Causal Map Elicitation
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67
68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84
85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101
102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118
119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135
136 137 138 139 140 141 142 143 144 145 146
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.