 |
| FIGURE 5.1. Example of how sub-concepts are related to a higher-level concept in a cognitive map. |
| RobertCutler.org » Site Map » [ Epistemology | Decisionmaking ] » This document | 1982 |
| C.v. | Resumé | Contact author | Search | “Top 10” Lists |
[ page 91 ]
The purposes of this chapter are to assimilate the cognitive map and operational code (OC) approaches to a mutually compatible framework, and to demonstrate why this is useful. The framework for the synthesis is conceptual dependency analysis, a cognitive model that systematically represents the existence of a hierarchy of levels of abstraction in thinking. Because the cognitive-map and OC approaches can each be associated with a distinct level of human information processing, conceptual dependency analysis can provide the basis for combining the two approaches. The utility of such a combination is illustrated by an example that applies the synthesis to the prediction of a policy decision.
The first part of this chapter summarizes the salient features of the conceptual dependency framework. The next part discusses its relationship to the cognitive-map approach on the one hand, and to the OC approach on the other. An integrated notational system permits a given cognitive map to be transformed in such a way that higher-level cognitive structures immanent in it, analogous to operational codes, are revealed. Simulation techniques, originally developed within the cognitive map framework, are also adapted to the unified framework that is based on the conceptual dependency approach; it thereby becomes possible, when the OC immanent in the cognitive map is made explicit, to specify how the OC may constrain relationships predicted from the simulation based on the original cognitive map. The logic of such a simulation is developed for one class of OC in particular. The third part of this chapter applies the proposed framework to a particular example, drawn from this writer's research into Soviet foreign policy formation, and
[ page 92 ]
evaluations the results. The concluding part of the chapter assesses the pragmatic value of the new method, suggests how it may be refined, and explores some of its implications.
The conceptual dependency framework, as formulated by Abelson (1973, pp. 287–339), posits six levels of conceptual abstraction in human information processing (i. e. six levels of complexity in the structure of belief systems). Specific rules govern the mutual relations of units at every level, and the operation of those rules at each level produces the units at the level next highest. Abelson's six levels are:
Heradstveit and Narvesen (1978, p. 84) note that both cognitive maps and OCs address the same 'dependent variable': policy preferences. In Abelson's scheme (1973, p. 295) a policy preference, being a means that is designed to accomplish a desired end, corresponds to the plan level. It seems clear that the cognitive map approach corresponds to the analysis of plans 'upward' from the element, atom and molecule direction, while the OC approach attacks the same problem 'downward' from the theme and script
[ page 93 ]
direction. In order to operationalize a synthesis that would integrate cognitive-map techniques into a conceptual dependency framework, and in order to understand higher-level structures in that framework as OCs, the conditions that Abelson posits for the existence of molecules and plans must first be summarized.
Recall that there are three kinds of atoms: an (A) atom, for 'action', is formed by an actor–action bond plus a specification of the action; an (S) atom, for 'state', expresses an outcome condition of some actor or thing, in relation if appropriate to some other object or thing; a (P) atom, for 'purpose', is formed by the action 'want' by some actor plus a specification of the state wanted. Abelson specifies three conditions for the existence of a molecule formed by these atoms:
Only on the molecular level does causation appear in conceptual dependency analysis: a (P)−(A)−(S) molecule is 'an action undertaken to attain a goal desired by the sponsor of the action'. In conventional cognitive map notation, with causation flowing from left to right and ending in a utility concept, such a molecule would be written
(A)—(—)—>(S)—(+)—>(P) .
In conceptual dependency analysis a plan is a molecule that consists of alternating (A) and (S) atoms, following an initial (P) atom:
(P)—(An)—(Sn)—(An−1)—(Sn−1)— … —(A1)—(S1)—(A0)—(S0)
such that four rather intuitive formal conditions are satisfied:
[ page 94 ]
It should be noted that an Abelsonian plan is only an extended molecule, which can however be a more complex network involving branching structures and loops. If two simple molecules are interrelated in the manner specified by the definition of a theme, then they are functionally indistinguishable from plans. A conceptual-dependency plan that is limited to causal relations is nearly isomorphic to a cognitive map.
To test the utility of the approach being outlined in this chapter, we should:
This part of the chapter establishes a systematic procedure for executing the third and fourth of these steps.
