THE NEW FACE OF STRATEGIC MANAGEMENT:



ENDOWING COGNITIVE MAPPING WITH COMPUTATIONAL PROPERTIES

FOR STRATEGIC ANALYSIS

William Acar (

Grad. School of Management

Kent State University

Kent, OH 44242, USA

330-672-1156

fax: 330-672-2953

wacar@kent.edu

Douglas Druckenmiller

Dep’t of Information Systems

Western Illinois University

Macomb, IL 61455-1390, USA

563-332-5017

fax: 309-298-2500

DA-Druckenmiller@wiu.edu

FEBRUARY 2005

Based on papers presented at the

2nd International Conference on Organizational Foresight

UNIVERSITY OF STRATHCLYDE

26-28 August 2004

Submitted to FUTURES

Endowing cognitive mapping with computational properties

for strategic analysis

Abstract

A number of cognitive, causal mapping and simulation techniques exist for dealing with the growing importance of environmental uncertainty. After briefly commenting on some of the more salient extant approaches, this paper offers a new one for consideration by the scenario planning community. Comprehensive Situational Mapping (CSM) is a powerful analytical tool combined with a process for framing and debating strategic situations. The CSM approach combines the problem framing features of causal mapping with a dialectical inquiry process patterned after Churchman’s. Like the better approaches to planning through cognitive mapping, it facilitates the “backward analysis” of the underlying strategic assumptions. Its novelty is that it also allows the “forward analysis” of a situation by computing the potential change scenarios. Initially developed for manual application, the principles of CSM were originally tested in appropriate case studies. The contribution of the present paper is to present its theory and point out that its future potential is even greater: in concluding we indicate that, by using recent distributed artificial intelligence (DAI) technology, a fully computerized and interactive prototype is now being set up for commercial applications.

Introduction

Foreshadowing the evolution of the dominant paradigm of strategic management toward the theory of dynamic capabilities [1], two parallel content-oriented thrusts have been aiming at facilitating adaptive strategy design. These are the broad categories of Dialectical Inquiry for assumption analysis (DI) and scenario analysis (SA). However, the same content can espouse different methodological forms. Methodologically speaking, a range of approaches has been proposed. This spectrum encompasses quantitative approaches such as System Dynamics (SD) at its top end, and a cluster of informal approaches broadly designated as cognitive mapping (CM) at its informal end. Separate literatures exist within DI, SA, SD and CM. The contribution of this paper pertains to the emerging effort of attempting to bring them all together [2,3].

Bringing together the work of two authors, one involved with the above four thrusts and one involved with evolution of information technology toward artificial intelligence, this paper presents the theory for a simultaneous approach to assumption analysis scenario planning capable of providing its users with dynamic and interactive analytical capabilities for handling strategic uncertainty. It does so by adopting a causal mapping technique capable of bridging the gap between quantitative and qualitative methods.

After presenting the origins of the method, the paper will contrast it with other approaches and discuss how its analytical processes offer modeling support for strategic problem formulation and scenario planning. It will conclude by showing how this branch of the evolution of strategic theory could mesh in with the recent developments in information technology, namely the latest developments in distributed artificial intelligence. The development of a web-based distributed tool for simulating changes within causal maps can help integrate the DI and SA desiderata, generate new types of applications of the sort favored by multinational corporations and international businesses, and thus take strategic theory to a new level.

2. Two desirable features for strategic inquiry

2.1. Dialectical Inquiry and the surfacing of hidden assumptions

Let us first address the content orientations desired from approaches to strategy facilitation. Although modern strategic theory harks back in a sense to the resource orientation of its founders, its current recognition of the import of dealing with uncertainty has affected the way in which strategy design is undertaken [4-6]. While the debate between the proponents of the “design school” (once led by Ansoff) and the incrementalists of the “strategy emergence” school (maybe still led by Mintzberg) have pedagogical value, they are no longer crucial to the field. The reason is that the modern reckoning of the pervasiveness of environmental uncertainty in a sense dissolves the dilemma.

Much in the way of Hegelian dialectic, a synthesis of the design and emergence concepts has been provided by the philosophical insights of Churchman [7] and his followers [8,9]. In their Challenging Strategic Planning Assumptions, Mason and Mitroff [10] brought Churchman’s ideas to the managerial public. Their book redirected strategy design away from attempting to optimize on the basis of shaky assumptions to a dialectical search for “surfacing” the hidden premises and underlying assumptions on which the strategic options rest. This notion is now broadly accepted.

What is not generally well understood is that Churchman’s idea goes even deeper. While the age-old devil’s advocacy approach to debating still finds supporters [11], Churchman’s concept of Dialectical Inquiry (DI) does not limit itself to questioning the validity of basic underlying assumptions – or even coming up with a defensible counterplan. Rather, projecting Hegel’s thought onto the realm of management, Churchman [12,13] informs us that a fruitful dialogue should be a multi-party interchange among the major stakeholders of a plan or an issue, thus raising the notion of assumption analysis from a traditional devil’s advocacy process to the level of a collaborative search process.

The works cited in the above paragraphs provide mild examples of DI. More elaborate and truer-to-form examples are provided in Ackoff’s writing [14] on the way his Interactive Planning form of consultancy leads the major stakeholder groups of an organization to agree on implementing those configurations ideally desired by most of them. Outside the circle of these systems researchers, DI is still awaiting full methodological integration into modern strategic planning approaches. It will be shown that the proposed CSM method contains a “backward analysis” feature into which a DI component is built.

