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 

The main purpose of this chapter is to demonstrate the use of evidential reasoning

approach under Dempster-Shafer (D-S) theory of belief functions (Shafer, 1976; see also,

Srivastava & Datta, 2002; and Srivastava & Mock, 2000, 2002) to analyze revealed causal

maps. The Revealed Causal Mapping (RCM) technique is used to represent the model

of a mental map and to determine the constructs or variables of the model and their

interrelationships from the data. RCM focuses on the cause/effect linkages disclosed by

individuals intimately familiar with a phenomenon under investigation. The researcher

deliberately avoids determining the variables and their associations a priori, allowing

both to emerge during the discourse or from the textual analysis (Narayanan & Fahey,

1990). In contrast, other forms of causal mapping begin with a framework of variables

based on theory, and the associations are provided by the participants in the study (cf.

Bougon, et al., 1977).

While RCM helps determine the significant variables in the model and their associations,

it does not provide a way to integrate uncertainties involved in the variables or to use

the model to predict future behavior. The evidential reasoning approach provides a

technique where one can take the RCM model, convert it into an evidential diagram, and

then use it to predict how a variable of interest would behave under various scenarios.

An evidential diagram is a model showing interrelationships among various variables in

a decision problem along with relevant items of evidence pertaining to those variables

that can be used to evaluate the impact on a given variable of all other variables in the

diagram. In other words, RCM is a good technique to identify the significant constructs

(i.e., variables) and their interrelationships relevant to a model, whereas evidential

approach is good for making if-then analyses once the model is established.

There are two steps required in order to achieve our objective. One is to convert the RCM

model to an evidential diagram with the variables taken from the RCM model and items

of evidence identified for the variables from the problem domain. The second step is to

deal with uncertainties associated with evidence. In general, uncertainties are inherent

in RCM model variables. For example, in our case of IT professionals’ job satisfaction,

the variable “Feedback from Supervisors/Co-Workers” partly determines whether an

individual will have a “high” or “low” level of satisfaction. However, the level of job

satisfaction will depend on the level of confidence we have in our measure of the variable.

The Feedback from Supervisors/Co-Workers may be evaluated through several relevant

items of evidence such as interviews or surveys. In general, such items of evidence

provide less than 100% assurance in support of, or negation of, the pertinent variable.

The uncertainties associated with these variables are better modeled under Dempster-

Shafer theory of belief functions than probabilities as empirically shown by Harrison,

Srivastava and Plumlee (2002) in auditing and by Curley and Golden (1994) in psychology.

We use belief functions to represent uncertainties associated with the model variables

and use evidential reasoning approach to determine the impact of a given variable on

another in the model. This combination of techniques adds the strength of prediction to

the usefulness of descriptive modeling when studying behavioral phenomena. Evidential

reasoning under Dempster-Shafer theory of belief functions thereby extends the

impact of revealed causal mapping.

The chapter is divided into eight sections. Section II provides a brief description of the

Revealed Causal Mapping (RCM) technique. Section III discusses the basic concepts

of belief functions, and provides an illustration of Dempster’s rule of combination of

independent items of evidence. Section IV describes the evidential reasoning approach

under belief functions. Section V describes a causal map developed through interviews

and surveys of IT employees on their job satisfaction. Section VI shows the process of

converting a RCM map to an evidential diagram under belief functions. Section VII

presents the results of the analysis, and Section VIII provides conclusions and directions

for future research.