There are two notational difficulties in assimilating cognitive map methodology to the epistemological field defined by the conceptual-dependency hierarchy. Both difficulties concern causation. Their significance is directly proportional to the degree to which the epistemology of cognitive methodologies figures in one's project, and inversely proportional to the degree to which one's project is to operationalize a synthesis of those methodologies. This chapter sets for itself the latter task and resolves the difficulties on the criterion of notational parsimony. Readers who are stimulated to the former project by the empirical results tentatively presented here, may find Roig (1977, esp. pp. 87–109) a fertile point of departure for an epistemological critique.
[ page 95 ]
The first difficulty is that a simple isomorphism between the variables in a cognitive-map string and the atoms in a conceptual-dependency plan is specious. A cognitive map begins with a 'policy variable', ends with a 'value variable' (which is formally associated with the utility concept), and is otherwise composed of intermediary 'cognitive variables'. Directly to equate policy variables with A−atoms, cognitive variables with S−atoms, and value variables with P−atoms, would lead to the projection of policy variables into intermediate positions in cognitive map strings. For example, if the conceptual-dependency plan
(P)—(A1)—(S1)—(A0)—(S0)
were transformed into a cognitive map according to the supposed rule, then it would begin with a policy variable (the analogue to (A1)) and terminate with a value variable (the analogue to (P)), but it would include a policy variable (the analogue to (A0)) in an intermediate position.
The second difficulty is that conceptual-dependency plan-structures can be hidden within cognitive maps. For example, if the concept 'forces of counterrevolution' subsurned those of 'haute bourgeoisie', 'high-ranking civil servants', and 'monopolistic circles', then these relationships would be portrayed by a cognitive map as in Figure 5.1; but in the absence of specification of the three antecedent concepts, under conceptual dependency analysis 'forces of counterrevolution' would be the actor with the plan.
|
[ page 96 ]
In order to resolve these difficulties together, we may propose to categorize the variables in a cognitive map into 'chunks', this name deriving from the use of the term 'chunking' in psychology to connote the representation of large amounts of information by concepts requiring relatively small storage space. (Acronyms are an everyday example.) Any categorization of cognitive map variables into chunks should be based on concepts that derive from cognitive map methodology: in particular, on the distinction among cognitive variables, value variables, and policy variables. Within each of these categories, topical classes of variables (e.g. political, economic) may be distinguished; within each such class, chunks of variables to which the relevant actor attaches positive affect may be distinguished from chunks to which negative affect is attached.
Table 5.1 gives an example of this chunking procedure, taken from the example presented later in this chapter, concerning Soviet policy toward the 1974 revolution in Portugal. The assignment of concepts to categories, as well as their arrangement along hardline-moderate or 'intensity' continua, is made by this analysis on the basis of familiarity with the evolution of the Soviet policy debate on the situation in Portugal. A reliability check was not possible but would be desirable in future studies.
According to Abelson, a theme in conceptual dependency analysis is determined by the interdependence of two actors' plans, the nature of the theme depending on the actors' role in each others' plans. Let us consider that the causal link between two chunks is of the same direction and sign as the link(s) between their constituent variables; links between variables in the same chunk do not appear on a 'chunked map'. Such a higher-level map, constructed out of the chunks from a simple cognitive map, is analogous to the theme-level structure in conceptual dependency analysis. In order to distinguish conveniently between conceptual-dependency plans and the analogous structures appearing in chunked cognitive maps, the latter may be called cognitive plans. The relationship between two cognitive plans may then be said to define a conceptual-dependency theme.
In general it is desirable to regard chunks as bonded (S)−(A) atoms. The two exceptions to this rule involve chunks of policy variables at the beginning of a cognitive plan network, where they may be considered bonded (P)−(A) atoms, and chunks of value variables at the end of a cognitive plan network, where they may be considered (S) atoms. This proposed correspondence, to which idiosyncratic counter-
[ page 97 ]
|
||
[ page 98 ]
examples may exist but which is itself the inductive fruit of empirical analysis, permits a two-fold simplification:
The theme immanent in a given cognitive map may then be constructed by first separating 'positive' and 'negative' chunks into mutually distinct cognitive plans, and then connecting chunks within and between plans according to the connections between their member-variables in the original cognitive map (omitting connections between variables aggregated in the same chunk). research has revealed that two fundamental dimensions of the OC construct may in fact be designated image-of-self and image-of-other (e. g. Heradstveit, 1979; for another typology, cf. Holsti, 1977). If the actor whose cognitive map we are examining sees himself in the role of one of the actors with a cognitive plan, then that actor, in the thematic analogue constructed out of cognitive plans, may be labelled 'Self'. The other actor may be labelled 'Other'.[2] Since OC studies explicitly using image-of-self and image-of-other categories most often dichotomize each of them (see the review in Jönsson, 1982, for example), it should be possible to map the set of themes enumerated by Abelson (1973, pp. 318–21) into a two-by-two matrix, the dimensions of which are dichotomous image-of-self and image-of-other variables. The temptation to suggest how this might be done is irresistible; Table 5.2 is such a possible classification.
