Atypical buildings can give users a variety of spatial experiences compared to existing typical buildings. Thus, the ability to design atypical structures has become a vital aspect for architects. Users interact not only with the physical context of atypical spaces and other users but also with changes in the surrounding context. External circumstance changes are diverse, such as weather, fire, riots, and gatherings. Thus, to realize user simulation in a more realistic atypical space, it is necessary to develop an intelligent agent that responds to changes in various external situations. This study is part of a research project on the development of intelligent agent technology that responds to changes in external circumstances. Specifically, users' behavioral changes according to external situation changes in architectural spaces were investigated, and the psychological mechanism was theoretically analyzed. In addition, according to the specificity of an atypical space, the difference in response behavior to changes in external circumstances was theoretically considered. Through on-site and media investigations, it was confirmed that users' actions responded to changes in external circumstances, and it was assumed that there was a psychological cause for this. Moreover, it was assumed that there would be differences in psychological changes and response behaviors due to changes in external circumstances because atypical spaces have different physical and social situations compared to typical architectural spaces. This study theoretically analyzed the psychological mechanisms of users' behavioral changes according to external situation changes in atypical architectural spaces. According to the specificity of an atypical space, the difference in response behavior to changes in external circumstances was also theoretically considered. And also, several strategies for the advancement of human behavior simulation can be derived. First, in the simulation of human behavior in an atypical space, the scope and method of the agent's perception should be adjusted based on the theory of direct perception so that the simulation can be highly realistic. Second, according to the positive and negative changes in the external situation, parameters that affect the agent's reaction behavior, such as pathfinding, should be developed and applied to the simulation. Third, it is necessary to develop an environment that can perform human behavior simulation by making changes in external circumstances within the design and developing agents that respond in various ways. The results of this study are meaningful, as they can be used as a basis for future human behavior simulation and to advance agent intelligence.