2. The second desideratum: scenario analysis

The ascent of scenario analysis (SA) method appears to have had many sources, most of which not unknown to the readers of Futures. Going all the way back to the work of the California “Futurists” (among whom Herman Kahn at RAND) in the 1950s, the scenario method gradually spread. An increasing number of texts describe the scenario method [15]][6,16-21], there is no reason to attempt here to encapsulate all this known history.

However, since we deem SA to be a valuable approach to reach for, we take this opportunity to briefly recall its principal advantages. One of them is the point often made by systems theorists that both quantitative operational research (OR/MS) and qualitative process-consulting approaches have, over time, proved disappointing when used in isolation. As Ackoff [14] and Mintzberg [22] have been quick to point out, quantitative methods appear extremely well suited to mid-level tactical decisions but ill suited to upper-level strategic issues.

Yet, however popular purely qualitative process consulting is with behavioral scientists, there is little evidence that it adds strategic clarity rather than just disseminating good feelings through facile and non-taxing solutions. Moreover, its focus is also on tactical issues, so little is gained by it at the strategic level. Thus one reason for promoting SA is that it holds the promise of blending qualitative and quantitative analytics in a unified approach as witnessed by work at the Battelle Institute.

Another reason is that scenarios can be used to deal with complex, interrelated real-world problems. One of the main challenges of strategic management is how to deal with pervasive uncertainty, and the classic broad environmental typologies do not suffice [15,23,24]. To be effective, one must analytically address the complexity of the entire situation rather than propose solutions to single problems [10,25].

As pointed out by Eden [26] and Ackoff [14], the messiness of reality requires a shift from problem formulation to expressing the messiness of the entire situation. Strategic uncertainty can be the crucible in which the organization’s future might melt away [4] but, opportunistically managed, it could also become the obstacle course in which to pass one’s competitors [6]. Current strategic thinking is that the analysis of change scenarios would allow a firm or institution to design winning strategies, and plan their successful implementation, in spite of situational messiness and uncertainty. The admission of SA into mainstream strategic thinking has become a reality to be acknowledged. It will be shown that our CSM method includes a “forward analysis” feature into which an SA component is built.

3. The quantitative and qualitative approaches to be bridged

3.1. Quantitative power of System Dynamics

The previous section dealt with the content-oriented desiderata. In this section, we turn our attention to the methodological range of potential approaches. The early treatises in OR/MS, dating back to the 1950s and ‘60s already provide for describing complex business situations mathematically as a set of equations. Ackoff [27] offers a clear terminology: diagram and set of equations expressing it constitute the model of the (business or problem) situation, and simulation of the propagation of change among the variables in the model (or “running the model”) is a way of bringing to life what could otherwise remain a set of dry, abstract relationships among the symbolic variables of a model.

Soon enough simulation languages aimed at describing operational and engineering tasks appeared on the scene. GASP, SIMSCRIPT and GPSS simulated well the occurrence of discrete events, and were ideally suited for use in scheduling or tactical management. No such decision aids were aimed at the strategic level until the seminal work of Jay Forrester at the engineering school of the Massachusetts Institute of Technology. Reasoning, like cognitive mappers, that management complications are often due to the presence of time of “feedback” loops and time lags, he devised a theory for simulating interconnected systems called System Dynamics (SD) and developed, with the Pugh-Roberts firm, a simulation language for strategic management situations called DYNAMO [28].

SD is a powerful analysis tool for examining the conditions under which change might occur as well as the extent of it. It was used in the important 1972 and 1992 “world dynamics” studies by the Club of Rome to judge the survivability of our planetary ecosystem. In addition, it has occasionally provided inspiration to authors looking for a new perspective on competitive dynamics [29]. And others [30] have found a way to use SD as a device for group model building. Yet the original DYNAMO software had retained from its roots unappealing and cumbersome graphics and notational system. Still, a number of authors, especially Morecroft [31] and Senge [32], tirelessly worked at promoting SD. However, the “techie” appearance and feel of the method militated against widespread diffusion.

Senior managers find cognitive mapping appealing but are generally indifferent SD and its offshoots, even though fairly simplified and user-friendly PC-oriented systems (DYSMAP2, VenSim, STELLA and “iThink”) have been available for a while now. In recognition of this, Nancy Roberts and a team of SD devotees initiated a qualitative approach to SD. In a very readable book, Roberts et al. [33] had the idea of preceding the complex flow-graphing of SD modeling with the drawing of causal-loop diagrams in which the arrows bear a direction and a sign, in the style of the “influence diagramming” promoted by Maruyama [34]. Other members of Roberts’s team, and Vennix, worked to extend her thrust by using SD to aid and systematize group model building [35]. Finally, Eden and some of his collaborators have devised a cyclical consulting process combining quantitative SD and qualitative CM components [36]. As will be seen in Section 5, this powerful combination of features is methodologically echoed in the design of our proposed CSM.

3.2. Qualitative problem framing and cognitive mapping

In spite of Roberts et al.’s and other valiant efforts to bridge the gap between the problem formulation and scenario simulation steps of SD, by and large the method has not taken hold. We attribute this lack of (otherwise well deserved) success to the fact that these approaches to popularize SD were proceeding from the upper, quantitative end of the problem formulation spectrum down toward its lower, qualitative end. Cognitive mapping (CM), on the other hand, has been far more successful with the general public because it proceeded in the opposite direction, namely from initially easier qualitative approaches to increasingly more complex ones, such as the one we promote in this paper.

