The quest for computational design assistants has long been cultivated in architectural design research, given human designers’ natural cognitive limitations. However, hitherto developed design computing systems can hardly fulfil the quest to aid the architectural creative design process effectively. One significant reason holding back the fulfilment of this quest is the disconnection between existing design computing systems and theories of design cognition. Meanwhile, the implication of conceptual space as an essential concept in developing both design cognition theories and design computing systems is significant. Nonetheless, prior research has had limited focus on the interpretation of the conceptual space per se, as significant knowledge gaps about feature representation remain, thus leaving the intrinsic structure and mechanism of conceptual space vague and indeterminate.
In her thesis, Jielin aims to bridge the gaps by pursuing three objectives listed as follows: first, re-examine the theoretical foundation of conceptual space from the perspective of design cognition; second, address the knowledge gap of conceptual space interpretation by leveraging the learning and interpretation capacity of data science-empowered deep representation learning models based on both Euclidean and non-Euclidean design data sources; third, provide a practical blueprint of a new paradigm of interactive design computing systems, aiming at broadening the application scope of current design assistant systems.
