Appendix: How can AI-HI collaboration improve the quality of emergency plan generation? ——Research on Stackelberg Game Model

With the development of natural language generative AI big model technology, AI with superb semantic understanding such as DeepSeek has become the infrastructure of the digital economy and society. The synergy between AI and HI has the characteristics of optimising time-intensive processes, reducing decision-making bias and mitigating all kinds of systematic risks, which is conducive to the enhancement of timeliness of emergency plan generation in highly dynamic environments, the effectiveness of and the operability. In emergency management practice, government departments need to weigh the costs and benefits of introducing AI. In this paper, we first consider the benefits and budgetary costs of human-computer collaboration between government departments and social organisations, and construct the Stackelberg game base model of AI-HI collaboration for emergency information sharing and emergency plan generation system, and obtain the equilibrium decision-making and optimal benefits of government departments and social organisations. Subsequently, the robustness of the model is enhanced by further extending the base model into three scenarios (customised AI, emergency plan amendment and data trust). The findings suggest that AI investment is not always beneficial and government departments need to choose whether to invest in AI technology based on their own budget. For government departments, customized AI should only be adopted when the marginal cost of AI technology investment is less than a certain critical value, and if the accuracy of an emergency plan is low, it is better to use manual preparation rather than human-computer collaboration followed by correction. For social organisations, the adoption of customised AI is always more beneficial than generic AI, and the amendment of the contingency plan has no impact on the optimal benefit of the social organisation. In addition, government departments should only adopt the data trust model if the benefits of improved digital governance capabilities outweigh the losses of human-computer collaboration.