As advanced by the theorists of the systems approach [10,12-14,37], the process of strategy making is an interactive, dialectical process that surfaces the emergent strategic assumptions present in an organization and allows them to become an explicit part of a rational strategic plan. A tactical system is then designed that implements the proposed strategies, and tactics are grouped into programmatic actions around which resources are committed. Implementation of programmed actions is evaluated using a structured process of reflection and the cycle repeats itself with a new round of vision, contradictions, etc. This cycle of organizational learning uses explicit rational models as a tool for organizing experiential learning in a continuous improvement process that is both participatory and structured [38-42].

The underlying methodological thrust to which our research subscribes is that traditional problem solving is giving way to problem “framing” or defining [37]. More boldly, some authors following the systems approach claim that single problems cannot be isolated from the surrounding messy realities. This no underhand or disguised incrementalism, but a flexible approach to strategy design whereby an attempt is made at capturing the entire strategic situation in one fell swoop. Such authors no longer talk of problem formulation but of problem framing or, better still, situation formulation [14,26,43].

Cognition is important in confirming and redesigning strategy, for managing complexity and organizational change, for strategic decision making and problem solving [5]. Intuition can be informed and creative processes can be structured. As used by management theorists, cognitive mapping has focused on the challenge of finding illustrative, often visual approaches to the synthetic and creative process of strategy conceptualization. Graphic representations are well known in the management and social science literature as both systems analysis tools and knowledge representation techniques [44].

Cognitive mapping integrates the naturally emergent strategizing of Mintzberg with the deliberately rational learning approach advocated by Ansoff; in recognition of this, it has been gradually moving to the mainstream of strategic thought [Eden & Spender, 1998]. In addition, Forrester [28], Maruyama [34], Weick [45] and Eden et al. [2] all show that complex management situations can be captured by some variety of CM. A number of specific approaches have been developed; they will be reviewed in the next section that will gradually introduce the specific causal mapping method we advocate.

4. Existing varieties of cognitive mapping and structuring

4.1. Flexibility of graphic representations in cognitive mapping

As the domain of strategy analysis shifts from algorithmic problem solving to capturing the essence of strategic situations, the variety of graphic representations of influence networks is thrown center stage. The early qualitative cognitive mapping (CM) derived from cognitive psychology. From Hinkle’s “implication grid”, Armstrong and Eden [46] derive an implications map in which single or double-directional arrows indicate which bipolar constructs imply or are implied by others. While the implication map could be duplicated with some effort by longer statements in prose, the concept map in which a central concept is seen to relate in several directions to many others provides in one glance an even more synoptic view of an entire mental view [47].

Even more flexible is the general-purpose cognitive map found in Eden et al. [2]. In one of their examples, the element or node “number of additional dwellings needed” positively affects the node “Conversion of upper floors of commercial buildings into flats”; also, both of these nodes positively influence the decision box or node “Do/ do not/ separate off upper floors on renewing lease”. At the same time, the “Conversion” node and another stating that “Building is in short supply” influence the result “Get/ do not get/ best out of existing stock of buildings”; and so on to a whole network of inter-connected events or variables.

Basically, cognitive mapping is very flexible because the elements of cognitive maps are freely selected without the strictures of the complex syntax of graph theory or formal logic. Rosenhead [48] and Huff [49] each review a number of CM approaches; between the two, a reasonably complete picture of the flexibility of cognitive mapping emerges. The nodes in CM could be statements, events, actions, impressions, quantifiable variables or even decision rules – and even sometimes restrictions or limitations. Sometimes the term cause mapping is used to denote cases in which the influences are presumed to be causal [49], but Eden, Ackermann and Cropper [50] point to the lack of established usage. In the remainder of this paper, we will use the more grammatically and euphonically acceptable expression “causal mapping”.

4.2. Formal graph and matrix methods

Even though graphic representations are now well known in the management and social science literature as both systems analysis tools and knowledge representation techniques [44], they seem to have originated in sociology and anthropology where there arose a need to model a number of social structures, such as the authority structure and communication structure. Since these are to be done one at a time, these “sociometric structures” were modeled by sociometric graphs. Contrary to the informal flexibility of the varieties of CM, sociometric approaches use formal or mathematical graph theory in which a rigorous and unique interpretation is ascribed to each type of node or line (link) in the graph.

Rigorous approaches to sociometry have been developed by using mathematical graph theory; Harary, Norman & Cartwright [51] provide an early yet impressive compendium of most of these in a single theoretical text devoted to the study of directed graphs or “digraphs”. Their results include the degree of reachability of an element from another, the number and severity of cycles in the structure, and so on. Using the indegrees and outdegrees, or number of lines going into or out of a node, and other layout characteristics, the analysis of the properties resulting from topological characteristics constitutes the structural analysis of a digraph.

Some of these results are obtained through the theorems of mathematical graph theory. But an easier approach is to notice that a digraph in which each link has a single interpretation can equivalently be represented by an adjacency matrix in which n entry (i,j) is 1 if there is a directed link from element or node “i” to node “j”, or 0 otherwise. As in their corresponding digraphs, the entries in adjacency matrices may or may not be symmetrical, thus allowing the modeler to express directional as well as mutual relationships. The existence of such an isomorphism between digraphs and Boolean matrices is very convenient because it allows mechanizing some of the elements of structural analysis. For example, theorems found in Harary et al. [51] show how reachability and distance information can be obtained from taking the higher powers of the adjacency matrix. In sum, the use of sociometric graphs and matrices is more rigorous but lends itself to less creativity than free-form CM.

