| The purpose of this research has been to develop a generic decision support system to aid managers in government Research and Development (R&D) organizations in the high level management decision concerning in-house versus extramural (contracted out) performance of research projects. This decision problem belongs to a larger class of unstructured, complex management decisions which cannot be addressed by traditional decision models.; The research adapts the systems approach for determining the set of projects to be conducted in-house and those to be conducted extramurally under the specific constraints imposed on the organization by its environment. The decision support system, LIBRAS (Logical Inference Based Resource Allocation System), was developed through enhanced versions with the active participation of three government R&D organizations. Artificial Intelligence techniques are used to elicit and represent decision makers' judgmental knowledge and preferences, and to incorporate intangible factors into the decision model.; LIBRAS software is implemented in IBM Personal Computer (or compatible) environment and consists of four interactive components written in Expert Choice, Personal Consultant Plus, and dBASE III Plus software. The first component determines the weights of the decision criteria. The second component evaluates and prioritizes the projects in terms of the desirability of in-house option based on a weighted additive model using the established criteria weights. Projects are evaluated on each criterion by a rule-based system which implements the preference structure of the decision makers formalized in the decision tables format. This format ensures complete, internally consistent, and coherent specification of preferences and is capable of addressing conditional criteria.; The Resource Allocation component allocates available resources to the list of prioritized projects consistent with the objectives, priorities, and constraints of the organization. It uses a heuristic approach that involves the use of an Artificial Intelligence search technique called dependency-directed backtracking. The fourth component of LIBRAS is the system database that serves as a communication pad between system component and stores system input and output.; LIBRAS has been designed, developed, implemented, and proves to be a comprehensive decision aid for the choice between in-house versus extramural performance of projects in government R&D organizations. The methodologies used in LIBRAS for preference structuring and resource allocation are recommended for use in modeling similar unstructured, multi-criteria decisions. |
