Spatial cognitive patterns of underground public spaces from a stimulus-organism-response perspective: evidence from LLM-empowered social media data mining

Current underground public space (UPS) renewal prioritizes engineering-oriented physical upgrades, neglecting users’ multidimensional cognitive experiences. Existing studies on underground space perceptions rely on small-sample surveys and fail to capture joint co-occurrence patterns among multiple stimuli. This study integrated the Stimulus-Organism-Response framework with large language models to decode spatial cognitive patterns. It drew on 76,078 social media reviews from 64 UPSs in 19 Chinese cities (2005–2024). DeepSeek extracted 27 spatial stimuli and their sentiments, serving as public perceptions. Generalized ordered probit regressions showed that management quality, environmental conditions, and atmospheric experiences exhibited the strongest associations with overall ratings. Negative perceptions showed stronger negative associations with ratings than positive perceptions showed positive associations. Association rule mining uncovered co-occurrence patterns across genders, city clusters, and review periods. Dissatisfaction was associated with shorter, more concentrated patterns of negative stimuli, whereas satisfaction was associated with the co-occurrence of positive functional and experiential perceptions. Co-occurrence patterns varied markedly across organismic attributes. Males showed simpler, function-oriented networks, while females displayed more complex, multi-dimensional ones. Regional expectation thresholds differed across contexts. Temporal patterns shifted from physical defects to operational management. These findings support a cognition-driven renewal paradigm, shifting from engineering-oriented interventions to weakness prioritization, multifunctional experience design, and inclusive context-specific strategies.