4.3. A synthesis? Influence diagramming

Of particular interest for bridging the gap between the rigor of structural sociometric analysis and free-form CM is a variety called influence diagramming (ID) [34]. In ID some rigor is imposed in that all the nodes have to be, if not quantitative, but at least quantifiable variables, with reference to which it would make sense to talk about increases or decreases in value or level. Similarly, all links have the same rigorous interpretation: they indicate the potential for transmission of change from an “upstream” node to one “downstream” from it. So that the notation U ―(+)―►V means changes in U will generate changes of the same sign in V; alternatively, U ―( ─ )―►V means that an increase in U induces a decrease in V and a decrease in U an increase in V.

Like SD, ID allows the representation of the considerable power of cycles or loops in a causal graph. Generally, there are more loops than initial inspection reveals. For instance, in an initial structural analysis the diagram below appears to have one single large loop, the cycle ABCDHGFEA.

A ▬▬▬► B ▬▬▬► C ▬▬▬► D

▲ │ ▲ │

│- │- │+ │-

│ ▼ │ ▼

E ◄▬▬▬ F ◄▬▬▬ G ◄▬▬▬ H

In reality, there are two more loops, ABFEA and CDHGC, adding to the transmission of influences.

When an arrow has no negative sign next to it, it is supposed to be positive. Clearly a loop in which only positive signs exist would tend to amplify the decreases or increases that enter it. Maruyama’s seminal [34] paper showed how could be used for policy making using cybernetic principles. Weick [45] and Diffenbach [52] concurred with his implicit advocacy of ID as a happy medium between flexibility and analytical power. However, over time these authors accepted the hypothesis that a loop with an even number of negative signs would also be deviation amplifying, presumably because the pairs of minus signs would cancel out each other. They also unquestionably accepted the implicit corollary that a loop in which there is an odd number of minus signs, or negative loop, would perforce be deviation correcting.

This belief that it sufficed to count the number of signs of the dominant loops of an influence diagram has become widespread. For instance, Eden, Ackermann and Cropper’s [50] paper on the analysis of cause maps that recommends simplifying a diagram by clustering its subsets into “islands” still accepts the above hypothesis relating to the influence of the configuration of the signs in the loops. However, a different approach to causal mapping was initiated by Acar [43] who showed that these results were unreliable as long as only the signs of the links were taken into account, while their magnitudes were being totally ignored. His proposal of CSM relates to that discovery.

5. CSM: the proposed approach to causal mapping

5.1. Introducing CSM – a synoptic view

The field has to move beyond such partial approaches, all the while retaining their most desirable properties. A powerful improvement on ID, and one with more richness that CM, is the comprehensive situation mapping (CSM – see Fig. 1) developed by Acar [43] that endows causal mapping of real computational properties. A synoptic judgement of it is that it is a collaborative process as well as an analytical framework. As a process it was conceived as a collaborative dialectical conversation, but one based on sharing the structural and scenario analyses of causal maps among the principal “actors” in a strategic situation. As such, it subsumes a process that provides the desiderata from both DI and SA sought in Section 2. In addition, it includes features of adapted from both CM and SD described in Section 3.

CSM’s first feature is that it builds on ID but improves on it by enriching its computational capabilities. By including in the method indications, not only of the signs of the presumed causal influences, but also of their intensities and the possible time lags, Acar developed a technique for simulating manually (on the causal map itself) the propagation of change through a causal network. Instead of designing strategies in vacuo or bemoaning uncertainty one can set out to profit from it by analyzing change scenarios [6].

As an added bonus, by having several actors produce several maps, compare them and discuss them, CSM embodies Churchman’s principle of DI. Specifically, CSM fleshes out further developments to the manner in which Mason and Mitroff’s [10] concept of assumption surfacing and analysis. To a greater degree than their procedure, CSM lends itself well to the kind of “backward analysis” of strategic assumptions such systems researchers deem essential to analytical strategic thinking in an uncertain world.

---------------------------------------

Insert Figs. 1 and 2 about here

---------------------------------------

A synoptic view of its capabilities is displayed in Fig. 2. It displays the stages a top management team would go through, from surveying their firm’s situation, to connecting its principal factors in a DI-like graph, to finally enriching the links with the notation specific to CSM, to mapping its goals. Fig. 2 also shows the array of analyses that could be undertaken at each of the CSM stages.

5.2. CSM notation and semantics

Improvements in information technology and the proliferation of automated decision support systems (DSS), group decision-support systems (GDSS) of the kind described by Heintz and Acar [53] and Phelan [54] were not always cherished by senior managers and their consultants. In fact, the prevailing view among practitioners of the soft systems approach was that personal contact and holistic mapping facilitates problem solving [14,37]. Acar’s [43] CSM approach to causal mapping was thus originally designed as a manual system because it aimed at direct use by top managers.

The originality of Acar’s approach was to create a bridge between SD and CM, not just by using SD qualitatively like Andersen et al., but also manually computing the transmission of change throughout a causal map or net. By including in the method indications not only of the signs of the presumed causal influences, but also of their intensities and the possible time lags, Acar (1983) developed a technique for simulating manually (on the causal map itself) the propagation of change through a causal network. As discussed in Section 4.3, ID offers over CM the analytical advantage of being able to qualify the influence of the causal loops, if not always accurately. And full-fledged SD has over ID the advantage of considerable computational power. As an approach that can be operated manually, CSM represents a happy medium between the simplicity of ID and the power of SD while avoiding some of their disadvantages in practice.

