Balzer - A Basic Model for Social Institutions

7/27/2019 Balzer - A Basic Model for Social Institutions

1/29


2/29

WOLFOANG BALZER

of one common basic theory-element is that the basic theory-element usually is em -pirically trivial or nearly triv ial while the whole net gets highly non-trivial by men mof the different special laws which hold only under very lim ited conditions and invery lim ited dom ains, This net p icture has proved correct for the most honouredtheories in physics like mechanics, or thermodynam ics.'

A theory of social institutions necessarily has to be rather comprehensive. Thereis only a narrow path between its becom ing unapprehendable because of too manyfeatures being included which are thought relevant, and between becom ing simplebut trivial because too many relevant features are left out. W e believe that we areon that path , and though we may be nearer to the trivial side, our account providesmany opportunities for refinement or specialization which w ill be indicated in someplaces.

In a first step we restrict ourselves here to social institutions of first order, and toa static model. By a first order iiistllufl6n we mean an institu tion all of whose actorsare individuals, In contrast, a higher order institu tion has among its actors at leastone corporate actor. The study of higher order institutions involves a marked in-crease in complexity w ith respect to the present approach and cannot be addressedhere. W e think, however, that our model w ill be useful for subsequent constructionof higher order models. Our model is static in the sense of not making fully explicitthe process of the origin and development of an institution. It is clear that ultim atelya satisfactory concept has to comprise these dynam ical aspects as well. The modelis not entirely static , however, it contains som e important dynam ic ingredients: lo-cal features of the orig in and development of "parts" of an institution (called sociolpractices in Sec. 3). Explicit reference to time can easily be introduced though wedid not for reasons of simplicity, and because mere inclusion of time does not byitself reveal new insights. The item s mentioned, origin and development of a wholeinstitution , did play an important role in the construction of our theory , but cannotfully be worked out here.

A lso there is no space for detailed comparison of our model w ith others. Thereader w ill recognize features sim ilar from the functionalist approach! in the formof hierarchies of power (Sec. 1), and from constructivisrrr' in the notion and roleof superstructures in our model (Sec. 2). The social practices in Sec. 3 are mod-elled along the lines of the picture of evolutionary theory", and our account ofpower in Sec. 4 was inspired by recent work of Wartenberg", What is new is theway of putting these item s together in a precise way, including a precise account ofpow er, in particular. W hat w ill be m issed is explicit reference to the game theoreticIThe bavic thcory..:lemcnts of the two theories mentioned can be shown to be empirically trivial. seeBalzcr, Moulincs and Sneed (\981), Chap. IV. Our methodnlogicnl approach is tt.,t of a recent schoolk.nown sometimes under the label of "structuralism". According to this approoch, 0 simple theory cssen-lially is given by a class of models and a das.~ of so cnlled intended applications or inteoded system~ (l.e.sets of data obtained from real system! by VlIriOU! systematic means). The empirical claim formulatedwith a tocory is that the intended system! fit into suitable models aod in this sense are explained by thetbeor y (see Scc. V). More comprehensive throrlc.s are conceived os f1(/J of simple theories. interrelatedby various link!. See Balzer, MoulirlCs and Sneod (1987) for [urther dCI"il~.lSee Parsons (1951).lFor in'tAnee Berger and Luckmann (1%6).'Se


3/29


4/29

WOLFOANO BALZER

uniquely, i.e. that X be one-one, though in most applications this w ill be the case.StatILS of the groups is treated in a weak, merely comparative form . A binary statusrelation ~ among groups indicates when one group, ,,(, has higher status than an-other, "(' : "(' ~"(. "Status" at this stage is a very vague notion, a "theoretical term "of our theory. W e speak of "status" because of certain sim ilarities to this term 'smeaning in network analysis (which cannot be worked out here, however.) The sta-tus relation is required to be transitive and antiref'lexive (A3 below), and such thatthere exists a group w ith highest "status" w ithin the institution modelled (A4).A core C of a social institution therefore is 8 structure

c'" (r,e,X,~where r is a set of "groups" and e is a set of "action types", X is a " ch ar ac te ris ticfunction" which maps each group into a set of action types (those "characteristic "for members of the group), [nd ~ is- a binary "status relation" among the group~.Moreover, the following axioms (Ire required to hold :A IThe sets of groups and action types are non-empty and disjoint. The set of

groups is finite.A2 Each action type in e belongs to some set X ( " ( ) characteristic for some group

of r.A3 The status relation is transitive and anti-reflexive.A4 There exists one group with highest status.In the following the status relation will get closely linked to the notion of powerso that a group's higher status basically is derived from it's members having moreopportunities to exert power over members of a group with lower status. However,this link to power does not serve as a definition of status: the status relation in thefinal model still has its status as an undefined prim itive.

Axioms A3 and A4 have empirical character. We can imagine possible counrerex-arnples which however, if our models are correct, do not occur in real-life institu-tions. Think of an anarchist society like, say, the Nuer.? Whatever grouping we mayimagine in such a society, we do not find a natural relation on the groups whichsrnisfies the above axioms. This does not show that the axioms are "wrong". Thepoint is that anarchic societies are not among the intended system s for a theory ofinstitutions: they are "uninstitutionalized". It has to be stressed that the claim as-sociated with A3 and A4 is hold up only for institutions of the kind mentioned inth e in tro du ctio n. W e have nothing to say about other kinds of social structures w ithstr atifications that m ight be described with a binary relation among groups. Noteth at ~ needs not be connected. There may be groups the status of which is notcomparable .

Groups, action types and X may be traced to the m icro-level of individuals andtheir actions (action tokens). Such connection provides meaning and partial oper -ational access to the macro-concepts occurring in a core. W e consider individuals.act ion-tokens (i.e. concrete, single actions in their historical uniqueness) and threerelations among individuals and actions: relations of PERFORMANCE, INTEND-ING and exerting POWER. Variables i,j will range over individuals, and a.b,c overs.,." Flap (\9R5) ror ,odo\ogic.t "udy or thi, example.


5/29

A BASIC MODEL FOR SOCIAL INSTITUTIONS

action-tokens in the following. While PERFORM has its standard syntax:i PERFORMS a

we use INTEND with three arguments, due to the special context in which it isapplied when we come to define power relations (in D9, AI7 below).'?iNTENDS that j should do b, or sim ply: Ir-rTENDS(i,j, b).The relation of power we use with four arguments.'!iby doing a exerts POWER over j so that j does b,for which we write: POWER(i,a, j ,b). This format allows for an easy definit ionof ex ercising pow er. We say that iexercises power over j iff there exist a. bsuch that POWER(i,a,j,b). In the following, with respect to a given relat ionPOWER(i,n, j ,b), i will be called the superordiuate agellt an d j the subordinateagent, ~ .

We define a m icro base MB for an institution to consist of individuals and i1ctions(=ac tion tokens), together with the relations mentioned:

M B = (J ,A,PERFORM,INTEND,POWER)where 1 is a set of "individuals", A is a set of "actions" (tokens), PERFORM ,INTEND, and POWER are relations of the above format, and the following axiomsare sat isfied:AS 1an d A are non-empty and disjoined. The set 1of individuals is finite.A6 Any two individuals i,j E 1 are involved in some POWER relation by means

of suitable actions (i.e. there exist a.b E A such that POWER(i ,n,j,b) orPOWER(j,a,i ,b,

There are no axioms about PERFORM and INTEND at this stag e. In Sec . 4, thes etwo relations will be needed in order to characterize the POWER relation, but theaxioms to be formulated there involve other notions in an unsepar able wily. So weabstain from formulating axioms here which would become redundant later OIL /\6requires that all individuals are involved in the POWER relation. POWER thereforecreates a connected network of ties between the individuals, a tie existing betweentwo individuals whenever one of them exerts power over the other (w ith respec t tosuitable actions a, b). In applications the sets 1an d A must not be treated as mer elyobservational. Though there is no doubt that individuals and actions in most case scan be determ ined by observation, not all observed items will be relevant fo r th esystem under inve stigarion, and the actions observed may vary with the observerSome choice and interpretation always will be involved. If we apply the theur y tl la grocer's shop, and observe a mother with child, shopping and wiping the ch ild'snose at the same time we may' well forget about the child and wiping its nose because these do not contribute to modelling the situation as all institution of theType "grocer's shop". Such problems of delimiting "the" correct sets of objects in asystem with respect to a given theory occur in every field (including the natural sc i-


6/29

6 WOLFOANO BALZER

ences), and the ultim ate criterion for a correct choice always is whether the processof application succeeds (compare Sec. 5).

