Causal Mapping

К оглавлению
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 
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 

Our search for an analysis approach that would assist this reality led us to causal

mapping. The purpose in selecting causal mapping is not oriented toward defining central

tendencies in the data, but rather toward documenting the range of possibilities. In other

words, we believe it is too early to test the relative frequency of three theorized causes

of UML success when we expect there may be four, five, six, or more potential causes that

have not yet been documented. We would consider it too early to even attempt to

conclude that in some precise percentage of situations UML increased positive development

outcomes by some specific amount. Rather, we use causal mapping to illustrate

the range of variables that could influence decisions to implement UML as well as the

outcomes should UML be implemented.

The employment of causal mapping in our research setting would differ from most others.

For example, Nelson et al. (2000) use the method to define a fairly narrow set of constructs,

variables, relationships, and relationship strengths. Although we expected to find

constructs, variables and relationships, it is not our goal to demonstrate that our causal

map analysis outcome would be reliable across research settings. Rather, we were looking

for indications of these as they emerge from the data. Also, we believe that causal

mapping would document missing connections (e.g., sometimes it is valuable to know

what elements should be absent as well as present in an equation) and bring to light

relationships that may not be linear, where a construct at one level may cause negative

results, at another level have no impact, and at still another level cause positive results.

In conclusion, although we expected to find the same elements as in other research

employing causal mapping our objective was not reliability but validity with regard to

mirroring the complexity of organizational approach to UML.

Our use of causal mapping has similarity with Fahey and Narayanan (1989). We seek to

explore a complex phenomenon (UML) in the wider organizational setting. Our use of

causal mapping can be understood in the context of the basic research method challenge

of the personal pronouns of we, I, and you (Bohman, 2000). “We” represents the objective

approach to research — as in a deductive approach. “I” represents the individualized

approach to research — as in an inductive approach. “You” represents the real world

subjects and their views, attitudes or behaviors. Obviously, researchers using the voice

of “we” or “I” attempt to express salient aspects of “you.” Additionally, researchers

employ a method within which aspects of “you” are studied and analyzed.

The particular concern in this research was that causal mapping strongly emphasizes a

clear definition of the two elements of construct/variable and causal relationship

between constructs. Used in a straightforward and narrow manner causal mapping

represents the voice of “we” — a deductive and objective approach to research, as in

Nelson et al. (2000). The present researchers argue that subjects — the “you” in research

— may possess clear views of constructs/variables and causal relationships. However,

the presumption, in Nelson et al.’s (2000) research is that these will be parsimonious on

the individual level and across individual roles. We believe, however, that our domain

of interest, UML within the organizational context, does not have the pre-existing

background of prior investigation to allow for identification of those constructs/

variables and causal relationships before examining data, but rather using the data as the

source for discovery of these.

Therefore, our data collection method was not intended to define a set of constructs/

variables and causal relationships, as in Fahey & Narayanan (1989). Doing so would have

the danger in leading respondents to fill in the suggested topics without considering

whether the set of topics is complete. That is, since very many constructs/variables and

relationships exist in the real world, directing subjects to talk about a pre-selected subset

of these phenomena may lead to definitions that reflect the pre-selection rather than

represent the real world. In this scenario, the outcome might appear clean, but actually

hide the fact of missing elements. Therefore, given the lack of prior research guidance,

in this research subjects were encouraged to talk freely while guided toward the present

research focus. Our approach should and must be closer to the spirit of having “you”

talk as “you” find appropriate. Data collected in this manner would, hopefully, be a good

source for defining constructs/variables and causal relationships with a high degree of

validity.