Designers often want to understand how a building will perform before it is constructed. However, feedback on human behavior and experience is often unavailable during the early stages of architectural design, despite existing research in environmental psychology and post-occupancy evaluation. This dissertation addresses this gap by developing a cognitive agent, an evidence-based computational model of human behavior that offers designers meaningful human feedback. The research focuses on the problem of wayfinding in unfamiliar indoor environments, a widespread challenge in architecture that can lead to wasted time, increased costs, and user anxiety. Central to this investigation is the concept of perceived uncertainty, a critical factor influencing human navigation. The dissertation presents one scoping review, two empirical studies and two modeling studies. The empirical studies introduce a continuous measure of perceived uncertainty and highlight the importance of the rhythm of information acquisition in shaping navigation behavior and experience. These studies also generate rich datasets that support the development and validation of computational models. The scoping review identifies methods and opportunities for modeling human behavior in unfamiliar spaces. Two agent-based models are proposed. The first relies on rule-based heuristics and data-driven predictions of perceived uncertainty. The second is grounded in utility and information theory. The latter model, called PATH U.2, shows strong alignment with human participants by accurately reproducing route choices and the diversity of navigational paths. Together, this work advances understanding of the human wayfinding process in unfamiliar environments and introduces computational models that represent this process in a concise and interpretable form. These models offer new tools to support designers in creating environments that better respond to human needs and behaviors.