A6 rem inds of network analysis, and is put forward in that spirit. W e believethat the possibilities of network analysisl2 are still far from being exhausted. Onedomain of application of network analysis that has been neglected up to now isthat of networks of theoretical, a bs trac t re lation s, like our relation of POWER . Themere fact that power relations are not observable in the same way as lire numbers oftelephone calls, say, does not indicate that they are not operationally accessible, Theleast we can say is that they are open to direct verbal investigation. M oreover, ineveryday life we have fine senses for determ ining power relations which m ight givefurther hints at operationallzation. F inally, 8 theoretical term of this kind needs notbe fully operationalized, it m ay 8S w ell be determ ined by means of our theory. W enote that our INTEND relation referring to two individuals also yields ties whichmay be considered as creating trnctwork. On our account, however, only those tiesof INTEND are important which have some connection w ith the POWER relation.A full characterization of POWER will be given only in Sec. 4.

Individuals and actions being available, groups of course can be treated as setsof individuals, and action types as sets of actions, respectively. G roup membershipand an action's being of a certain type then reduce to set theoretic m embership.In this way we may base any "macro" core on some suitable m icro base. O f course,such "foundation" requires further connections between the central notions on bothlevels. W e have to state how the characteristic function and the status relation arerelated to the underlying individual notions PERFORM , INTEND and POWER.No definition of the former in term s of the latter is to be expected. Concerning thecharacteristic function we first state the obvious condition that every action typecharacteristic for a group also is PERFORMED by some member of that group.Second, and more importantly, we require that PERFORMANCE takes place inthe frame given by the characteristic function (A B below). Any individual iwillperform only those actions b which are characteristic for one of the groups of whichiis a member. Since the group is characterized by a set of action types x C , ) ={T I, ... ,1 '"} this can be expressed by saying that the action b belongs to one of thetypes 1'1,... , 1 ' ' ' : bE ri, for som e i~1. This axiom is a cluster law binding togetherthree important concepts: groups (and group membership), PERFORMANCE andcharacteristic actions. Note that groups are not necessarily disjoined. If an individualbelongs to different groups then it's performed actions have to be charncteristicfor at least one of those groups. C learly, this requirement becomes stronger w ithcecreasing number of groups to which an individual belongs. The most frequentcase in applications is that of an individual's belonging to just one group. It hasto be emphasized that the groups considered here are only those occurring in oneinstitution.The connection between the status relation and the m icro concepts infor mally

may be stated as follows. A group, has higher status than group " only if allmembers of, can exercise power over members of ,', if a "big part" of the lowergroup is thus affected, and if the converse of this relation does not hold, i.e. not allmembers of,' can exercise power over m embers of" and a "big part" of, is thus"S


7/29

. ..,

A BASIC MODEL FOR SOCIAL INSTIn.rTIONS

left unbothered. This formulation may be emended as follows. F irst, we mily r e -place the modal aspect expressed in "can exercise" by the actual mode. This yieldsan over idealized form , however. Typically, not all members of one group actuallyexercise power over members of the other--even if such events are allowed to bespread over some reasonably long interval. Usually, however, a "big part" of thehigher group actually exercises power. So a more realistic version is obtained byrequiring "big parts" (and complementary "small parts") in all places of the stare-rnent. We do not formalize the notion of "big parts", this m ight be done in differenlways, for instance by referring 10 the actual numbers of members of the groups. andappropriate proportions. A lso, we m ight use different proportions on both sides inorder to obtain finer differentiation. Two groups 71 and 72, for instance, may bothhave higher status than group 7 with equal proportions in 71 and in 72 of membersexercising power over members of 7. If more members of 7 in this situation areaffected by members of 7t than are affected by members of 72 we may say thai/Ican be ranked "above" ;1 2 relative to 7.

We define a micro- based core for a social institution to be the rpult of foundinga "macro" core on some appropriate m icro base. Thus a micro-based core M BC foran institution is a structure MBC = (C ,MB) where C = (r,e,X,~ is a core for < lsocial institution, MB =J,A ,PERFORM , INTEND, POWER) is a m icro base. andthe following axioms are satisfied:A7 Each group in r is a non-empty set of individuals (elements of J), and each

individual in J belongs to some group in r. Each action type in e is a non-empty set of actions (elements of A). and each action in A belongs to someaction type in 8. Furthermore, for each action type characteristic for a groupthere exists an individual in that group PERFORM ING an action of that type .

AB For any action b PERFORMED by some individual i there is some group ofwhich i is a member such that action b belongs to one of the action typescharacteristic for that group.A9 For any two groups 7. 7 ' : 7 has higher status than 7' only if i) each memberof a big part of group 7 exercises power over some member of 7', and a bigpart of 7' is thus affected, ii) there are members of /' which do not exerci sepower over members of 7, and only members of a small part of 1re such thata member of 7' exercises power over them .

Note that A9 expres~es only a necessary condition for the status relation. Thisleaves us with the possibility of further constraining it in other ways. One posvibi!i-ty~pen for future investigation-c-would be 10 use the m ental superstructures 10 heintroduced below for further characterization: a group can have higher status thananother one, for instance, if it ranks higher in most individuals' images of soci alstructure and social ranks. The "surplus" requirements in A 7 exclude individualsand actions which do not occur in any group and action type of the structure, r e-specrively. Such individuals and actions do not contribute to the theoretical picture.


8/29

8 WOLFGANO BALZERBecause of the central role of axiom AB and its counterpart A13 to be introduced

below let us use some different terminology for the sets of actions characteristic fora group. We say that an action type is admitted for a group (and for each member ofthat group) i f f it occurs in the set of action types characteristic for that group, i.e, inthe set of action types assigned to that group by the characteristic function. In thesame way, each action occurring in an action type admitted for a group or a memberof a group also is called admitted for that group or its member. Axiom AB thenmay be restated as requiring that actions are performed only if they are admittedfor one of the groups to which their actor belongs. Thus the characteristic function,and with it the institution, provides !\ setting, or a space of possible actions, foreach individual. It may be objected that AS can be immediately refuted by real-lifecounterexampJes. In applying our theory to a firm we may be confronted, say, withan accountant performing a bank robbery which is not an action admitted for himin the firm under investiga1ion. However, this and similar examples cannot be usedas counterexarnples to AB for the action referred to does not belong to any actiontype relevant for the firm, and thus should not be included in the analysis. If anaction is included in a model then there has to be a corresponding action type, too(by A7). Yet the action type has to occur in a set of action types characteristic for agroup (by A2), so an appropriate group also has to be included in the model. Thusthe choice of observed actions as appropriate to occur in the model depends on thewhole process of application. Eliminating further possible ambiguities, the previousdefinitions may be formally stated as Iollows.P01 C is a core for a social institution iff there exist

F, e, X and ~ such that C =(r,e,X,~) andAO-\ r, e, X and ~ are sets, X : r -4 PO(e) and ~ ~ r x r,A \ rand e are non-empty and disjoint, and r is finite,A2 U{X(,)!t E r}=e,A3 ~ is transitive and anti-reflexive,A4 there exists 't' E r such that for all , E r. if, t 't' then 7 ~, .

02 MBC is a micro-based core for a social institution iff there exist r,e.X, ~.l.A,PERFORM, INTEND, POWER such thatMBC = (r.e.X, ..J,A.PERFORM.INTEND,POWER).(r,e.X,~) is a core for a social institution, andAO-2 l,A are sets. PERFORM


9/29

A BASIC MODEL FOR SOCIAL Ixsrrrunoss

9.1) Cor a big part,' of , all iE -y' exercise power over some i' E 't', an da big part oC,' is thus affected;

9.2) at most Cor a small part , 0 of -,', all i' E , 0 exercise power over someiE ,. and a big part oC, is not affected by this.03 Let MBC =(f,0,X , ~.J.A,PERFORM,INTEND ,POWER) be a m icro-base d

core for a social institution.a) For each - y E I', each T E x C , ) is called a type of actions adm it ted for (rnern-

ber s or) " and each a E UX(,) is called an action admitted fo r (membersof) - y .

b) ADMIT(f,0)= {(-"T)hEf,TE0 and TEX(-Y) is called the set of ad-miffed combinations (in M BC).

2 . SUPERSTRUCTURESPart of the rnncro core o r an institution: groups, action types, and the characteristicfunction, after a while get represented in the mental frames built .\J P in the individuals, they get internalized and often even explicitly represented by terms in the lan-guage spoken by these individuals. All structures thus built up. whether c on ccptu .ilor not. we subsume under the label of superstructures. There is a sim ple coridit iunfor their developm ent: the institution has to last for more than one (human) gen era-tion. This usually being the case, parents from each group will pass on their implicitknowledge about rules of behaviour, that is, about the core of the institution. totheir olfspring." But even those individuals which are involved in the emergence ofan institution (which thereCore is not present in their process of socialization) internalize the essential distinctions (groups) and terms of behaviour (action types andX )-though they may have no verbal expressions for them yet.