The OC approach as traditionally understood plays no direct role in the simulation of decision making: as Heradstveit and Narvesen have pointed out, the OC approach 'does not examine the process leading up to political choice', but holds only that 'beliefs will constrain the way in which events will be explained and choices made' (Heradstveit and Narvesen, 1978, p. 86, emphasis added).
[ page 99 ]
The structural analogue of the OC in conceptual dependency analysis, the theme, has a similar—but more formally modelled—function. In order to appreciate this, it is instructive to stop a moment to ask how we know what level we are on to begin with. As Alker (1975, p. 198) remarks, following Abelson, rebellion is both a theme and a script. To take a more radical example, 'John loves Mary' could be either an atom or a theme: perhaps even both an atom and a theme in the same script! When Abelson discusses the romantic-triangle script, in which (as two initial themes) John loves Mary but Jim does too, he notes that an outcome yielding mutual antagonism and finally conflict themes between Jim and John are likely to be unavoidable if all three actors share this same script. He emphasizes that such scripts are constructed from cultural examples through socialization and comments that 'the whole topic of the ways in which thematic content can influence script structure needs careful study' (Abelson, 1973, pp. 329–30). But since the definition of a theme in conceptual dependency analysis depends on the ways in which the atoms (chunks) of different (cognitive) plans are inter-related, decision making in this particular information-processing approach is really about the wide-ranging influence of some not very numerous linkages at relatively low conceptual levels.[3]
It is impossible to treat all conceptual-dependency themes in this chapter; let me instead give special attention to the one that appears in the illustrative example
|
[ page 100 ]
below, which is in fact not all that uncommon a theme in political analysis: conflict. That the goals of the actors involved in the conflict theme are mutually opposed, requires us to give special attention to lower-level linkages in the structure of this theme. When the value-variable chunks in the two actors' cognitive plans are mutually exclusive and directly connected by a negatively signed causal link, then the ability of one actor to prevent the other from gaining the upper hand depends upon their relative strength. This may be independent of the intensity with which each actor pursues his goal, as measured on an absolute scale tailored to the respective chunk. For example, if 'self's' goal is to intensify a revolutionary movement and 'other's' goal is to intensify a repressive state, then we may suppose that the latter will inhibit the former only if its relative strength remains greater. A sort of dialectical logic is in evidence here. Formally expressed, the relevant relationship is portrayed in Figure 5.2, where f and g are functions which standardize their arguments.[4]
|
It is possible that a change in some antecedent chunk may lead to the crossing of the threshold expressed in Figure 5.2, and so to the reversal of direction of the negative causality between the two actors' goals. The particular form of dialectical logic relevant to this conflict theme permits the intensification ('progress') of history to affect the intensity both of the revolutionary movement and of the
[ page 101 ]
repressive state, each measured on its own independent absolute scale. During a period when the repressive state is dominant, we would graphically express these relationships as in Figure 5.3.
|
However, if the intensification of history should result not only in the intensification of the revolutionary movement but also In the debilitation of the repressive state, as posited by the relations portrayed in Figure 5.3, then in the end the balance of forces between these two opposing goals (representing the opposing value variable chunks in the conflict theme) will be reversed, yielding Figure 5.4.
|
[ page 102 ]
This low-level linkage, idiosyncratic to the conflict theme, must be taken into account in the theme-level simulation that is meant to predict the constraints of such an OC structure on the lower-level cognitive map simulation. To execute the higher-level simulation, we need to do two things: first, to adapt by analogy the established cognitive map simulation procedure (Nozicka, Bonham and Shapiro, 1976) to this higher level; and second, to establish rules for modifying the prediction from the original cognitive map simulation according to higher-level constraints, in the event the direction of the inhibition (negative causation) between the goals of the opposing actors in the theme actually reverses in the higher-level simulation, as for instance in the manner described above. The first of these tasks may be performed as follows:
Then the second of these tasks requires us to do the following:
The purpose of this part of the chapter is to apply the proposed method of higher-level simulation to a well defined
[ page 103 ]
decision space and to compare the performance of the resulting 'constrained' prediction against the prediction emerging from a lower-level simulation based only on cognitive mapping. The case is drawn from this writer's other research (Cutler, 1982b). It concerns Soviet attitudes and policy toward the revolution in Portugal in 1974. A bit of background is necessary.
