Existing decision-making models are generally based on the classical normative paradigm, which seldom considers the limited cognitive and environmental resources of humans when making decisions in real environments. We draw on the theory of Artificial General Intelligence technology, Non-Axiomatic Reasoning System and heuristic-primed decision-making processes of psychology to propose a heuristic-primed decision-making model under the assumption of bounded resources. Firstly, we propose a spreading activation belief network based on experience-grounded semantics, which takes concepts as nodes and semantic relevance as edges, and changes the weights of logical reasoning links according to the evidence presented using multiple inferences rules so as to achieve the self-organization of its network topology. Next, multiple dynamically changing and conflicting goals are represented by a Non-Axiomatic Logic statement with two-dimensional desire-values. Finally, a controlled concurrent heuristic algorithm that can dynamically allocate limited computational resources and perform multiple inferences is proposed.

Existing fire emergency decision-making models often struggle with accurately handling multi-granularity uncertain linguistic information, loss aversion, and a lack of adaptability to dynamic fire evolution. To address these gaps, this study adopts two-tuple linguistic representation (2TLR) for quantifying multi-granularity linguistic information and combines the Analytic Hierarchy Process (AHP) with the entropy weight method (EWM) to determine the ability weights of the experts. Furthermore, a six-dimensional dynamic reference point is generated via the random forest algorithm, and the integration of prospect theory (PT) with a sequential decision-making framework (SDF) is implemented for the dynamic optimization of response plans. Validation through real-world cases demonstrates that the proposed Multi-stage Prospect Selection (M-PS) model outperforms both the TOPSIS method and the single PT model, compared with these two methods, the proposed M-PS model can effectively prioritize the avoidance of high-risk scenarios, accurately reflect decision-makers’ loss aversion tendency, and realize dynamic decision-making through updating the decision plan sequentially, thereby providing reliable support for urban fire emergency management. At the same time, in this study we conduct a comparative analysis of core metrics between existing methods and the proposed M-PS model. The evaluation across five dimensions demonstrates that the proposed M-PS model delivers superior performance.

Context: In the realm of Smart Firefighting, research on multi-attribute decision modeling has yet to develop a decision-making model for AI-assisted fire rescue plans that integrates both subjective and objective information. Meanwhile, while studies using intuitionistic fuzzy sets provide valuable support for multi-attribute decision-making, they often focus only on data fuzziness and neglect decision-makers’ preferences for alternative plans and attributes. Objective: The primary objective of this study is to develop a sophisticated multi-attribute decision-making model tailored specifically for urban fire emergency decision-making scenarios that integrates both subjective and objective information. Method: To overcome the limitations of previous research, we develop a multi-attribute decision-making model for firefighting that integrates entropy weights and preference weights with intuitionistic fuzzy sets. Specifically, we first preprocess the decision attributes and experts’ preferences, transforming real-valued attributes into intuitionistic fuzzy numbers and determining the attribute weights based on expert preferences. We then introduce an overall framework for multi-attribute decision-making, comprising two models: the Fuzzy Entropy Weighted Multi-Attribute Decision-Making model, FEW-MADM and the Intuitionistic Fuzzy Sets based Integrated Entropy-Weighted and Preference-Weighted Multi-Attribute Decision-Making model IFPS-EWMADM, which realize the harmonious balance between data-driven insights and human expertise. Results: We demonstrate an illustrative example in urban firefighting to evaluate the effectiveness of the proposed models. The results show that the FEW-MADM model is suitable for situations where the plan dataset contains no unexpected values, while the IFPS-EWMADM model is appropriate for scenarios involving complex and uncertain data. In addition, we conducted a sensitivity analysis, focusing on the conditions under which the original optimal solution would remain effective when key data in the case fluctuated within different ranges. Conclusion: The proposed models calculate optimal solutions for various fire area sizes, which not only aid emergency management personnel in devising effective dispatching plans but also enable the analysis and summarization of historical cases. This dual functionality significantly propels the development of “smart firefighting” and markedly enhances the overall efficiency and effectiveness of fire emergency management.