Once fully built uP. a superstructure covers various items: language. belief s. cis -positions, and representations of the components occurring in a core: groups andaction types (often expressible in the language) and the characteristic function andstatus relation (which sometimes may be so expressible but often are not). Wc re-strict ourselves here to those items really necessary in the static part of the theo rylanguage, causal beliefs, and representations for f. 0 and X . Other items that m ightbecome important in specializations w ill be suppressed here. W ith respect to lan-gunge substantial restriction is necessary in order to make applicable the technicalmeans available today. We represent language by a space of propositions. In tuitively.a proposition is a class of sentences (of possibly different languages) which 11;]\ 'e th esame meaning. Some philosophical objections notwirhsrandlng'? propositions arevery practical when applied with some awareness of the difficulties. U sing the mosteconornicnl approach we start with a binary relation ~ among propositions whichmay be interpreted as "implication in meaning". The proposition (represented bythe sentence) "I am walking" in this sense implies "I am moving" which is not alogical implication of course. Implication in both directions yields equality of me an -ing. so in n sense ~ is alre ady contained in the notion of a proposition. By a spaceof propositions we mean a set P together with a relation ~ on P such that (P.:::)


10/29

l a WOLFOANO BALZERis a distributive. complementary lattice16 with 0 and 1. In applications we may sim -ply work with sentences as representatives of propositions. and take the lattice op-erations ".11. V to correspond to the ordinary logical connectives. F ive additionalitems wilt be needed as occurring in any superstructure. F irst. we need a compo-nent representing a person's causal beliefs. For each individual. we use a binaryrelation 8 among propositions; reference to the individual w ilt be made explicit be-low . 8(p. p') expresses that propositions p an d p' represent events e.e' such thatthe individual under consideration beliefs e is a partial cause of e' , 8(p,p') may beread metaphorically as "the individual believes that p partially causes p'", Alterna-tively. we m ight work with a causal relation 8 directly established among eventswhich would lead to a more natural reading of 8'(e.e'): "the individual believes thatevent e partially causes e'", We opt for causality being represented at the level ofpropositions because social t~eory.pll.ts more weight on causal belief than on "real"causes and effects. and because of strategic reasons not to be defended here. Partialcauses are events which, together w ith other events (i.e. other partial causes) yield a"full" cause. The problems of causality cannot be discussed here." The relation ofcausal belief wilt play an essential role in our characterization of POWER in Sec. 4.Second, we need representations of groups, action types. and the character istic

function. which are denoted by G (groups). T (action types). and CH I. respectively.It would be most natural to assume G an d T to consist of sets of terms in the Inn-gunge used by the individual. and CH I to be given by a set of propositions definingwhich action types are characteristic for what groups. Such treatment is . in fact.possible. but the technical complication implied is not balanced by direct benefit inthis paper. So for the moment we prefer a co arser a pp ro ac h, treating G and T justas unstructured sets. and CHI as a function. mapping each element of G in a setof elements of T. In addition. we use a binary relation l among members of P an dT. The interpretation is this. Elements of G are internalized representations of thedifferent groups in the individual's superstructure, and elements of T are in tern al-ized representations of the different action types. CH I is some internalization of thecharacteristic function in the individual to which the superstructure is assigned. CHImay be regarded as that individual's disposition to associate adm itted types of actionwith the different groups, an d l as it's disposition to subsume some (representationof an) action under some (representation of an) action type. "CH I(g) ={II ... I"} .may be read as "the individual internally associates representations 11... 1" of ac-tion types w ith the representation g of a group", and "p EI" as "the individualsubsum es proposition p, which represents some action. under it's representation Iof an action type". If all representations are verbal, we may think of P as a sentence(describing some action). of '10 ... t; as expressions for action types. and of g as aterm denoting some group. In general. however. we must not assume that nil thoserepresentations "are" verbal, or can be verbalized. To the present account verbaliza-

"I.e. ~ is tr arwitive, reflexive, and antl-symmeuic (.r ~ y.nd Y ~ x implie.\ x ~ y) . for any two o. b E P,their inftmum a /I. band supremum a v b with respect to ~ exist. and these infima ond ,uprema sot;,fy theusua] I,,,,, or distr ibutlon, Furthermore, there e";SI> the infimum (rcsp. supremum) or oil 0 E P, denotedby 0 [rcsp. I) nd ror each DE P there Is elClClly one b in P (denoted hy ~o) such thot D /I. b ~ 0 anda v b - I. See Graetzer (1971) for basic notion s.11A cornpr eberwive, non-technlcal account of cl1u'~lity i~ound in M:tckic (1974), for" technical. pr oha-hili\tic ~rrr()


11/29

A,

A BASIC MODEL FOR SOCIAL INSTITUTIONS 11

tion is inessential, in particular in connection with CHI. A ll we need is that CH Ibe internalized so that it may guide the individual's behaviour. CH I also may be re-garded as a rudimentary (representative of a) norm . In a more elaborate version ofour models norms would have a natural place in the superstructures, together w ithsome constraint requiring identical norms in members of the same group. This isanother point where the present model may serve as a basis for further specializa -tion.

We do not state special axioms for the relation of causal belief. Some generalaxioms might be found by studying philosophical literature!" while more interestingaxioms will not hold in general but w ill be restricted to particular forms of socia lsystems, Not only will particular causal beliefs in a society being thoroughly orientedtowards magic like the Azande ", for example, be different from ours in the age ofscience, but the structure of the whole belief system is likely to be different. There-fore it seems better to leave axioms for causal belief to be studied in speciatizationsof the present theory. .

We define a superstructure to be a structure .r the form.r = (p,~,B,a,T,CHI,()

such that (P,~) is a ~pace of propositions, B is a binary relation among elementsof P denoting an individual'? i's relation of causal belief, a and T are sets ofinternal representations of groups and action types in i, respectively, CH I : G -PO(T) is a function mapping representations of groups on sets of representations ofaction types which denotes the characteristic function as internalized by i, and ( isa relation of subsurnption between elements of P an d T. The only axiom requiredto hold is that (p,~) be a distributive, complementary lattice with 0 and 1.

In principle each individual imay have its own superstructure. We use a func-tion .r to assign that superstructure to each individual in an institution. Thus r(i)denotes the superstructure assigned to individual i, or simply: i' s superstructure. Inorder to keep things legible we refer to the components of x(i) by an upper index"i", S020 pi, ~i et c. w ill denote nl(x(i, n2(.r(i etc. In addition to this as sign-ment, in Sec. 4 we will need a more fine-grained representation [unction, r ep., which(depending on each individual i) m aps actions into propositions, groups into repre-sentatives in c'. and the character istic function X into a function CHI'. Formally,repi may be defined on the union of the sets A o f action s, r of groups, arid thesingleton { X ) (see A~5 below), and be required to map each kind of argumentinto an appropriate value, i.e. each action into some proposition represenring thisaction, each group into some member of ai representing this group, and the func-tion X into the function CH l' occurring in i's superstructure. Wc agree that rep,(b)denotes the proposition (sentence) describing action band repi(l) the representa-tion of group "I in the language. In Sec. 4 representations as given by the functionsrep; w ill be used to formulate the central axiom for POWER relations.Starting from a m icro-based core we add one superstructure for each individual

in the core, and we use functions .r and rep to assign the whole superstructure and"Sce Evans-Pr itchard (\937), [or example,


12/29

12 WOlFOANO BALZER

their components to the different individuals, respectively. The resulting structurewe call a social schema.A social schema SS therefore is a structure

SS = (r,e,X, .J,A,PERFORM,INTEND,POWER,M,x,rep)such that (r,e,X, .J,A,PERFORM,INTEND,POWER) is a micro-basedcore for a social institution, M Is a set, x is a function mapping J into Mand rep is a function which, for each iE J maps actions, groups, and Xinto respective elements from some superstructure. Moreover:

AIO Each element in M is a superstructure,All x is onto,AI2 For all groups 1 E r and for any two individuals i,j, if i, j both are members

of 1 then i and j liav-e the same superstructure (x(i) =x(j)) and the samerep-function (rep, =rep,),

A!3 For each individual i E J and each action a E A, i PERFORMS a only if i'srepresentation repi(a) can be subsumed under one of i's representations of theaction types in CH I1(rePi(1)) characteristic for one of the groups 1 to which ibelongs.