The Portuguese army had fought a stalemated anti-guerrilla war in Africa since 1960. On 25 April 1974 a widely based group of rebel army officers calling themselves the Armed Forces Movement seized control of the government in a bloodless coup which found widespread widespread popular support, and a National Junta of Salvation was formed. The in late May a series of strikes swept Portugal: in the textile factories, on the subways, buses and trains, and in the shipyards. The Portuguese Communist Party (PCP) had to decide whether to encourage the unrest or to restrain its natural constituency. (The best day-by-day chronology of the revolution in Portugal is: Sobel, 1976.)
It is useful to distinguish several consecutive phrases in Soviet commentary leading up to the decision. During the first week of the month, debate in the Soviet press centred on the question whether the revolution in Portugal was due primarily to international Portuguese political and social conflicts or ot the failure of the colonial wars in Africa. During the second week of May, that debate evolved into a controversy over how the revolution might best be consolidated, and two opposing positions began to emerge: one came out in favour of promoting 'left democratic unity' in the country; the other tendency, which gave primacy to colonial and other international conditions, stressed the necessity of 'preventing a counter-revolution'.[5]
In the third week of May, the proponents of the latter view specified that such a counter-revolution would be launched with the aid of international fascism at the service of monopoly capitalism, which was therefore the primary enemy. The adherents of the other tendency explained that support for 'left democratic unity' meant that the struggle against international imperialism was secondary to the domestic antifascist struggle in Portugal, and that the PCP's efforts in this latter regard deserved principal attention. At the end of the third week of May, a[ page 104 ]
compromise between the two tendencies seemed to emerge, emphasizing both the 'political front' (meaning PCP mass propaganda work) to counteract that threat of a counter-coup and the need for mass vigilance against 'the monopolies, [which] still hold the real economic power in the country.'
With the beginning of the fourth week in May, the Portuguese state—and the PCP with it—dentered the crisis, exacerbated by strikes that resulted from economic and political agitation among the workers. At just this time the Soviet journalist Oleg Ignat′ev arrived in Lisbon.
Under these conditions, the Soviet compromise began to break down; however, tkat compromise outlined the bounds of permissible debate. The harder line reappeared in Portuguese-language broadcasts to Portugal and Africa as well as in signed and unsigned TASS reports. The most articulate and concise example of this tendency of articulation is an anonymous Portuguese-language commentary, broadcast from Moscow to Portugal. The more moderate line was reflected in Ignat′ev's initial dispatches. Figures 5.5 and 5.6 are, respectively, the anonymous Soviet radio commentary in Portuguese beamed to Portugal and Ignat′ev's 25 May article (FBIS, 1974a; Ignat′ev, 1974). These two sources represent the Soviet policy tendencies that reappeared after the compromise line had disintegrated under the pressure of events. a flurry of commentary, as the May crisis entered its acute phase, Soviet as well as PCP press organs became suddenly silent. It is possible that Ignat′ev and the PCP leaders consulted during these several days. On 29 May the silence was broken by a communiqué issued in the name of the Executive Committee (Politburo) of the PCP Central Committee (FBIS, 1974b) its cognitive map is portrayed in Figure 5.7. Portuguese and Soviet Communists communicate directly with one another, have a common conative environment, and share cognitive predispositions, the combination of Figures 5.5 and 5.6—that is, the set-theoretical union of the relations composing the cognitive maps—would represent the analytical structure shared by the PCP and their Soviet comrades.[6] That union map may be used as the basis for one set of predictions about thb tactics that the PCP would choose to deal with the crisis on hand
[ page 105 ]
|
[ page 106 ]
|
[ page 107 ]
|
[ page 108 ]
|
[ page 109 ]
at end of May. Figure 5.8 is the union map of the two tendencies.
