Individual differences and career decision factors: A Bayes net representation

The theoretical basis for this research, provided by Janis and Mann (1977), indicates the plausibility of identifying career decision styles based on personality constructs and individual differences. The focus of this dissertation is three-fold: (1) the proposal of a dynamic decisions career model (DDCM) that incorporates a Bayesian Belief Network (BBN) for the presentation of career decision factors. The BBN is a graphical system that allows direct manipulation of a suggested model by providing the means to explore the impact of various plausible scenarios on that model (2) the description of various modeling techniques used in the development of the proposed career decision model, including cluster analysis, and path associative networks and (3) an empirical study designed to test the predictive ability of the Individual Differences Module in the DDCM. The presented research describes procedures used to develop a preliminary model and decision support tool within the single problem domain of career decision making.Results of this study provide evidence for the use of cluster analysis and path associative networks in the development of a BBN structure when domain expert knowledge is unavailable. The BBN structure, developed as part of the DDCM, represents the integration of individual differences that influence career decision styles. Entry, into the BBN, of individual difference variable values results in a set of decision style probability scores. The models ability to predict the most probable or dominant decision style (identified with an independent criterion measure) was confirmed with a Chi-square goodness of fit test.The developed BBN will update values for all variables in the structure when new information is added. This feature reflects the dynamic component of any decision process and can provide decision makers with important information regarding the impact of new information on the value of all components in the decision process. The research presented in this dissertation is not the first to support use of a Bayesian Belief Network, but it is the first research to demonstrate an alternative set of techniques for development of a BBN structure and provide evidence for the predictive validity of the developed structure.