Function x being onto as required in All restricts the set M of superstructures tothose really needed. We call x(i) individual i's superstructure. Two individuals i, jin general may have different superstructures: x(i) f x(j). By AI2 this possibility isruled out for individuals belonging to the same group. In other words, superstruc-tures within one group are homogenous. and so are the ways different membersrepresent actions and groups in the language. In its present form this axiom is per-haps too strong and idealized. It implies, for instance, that in an institution in whichall groups are overlapping all individuals have identical superstructures. This prob-lem of the formalism points to a real problem, however. There is some balancebetween the degree in which groups overlap on the one hand. and the degree towhich the groups' languages are similar or equal on the other hand. There are twoways to weaken this axiom. First, we simply may blur it, and require that in eachgroup the superstructures and the rep-functions are only approxirnatively equal. Theprecise way of blurring here is not obvious and will depend on the concrete case.Another, theoretically more interesting way consists in assigning superstructures notto individuals but to pairs of individuals and groups. This allows to speak of thesuperstructure of an individual as far as it is member of some group, and thereforeof one individual "having" several superstructures, each one being used in situationsgoverned by a corresponding group. We do not pursue this possibility here but notethat the final definitions in Sec. 4 can be easily adjusted to such a treatment.AD mirrors AB on the level of superstructures. Roughly, it says that each individ-

ual's PERFORMANCE has to be compatible with the characterization of groupsin the institution, but now with the characterization as internalized by that sameindividual. This axiom is important for the stability of institutions. One major rea-son for stability is that the individuals have internalized the institution's character-istic function, and because they behave in the frame given by that function in theirindividual internal representations (AD). Using the notion of admissible act ions


13/29

"

A BASIC MODEL FOR SOCIAL INSTIrunoNs 13

we may restate A13 as saying that each individual PERFORMS only those actionswhich are adm issible according to the individual's superstructure. Thus the super-structures restrict and guide the possibilities of individual behaviour-in line w ithconstructivism . Adm ittedly , little is known by now about the nature of these internalrepresentations of characteristic functions, and our CH I-functions are just a dummyto be filled by future research. However. even in this crude form the role of theCH I-functions in our theory as expressed in A13 is crucial. H ere are the formaldefinitions of superstructures and social schemata.04 ;r is a superstructure if f there exist p,~. B. C. T. CH I.l such that ;r = (P, ~.

B.C,T.CHI.l) andI) (p.~) is a distributive. complementary lattice w ith 0 and 1.2) B ~ P x r.3) G and T are disjoint sets, C is fin ite .4) Cm.: C -+POeT) .5) e ~ P x T.

05 S is a social schema i f f there exist r,e.X .. ,J.A .PER FORM,INTENO ,POWER , and M ,x,rep such thatS = ( r, 8,X , .. J ,A ,PERFORM, INTENO.POWER,M,x ,r ep ),(r,8 ,X .. J,A ,PERFORM ,INTENO .POWER) is a m icro-based core for a so-cial institu tion, andA0-5 M is a set. x : J . . . M. and r ep :J -POAUrU{X)x(U{P ;/ x( i)E

M)U{C;/x(i)EM)U{CHli/x(i)EM))) is such that for nil iEJ:repi :'" ' rep(i) is a function. repi : (A U r U (X )) .... (UP; U UC ; U (CH I; I.t(i) EM ) such that reP i(u) is in pi (resp. in Ci) if 11 is in A (resp, inI'), and repi(X ) '"' cur,

AlO Each element in M is a s up ers tru ctu re .A ll x is o nto,AI2 for all , E r and all i,j : if iE, and j E, then xCi) = x(j) and rep; =

repj,A l3 for all iE J and all a E A: if (i PERFORMS a) then there exist jE r andu E CH li (reP i(' such that i E1nd rep;(a)t u.

3. SOCIAL PRACTISESRoughly, a new socially relevant type of actions originates and develops much likea new species. A new kind of action is performed w ith or w ithout reasons, and ifthe surrounding is favourable. if other people around find it interesting, or impor-tant or exotic or chique, they will im itate it thus starting an avalanche of im itations.The original action (or actions) plus these im itations then form a new action typein our technical sense. A sim ilar structure we find in the origin and developmentof groups of actors which perform a new type of action. A t the beginning there areone or more "founders", people perform ing the new kind of action for the first tim e.O ther people im itate the actions and in this sense become "disciples" of the original


14/29

14 WOLFGANG BALZER

these processes constitute genidentical entities, for they both spread from a respec-tive source by means of a relation of imitation, and it is just this which providestheir unity. In the case of groups it is obvious that different persons in a group maybe quite different, they have nothing in common except their imitating the "inven-tors"of a new action type. But also for actions it will be hard to hold up a thesis oftheir having common features besides their being copies of the original actions. Itseems hard to identify, say, different forms of greeting one another only on the basisof the observed events in space-time. The important clue for identification is thatthey can be traced back to other, previous events which are imitated ("learned').We define the auxiliary notion of a genidentical structure to consist of an abstract

set D of "carriers", a subset SOURCE of D of "originals" or "founders", and arelation COPY among carriers. (6 COPIES 0') may be circumscribed as "0 is animitation of 0'" in case of ~ctioIlS, ~n~ _as"6 is a disciple of 6'" in case of individuals(children count as disciples).-D6 GS is a genldentical structure iff there exist D,SOURCE and COpy such that

GS =(D,SOURCE,COPY) and1) D is a non-empty set,2) SOURCE ~ D is not empty,3) COPY ~ D x D is reflexive and anti-symmetric,4) each 6 E D can be traced back through a chain of COPIES to some element

of SOURCE,5) SOURCE contains much less elements than D

D6-4 may be formalized as requiring for each 6 E D the existence of some O t ESOURCE and of 62, ... ,6" such that 6" =6 and each 6i1 COPIES 0 ; (i < 11). D6-5 has to be made precise in the respective context. Different ratios and speeds ofpropagation are studied in the evolutionary branches of game theory.?' If time ismade explicit the number I D , ! of carriers at a given instant Imay be studied as afunction of time (often an exponential one).

By combining the two genidentical structures associated with an action type andthe group of agents performing actions of that type we obtain the fundamental con-cept of a social practise. It is fundamental because it is concerned with the smallestunit of socially relevant behaviour, a type of net ions, and the W


15/29

>.,


tained from originals and founders. W e need two different COPY -relations, one foreach set. (6 COPY('y)6 ') applies whenever a new actor 6 definitely takes over thenew behaviour And in this sense becomes a new member of" and (6 COPY(T)O ')applies whenever action 6 is a copy, an im itation of action 6'. Two axioms may beformulated connecting the two basic sets, and r. F irst, they are restricted to con-tni:1 only elements which are involved in some PERFORMANCE relation (D7.4below). Actions not PERFORMED by any member of the group can be excluded.even if they are sim ilar to those occurring in T. In the same way we exclude indi-viduals which do not PERFORM any of the actions in ,. Such individuals are notsocially relevant in constituting a social practise-though they may be quite relevantin other respects (for instance in providing the physical means of life for the wholegroup). A second axiom (07-5) connects the original actions with the founders ofthe group. Each original action, i.e. each element of SOURCE(r), has to be PER FORMED by some "founder", i.e. some member of SOURCE(,), and conversely,each foundef'has to PER 'FORM at least one original action of the type under con-sideration. W e note that the long historical development of social.practises resultsin the individual superstructures' being firm ly and deeply implanted, which in turngives heavy weight to the "frame o f admissibi li ty ".D7 P is a social practise iff there exist "r,PER FO RM ,SO UR CE(,),COPY (-y),

SOURCE(r),COPY(r) such thatSP =(1 ,r ,PERFORM ,SOURCE(, ) ,COPY (,) ,SOUR CE(r) ,COPY (r) and1) "( and rare non-empty sets, and disjoint,2) PERFORM ~ "( x r ,3) (f,SOURCE(,),COPY(, and (r,SOURCE(T),COPY (r) are gcnid cnticn!

structures,4) for all a E T there is some i E, such that ( i PERFORMS a), and for all

i E "( there is some a E r such that ( i PERFORMS A),5) for nil a E SOURCE(r) there is some i E SOURCE(,) such that (i PER

FORMS a), and for nil i E SOURCE(,) there is som e a E SOURCE(T)such that (i PERFORMS a).

Further axioms concerning the COPY relations may be Ior mulated, but are no!needed here. The concept of a social practise has numerous applications. like "con-ferring a doctor's degree", "taking the holy communion" (Rom an Catholic, s ay),"burning a w itch", "sieging a town", "perform ing a campaign (in war)", "electing aleader" (say, the US president).

It is clear that the components of a social practise may be difficult if not irnpos-sible to determ ine. The SOURCES often are lost in history, an d the COpy rela-tions also may be difficult to trace historically . This may create the impression thatthe notion is empty and irrelevant to social institutions. To this possible obj ectionthere are two replies. F irst, as already mentioned, there are no natural standards ofsimilarity for actors and actions. As long as actors and actions are not formally de-fined in an institution, the basic approach towards their sirnilar i r y , and thus towardsthe notions of groups and of action types themselves, is via genidenticnl structures.


16/29

16 WOLFOANO BALZER

which various different forms of growth and of grow th conditions may be Iorrnu-lated and studied. By giving further, special inner structure to the actors and actiontypes, by introducing the notion of special external conditions together w ith spe-cial laws governing the COPY relations we may obtain quite substantial structures.H owever, just as in the theory of evolution, such specialization is possible only atthe cost of considerably narrowing down the range of applications. As stressed inthe introduction, our aim here is only to present the general model.

In order to incorporate social practises into social institutions, two adjustmentshave to be made. F irst, actors have to be interpreted as individuals for we deal herew ith first-order institutions only . Second, we must not always identify a "group" ofa social practise with a group in an institution. There are important social practisesthe actors of which are distributed over different groups of an institution. Think ofthe holy communion, say, in ~eurlaLF(ance. M oreover, a social practise may be mucholder than an institution of which it becomes an ingredient. In this case the "group"of the social 'practise w ill contain many more individuals than each correspondinggroup of the institution into which the practise has found entrance.