Reasons of space prevent us here from going into detailed examples with large causal maps, but the sketch in Fig. 1 exemplifies the nature of CSM. While not shown on this small-size diagram, the nodes in CSM have to be quantified variables (e.g. “quantity sold”) or at least quantifiable ones (e.g. “the degree of effort”). When it comes to the links expressing the presumed causal influences, they are rigorously specified as being of only three types. Fig. 1 shows the links as being either full channels (drawn as double arrows), partial channels or half channels (drawn as single arrows), and restrictions or constraints on the transmission of change (shown as dashed or dotted arrows).

In addition, whenever the change transmitted from an upstream to a downstream node is not strictly proportional to the upstream change, an indication of the sign and transmittance coefficient expressing the correct proportion of change transmitted is written next to the arrow. Similarly, the minimum thresholds and time lags are also shown next to the appropriate links in CSM mapping. The nodes from which change originates (the “sources of change”) are noted by the name of the source in parenthesis and, in order to complete what is needed to endow CM or ID with computational properties, goal vectors when applicable are noted on the appropriate node by the sign and percentage of change desired. Fig. 3 reproduces with our current software a CSM map constructed manually in the original case study that tested the method [43].

-----------------------------

Insert Fig. 3 about here

------------------------------

5.3. Principles and principal stages of CSM

Full channels transmit change as dictated by the signed magnitude, direction and time lag parameters. The basic idea proposed by Acar [43] for endowing CM with a computational property is to look at links as change-transmission channels, and at a full channel as a linear operator on its downstream variable. In other words that it would be equivalent to write:

a

X ═════►Y or Y t+k = a X t + b or ΔY t+k = a (ΔX t )

t=k

So the full channels are sufficient for the transmission of change. The partial channels, on the other hand, model the necessity condition in causal reasoning: in order for half-channel linkages to activate, all half-channel inputs to a node must be activated. The flow of change through the causal graph is thereby visually quantified and may be used for a more in-depth analysis. Causal maps can therefore become the basis of a simulation study of the propagation of change through the network of presumed causal relations.

As a process-oriented method, CSM progresses through several stages or levels in the development of causal graphs. As in any variety of CM, its first stage (Level 0) is a survey stage that gathers specific information from individual stakeholders on the principal factors or variables (nodes A, B, C,… in Fig. 1) that constitute the problem situation. The next stage (Level 1) is a structuring stage in which those principal elements are graphically connected by arrows to reveal the qualitative structure of the strategic situation. The final stage (Level 2) or full graphing entails complementing the causal structure by specifying the parameters (transmittance coefficients and time lags) without which no computations can be performed (e.g. markings next to the arrows in Figs. 1 and 3).

5.4. Structural and “Backward” Analyses

In practice, the graphically based CSM method uses causal maps developed by opposing managerial groups (such as the archetypical nemeses, the Production and Marketing divisions of a manufacturing concern) to model their specific view of their organization’s problématique. The synoptic layout of Fig. 2 illustrates that the successive stages of CSM support three types of analyses: structural analysis, backward analysis and forward analysis.

At the Level-0 Survey stage, one simply takes stock of the factors considered and assesses them for the extreme cases of incompleteness of scope or overlaps. At the Level-1 Structuring stage, linking up the nodes with directed arrows provides a digraph. Even though the full CSM graphing is still to be completed, structural analysis can already be undertaken in Acar’s CSM framework. The “structural analysis” of a digraph actually consists of a cluster of structural analyses aimed at assessing graph change, reachability, graph scope and connectivity. They are carried out by using the algebra of “adjacency matrices”, the theorems and results of which are described in the classic text by Harary et al. [51] as noted in Section 4.2.

In addition, analyses pertaining to CM and problem framing can be performed at the Level-1 Structuring phase. This is the backward analysis of key assumptions that underlie the modeling of the factors, variables, actors and their relationships, and details on the directions and quantifications of those relationships (see Fig. 3). The theory of backward analysis is based on the “assumptional analysis” process of Mason and Mitroff [10]. Major and minor assumptions about the situation are surfaced and tested, resulting in a better understanding of the situation by all participants. Once surfaced, the formerly tacit assumptions are evaluated through a DI-type debate among the firm’s participating groups (in a manufacturing firm, its Production, Marketing, International and Corporate divisions, etc.).

The process proceeds in stages starting with a divergence phase in which the participating groups create separate causal maps, to be followed by a convergence phase where a common causal map is collaboratively attempted using the insights from the individual maps. This alternation may result in agreement through dialectical inquiry. Acar [43] argues that his system can be classified as a Singerian inquiring system [12] because the different perspectives on a situation are merged through a “sweeping-in” process providing an inter-subjective view of it.

5.5. The “forward” analysis of scenarios in CSM

The Level-2 Full Graphing and Goal Mapping stages of CSM link situation formulation with a forward-analysis capability. Strictly defined, forward analysis denotes the computation of the propagation of change through a causal network, or the simulation of scenarios to use the language of SA. The sources from which change can enter a network are noted by the name of the source in parentheses on the causal map, and goal vectors are indicated on the appropriate node by the sign and percentage of change desired.

