelung;therepetitive cyclesofag-
riculturalproduction can exhaust the the
soil; periodicity ofthetides

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

also has itslong-termeffectoflengtheningthedaythroughtidalfric-

tion; the buyingand sellingofcommoditiescan resultin theconcen-
trationofcapital.Long-termchangesalterthecircumstances towhich
theshort-term systemresponds as wellas themeans availableforthat
Here mathematicalsystemstheoryis less useful,since the math-
ematicsis much betterdeveloped forstudyingsteady-state systems
thanevolvingones. (The workofllyaPrigogineon dissipative systems
is onlya partialexception to thislimitation.)


Systemsanalysisis one of the techniquesforpolicymaking.As

itstechnicalside becomes moresophisticateditalso is usuallylessac-
cessibleto thenon-specialist.Thereforeitoftenreinforcesa techno-
craticapproach to public policy,and does thatin theserviceofthose
who can affordto contractitsservices.The rulingclassand itsrepre-
sentativesare referredto in the tradebythe more neutralterm"de-
cisionmakers."This is ofcoursenotunique to appliedsystems theory,
but is a common correlateof itsincreasinguse withina managerial
framework. A special efforthas to be made to counteractthisten-
dency,to demystify the studyofcomplexityand to democratizeeven
complex decision making.The SovietauthorAfanasyev, beforehe
embracedthe "freemarket,"wrotean interesting book, TheScientific
Management whichemphasizedthesystems-theoretic
ofSociety, aspects
of planningas a technocraticprocedurewithonlyperfunctory nods
in thedirectionofpopularcontroloftheplanningprocessas a whole.
Systemstheorycan be understoodas a "moment"in the investi-
gation of scientificproblemswithincomplex systemsby means of
mathematicalmodels. Its value depends in large measure on the
contextof itsuse, and here dialecticshas a broaderrole thatcan in-
1. The posingof the problem,the domain to be explored,what
is takenas the "fundamentalelements"and whatas thegivensof the
problem,the boundaries thatare not questioned. To do thiswell
requiresnot onlya substantiveknowledgeof the objectsof interest,
theirdynamicsand history, and an understandingofprocess.There
is also frankpartisanship,since what is taken as givenand what is
assumed to be "fundamental"is a political as much as a technical

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

problem.For instance a model of a societythatconsistsof atomic

individualsmakingdecisionsin thevoidcan notescape thedead end
ofbourgeoisindividualistreductionismno matterhowelegantlythe
mathematics is developed.An economicmodel thatconsistsofprices
and productionand profitsand such can giveprojectionsof trajec-
toriesofpricesand productionand profitsand such (at best;in real-
itytheydo thisverybadly). But itwillneverlead to an understanding
of economicsas social relations.
Sometimesthevariablesare givento thesystems analyst:thespe-
cies in a forest,the networkof productionand prices,the gizmosin
a radio, the molecules in an organism.That is, the "system"is pre-
sented to us as a problem to be solved ratherthan as an objective
entityto be understood.But oftenitis presentedmorevaguely:how
do we understanda rainforestor the healthofa nation?The wayin
whicha problemis framed,theselectionofthesystemand subsystem
is priorto systemstheorybut crucial to dialectics.A dialecticalap-
proach recognizesthatthe "system"is an intellectualconstructde-
signed to elucidate some aspects of realitybut necessarilyignoring
and even distortingothers.We ask whatthe consequenceswould be
of different waysof formulatinga problemand of bounding an ob-
ject of interest.
2. Selectionoftheappropriatemathematicalformalisms(equa-
tions,graphdiagrams,randomor deterministic models,and so on) .
Whiletechnicalcriteriainfluencethesechoices thereare also issues
ofthepurposesof the model, the partiallyconflicting goals ofpreci-
sion,generality, realism,manageabilityand understanding.The im-
portantthinghere is not to be limitedbythe technicaltraditionsof
a fieldbut to examine all thesechoices not onlyforhidden assump-
tionsbut also fortheirimplications.
3. Interpetationofresults.Here qualitativeunderstandingis an
importantsupplementto numericalresults.In the course of an in-
vestigation we maygo fromvague qualitativenotionsthroughquan-
titativeexplorationsto more precisequalitativeunderstanding.This
is onlyone example of non-progressivist, nonlinearthinkingthatis
captured in our "mysterious"negation the negation.
Progressis not fromqualitativeto quantitative.Quantitativede-
scriptionof a systemis not superior to qualitativeunderstanding.
When approachingcomplexity,it is not possibleto measure "every-
thing,"plug it all into a model and retrieveintelligibleresults.For