4. SOCIAL INSTITUTIONSWe now have prepared the ground for introducing a comprehensive definition ofsocial institutions. W e start from a social schema, which is enriched in two steps.In the firs! step, we assume that each adm itted pair (J ,T) consisting of a group1 and an action type T in the schema's core is "given" by, or embedded or an-chored in, a social practise. Recall that a core was defined to consist essentially ofa set of groups each of which is characterized in terms of the action types per-formed by it's members. An adm itted pair consists of one such group and one ofthe action types characteristic for that group. Our assumption of embedding thusamounts to regarding each such action type as having originated from some his-torically first events of actions of that type performed by individuals perhaps longago through sequences of im itations in which the number of individuals acting alsoincreases. U sually, the group and action type making up one admitted pair in aninstitution w ill not correspond to a full social practise. In a typical case, a group inan institution originally is formed by individuals which already are used to one orseveral social practises existing before the institution is formed, but there will beother individuals used to these social practises which do not become members ofthe group in question. O ften, one social practise in this way contributes to adm it-ted pairs in different groups of one or several institutions. For this reason we mustnot identify an adm itted pair (1, T) w ith the full "base sets" of a correspondingsocial practise. W e say that an adm itted pair (1, T) is anchored in a social pr ac-rise if ., and Tare subsets of the corresponding sets of actors and actions in thatpractise. This relation is best seen from the point of view of a given social prac-tise p ( . , ' ,T ' ,PERFORM',SOURCE(1') ,COPY(-, ' ) ,SOURCE(T') ,COPY(T)}.I n the formation of a new institution it may happen that some of the individualsinvolved are practitioners of the social practice, i.e. members of 1'. M oreover, itmay happen that the action type of the social practise is relevant for the institution.In thnt case it Is likely that the set of rnernber s of I' which occur in the system will


17/29

A BASIC MODEL FOR SOCIAL INSTITUTIONS t7

form one of the evolving institu tion's groups, and the set of actions in r: performedby those members w ill form one of the institution's action types. If this is so, wesay that the adm itted pair (f, r ) of the institution is anchored in the social practiseP, provided two further technical conditions are satisfied. First, we require that noother action type from the institution is contained in the set iO of the practise'sactions, i.e, that the actions occurring in the social practise determ ine a unique ac -tion type in the institution. Second, we require that both PERFORM relations, thatoccurring in the social practise, and that occurring in the institution, are identicalfor adm itted pairs, i.e. for pairs in 7 x T.D8 If (r,e,X ,~,J,A ,PERFORM ,INTEND ,POWER ,M ,x,rep) is a social schema,

P = (7,i ' ,PERFORM',SOURCE(7') ,COPY(7') ,SOURCE(iO),COPY(iO)is a social practise, and 7 Er, TEe are such that r E X(7) then ( - y , i) is an-chored ill P if f1 ) T ~ ~ and for all Tt E 8: if Tt 'f T then Tt ri TO = 0,2) 7 ~ 7"3) PERFORM and PERFORM" are identical when restnctedto 7 x r ,

Note that the analogon to 08 1 fails to hold for groups. Individuals from different groups may well eng< lge in a common social practise. In fact, if there are nocommon practises at all, an institution will not last for long.

For a social institution we require that all the institution 's adm itted pairs be an-chored in the sense of 08 in suitable social practises. W e use a function y to assignthese social practises to the admitted pairs, so Y ( 7 , T ) ) denotes the social practisein which the admitted pair (7 , T ) is ancho red.The second feature by which social schemata are enriched in order to obtain

institutions consists in a detailed characterization of the POWER relation. Considertwo in div id ua ls i,j and two actions a.b. Then our characterization of POWER isexpressed in the follow ing axiom :(AP) ('by doing (I exerts POWER over j to do b if and only if the four follow ing

requirements are satisfied:a) actions nand b are actually PERFORMED by ind j,b) iNTENDS that j should do band j does not INTEND to do b,c) the individuals believe that action a p artially c au ses b,d) ind j are members of groups 7 , 7 ' such that actions a an d b are adm itted

fo r ind j as members of groups 7 and 7', and such that the admissibilityof a and b for these groups is represented in i' s an d j' s superstructure,respectively.

Our characterization has the form of an explicit definition but is not intendedto serve as a mere definition. Rather, we regard it as an ordinary axiom of thetheory which happens to have the form of a biconditional. Such axioms frequentlyoccur in respected empirical theories; th ink of Newton's second law . Not regarding(AP) as a definition implies two things. First, the axiom is open to variation in the"defining" conditions. W e may add further requirements to a)-< 1) in order to obtainmore special characterizations which do not hold in all cases of exertion of power


18/29

18 WOLFGANG BALZER

a little self-contained theory of power in small groups which can be specialized inseveral ways to deal with different important forms of power like force, coercion,and manipulation.22 Second, the notion of POWER is not fully reduced to the otherconcepts occurring in conditions a}-d). Rather, it is taken as an ordinary primitivereferring to some feature of reality, and accessible by means independent of theabove axiom. A5 the concepts occurring in b) and c) are of the same difficult care-gory as POWER itself with respect to operationalization nothing would be gained byinsisting that the axiom defines POWER. On the contrary, there are direct-thoughnot very reliable-means to determine POWER, for instance by appropriate scales.Our characterization of POWER intuitively may be split up into two parts, one

part (conditions a-c) concerning the micro level of actions, performance, intentions,and causal beliefs, the other part (condition d) exploiting the frame given by a socialschema (characteristic functica, superstructures), The first part of b) is necessary inorder to exclude actions with unintended causal consequences (like a car accident)from the range of exertions of POWER. the second part of b) captures the in-sight that POWER exists only where there is some form of resistance.P The third"micro" condition c) deals with the causal connections between the two agents' ac-tions. Here the easy account would be to refer to an "objective" causal relation andto hold that j' s action b is causally determined, at least partially, by j's action n.However. causal beliefs may differ between individuals from different social groups(think of magic beliefs, or belief in witches). We do not want to decide here whosecausal relation is "the correct" one. In social reality there are frequent cases inwhich power is exerted on the basis of beliefs on the side of the subordinate agentwhich the superordinate agent regards as wrong or superstitious. This is part of thereason why we chose causal beliefs rather than an "objective" causal relation to fig-ure in superstructures. Now we are in the position to use causal beliefs efficiently.In c) we require the individuals to believe that j' s action b is partially caused by i'saction a. It is not necessary that both individuals have such belief. Varying with theparticular form of power it may suffice that either the superordinaie agent or thesubordinate agent has it2(. Of course, very often, both of them will have it. An im-portant example of a form of power in which the causal connection may be hiddento one of the agents, is manipulation.The second part of conditions for the POWER relation in (AP) refers to theframe given by the institution, or rather the social schema. in which the events take

place. As stressed twice already. the characteristic function occurring in the core aswell as its representations CH li in the superstructures provide a frame of admittedactions. All actions PERFORMED by an individual in a social schema have to beadmitted. on the object level (AB) as well as on the level of superstructures (A 13).Condition d) in (AP) is intended to make this explicit for the actions involved ina POWER relation. Thus (A P) has to be seen as a characterization of the notionZ1Compare Wartenberg (1968) Tor 8 recent account of these (orms of power. Though the syntax of ourPOWER relation is the same .1 IMI of Dahl (1957) our characterization of the notion in 09 is nOIintended u. reconstroction of .ny existlng account. It wn1 developed Independenlly, in an attempt lapre.5CfVe tbe imighl! of 01her definition s, of course.llWcbcr (1960), p. 28.HThis comes out onc e special rorm~ or power, a~ descr ibcd in War ienber g (19R R ). nrc


19/29


of "power in a social institution", rather than of power in general. However, sincePOWER(i ,a,j ,b) , by a), implies that (i PERFORMS a) an d (j PERFORMS h).AB and AI3 automatically imply that actions a an d b are admitted on both levelsin any exertion of POWER. So adrnissability needs not be explicitly stated as acondition for POWER. In other words: in a social schema (AP) above is equivalentto a characterization in which d) is omitted. Using AB and AI3 we prove withoutdifficulty:

LEMMA. If S = (r, ... , r ep) is a social schem a then, il l S, (AP) is equivalent t o: forall i,j,a,b : POWER(i,a,j,b) if! conditions a). b) and c) of (AP) ab ove are satisfied.So in the following final definition of a social institution d) can be omitted.SI is a social institution i f f there exist r,e,X , ~,J, A , PERFORM , INTEND .POW ER,M ,x,rep,SP,y such that (r,e,X ,~,J,A,PERFORM ,INTEND ,POW ER.M ,x,rep) is a social schema, y is a function assigning a social practise y ( J , T toeach admitted pair ( - y , r} in the core (I', e , X , -e), an dA14 SP is a set of social practises,A15 y is ontoAI6 each pair ((,r) admitted in the core (r,e,X , .. is anchored in the cor respond-

ing social practise y ( { , T ,AI? For all i, j E } and all a.b EA: iby doing a exerts POWER over j to do b iffa) (i PERFORMS a) and (j PERFORMS b),

b) i INTENDS that j should do band j does not INTEND to do b.c) at least one of the individuals i, j believes that a causes b.