In order to design a technique that could perform manually what it takes computing power for simulation languages to accomplish, Acar [43] introduced some simplifications. Among them is the assumption of linearity described in Section 5.3, and the rather neat trick of expressing changes in terms of the (fixed) status quo levels, thus rendering them are additive and cumulative. This provides CSM with an elegant simplicity: when changes “arrive” at a node after the time lag indicated on each link, the state of each node can be easily updated by adding the amount of new change adjusted by the transmittance coefficient of the link to that node.

Nonetheless, his model accommodates the complexity that, in order for half-channel linkages to activate, all half-channel inputs to a node must be activated. The flow of change through the system can thereby be computed manually in CSM, in spite of the nonlinearity due to the presence of loops discovered by Maruyama, and also those due to production function nonlinearity that would normally require computerized simulation. As discussed by Acar [43] and Georgantzas and Acar [19], CSM’s quantitative forward analysis goes beyond the traditional qualitative scenario analysis, and can also assist in the evaluation of goal compatibilities and feasibilities.

6. Current developments: combining manual and computerized methods

1 Computerized implementations and combination processes

It has been almost a couple of decades that Eden and his associates have developed a process around the use of cognitive mapping. They called this process Strategic Options Development and Analysis (SODA). However, while most systems researchers continued to rely on manual methods, Eden and his colleagues at the University of Bath (such as Smithin, Wiltshire, Ackerman and Cropper) started developing in the late 1970s the computerized Graphics COgnitive Policy Evaluation (COPE) system they used for many years [55,56]. In time, they were joined by others. To cite but a few: the CMAP2 of Laukkanen [57], the DISTRAT of Markóczy and Goldberg [58], soon joined by CAFE or Decision Explorer of Brightman, Eden, Langford and van der Heijden [59].

Even though strategists had long been suspicious of computer modeling tools because they find them time consuming and expensive to develop, and frequently result in models that do not fit the decision maker’s conception of the problem [10], the benefits of automated support for cognitive mapping are becoming well established [53,56,60].

6.2. The partial automation of CSM

In line with these developments, Heintz and Acar [53] initiated an object-oriented approach to computerizing CSM. This early prototype developed an object model that only supported causal mapping semantics and process. While that prototype (implemented as a Smalltalk application) demonstrated the ease of development with object-oriented techniques, it was limited to the graphic editing of causal maps entailed in backward analysis and did not attempt to incorporate any part of forward analysis. Yet this initial research was useful in that it identified the major technical problems to be overcome for successful implementation.

Thus the scope and complexity of the CSM system have dictated a partitioned approach to development. The natural extension would be a prototype not limited to Levels 0 and 1 of CSM, but that might undertake the Level-2 forward analyses. The complexity of causal loops in the models and calculations of successive waves of change through the model when simulated presented a substantial design challenge for developing automated scenario support using object-oriented techniques alone. That initial prototype study suggested an artificial-intelligence solution; however, only recently are the relevant programming aids becoming available. Section 7.2 will introduce research still in progress.

7. Looking to the future: computerized and web-based aids

7.1. The advent of distributed artificial intelligence

The emergence of the electronic marketplace and the pervasive use of information technology in business have radically changed the way firms operate in the 21st century. While the “actors” discussed above are the live stakeholders of an organization, distributed artificial intelligence (DAI) now relies on “agents” that are software systems capable of flexible and autonomous action. These multi-agent systems (MAS) are groups of agents that collaborate to solve complex real-world problems.

MAS are currently being used in intelligent search engines on the Internet, flexible-manufacturing systems, and as third-party negotiation systems in E-commerce market prototypes. They are also being used in intranet collaboration systems, groupware, and knowledge management applications [61]. According to Phelan [54], as the new century unfurls, the MAS technology can become the latest potential addition to a firm’s dynamic capabilities based on artificial intelligence research.

7.2. On-going research for enriching our current prototype

Using Java-based routines and MAS technology not available during the first attempt at computerizing CSM in the 1990s, a prototype has been constructed that does support both backward and forward analyses. This goes well beyond just tallying the number of negative signs in loops and approaches SD in effectiveness as a management aid, while being user-friendly and well suited to strategic rather than tactical issues.

Our current prototype links a user-friendly system for individual editing of causal maps with an agent-based modeling and simulation system developed for public use by the University of Chicago and called RePast [62]. Individually developed maps can be put into motion for forward analysis. We find agent-based modeling and simulation to be uniquely suited to analyzing the complex loops found in causal maps. The complexity such loops embody is analytically intractable to “unintelligent” systems not using the MAS approach to DAI; and thus precludes the use of a single-shot approach to strategy design. An iterative process of backwards assumption modeling and forward simulation analysis would allow decision makers to understand the nature of their strategic situation, and hence how to remedy the interconnectivity of the problems it entails.

Space prevents us here from describing the way the aggregation of the successive waves of change can be present on a link at the same time due to the presence of time lags and loops. Such a description would perforce be quite technical. Let us simply say here that changes in initial conditions are made in nodes that represent either environmental triggers (external agents beyond the control of anyone system) or decision levers (agents under the control of the focal firm or some competitor). Thus the user can compute scenarios by changing the initial level of trigger or lever nodes that represent different sets of initial conditions. Fig. 3 provides an example of an actual CSM map adapted from Acar’s original case study [43] with the current prototype, and Fig. 4 illustrates a scenario in which the oscillations in a particular goal node are oscillating exponentially because of a deviation-amplifying loop.

