Decision-making in urban infrastructure management during extreme events relies heavily on human operators, yet current computational support systems often fail to account for non-monotonic human adaptation and latent psychological biases like overconfidence and defensive overcorrection. This study addresses this gap by integrating Instance-Based Learning Theory (IBLT) into the domain of civil engineering computing. We establish a computational cognitive architecture that simulates operator decision processes through the mathematical mechanisms of memory retrieval and utility blending. This model functions as a computational baseline, representing boundedly rational adaptation driven by experiential priors, thus allowing for the algorithmic isolation of latent psychological biases from the baseline dynamics of memory-based learning. We demonstrated this framework using a human-in-the-loop microworld experiment simulating subway flood-induced track suspensions, where dispatchers must balance passenger safety against service efficiency. Analysis revealed a complex, non-linear human adaptation cycle consisting of four phases: acquisition, overconfidence, overcorrection, and recalibration. Specifically, the computational model exposed a significant divergence during the post-accident "overcorrection" phase: while human operators exhibited immediate, defensive risk overestimation, the model maintained a stable trajectory based on accumulated experience. This strategic divergence confirms that operational instability following failure is often attributable to acute psychological bias overriding stable memory-based adaptation, a pattern theoretically expected to recur across analogous high-stakes environments and validatable through multi-modal behavioral and sensor data from professional operators.

翻译：极端事件期间城市基础设施管理的决策高度依赖人类操作员，但当前计算支持系统往往未能考虑非单调的人类适应性及过度自信、防御性过度修正等潜在心理偏差。本研究通过将基于实例学习理论（IBLT）整合到土木工程计算领域来填补这一空白。我们建立了一个通过记忆检索与效用混合的数学机制模拟操作员决策过程的计算认知架构。该模型作为计算基线，表征由经验先验驱动的有限理性适应性，从而允许从基于记忆学习的基线动力学中算法性分离潜在心理偏差。我们采用模拟地铁水灾导致轨道中断的人机环微世界实验验证该框架，实验中调度员需平衡乘客安全与服务效率。分析揭示了由四个阶段组成（获取、过度自信、过度修正与重新校准）的复杂非线性人类适应周期。具体而言，计算模型在事故后"过度修正"阶段暴露出显著偏差：人类操作员呈现即时的防御性风险高估，而模型则基于累积经验维持稳定轨迹。这种策略性偏差证实，失败后的操作不稳定性常源于剧烈心理偏差压倒稳定的基于记忆的适应性——该模式理论上会在类似高风险环境中重现，并可通过专业操作员的多模态行为与传感器数据加以验证。