Causal Map Elicitation

As mentioned previously, we use a method where participants identify congregate labels

before the causal relationships are identified at the individual map level (Tegarden and

Sheetz, 2003). This approach minimizes the potential for researcher bias and provides an

efficient means for deriving collective maps in a comparatively short period of time.

However, once congregate labels are derived, any causal mapping approach can be

adapted to construct cognitive factions using the cluster analysis based approach

described here. In our approach, both the individual causal maps and collective maps

embody the congregate labels. As such, we can analyze differences as well as similarities

among the individual causal maps to create an overall collective map and collective maps

representing cognitive factions.

The causal mapping methodology used is a modification of the Self-Q Technique

(Bougon, 1983; Sheetz, Tegarden, Kozar & Zigurs, 1994). The methodology is supported

by a distributed system that runs in a WWW-based environment.1 The software couples

group support systems (GSS) technology with causal mapping to provide a mechanism

for the group to identify their own congregating labels. Like most systems that use GSS

technology, the software supports anonymity. Anonymity minimizes the effect that the

more powerful members of the top management team can have on the other members

(Valacich, Dennis & Nunamaker, 1992). Using the software, no one member has any more

or less influence on any other member.

The causal mapping elicitation procedures are supported using an agenda of activities

implemented in the software. Each activity is supported with an individual tool. Throughout

the session, a facilitator provides procedural guidance to the group, e.g., administrative

activities such as reading instructions and keeping time. To avoid potential

researcher bias, at no time does the facilitator provide feedback on group responses.

Individuals first log on to the system to begin a group causal mapping session. After

successfully logging on, individuals identify concepts, define categories from the

meanings of the concepts, determine the relative importance of the categories, and

Causal Map Elicitation

Collective Causal Map Derivation

Cognitive Faction Identification

Figure 2. Cause map elicitation and cognitive faction identification steps

permission of Idea Group Inc. is prohibited.

indicate the influence of each category on other categories (Table 2). Each of the steps

is described.

Step 1: Concept Identification

The purpose of this step was to allow the participants to identify and exchange their

beliefs about the future direction of their firm. A framing statement was presented to the

participants to set the context for the brainstorming of concepts. The framing statement

used in this case focused on eliciting ideas about the future direction of the firm.

Specifically the statement listed four questions:

1. What do we want to accomplish in the next five years?

2. What is it that we do especially well?

3. What other things should we be doing especially well?

4. What present and future constraints do we face in our operations?

During this activity, the participants occasionally experienced a mental block. A stall

diagram also was used to alleviate this situation. The stall diagram allowed the participants

to cue themselves by presenting ideas associated with the future direction of their

firm (Figure 3).

Activity Description

1. Elicit Concepts

Introduction

Log-in screen and presentation of the framing statement and stall

diagram.

Concept Identification Elicit characteristics, concepts, and/or issues that contribute to strategic

situation of the firm in the case. Comments are shared among all

participants as they are entered.

2. Identify Categories

Category Identification

Elicit categories that group concepts by similarity; agree on category

definitions and names. Each participant verbally suggests a category

name and definition. Other participants comment on the name and

definition. The facilitator lists the names and records the definition using

the system.

3. Classify Concepts

Concept Categorizations

Each participant classifies the concepts into categories.

4. Rank Categories

Category Rating Step

Each participant rates each category on a 9-point scale, from important to

extremely important without knowing the responses of other group

members.

5. Define

Relationships

Identify Relationships

Each individual selects from list uses the system to identify causally

related categories. Each causal relationship is assigned a direction

(positive + or inverse -) and a strength from 1 to 3 for a scale of -3 to +3,

from strong negative influence to strong positive influence of one

category on another category.

Table 2. Causal mapping elicitation procedures

Step 2: Identify and Define Categories

The purpose of this step was to identify and define a set of categories (congregating

labels) that group the similar concepts identified in the previous step. Participants looked

through the list of concepts to identify those that address a similar issue or idea. The

participants then voluntarily proposed a category name and definition to the group. The

facilitator recorded the proposed category name on a chalkboard. This continued until

the group was satisfied with the group of the proposed categories. At no time did the

facilitator provide any guidance as to the completeness or correctness of the group of

categories suggested by the strategic planning team. The only types of comments by the

facilitator were to ask whether the team wanted to add, delete, and/or merge categories.

The facilitator, however, did not provide and suggestions as to what should be added

deleted or merged. Since this step was not anonymous, the facilitator took great care to

manage the power relationships that existed in this top management team2. This process

continued until the group was satisfied with the list of categories and their definitions.

This step is intended to allow the group to identify the sufficient congregating labels for

identifying the causal maps.

Step 3: Classify Concepts

The purpose of this step was to allow the participants to deepen their shared understanding

of the categories. In this step, the participants placed each concept into one of the

categories defined in the previous step. Concepts that a participant did not feel belonged

in one of the categories were placed into an Unknown category. This step was completed

when all participants had placed the concepts into categories.