-----------------------------

Insert Fig. 4 about here

-------------------------------

3. Potential future extensions

We plan to embed the current prototype in a web-based GDSS. This development would allow users to collaboratively develop causal maps in a network (web-based) environment through the process of backward analysis and negotiation. As an improvement on the present prototype, alternate scenarios could be created as separate maps and stored on the web for use by the far-flung divisions of any large, even international, corporation. In time, our improved prototype would support merging individual maps into a composite map representing the consensus of participating decision makers.

8. Conclusion

Strategic management is a process of continually adjusting the fit between the firm’s capabilities and its changing environment. The interplay between learning and planning is central to managing organizations in today’s turbulent and uncertain environments. The Mintzberg-Ansoff debate exposes a gap in the strategic process literature. Can the gap be closed? The systems approach provides a holistic solution inclusive of both rational and existential approaches to learning, one that achieves planned organizational learning [19,55] through reflective strategy design.

After briefly presenting Churchman’s Dialectical Inquiry (DI) and scenario analysis (SA) as two desirable features of a strategy analysis system, we presented the methodological approaches of System Dynamics (SD) and free-form cognitive mapping (CM) as two opposite approaches along the quantitative-qualitative spectrum. Following that, influence diagramming (ID) and matrix methods were introduced as a potential synthesis.

Having laid out those preliminaries, we presented Acar’s [43] Comprehensive Situational Mapping (CSM) as a better synthesis. It is an integrated approach that competes in rigor with ID and even somewhat SD, because its forward analysis allows for the computation of change scenarios. It also competes in richness with CM, because CSM is a dialectical inquiry and scenario planning method that facilitates the backward analysis of key assumptions.

The advent of CSM as a comprehensive method for both the “backward” analysis of hidden assumptions and future-oriented strategic analysis through scenarios provides a competitive advantage to firms in turbulent and uncertain environments. Our current development of a user-friendly web-based tool for simulating changes within causal maps will improve the theory and practice of scenario analysis. It should prove helpful to taking strategic theory to its next level, the level of enriched intellectual capital and dynamic capabilities to supplant an over-reliance on scarce resources.

CAPTIONS & LEGENDS

Figure 1. An Example of CSM [43]

Figure 2. A synoptic view of Acar’s system: Components of Causal Mapping. [43]

Figure 3. “Jay’s Map” according to Acar’s [43] case study at the Philadelphia VAC.

Figure 4. Example of forward analysis – a scenario of widening oscillations.

Reference List

[1] D. J. Teece, G. Pisano, A. Shuen, Dynamic capabilities and strategic management, Strategic Management Journal 18 (7) (1997) 509-533.

[2] C. Eden, F. Ackermann, S. Cropper, Images into models: the subjective world of the policy maker, Futures (1979) 56-62.

[3] C. Eden, J. C. Spender (Eds.), Managerial and organizational cognition: theory, methods and research, Sage, London, 1998.

[4] M. E. Porter, competitive strategy: Techniques for analyzing industries and competitors, Macmillian Publishing Co., Inc., New York, 1980.

[5] C. Eden, F. Ackermann, Making strategy, Sage, London, 1998.

[6] P. J. H. Schoemaker, Profiting from uncertainty: strategies for succeeding no matter what the future brings, Free Press, New York, 2002.

[7] C. W. Churchman, The systems approach, Dell, New York, 1968.

[8] R. O. Mason, A dialectical approach to strategic planning, Management Science 15 (8) (1969) B403-B414.

[9] I. I. Mitroff, J. R. Emshoff, R. H. Kilmann, On strategic assumption-making: a dialectical approach to policy and planning, Academy of Management Review 4 (1) (1979) 1-12.

[10] R. O. Mason, I. I. Mitroff, Challenging strategic planning assumptions, John Wiley & Sons, New York, 1981.

[11] R. A. Cosier, Dialectical inquiry in strategic planning: a case of premature acceptance?, Academy of Management Review 6 (1981) 643-648.

[12] C. W. Churchman, The design of inquiring systems: basic concepts of systems and organization, Basic Books, New York, 1971.

[13] C. W. Churchman, The systems approach and its enemies, Basic Books, New York, 1979.

[14] R. L. Ackoff, Creating the corporate future, Wiley, New York, 1981.

[15] Amara, Lipinski, Business planning for an uncertain future: scenarios and strategies, Pergemon, New York, 1983.

[16] L. Fahey, V. K. Narayanan, Macroenvironmental analysis for strategic management, West Publishing, St. Paul, 1986.

[17] P. Schwartz, The art of the long view, Currency, 1991.

[18] P. J. H. Schoemaker, Multiple scenario development: its conceptual and behavioral foundation, Strategic Management Journal 14 (3) (1993) 193-213.

[19] N. C. Georgantzas, W. Acar, Scenario-driven planning: learning to manage strategic uncertainty, 1st ed., Quorum Books, Westport Connecticut, 1995.

[20] K. van der Heijden, Scenarios: the art of strategic conversation, Wiley, Chichester, 1996.

[21] G. Ringland, Scenario planning: managing for the future, Wiley, New York, 1998.

[22] H. Mintzberg, The rise and fall of strategic planning, Free Press, New York, 1994.

[23] J. Rosenhead, Planning under uncertainty: II. a methodology for robustness analysis, Journal of the Operational Research Society 31 (1980) 331-341.

[24] R. A. D'Aveni (with R.Gunther), Hypercompetition, Free Press, New York, 1994.

[25] R. L. Ackoff, The art of problem solving, Wiley, New York, 1978.

