In large-scale three-dimensional environments, gas diffusion exhibits pronounced anisotropy, spatial non-uniformity, and temporal intermittency. These characteristics make it difficult for traditional infotaxis-based strategies to effectively characterize the dynamic evolution of uncertainty in environmental cognition and motion decision-making, thereby affecting search efficiency and decision stability. To address these challenges, a multi-agent cooperative search decision-making strategy based on cognitive uncertainty modeling is proposed. Within a unified Bayesian inference framework, the weighted particle variance of the posterior particle distribution is used to quantify environmental cognitive uncertainty, enabling dynamic representation of the cognitive state during the search process. On this basis, an information-potential-field jointly driven decision-making model is constructed. Furthermore, a variance-adaptive weighting strategy and an anisotropic adaptive step-size strategy are designed based on the cognitive uncertainty measure, allowing the search strategy to dynamically adjust the balance between exploration and exploitation, as well as the motion scale in different directions, in response to changes in the environmental cognitive state. In addition, a constraint update negotiation mechanism is introduced to enable decision exchange and coordination among agents, thereby reducing cooperative conflicts and path redundancy. Simulation results demonstrate that the proposed multi-agent cooperative decision-making strategy effectively improves search efficiency, decision stability, and robustness in odor source localization tasks.