A formal model for maintaining consistency of evolving bureaucratic policies: a logical and abductive approach
KayLiang Ong · 1992
A bureaucracy can be viewed as a set of policies governing the actions of the people within an organization. This set of policies tends to grow sticky upward and more complex through time with the addition and deletion of policies. Thus, eventually, many unexpected side effects introduced by these updates lead to different forms of inconsistencies (conflicts) in the policies. In fact, Meyer (1985) has shown that such phenomena occurs because bureaucratic systems lack the mechanisms for detecting and removing such inconsistencies during the evolutionary process. In this study, we investigate a formal approach using information technology to assist policy makers managing and controlling the evolution of policies in a bureaucracy. Automated detection and explanation mechanisms are provided to eliminate inconsistencies in policies incrementally as new policies are added or old policies are removed. A logic model based on horn clause logic and deontic logic was developed that represents policies as first order logical predicates and rules. Consistency of the policies is defined based on a set of meta-level logical statements call integrity constraints that must be satisfied in all evolutionary phrases of the policies. Thus, whenever there is an update, inconsistencies of these policies can be detected if any of the integrity constraints is proven false and explanation can be provided based on the proof tree. Moreover, a more powerful inference mechanism based on the theory of abduction extends the detection process to explore inconsistencies of policies in future scenarios and thus, maintaining potential consistencies of these policies. The outputs of this study have both theoretical and practical contributions. Theoretically, a formal framework for modeling policies is developed. The framework provides a sound logical inference mechanism for detecting and explaining inconsistencies. Furthermore, the inference mechanism is extended with the abductive capabilities to detect potential inconsistencies through simulation of future scenarios. On the practical side, a prototype system called the ALP system is implemented. It is used to develop two sizable applications for validating the feasibility of the approach used in this study. One of the applications is based on a subset of the actual policies in the library lending code. In short, this study has demonstrated that bureaucratic policies can indeed be automated and that it is indeed feasible to provide information technology to assist policy makers in eliminating inconsistencies in these policies.