[26] C. Eden, Operational research and organization development, Human relations 31 (8) (1978) 657-674.

[27] Ackoff R.L., Scientific method, Wiley, New York, 1962.

[28] J. W. Forrester, Industrial dynamics, M.I.T. Press, Cambridge, Mass, 1961.

[29] K. Warren, Competitive strategy dynamics, Wiley, Chichester, 2002.

[30] J. A. M. Vennix, Group model building: facilitating team learning using system dynamics, Wiley, Chichester, 1996.

[31] J. D. W. Morecroft, Strategy support models, Strategic Management Journal 5 (3) (1984) 215-229.

[32] P. M. Senge, The fifth discipline: the art & practice of the learning organization, Doubleday, New York, 1990.

[33] N. Roberts, D. Andersen, R. Deal, M. Garet, W. Shaffer, Introduction to computer simulation: a system dynamics modeling approach, Addison-Wesley, Reading, MA, 1983.

[34] M. Maruyama, The second cybernetics: deviation-amplifying mutual causal processes, American Scientist 51 (1963) 164-179 & 250-256.

[35] D. F. Andersen, G. P. Richardson, J. A. M. Vennix, Group model building: adding more science to the craft, System Dynamics Review 13 (2) (1997) 187-201.

[36] F. Ackermann, C. Eden, T. Williams, Modeling for litigation: mixing qualitative and quantitative approaches., Interfaces 27 (2) (1997) 48-65.

[37] P. Checkland, Systems thinking, systems practice, Wiley, London, 1981.

[38] H. I. Ansoff, Corporate strategy, McGraw-Hill, New York, 1965.

[39] H. I. Ansoff, Implanting strategic management, Prentice-Hall, Englewood Cliffs, NJ, 1984.

[40] R. L. Ackoff, A concept of corporate planning, Wiley-Interscience, New York, 1970.

[41] K. R. Andrews, The concept of corporate strategy, Irwin, Homewood, IL., 1971.

[42] C. W. Hofer, D. Schendel, Strategy formulation: analytical concepts, West, St. Paul, MN, 1978.

[43] W. Acar, Toward a theory of problem formulation and the planning of change: causal mapping and dialectical debate in situation formulation, Ann Arbor, Michigan:U.M.I., 1983.

[44] J. L. Sowa, D. Dietz, Knowledge representation: logical, philosophical, and computational foundations, Brooks/Cole Publishing Inc., Pacific Grove, 1999.

[45] K. E. Weick, The social psychology of organizing, Addison-Wesley, Reading, MA, 1969.

[46] T. Armstrong, Eden C, An exploration of occupational role: an exercise in team development, Personnel Review 8 (1) (1979) 20-23.

[47] S. Jones, Eden C, OR in the community, Journal of the Operational Research Society 32 (4) (1981) 335-345.

[48] J. Rosenhead (Ed.), Rational analysis for a problematic world, Wiley, Chichester, 1989.

[49] A. S. Huff (Ed.), Mapping strategic thought, Wiley & Sons, England, 1990.

[50] C. Eden, F. Ackermann, S. Cropper, The analysis of cause maps, Journal of Management Studies 29 (3) (1992) 309-324.

[51] N. Harary, R. Z. Norman, D. Cartwright, Structural models: an introduction to the theory of directed graphs, Wiley, New York, 1965.

[52] J. Diffenbach, Influence diagrams for complex strategic issues, Strategic Management Journal 3 (1982) 133-146.

[53] T. J. Heintz, W. Acar, Causal modeling as a tool for problem framing within a GDSS: an object-oriented approach., Information Systems Journal 4 (1994) 291-310.

[54] S. E. Phelan, Using artificial adaptive agents to explore strategic landscapes, La Trobe University, Bundoora, Australia, 1997.

[55] C. Eden, T. Smithin, J. Wiltshire, Cognition simulation and learning, Journal of Experiential Learning and Simulation 2 (1980) 131-143.

[56] F. Ackermann, S. Cropper, C. Eden, Graphics COPE user guide, University of Strathclyde, Glasgow, 1992.

[57] M. Laukkanen, Comparative cause mapping of management cognitions: a computer database method for natural data, Helsinki School of Economics, Helsinki, 1992.

[58] L. Markoczy, J. Goldberg, A method for eliciting and comparing causal maps, Journal of Management 21 (23) (1995) 305-333.

[59] J. Brightman, Eden C, D. Langford, K. Van Der Heijden, CAFE: constructing alternative futures explorer, University of Strathclyde, Glasgow, 1997.

[60] B. Chaib-draa, Causal maps: theory, implementation, and practical applications in multiagent environments, IEEE Transactions on Knowledge and Data Engineering 14 (6) (2002) 1201-1217.

[61] C. Ellis, J. Wainer,G. Weiss (Ed.), Multiagent systems, MIT Press, Cambridge, Massachusetts, 2001.

[62] University of Chicago: Social Science Research Computing, RePast 2.0. 12-19-2002, ,

Ref Type: Computer Program

|[pic] |

|Figure 1. An Example of CSM [43] |

|[pic] |

|Figure 2. A synoptic view of Acar’s system: Components of Causal Mapping. [43] |

|[pic] |

|Figure 3. “Jay’s Map” according to Acar’s [43] case study at the Philadelphia VAC. |

|[pic] |

|Figure 4. Example of forward analysis – a scenario of widening oscillations. |

( Corresponding author

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download