Government

IT Services

Firm

Customers Competitors

Year 1 Year 2 Year 3 Year 4 Year 5

Figure 3. Stall diagram

Step 4: Rank Categories

The purpose of this step is two-fold. First, to understand the relative importance of

categories to the issues contained in the framing statement. Second, but equally

important, this activity is to increase the shared understanding of the meanings of the

categories for the participants, i.e., an attempt to ensure that the categories are indeed

sufficient congregating labels. In this step, the participants rated the importance of each

category to meeting the issues contained in the framing statement.

Step 5: Relationship Identification

Causal relationships are identified between categories. These relationships provide the

final component of creating individual causal maps. Relationships are identified without

viewing the relationships of other participants. In a causal map, an arrow indicates that

a participant perceives that a change in the originating category affects the terminating

category. To identify a causal relationship, the participant selects: (1) the origin category,

(2) the destination category, and (3) the direction (positive or negative) and amount of

influence (strong:3, moderate:2, or slight:1) that the origin category has on the destination

category. If the participant decides that they should not have included a relationship

that is currently in their map, they may remove it. The participant repeats these steps until

the participant is comfortable with the displayed map. At that time, the participant saves

the map to the system. This activity is completed when all participants have saved their

maps.

Collective Causal Map Derivation and Analysis

In our case, the process of deriving a collective causal map involves determining the

number of participants that identified each possible relationship between the categories.

This process is possible since the nodes of the derived collective maps have already been

agreed upon by the group. As such, the system simply derives a series of collective

causal maps from the individual maps by examining each of the possible relationships

among the categories. The number of participants that identified the relationship and the

average strength of the relationship are computed. The series of collective maps begins

with a map containing only the relationships identified by all participants (a consensus

map) and ends with a map containing relationships identified by any participant (a total

map).

There are many techniques in the literature for analyzing causal maps. To begin with, in

this study we analyze the complexity of the maps. A simple analysis of complexity is the

number of nodes and links in the map and the ratio of links to nodes in the map. Simply

put, the higher the ratio of links to nodes, the more complex the map (Eden, Ackermann

& Cropper, 1992; Knoke & Kulkinski, 1982; Narayanan & Fahey, 1990). Cognitive

centrality has also been used as a measure of the importance of a node in addressing the

issues in the framing statement (Eden et al., 1992). As such, we converted the cognitive

centrality of each node into a rank ordering of importance for each participant. This is

similar to the approach we used in converting the category ratings to rank orders.

Cognitive Faction Identification and Description

To identify the set of cognitive factions within the top management team, we used cluster

analysis to group the individual participant maps together. To compute the similarity of

one map to another, we computed Jaccard coefficients (Boyce, Meadow & Kraft, 1994)

based on the shared causal relationships among the participants, i.e., the structural

properties of the maps. The Jaccard coefficient was computed as:

# of causal relationships in Map 1 # of causal relationships in Map 2 # shared causal relationships in Map 1 and Map 2

# shared causal relationships in Map1 and Map 2

−

Since the nodes in the individual participant maps are identical, we only have to measure

the similarity of the relationships contained in the individual causal maps. In this case,

the more shared causal relationships, the greater the similarity is between the maps. For

cognitive faction identification purposes, we ignored the strength of the relationships,

e.g., if a positive relationship existed between two nodes in two different maps, but they

had different strengths, we decided the two maps shared a causal relationship. However,

we differentiated the relationships if they had different causal direction (positive vs.

negative/inverse). Based on the computed Jaccard coefficients between every pair of

individual maps, we clustered the maps together using Ward’s method. Ward’s method

was chosen to attain increased coverage of cases, improved handling of outliers, and to

minimize the effects of cluster overlap (Aldenderfer and Blashfield, 1984).

To provide further evidence of the cognitive factions uncovered, we used a set of

analytical techniques, beyond that of cluster analysis, to provide independent justification

of the identified cognitive factions. The analytical techniques included a productivity

measure based on the number of concepts generated by the members of the

respective cognitive factions. By looking at the productivity of each cognitive faction,

we ensure that we are not simply clustering together the members of the team with similar

levels of concept generation, e.g., those with a high-level of commitment to the process

and those with a lower-level of commitment. We also used the techniques described

previously with the overall group analysis: the complexity analysis, ratings, and cognitive

centrality. In this case, we applied the techniques to the cognitive factions and

compared the factions among themselves and to the overall group.

We also used Givens-Means-Ends (GME) analysis on the different cognitive faction

consensus maps to demonstrate that each faction had different causal insights. GME

analysis is an analytical technique that interprets the flow of causality as measured by

the ratio of the number of inflows divided by the number outflows for the categories in

the causal map (Bougon, Weick & Binkhorst, 1977; Eden et al., 1992; Weick & Bougon,

1986). Givens are identified by having more outflows than inflows of causal influence

 (ratio < 1), means as having approximately the same number of inflows and outflows (ratio

close to 1), and ends having more inflows than outflows (ratio > 1). Viewing the categories

in increasing order of this ratio shows the direction of causality in a causal map (Bougon

et al., 1977; Weick & Bougon, 1986). Givens represent the variables in the map that can

be manipulated to influence the Ends, which represent the goals of the participants. The

Means are moderators of the Given’s influence. GME analysis also can be used to identify

causal themes contained in a causal map.

Results

This section is organized along the steps of the methodology used: causal map

elicitation, collective causal map derivation and analysis, and cognitive faction identification

and description. The results from each step are given.