In Figure 5.8 the cognitively most central concept—the one to be 'highlighted' in the cognitive-map level simulation—is the issue confronting the PCP, that of 'the disruption of the economy.' The task is to inhibit this variable (i.e., the create negative causal effect upon it). A search of the decision space defined by Figure 5.8 reveals that new paths with negative effects on 'the Caetano regime' or on 'the provocations of reactionary forces' satisfy this imperative. Although the former concept is an element in the present cognitive glossary, it is a past event and cannot be manipulated in this way. Since no strategy exists, in the map itself, for creating a negative causal effect on the concept of 'the provocations of reactionary forces', the development of such a strategy becomes the actor's task. Following the traditional emphasis on centrality in cognitive-map simulation, we see that the cognitively most central concepts (after the two already considered, those of 'disruption of the economy' and 'the provocations of reactionary forces') and 'reinformcement of (left) unity of working-class and democratic movement', 'consolidation of alliance of popular masses with armed forces', and 'the process of democratization'. We may hypothesize, then, that the cognitive system portrayed in Figure 5.8 would respond to the defined problem by innovating strategies that can be represented as paths beginning with members of this triad and ending, with overall negative effect, on 'the provocations of reactionary forces'. The concepts involved in these paths are (9), (13), (14), and (18). We will want to add to our prediction any links among these concepts that appear in Figure 5.8, the union cognitive map of Soviet tendencies. Figure 5.9 portrays the ensemble of these relationships: the three links ending on (9) are the predicted innovated paths, and the two links ending on (18) are supplemented by the union map.[7]
The higher-level simulation proceeds in a manner analogous to that at the cognitive map level, as described at the end of the second part of this chapter. Figure 5.10 is the chunked union map (i.e. it is Figure 5.7 represented at
[ page 110 ]
|
the higher level of information processing). The chunk with the highest mean centrality over its member variables, NEVV, is highlighted (step 1). The antecedent inter-chunk path runs backward through the NPVV chunk to the PPV:CU chunk of policy variables (which is cognitively more central than the PPCV chunk, also antecedent to the NEVV chunk). The antecedent path on the chunked map, therefore, is
PPV:CU—(−)—>NPVV—(+)—>NEVV ;
there are no consequent paths because the originally highlighted chunk is itself composed of value variables. Therefore the search for relevant policy options is confined to the PPV:CU chunk (step 2).
In order to achieve the desired value (diminution of 'disruption of economy'), the most intense policy variable within the PPV:CU chunk—'consolidation of alliance of the popular masses with the armed forces'—is chosen (step 3). Because the PPV:CU chunk is intensified, the strength of the PPVV chunk relative to the NPVV chunk increases (step 4). This leads to the reversal of the direction of causality between them (step 5).
The constraints that this reversal effects upon the predictions from the cognitive map level, portrayed in
[ page 111 ]
|
[ page 112 ]
Figure 5.9, is as follows (step 6). Because the (18)−(−)−>(9) relationship does not appear in the union map of Soviet tendencies (Figure 5.8), and because concept (16) is the most intense member of the PPVV chunk, the (18)−(−)−>(9) link is over-ridden by the (16)−(−)−>(9) . (Because the (13)−(−)−>(9) link does appear in the union map, it survives even though the (14)−(−)−>(9) link provisionally overrides it, (14) being the most intense concept in the chunk containing (13).) This constraint, expressed as a fragment of a cognitive map, is portrayed in Figure 5.11.
|
The combination of Figures 5.9 and 5.11 appears as Figure 5.12. There, the constraint derived from higher-level processes, expressed in Figure 5.10, is applied against the lower-level prediction in Figure 5.9. Figure 5.13, which is the actual policy result, is a fragment taken from Figure 5.8. It will be seen that Figures 5.12 and 5.13 are nearly identical. The constrained prediction is clearly better than the simple cognitive-map prediction, but it is hard to say how much better. The remainder of this part of the chapter presents a metric with which we may be able to say how much better it is: a way to measure distances between cognitive maps.
The Index of Relationship Similarity (IRS), developed by Bonham et al. (1978, p. 103), is not quite satisfactory for
[ page 113 ]
|
|
[ page 114 ]
the desired purpose. According to Bonham, the IRS between two maps is calculated only on the set of direct and indirect relationships that they have in common, and it is sensitive only to relationships that appear with different signs in both maps' valency matrices (and exponentiated valency matrices).[8] This means that the IRS is indifferent to the size of the two maps, and thus also to the number or relationships that they do not share.
