Built environments are influential in shaping human experiences, such as motivation to work, stress/anxiety, and pleasure. Subtle differences in architectural design feature configurations (e.g., texture of surfaces, symmetry of building components) influence the resulting human experiences. This paper is part of a larger study that aims to quantify human experience in designed spaces, and evaluates how these subtle differences are perceived by people with different demographic backgrounds (e.g., age group and occupation). Through a data-driven approach, this paper examines how different configurations of design features related to stress and anxiety can result in people's decision of preferring one space over another. A crowdsourcing study was designed and administered on a platform with 296 participants without informing the participants about the differences in these spaces. During the study, the subjects were asked to select a preferred space out of two options, presented as dual-images rendered from 3D models of real buildings. Each image for a space presented a design feature configured differently, while keeping other features constant, to give alternate experiences (i.e., positive or negative), depending on the aspect attributed to that feature in the literature (e.g., poor lighting being attributed to negative feelings and vice versa). A total of six architectural design features, resulting in twelve paired spaces, were evaluated. This paper analyzes the collected data to identify configurations of preferences across design features using unsupervised learning algorithms. The results showed four clusters of preferred configurations of the design features. When the demographics of participants across clusters are analyzed, it is apparent that age and education level have little influence on the preferences of design features, while occupation is impactful for people's selection of desired spaces. The outcome can be used as a design guidance for architects, given the demographics of the prospective occupants.