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

one thing,"everything" is too big.Qualitative understanding is es-

sentialin establishing quantitative models. It intrudes into the inter-
pretation oftheresults. The taskofmathematics istomakethearcane
obviousandeventrivial. Thatis,itmusteducatetheintuition so that
confronted witha dauntingcomplexity wecangraspthecrucialfea-
turesthatdetermine itsdynamics, knowwhereto lookforthefea-
turesthatmakeitwhatitis,suspectmainstream questionsas wellas
A dialecticalunderstanding ofprocessin generallooksat the
opposingforcesactingon thestateofa system. Thisisnowaccepted
moreorlessin ordinary scientific practice. Excitatory and inhibitory
neurons,sympathetic parasympathetic stimulation, opposing
selectionforcesoran opposition betweenselective andrandompro-
cessesare all partof thetoolkitofmodernscience.However,this
hasstillnotbeengeneralized tothinking ofprocessas contradiction.
4. Whendoesthesystem itselfchangeandinvalidate themodel?
Weneeda permanent awareness ofthemodelas a humanintellectual
construct thatis moreor lessusefulwithin certainboundsand then
can becomenonsense.The internalworkings of thevariablesin a
model, dynamics of the model or
itself the development ofthe
scienceeventually revealsall modelsas inaccurate, limited, mis-
leading.Butthisdoesnotdestroy thedistinction between modelsthat
areterribly wrong from the startand those that have relative validity.
5. Structures doubts.Doubtis an essential part of the search for
understanding. There are areas of science that have been consolidated
tothepointofnearcertainty. Othersareborderregions ofourknowl-
edge where there is a plurality insights opinionsand conflict-
of and
ing evidence.Here doubt andcriticism areessential. Andbeyondthat
theunknown, wherewehavedivergent intuitions andwhereourbiases
canroamfreely. Butwherewehavethesamedoubtspersisting forlong
periods thisis nota sign of a postmodern pluralist democracy butof
stagnation. Usefuldoubt isnot the expression of an esthetic of indeci-
sionora responsetothepetulant reproach of"you'reso damnsureof
yourself!"oran acknowledgement thattruth is"relative," buta histori-
cal perspectiveon error,bias,and limitation.
The artofmodelingrequiresthesensitivity to decidewhenin
thedevelopment ofa sciencea previously necessary simplification has
becomea grossoversimplification and a braketo further progress.
Thissensitivity depends on an understanding of science as a social

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions

processand ofeach momentas an episode in itshistory, a dialectical

thatis not taughtin the "objectivist"
sensitivity traditionsof mecha-
Thus systemstheoryis best understood as reflectingthe dual
natureofscience: partofthegenericevolutionofhumanity'sunder-
standingoftheworld,and a productofa specificsocial structurethat
supportsand constrainsscience and directsittowardthe goals ofits
owners.On the one hand it is a "moment"in the investigationof
complex systems,the place betweenthe formulationof a problem
and theinterpretation ofitssolutionwheremathematicalmodeling
can make the obscure obvious.On the otherhand it is the attempt
ofa reductionistscientifictraditionto come to termswithcomplex-
ity,non-linearityand changethroughsophisticated mathematicaland
computationaltechniques, groping towarda more dialecticalunder-
standing that is held back both by philosophicalbiases and the
and economic contextsof itsdevelopment.

677 HuntingtonAvenue
Boston,MA 02115


Ashby,W. Ross. 1960. Designfora Brain.New York:Wiley.

James,H. M., N. Nicholsand R. S. Phillips,eds. 1947. The Theory
New York: McGraw-Hill.
Levins,Richard. 1966. "The Strategyof Model Building in Population Science."
American 54, 421-431.
MaynardSmith,John. 1986. "MoleculesAre not Enough" (reviewof TheDialectical
byR. Levinsand R. Lewontin). LondonReviewofBooks,6 (February).
Meadows,Donella H., Dennis L. Meadows,and JörgenRanders. 1992. WorldOver-
New York: Routledge.
Oilman, Berteli.1993. DialecticalInvestigations.
O'Neill, R. V., D. L. DeAngelis,J. B. Waide and T. E. H. Allen. 1986. A Hierarchical
ConceptofEcosystems. Princeton,NewJersey:PrincetonUniversity Press.
Therborn,Goran. 1978. WhatDoes theRulingClassDo WhenitRules?London: New
von Bertalanfíy, L. 1950. "An Outline of General SystemsTheory."TheBritish Jour-
nal for thePhilosophyofScience,1:2.

This content downloaded from on Fri, 27 Nov 2015 03:20:30 UTC
All use subject to JSTOR Terms and Conditions