.~

A 15 in analogy to All guarantees that all social practises occurring in SP arereally needed. W ith respect to A) 7 we have several further remarks. F irst, our non-standard syntax for INTEND pays off here.

A17-b may be written formally as follows: INTENDS(i,j,b) and not INTENDS(j,j,b). In the first conjunct i and j are different, ih as in ten tio ns about an othe rperson to do something. In the second conjunct both arguments for individuals arefilled in by (the name of) the same individual. Here, INTENDS( j , j ,b) of coursemeans that j "intends" to do b in the ordinary sense. Second, in A17-c the relationof causal belief operates on the level of propositions, as stared in Sec. 2. Thereforethe actions a,b to be related as cause and effect first have to be represented in theform of propositions. If k denotes any of the individuals i, j ho lding a caus al be liefthen we have to look into k 's superstructure x(k) =(P',~', 8',...) in order 10 getk 's relation 8k of causal belief which relates k's representations of actions a and h.rep.(a) and rep.(h): B (rep.(a),rep.(b. This way of stating a causal connectionin the superstructures (as opposed to the level fo material reality) does not perfectlyagree with causal talk in ordinary language which always proceeds in the realisticmode. But as stated above it is causal belief rather than real causes and effects thatmatter in social theory, and a causal relation among the propositions which are atan individual's disposal is well suited to express such beliefs.

It has to he stressed that condition AI7 above covers only the mode of actually


20/29

20 WOLFGANG BALZER

w ould PERFORM a then i by doing a would exert POWER over j to do b. Thereare standard ways to analyze such counterfactuals in possible world sernanucs-'. Inthe present case such analysis would require to introduce sets of social institutions"similar" to a given one.

If in condition AI7 we look at part c) being satisfied for individual ian d


21/29

A BASIC MODEL FOR SOCIAL INSTIruTIONS 2tIn order to make precise all the details of the model, let us state the definition

in a completely formal way. The theory of social institutions introduced in thi s wavconsists of the class of all possible social institutions (as defined in D9) plus the se tof all real sy stems to which it is intended to apply. This set of intended S)"-Hcrru wasroughly described in the introduction. The claim associated with the presen t the oryis that each intended system is a social institution in a sense still to be specified.D 9 SI is a social institution iff there exist r,8,X , ~,J,A, PERFORM,INTEND.

POWER,M,.l",rep,SP ,y such thatSI = (r,8,X,~,J,A,PERFORM ,INTEND ,POWER,M ,.l" ,rep,SP,y), an dA0-9 (r,8,X , ~ ,J,A ,PERFORM ,INTEND,POWER,M x.rep) is a social sche-ma, A14 SP is a set of social practises,A15 y : ADMIT(r ,8) -+SP is onto, (compare D3-b)A16 for all (( ',T) E ADMIT(r ,8): ( t ,T) is anchored in y(-p),AI7 fqr all i.] E J and nil a,b EA:

[POWER(i,a, j ,b) if fa) (i PERFORMS 0) and (j PERFORMS b) ,b) INTENDS(i, j ,b) and not INTENDS(j, j ,b),c) there is k E {i,n such that Bk(repk(a),repk(b].

S. APPLICATIONThe process of application to some real system of a theory as given by a clas s ofmodels in all areas of empirical science has the following form . F irst, dat a are co l-lected and formatted in the theory's vocabulary!". Second, it is tried to fit these datawith the theoretical hypotheses. Identifying hypotheses and modelsU l such fit e sscn -tially amounts to an existential c1aim 29 The data fit w ith the hypotheses if there(lists some (hypothetical) model into which the data can be consistently embedded ,i.e. which contains "parts" corresponding to the data in a natural way. If the theorycan be successfully applied to some intended system in this sense we may claim Ih~1th e sy stem investigated is a so cial in stitu tio n. Accordingly, the claim associated withthe present theory is that all the intended systems described in the introduction aresocial institutions in the sense just explained. In other words. the data which can hecollected from those systems all are ernbeddable in corresponding models.

Due to the complexity of our models it is impossible to provide a two or thr ee-page example based on proper empirical investigation or data. Instead. let us con-sider an unspecific system which by appropriate historical studies could give rise toa real application. Three aims are pursued by these considerations. First, we specify what kind of historical data and methods are required to do n proper empiricalstudy of an institution. Second, we want to show that all our models' componentsare present and capture important features of real institutions. Third, we want 10examplify our general view of application sketched in the previous par agr aph.27Note that we describe the process of .pptying an already e.isting theory. not the process of inventingil. The theory is given beforehand.


22/29

22 WOLFGANO BAlZER

One feature of this view which is particularly relevant to the present theory is thatit does not presuppose a distinction between theoretical and observational termswhich is often made in order to separate "reliable", "objective", observational datafrom "merely" hypothetical hypotheses. Such a distinction being very problematiceven in the natural sciences'? we think there is no reason to insist in austere ob-servational foundation which simply is not feasible. A theory T's data consist of allatomic sentences of T for which there are sufficiently reliable means of determi-nation, sufficient reliability often being a matter of agreement in the respective sci-entific community. This view goes together with a very liberal conception of data:by a datum we understand every atomic statement which can be obtained in a sys-tematic way. Roughly, this means that it be obtained from other data or hypothesesin a unique way as guaranteed by some regularity.31 This notion does not insist inreproducability as it occurs- in measurement in the natural sciences (which cannotbe achieved in sociology) but keeps enough substance to make data a non-trivialmatter of intersubjective (and in this sense objective) agreement. In particular, wedo not insist in methods of determining the "objects" occurring in a system (likeaction tokens, action types, individuals and groups) in a way completely neutral andindependent of the language and intention of the investigator. In the social sciencesit seems necessary and adequate to admit for a moderate amount of antecedentunderstanding to provide the investigator with a first rough guide for application.

We begin our example by looking at a realistic set of empirical or historical data.Consider a system with three groups as realized many times in medieval Europeanvillages: one group consisting of the local nobleman (a count, say) plus his family,a second of the peasants and their families, and a third of "intermediate" persons:priest, teacher, servants.

It seems relatively easy to determine the individuals and actions occurring in thesystem 8S well as the PERFORMANCE relation. By direct inspection as a cornpe-tent speaker of the language or by historical studies we may collect a set of descrip-tions of action tokens (printed in italics below) together with a list of statementsof the form (i, PERFORMS Q,), i= l, ... m about which person performs whichaction. Also, the determination of action types does not seem to pose any particu-lar problem for our theory. Things are different for the remaining macro concepts:groups, characteristic function, and status relation. How can these be determined?If we try to determine each of these notions on its own, and independently of ourtheory, we run into difficulties. Concerning the groups an investigator with differ-ent intentions (biological or medical. say) would perhaps arrive at a very differentgrouping. Even the sociologist who understands the system along our lines has dif-ferent possibilities of grouping, corresponding to different levels of detail. She maytake a coarse grained group structure lumping together nobility and clergy. or theone indicated above, or proceed even more fine grained differentiating, sny. be-tween male, grown-up peasants. women. and children inside the larger "group" ofpeasants. Concerning the characteristic function it is not adequate to take all actiontypes observed as being realized by members of a group to be characteristic for thatgroup for in this way we would arrive at many types which simply are irrelevant in~Stt Balzer (1986) for. recent di'cu


23/29

.,

A BASIC M OD EL FOR SOCIAL 11-151111.]1101-15 23

the institution under study or do not contribute to any differentiation of the groups.Statistical considerations and/or techniques from network analysis are required inorder to achieve a simultaneous determination of groups and the characteristic func-tion. We do not want to look at a particular method here, the important point isthat such methods do not rely on the present theory. By contrast, it is difficult toimagine any method to determine the status relation which would not use A9, andtherefore would be independent of our theory. So in order to avoid circularities itseems wise to avoid statements about . in the data.

In the system of a village we might obtain action types of the following kind:7"1 : HUNTING7"2 : EXERCISING for fight on horseback7") : ORDERING, . .7"4 : WORKING in the fields ."7"5 : FIGHTING with the fist1'6 : SERVING as a beater1'7 : SENDING one's children to the Sunday school1'8 : READING1'9 : TRANSMITTING orders, etc.,

and by applying some statistical method we might get a grouping of the followingform:

71: {COUNT, COUNTESS, GHLD_l, CHILD_2, MOTHER-IN.-LAW}72: {PRIEST,TEACHER,SERVANT_l,SERVANL1}73: {PEASANT_I, PEASANT.-2, , WlFE...l, WlFE...2,

.. ,' CHILD" _1, CHILD"..J., }and facts about the characteristic function, like:

1',,1'2,1') E X ( 7d7"4,7"\,7"6,7"7 E X(72)

7"8,1'9 E X(73).The remaining two notions, INTEND nnd POWER, are of a different kind. In

contrast to the physics paradigm there is no measuring apparatus for these notionsfunctioning independently of the observer. There is no hope of achieving such


24/29

WOLFOANG BALZER

the language used in the system has acquired together with learning the language.These means are communicable, investigators may disagree about the intentions asexpressed by observed verbal and non-verbal behaviour, or corresponding histori-cal data, and they may argue systematically about them (as done by historians andpolitical scientists). We think that our abilities of ascribing intentions and power ac-quired through language competence may be used as a basis in order to determinenotions like INTEND and POWER for the aim of applying our theory to ernpir icalsystems. We admit that these means are not the most reliable ones, but we holdthat they are not unscientific apriori. 1b exclude them dogmatically (i.e. by pointingto the big brother of physics) would mean considerable impoverishment of socialscience.

