An Agent-based Method For Artificial Society Emergency Management Policies Analysis

Complex systems(e.g., human social systems, software-intensive sociotechnical systems) typically consist of a number of autonomous individuals that interact with each other and therefore result in macro behaviors with such characteristics as dynamic, uncertainty, and hard-to-reproduce or nonrepeatability, etc. Some emergencies(such as spread of an infectious disease, hacker attacks) in complex systems may result in ‘bad’ emergences at macro-level, e.g., the outbreak of the infectious disease and the breakdown of the system. These emergences may be seriously harmful, and have negative impacts on the stakeholders of the system and its individuals. Thus, how to provide effective method to support emergency management for such complex systems has become an important research subject in the areas of social science, management science, control science and computer science, etc. Social computing borrows the idea and technology of multiple disciplines like computer science and social science to study the social and computing behaviors of complex systems. It is an important method and means for the emergency management research.Aiming at addressing the emergency management issue in society, the thesis employs the social computing method to support the evaluation, analysis, refinement or selection of emergency management policies. Social computing supports computer experiments of emergency management through constructing computing model(i.e., artificial society) of the social system in the computer world, which borrows the idea of multi-agent system to abstract, model and simulate the artificial society. In term of the repeatable experiments the effectiveness of emergency management policies can be explicitly proved and these policies could be applied to the emergency management in real social systems. The main contributions of the thesis are described as follows.1. In order to support the evaluation, analysis and selection of emergency management policies, this thesis proposes a lightweight social computing approach- PZE(Policies selection, Zombie-city for artificial society modeling, and Evaluations of policies).To deal with the great challenge of emergency management and the shortcomings of the existing social computing methods such as being cumbersome and lacking of flexibility, this thesis focuses on the core work of emergency management(i.e., the selection of emergency management policies), borrows the ideas of iterative development and closed-loop control, and introduces the scenario/response emergency management mode. It proposes a lightweight social computing approach- PZE for emergency management policies selection, evaluation and refinement, which has such charateristics as simpleness, flexibility, easy-to-use, low-cost and so on. PZE approach consists of three parts:(P) Select appropriate policies based on the specific emergency scenario, i.e., policies selection;(Z) Construct artificial society with the Zombie-city model;(E) Perform experiments of emergency management with artificial society model to evaluate the effectiveness of emergency management policies and present the evaluation results. As the feedback of policies selection, these evaluation results provide effective supports for policies selection.2. To provide a general analysis model for emergency emergency management, this thesis proposes AVI model that takes agent, virus and interaction as its core concepts, and proposes the extended model of AVI-Zombie-citymodel.Due to local nonlinear interactions between individuals in the social system, an emergency(like infectious disease and rumor diffusion) will continually evolve, and the propagation is the intrinsic issue of the emergency evolution. In order to reveal and describe the nature of propagation issue, this thesis proposes a core model of artificial society- AVI model, including three core concepts: agent, virus and interaction. Agent is the propagation carrier that can carry viruses; Virus is an abstract concept that represents the propagation content; Interaction expresses the propagation channel, i.e., virus can be spread through interactions between agents. The thesis proposes an artificial society model for emergency management – Zombie-city model. Besides the above three core concepts, this model also introduces concepts like role, environment(physical environment and social environment) and rule. Role represents the abstract of a kind of agent, and agent can acquire different behaviors and attributes through playing roles. Meanwhile, agent can dynamically play roles to adapt to changes of the environment and itself. Physical environment is the space where agents live. Social environment mainly represents the social relationships between agents. Behaviors of an agent are affected by the environment. In addition, rule defines the behavioral and attribute norms of agent, virus, role and environment, and it can also be used to describe the emergency management policy. This model systematically describes artificial society for emergency management, and provides reference and conduct for artificial society modeling.3. To support and simplify artificial society modeling based on Zombie-city model, this thesis proposes a graphical artificial society modeling language – ASML(Artificial Society Modeling Language) and develops the corresponding artificial society modeling support tool- ASMLTools.To support and simplify artificial society modeling, this thesis proposes artificial society modeling language ASML, including meta-model, models, graphical notations and formal semantic, etc. ASML is based on the Zombie-city model and borrows the idea of social organization and mechanism of dynamically playing roles. ASML provides five kinds of models with different viewpoints, levels, and modeling meachanisms, including demographics model, environment model, role model, group structure model and interaction model. These models can facilitate the construction of artificial society models. ASML has the charateristics of high abstraction, natural modeling, multiple viewpoints, virtual modeling, easy-to-understanding, and strong expression ability, etc. Meanwhile, to support artificial society modeling with ASML, this thesis designs and develops artificial society modeling support tool – ASMLTools, and it can support for graphically editing model, storing model, loading and exporting model, model transforming, and so forth.4. To support for evaluating emergency management policies and then support policies refinement and selection, this thesis proposes an evaluation method combining with quantitative evaluation and qualitative analysis.The evaluation of emergency management policies plays a key role in finding whether emergency management policies are effective. It provides the feedback for policies selection or refinement. The thesis proposes a method combining quantitative evaluation and qualitative evaluation methods. As for the quantitative evaluation methods, the thesis provides a general method framework with structured steps: first extracting macro artificial society attributes based on scenarios, then integrating macro attributes with evaluation parameters to quantitatively evaluate the effectiveness of policies with numerical methods. Meanwhile, the degree of emergence indirectedly reflects the effectiveness of an emergency management policy. We classify emergences in artificial society, and propose three metrics for measure the degree of emergences based on information entropy. With simulations and experiments the effectiveness of these metrics is shown, and these metrics can be also applied in quantitative evaluation of policies. In the qualitative evaluation methods, the thesis proposes qualitative evaluation method for emergency management through formal reasoning based on theorem and rules. Firstly, we should construct artificial society formal model based on Zombie-city model and transform the policy into rules in this formal model. Then, we can make use of formal reasoning to qualitatively evaluate and analyze the effectiveness of policies.Thus, in order to support formal reasoning of artificial society, this thesis provides formalizations of Zombie-city model and formal specifications of scenarios in the artificial society. The qualitative evaluation method can assist the quantitative evaluation method, which can further confirm and show the evaluation results of the quantitative evaluation method. The evaluation method combining quantitative with qualitative ways can explicitly and correctly show the effectiveness of policies.The thesis studies several cases related with the emergency management in society, including the H1N1 spread in a campus and information spread in social media. The purposes of the case analysises and experiments are to show how to construct artificial society with Zombie-city model and the modeling language ASML, demonstrate how to use the PZE approach for emergency management and show the effectiveness of PZE approach.