The work of Grofman and Hyman (1973) concerns systems of belief that involve subjective probabilistic judgments. They develop not only a measure of distance between belief systems but also measures of the connectedness, reliability, congruence, and polarization within a single belief system, all formally defined. Enterprising psycho-logicians should be able to adapt, more systematically than is done here, such measures to systems where, as in cognitive maps, there are no assignments of fractional numbers to the association between two concepts. Such work would be extremely valuable for the systematic development of research using psychocartographic techniques.
The measure of distance between cognitive maps, proposed here, is related to Grofman and Human's measure of distance between subjective probabilistic belief systems, but it is also related to their measure of the connectedness of a single belief system. The basis for this measure is a determination of the number of direct and indirect relationships shared by two cognitive maps, as a proportion of all direct or indirect relationships in either map. It is defined on a unit metric from 0 to 1, with 0 representing the identity of the maps and 1 representing their lack of any relationship in common. Formally expressed, the measure is:
where [VX] is the valency matrix of cognitive map X; N is the number of concepts in cognitive map X; F is a function which counts the number of non-zero entries in a matrix; and S∪T is the union of the set of relationship composing map S with the set of relationships composing map T, under the condition that a relationship is defined as non-
[ page 115 ]
|
[ page 116 ]
zero if it appears in both maps S and T but with different signs.
Using this metric, the distances among the various predictions may be calculated, and their matrix appears as Table 5.3.
Although the union map of the two Soviet tendencies is treated as if it were the cognition of a single actor, this does not necessarily imply that Soviet foreign policy decisions can always be considered as rational-actor calculations. One of the reasons for using the union-map technique here is that we lack interview data that would permit the application of more advanced techniques for the analysis of collective decision making. The possibility of discerning a 'master script' that guides the resolution of Soviet foreign policy debates is not denied. However, as Abelson (1973, pp. 333–7 passim) notes, the plan and theme levels can accommodate significant changes in content while the script remains unchanged. In such an instance, knowledge of the script alone would be insufficient for making predictions about individual Soviet foreign policy decisions. For this reason, the totalitarian model of Soviet politics tends to be unsatisfactory for the analysis of individual Soviet foreign policy decisions.
Much confusion has arisen, in the study of Soviet foreign policy, from the lack of distinction between two levels of abstraction: philosophical attitudes on the one hand, and specific policy positions on the other. These represent, in fact, two distinct levels of information processing. The cognitive-map method is appropriate to the 'interest group' approach (Skilling, 1966), and the OC method is appropriate to the 'tendency of articulation' approach (Griffiths, 1971, 1972). These methods are also appropriate to the investigation of the role of the press in the organizational process of Soviet foreign policy formation, and they offer an avenue for systematic investigation of collective decision making. It is worthwhile to combine these approaches, and the framework presented in this chapter may possibly be one way of doing that. Distinguishing between philosophical attitudes and particular policy preferences will contribute to eliminating the confusion between the 'interest group' and the 'tendency of articulation' approaches to the study of Soviet foreign policy formation. (See Cutler, 1982a, for a further
[ page 117 ]
discussion of these issues.)
Several features of the simulation deserve further study. One of these concerns the categorization of concepts into chunks. Some rules for a typology of chunks have been suggested here, but what could be the rules for classifying concepts into them? A similar observation could be made about the notion of an intensity scale within all chunks. Although it is intuitively plausible that such scales may exist, how are they to be constructed? Universally applicable rules for such tasks can be elaborated no more easily than a universal grammar can be created; even if such a system could be constructed, capable of regulating every possible case we might conceive, still we would not know and could not know whether that system 'naturally' existed a priori to our construction of it. So long as the simulation concerns only a single case or small number of cases, the construction of chunks and their intensity scales can only be based on an internalized contextual understanding of the situation. This fact does not exclude the possibility of inter-coder reliability checks. Multiple case studies, closely related and bearing on the same cognitive space, may be able to validate inductively constructed categories.