From observing verbal and other kinds of behavior (or from corresponding histor-ical facts) we might obtain a lrst of statements about INTEND and POWER. Thecount intends to hunt, and intends that peasant_I serves as a beater. Peasant.i l,on the other hand, intends to work on his field in the same time. The teacher in-tends peasant_2 to send his daughter, child_3, say, to the Sunday school. Peasant_2intends his daughter to help him working in the fields etc.:

INTENDS(COUNT, COUNT,I!IIIII)INTENDS(COUNT, PEASANT _I,bealillg)INTENDS(PEASA NL I. PEASA NT_I. work)INTENDS(TEACHER, PEASANT -2,sl'lIdillg)INTENDS(PEASANT -2, CH ILD _3. work) etc.

POWER(COUNT,IIUlrI, PEASANT _) ,beatillg)POWER(COUNT,orrierillg, TEACHER,lratlJmillillg)POWER(TEACHER,lratlJmillillg, PEASANT .-1,sl!lIdillg) etc.

It is much more difficult to get data, or to agree on data, about the superstruc-tures which are likely to differ for members of different groups. If we think of asystem in which nobility is in close contact with the court then very likely its lan-guage will be refined and contain many terms unknown or at least not used bythe peasants. By blurring some idiosyncrasies present in each individual's languagelinguistic studies might yield spaces of propositions (pi, ~) for each individual ioccurring in the village. Also, causal beliefs are likely to differ for members ~f thedifferent groups. The peasants may believe, say, that an old woman living in theforest is a witch and may cause certain unusual things to happen while members ofgroups 1'1 and 1'2 do not have such causal beliefs. Also the priest may hold somecausal beliefs involving his God which are not shared by the very mundane count.Though all these causal beliefs are hypothetical from the sociologist's standpointthere are methods of different degrees of reliability in order to infer them: verbalinterrogation and observation of behaviour (or corresponding inferences on the ba-sis of historical sources). It seems sound to assume thot at least some of the causalbeliefs Bk(rep~(ai),repi(bi)) (for appropr inte i, j,k ) can he obt a incd in this wnv


25/29

A BASIC MODEL FOR SOCIAL INSTlnJTlONS 25

The representations of groups, action types and characteristic function in the suoper structures can be determ ined if these items have verbal representations. In thiscase the terms in the language as used by the individuals themselves may be takenas representatives. O therw ise those representations are strongly hypothetical. In theexample not all individuals may have terms for denoting group /2 ("clergy" is notvery appropriate), the relations between groups and action types as captured by thecharacteristic function are not directly represented by terms, neither are the rcla-tions of subsurnpt ion (j in Sec. 2) of an action representation repj(oj) under anaction type representation from T. Where appropriate representatives I,can he as-sumed in the form of known terms of the language we may collect some statementsabout the individual representation functions repj : repj(oj) =I,and rerJ;(fd = '"(for appropriate indices i,j,s,s').The last feature to be considered are the social practises from which the relevant

action types stem . Thesepractises are difficult to trace. How and where did patternsof feudal behaviour typical for the noble persons in the system originate? They musthave originated at some time, the medieval patterns did not exist-in antiquity. W ehave to go back to the early m iddle ages when the first cavalry arm ies were formed,and the European type of the knight made its first appearance. We have to lookat the formation of the catholic church in order to find the practises relevant forthe priest and, Inter on in 11th and 12th century, for the teacher. The church alsoyields some practises for the other groups, like the holy communion, or the ways ofdealing with birth, marr iage, and death. Even at the side of the peasants we m~y findsocial practises, for instance in connection w ith ways of farm ing, of growing c attle.of dealing with sickness, or just of cooking. A ll these action types once had been"invented" and were delivered from then generation after generation. It is clear Ih;]ta full statement of nil the knowledge available about the different social pr a ciiscsinvolved here would blow up the set of data without end. By spending enough energyit certainly is possible to provide substantial collections of relevant dat a about theactions, actors, and the respective SOURCE and COpy relations connected w it hthe action types considered above. Realistically, however, application of the presenttheory will not go into much detail concerning the social practises.

H aving shown what kind of data are needed for the present theory, and howthey can be obtained, let us now consider the question whether all our prim itivesare really important. No argument seems necessary here for the notions of action.action type, individual, characteristic function, the status relation, and PERfORMand POWER. First doubts m ight occur w ith respect to INTEND. Intentions are anessential ingredient in our character ization of power (as well as in human beingsgenerally) because without intentions and the corresponding requirement A 17 b wewould be left w ith mere causal influence instead of power. This, in turn, woulddevaluate the use of POWER in determ ining the status relation via M, and leavethe status relation without link to the m icro base. So INTEND, in fact, is importantto our theory.A second doubt m ight arise for the superstructures. Omitting them would yield an


26/29

26 WOtFGANG BALZER

only together with internal representations. Second, superstructures are the carriersof education and ideology. Institutions typically get "fully expressed" only a gener-ation or more after their first appearance. This is so because for later generationsinstitutions are a "natural" part of the system. They get firmly impressed in the su-perstructures of the individuals by the process of education. As a consequence ofthis, thirdly, superstructures are crucial for the explanation of an institution's sta-bility. Without recourse to superstructures we simply could not understand why inmany cases the "lower" groups bear an institution for long periods. Though we donot focus on the explanation of stability in the present paper it is clear that our the-ory is able to provide such explanation, and that such explanation cannot be givenwithout the superstructures.The final items to be checked for importance are those occurring in the social

practises. It might be objected -that these are not only superfluous but even hinder-ing because they introduce an element which practically escapes empirical investi-gation. There are two reasons why we think that none the less social practises areessential in a theory of social institutions. First, (in the absence of legal or formaldefinitions) they provide the major means for an identification of social groups, andoften also of action types. In the example, the group of nobility even formally isidentified by genidentity. Second, as already mentioned in Sec. 3, social practisesprovide a general basis for specializations in which conditions of an institution's fitto its surrounding may be studied. The dynamical part of an explanation of why aparticular institution did develop and spread in a particular setting ultimately hasto refer to things like our COpy relations: why do individuals take up and stick tocertain kinds of behavior while they do not take up other kinds. Very roughly, wecannot ignore the immense historical depth of many of our most important socialpractises if we want to understand our most complex and important institutions.Turning now to an exemplification of our general view of the process of appli-

cation we have to ask whether a set of data as described above can be fitted witha model? We have to go through the various axioms, and see whether the datasatisfy them or can be shown to be embeddable into a structure satisfying them.Since the status relation is not represented in the data an existential claim has tobe made: there exists a hypothetical status relation which satisfies axioms A3, A4and A9. On the basis of the data in the example such a relation indeed exists. Wemay define it by setting /3 ..2 ..1 and /3 ..10 no other pairs of groups being re-lated by ...Clearly, .. is transitive, anti-reflexive, and has a maximal element asrequired in A3 and A4. Moreover, the full list of POWER relations available willverify--r at least be compatible with-A9. Most noble individuals exert powerover clergy and peasants, most individuals in /2 exert power over peasants, but nei-ther o C these quantitative relations holds in the other direction. A 2 and A7 can besatisfied conventionally, A6 stating that all individuals are involved in POWER rela-tions is satisfied in the data, and the same holds C o r the axiom A8 of adrnissibility.f?"It h not e>'Y to see how the colloctlon of do," 1\.,. to proceed", that AR will come out Ialse. Bruically,lhe data about the character lst!c tunctlon ",ill be obtained by observing many performances of differentactlons, end use tbese as a besL! for abm.Cling group' 8nd cbarncter ist ic action Iype.,. Therefore aperformed action ha, 10 be to an outlyer in the ".ti"lcnt sense in order to conf lict with the requiremenlof beinR admi. ..Ihle.