On the operational level, one can only say that the framework needs further work. The logically complex but important relations of enablement and gating, from conceptual dependency analysis, need to be incorporated. That will exacerbate the incompatible marriage of the mathematical model of logic (cognitive mapping) with the linguistic one (conceptual dependency analysis). The offspring of this marriage is already a problem child that requires better defined rules of conduct. Such issues are therefore best addressed in the context of further formalization of the model. Here I have only tried, first, to suggest the possibility of integrating, within an analytical framework informed by the existence of multiple levels of abstraction in thought, two approaches to the study of cognition and choice that already exist in the discipline of political science; and second, to apply that integrated approach to the empirical analysis of one sort of cognitive activity that is related to the behaviour of actors in world politics.
I wish to acknowledge the support of the Albert Gallatin Fellowship in International Affairs at the Graduate Institute
[ page 118 ]
of International Studies (Geneva, Switzerland), without which neither the ideas in this chapter would have been formulated, nor the paper on which the chapter is based, prepared. Thanks are due to Professor Christian Dominicé, Director of the Graduate Institute, and to Professor Urs Luterbacher, Director of its Center for Empirical Research in International Relations, for facilitating both this work and its presentation at the 1980 Joint Sessions of the European Consortium for Political Research. For sharing and criticizing the ideas expressed here, I am indebted to Pierre Allan, Robert Axelrod, Charles Roig, and Michael Shapiro, who are responsible only for my awareness of remaining problems and not for the existence of these.
[Note 1]. The lists in this part of the chapter draw heavily on Abelson (1973, pp. 294–89, 305).
[Note 2]. A potential problem is that a given situation may involve more actors than 'self' and 'other'. Operationally, this could be resolved by constructing multiple themes from the cognitive plans of all possible dyads, for several themes may exist simultaneously in the context of a script. This issue deserves closer examination but may not be too crucial, in view of Abelson's observation that 'plan rules allow an agent to be absorbed, as it were, within the molecule of another actor' (Abelson, 1973, p. 321).
[Note 3]. The issue, on which level such non-causal relations as enablement and gating may operate, also requires further study. Rosenstone's (1974, esp. pp. 16–19) experiment on changes in cognitive maps over time suggests the inference that such relations may operate at the cognitive map level, and Alker's (1975, pp. 194–5) operationalization of conceptual dependency analysis redefines an 'intentional action molecule' as 'any collection of atoms bonded in a permissible fashion by molecular bonds: positive or negative, intentional or unintentional varieties of action, causation, enablement, or equality ties', which list 'includes, but goes beyond causal relations'. It may turn out that the answer depends in part on how the individual research problem is approached; the level of abstraction on which an investigator who uses conceptual dependency analysis finds himself, is not independent of how he goes about establishing intersubjectivity with
[ page 119 ]
the entities whose consciousnsesses he is analysing in the particular instance.
[Note 4]. I leave it to logicians and linguists to explore the assumptions emboded in this manner of presenting the issue. I have been influenced by Elster (1978, esp. chaps. 4–5). In the present instance, the union (see Figure 5.8) could be considered the definition of the bounds of permitted discourse in published commentary.
[Note 5]. Fuller analysis of this debate, and references to it, may be found in Cutler (1982b).
[Note 6]. Although no figure in this chapter is a complete and exhaustive map of its respective text, each portrays the relations among the concepts which they all share as most central. No relation involving any of these concepts has been omitted from any map; the networks are complete and have only been isolated from other, peripheral networks that are unconnected with the ones portrayed, which are the principal ones most important to the analysis.
[Note 7]. None of the newly predicted path either antecedent to or consequent from the highlighted concept; no antecedent path in fact led back to a policy concept. The simulation technique as operationalized here, in contrast to methods applied in previous studies, allows for path innovation within the decision space defined by the cognitive map. See Bonham et al. (1978, p. 96) who combine pre-crisis and post-crisis maps in order 'to replicate adjustments that policy officials make … to accommodate perceptions of new events'.
[Note 8]. The formula for the IRS between two cognitive maps S and T is:
where [VX] is the valency matrix of cognitive map X; N is the number of concepts in cognitive map X; F is a function which counts the number of non-zero entries in a matrix; S∩′T is the union of the set of relationships
[ page 120 ]
composing map S with the set of relationships composing map T, under the condition that a relationship is defined as non-zero if it appears in both maps S and T but with different signs; and S∩T is the intersection of the set of relationships composing map S with the set of relationships composing map T, under the condition that a relationship is defined as zero if it appears in both maps S and T but with different signs.
I wish to thank Professor Bonham for his help in clarifying this matter. Note that using reachability matrices in the numerator and denominator will not work, because they do not distinguish between direct and indirect relationships between the same two concepts. For definitions of valency matrices and reachability matrices, and other background, see Nozicka et al. (1976).
ABELSON, R.P. (1973), The structure of belief systems, in Schank. R.C. and Colby, K.M. (eds), 'Computer Models of Thought and Language', San Francisco, W.H. Freeman.
ALKER, H. R., Jr. (1975), Polimetrics: its descriptive foundations, in Greenstein, F. and Polsby, N. (eds), Handbook of Political Science, Vol. 7, Reading, Mass., Addison–Wesley.
BONHAM, G.M. et al. (1978), 'A cognitive model of decision-making: application to Norwegian oil policy', Cooperation and Conflict, 13, 93-108.
CUTLER, R.M. (1982a), 'The formation of Soviet foreign policy: organizational and cognitive perspectives', World Politics[, 34, 418–436].
CUTLER, R.M. (1982b), 'Soviet foreign policy towards Western Europe, 1971-1975: case studies in formulation and substance', University of Michigan, PhD dissertation.
ELSTER, J. ( 1978), 'Logic and Society: Contradictions and Possible Worlds', New York, Wiley.
F[oreign] B[roadcast] I[nformation] S[ervice] (1974a), unattributed commentary in Portuguese to Portugal, 'Daily Report: Soviet Union', 23 May 1974, p. E4.
F[oreign] B[roadcast] I[nformation] S[ervice] (1974b), communiqué of the Executive Committee (Politburo) of the Central Committee of the Portuguese Communist Party, as broadcast by Radio Free Portugal (Clandestine) on the morning of 30 May 1974, 'Daily Report: Western Europe', 4 June 1974, pp. E8–E10.
GRIFFITHS, F. ( 1971), A tendency analysis of Soviet
[ page 121 ]
policy making, in Skilling, H.G. and Griffiths, F. (eds), 'Interest Groups in Soviet Politics', Princeton, Princeton University Press.
GRIFFITHS, F. (1972), 'Images, politics, and learning in Soviet behavior toward the United States', Columbia University, PhD dissertation.
GROFMAN, B. and HYMAN, G. (1973), 'Probability and logic in belief systems', Theory and Decision, 4, 17–95.
HERADSTVEIT, D. (1979), 'The Arab-Israeli Conflict: Psychological Obstacles to Peace', Oslo, Universitetsforlaget.
HERADSTVEIT, D. and NARVESEN, O. (1978), 'Psychological constraints on decision-making. A discussion of cognitive approaches: operational code and cognitive map', Cooperation and Conflict, 13, 77–92.
HOLSTI, O.R. (1977), 'The "operational code" as an approach to the analysis of belief systems', unpublished manuscript.
IGNAT′EV, O. (1974), 'Portugaliia: 30 dnei novoi zhizni' [Portugal: Thirty Days of the New Life], Pravda, 25 May 1974, p. 4.
JÖNSSON, C. (1982), Foreign policy ideas and groupings in the Soviet Union, in Kanet, R.E. (ed.) 'Soviet Foreign East-West Relations', Oxford, Pergamon.
NOZICKA, G., BONHAM, G.M. and SHAPIRO, M.J. (1976). Simulation techniques, in Axelrod, R. (ed.) 'Structure of Decision', Princeton, Princeton University Press.
ROIG, C. (1977), 'Symboles et société', Bern, Peter Lang.
ROSENSTONE, S.J. (1974), 'The stability of cognitive maps: a case study', unpublished manuscript.
SKILLING, H.G. (1966), 'Interest groups and communist politics', World Politics, 18, 435–51.
SOBEL, L.A. (ed.) (1976), 'Portuguese Revolution, 1974-1976'. New York, Facts on File.
Dr. Robert M. Cutler [ website — email ] was educated at MIT and The University of Michigan, where he earned a Ph.D. in Political Science, and has specialized and consulted in the international affairs of Europe, Russia, and Eurasia since the late 1970s. He has held research and teaching positions at major universities in the United States, Canada, France, Switzerland, and Russia, and contributed to leading policy reviews and academic journals as well as the print and electronic mass media in three languages.
| C.v. | Resumé | Contact author | Search | “Top 10” Lists |
| RobertCutler.org » Site Map » [ Epistemology | Decisionmaking ] » This document | 1982 |