27/29

t

A BASIC MODEL FOR SOCIAL n-snronoss 2 7Altogether the axioms for macro core and micro base come out true. The axiomfor proposition ~paces (AIO, 04-1) is rather idealized, and subject to doubt, but notvery essential for the overall claim that the system is a social institution. The otherparts of superstructures lire represented in the data only very partially. So they haveto be completed in a hypothetical way. It is not difficult to find a set of hypotheticalsuperstructures which satisfies AlO and All.Similarly, the axioms for social practises have to be satisfied essentially in a hypo-

thetical way, by referring to hypothetical entities extending the few available data tothe full structures required. There remain the two central axioms of 09. The contentof the first axiom, AI6, is that to tvery admitted pair there exists ("we can find") a"corresponding" social practise in which the pair is anchored. Consider for examplethe group of noble individuals and the action type of huntingJ3. Clearly, hunting isa social practise even though it is impossible to specify the complete sets of individ-uals, actions. and the SOURCE and COpy relation. There must be historically firstevents of hunting and there is a tradition in which the techniques are inherited. Itseems realistic to consider one of several different social practises 'here which m~yhave been invented independently of each other in different periods and differentregions. Anyway, by combining sparse historical data with the given admitted pair.it seems possible to claim that there exists some social practise into which thesedata can be embedded. The same holds for the other admitted pairs-with varyingdegree of plausibility, The axiom for the POWER relation, AI7, finally seems to besatisfied as far as the available data are concerned, We see no problem in addinghypothetical entities at places where data are lacking (as for instance data "boutrePk(a) in AI7-e so that the axiom comes out true, Altogether, we think the claimthat the system considered is a social institution can be seen to be correct.

In our example we have social practises common to nil the groups involved, forinstance the holy communion (as long as the village is small enough and nobilitydoes not celebrate separately), One might suggest that such practises are irrelevantfor they do not serve for any differentiation. They play an important role, however,in the internalization of the different types of actions and the characteristic func-tions and thus mny be quite essential for the institution in question. Our examplealso shows that POWER relations may exist "from bottom to top", The priest, forinstance, by instructing the countess appropriately, may exert power over the count.Such mutual relations of POWER suggest to apply some notion of equilibrium tothe net of POWER relations. Systems closer to equilibrium, so the correspondinghypothesis, are more stable over time.

Let us finally turn to questions of explanation, There are two basic notions ofexplanation, The first notion, called the instance V;(W of explanatiorr'", deals withexplanation of more complex entities, like sets of data, or laws, Such an entity isexplained by successfully applying to it a theory in the way described above. Expla-

"The r.ct that onty the grown up, male Individuals engage into actioru or th~t type doe> not yieldincorwis tercy wilh the axioms of 8dmi"ibilily for the Jailer are only necessary condit ions of performance.


28/29

28 WOLFOANO BAUER

nation thus am ounts to system atizing com plex data into one com prehensive patternor "whole", to see how the data fit together in some particular way. In this sense ofexplanation our theory explains sets of data w hich are available about social system sof the kind Intended. It explains, for instance, the set of data described before. Inmore realistic terms we may say that the theory provides a consistent picture orpoint of view from which all the actions and relations observed fit together andmak e sen se.The second notion of explanation is that of deductive nom ological explanation. It

aim s at explaining an atom ic proposition ("a fact") by means of deducing it C romthe theory plus appropriate initial conditions. C learly, the latter type of explanationis just a special case of the former. D educing an atom ic sentence from Initial con-ditions is a special case of show ing that the set of both is explained as an instanceof the theory. A ccordinglo"1he d~ dtl~tive nom ological view V arious explanations ofconcrete behaviour can be given in the present theory. W e can explain power re-lations in ierm s of Intentions, perform ance and causal belief, we can explain singleactions in terms of power, we can explain Intentions and even causal beliefs in thesam e w ay. W e can explain statistical differences in exertions of pow er am ong differ-ent groups, and so on. W e may explain, for instance, why groups of peasants obeythe court's order under conditions in which they could easily overcome him , andin which execution of the order is rather unpleasant for them . Or we may explainwhy the count's children are educated In a way utterly different from that of thepea sants ' c hild ren.In order to obtain more comprehensive. far reaching, or critical explanations of

social phenomena the theory either has to be joined w ith other social theorlesls orto be further refined. Joining it w ith some form of decision theory we m ight ob-tain deeper explanations of "subordinate" behavior in terms of adm issibility. Thebasic Intuition here is that possible "subordinate" actions or reactions to be evalu-ated in a model of decision theory in an institution are constrained by the frame ofadm issibility. In the decision m odel relative to an institution the subordinate agentconsiders and evaluates only alternatives which are adm issible, so her set of actionalternatives is severely narrowed down in comparison to what would be feasiblein the absence of the institution. On the basis of this restricted set of alternativessh e ch oo ses rationally, I.e. as d escrib ed by the decision model, but the action cho-sen m ight look quite irrational if the institution would be left out of consideration.By further refinement, on the other hand, we can achieve a real alternative to thegame theoretic account of how and why Institutions emerge. The basic "mecha-nism" is present in the models already: Social Institutions emerge as the result ofnew ways of exercising power which are invented and found to work successfullyin favour of the superordinate agents. O ften, the full final pattern of actions andreactions develops from one single new action type which is invented as a new wayof exerting power. Therefore it Is riot necessary to see the emergence of an insti-tution as the introduction In one step of a whole finished pattern of action types.The pattern itself may develop in different possible ways (e.g. by trial and error))Itn line ..,tlll the thais or unity of the social scknc:es oC\"n put fOfWII,d by gn:", scbolar s, See Brsudel


29/29

A BASIC MODEL FOR socrxt, I~STInmO~Sas reaction to just one new action type. Once the resulting pattern gets stable theinstitution originates and grows In the interest of the groups in the "upper part" oftheir core structure. These groups therefore are interested In having correspondingsuperstructures built up In the other individual~d that is why institutions remainrelatively stable even when the conditions favourable for their emergence are gone.

REFERENCESAxelrod. R. (1984) Tht Evolulu", of CboptrtJl/c". New York: Basic:Books.Balzer. W. (1986) Tbeorettcal terms: A nc;wperspeetbe, Tht Joumlll of 1'1lilosop/ry SJ: 71-90.B~ller. W. (1990) The struc1urallst view or measurement. to appear inMi""uO(/J Snufiu in lilt Phi/mop/'fof Sc~"Ct.Balzer, W~ MouUnes. C: U. and Soeed, J. D. (1987) A" ArdtiltCfOflic for Science, Dordreeht: Reidel.Braudel, F. (1980) Ort History, Chicago: UP.Berger, P. L.. and Lockmann. T. (1966) n,t Social Connrualo of Rt,,/il},. New York: Doubleday.Burt, R. S. (1980),Models or network structure. Aftft. Rev. Sociol. ,: 79-141.Dah], R. (1957) The concept or power. Bt/lIsviourtJ! Scit"Ct 2: 201-1S.EvwNPrhchard, E. E. (1937) Wilcn=f4 Omcles "ftd MaKic A"'0fI8'/lt Az.",dt, Oxford:-Clarendon Press.

Fararo, T. 1.. and Skvoretz. J. (1984) INtitulioN IS prodoetlon systems. JOUf1ldl of Mdlnt",alical Sociolor>'10: 117-\81.Flap, H. (1985) Conlfiet, loyaliteit en geWllld,Dissertation. University or Utrecht.Forge, J. (1986) The iNtana: theory or explan:lllon, AWlrtJlasiM loumel of Plli/osoP/'}' 64: 127~2.Graetzer, G. (1971) Lastlce n,tory, San Francisco: Freeman &. Co.Lewis, D. K. (1973) COU,,'trfdclUalJ. Cambridge. MA: UP.Lukes. S. (1974) P~" A Radical Vit London: Macmitlan.Mac:1dc,J. 1.. (1974) T h Ctmt'" of Int Univene , Oxford: Clarendon Press.March, J. 0.and Slmon. H. A. (19S8) Or-gcniza.icru. New York: Wiley.Meynard Smith, J. (1982) EWJlul;"" d"d Ilrt n.tory of a_u. Cambridge: UP.Panons. T. (19S1) Tht Social SySltm. Olencoc, It.; Free Press.SchiITer. S. (987) Rtm"/J1t1S of Mt/J1t;"8. CambrldJe, MA: UP.Sebouer, A. (1981) T h Econom;c 71ltory of SocW fflSl;IUI;/HU. Cambridge: UP.Scott, W. R. (1981) Orr/J1tiUllicru: RIJI;ond/, "'dlUral, dftd Opt" Syrrtml. Englewood Clilr, NJ: PteruiceHall.Suppes. P.(1970) A Probdbilislic 71.(ory ofCawa/iIY. Actl Philosophic. Fennica, Amsterdam.Thylol. M. (1976) A""rtlry Md Cooperation, London: Wiley.Thlbaut, J. W. and Kelley. H. H. (1959) n,t Sociat PS)'Clro/"8Y of Groups, New York: Wilc:y.W.rtenbers. T. (1968) Tbe rorms or power, A""I)~( und !Vilile 10: }-J1.

Weber. M. (1980) lV;mcnllft uftd Gesellschoft, Tilbingen: Moht.

r

29

Balzer - A Basic Model for Social Institutions

Documents

Transcript of Balzer - A Basic Model for Social Institutions