Enactive Genesis: Toward Generative Architecture with Human-Centric Artificial Intelligence
Enactive Genesis-Towards Generative Architecture with Human-Centric Artificial Intelligence.pdf (7.158Mb)
Access StatusFull text of the requested work is not available in DASH at this time ("dark deposit"). For more information on dark deposits, see our FAQ.
MetadataShow full item record
CitationTian, Runjia. 2021. Enactive Genesis: Toward Generative Architecture with Human-Centric Artificial Intelligence. Master's thesis, Harvard Graduate School of Design.
AbstractArtificial intelligence (AI) algorithms are gaining increasing popularity in the domain of architecture, urban design, and landscape architecture. However, most of recent generative design workflows using image-based AI such as generative adversarial neural networks do not incorporate human-centric evaluation metrics and are prone to potential bias embedded in the dataset researchers used to train AI agents. Moreover, the outcomes of such approaches are pixelated images that are not directly useable in real world scenarios.
Inspired by enactive learning in developmental psychology, the machine learning community has developed increasingly powerful AI agents that learn emergent behavior through unsupervised and reinforcement learning approaches such as self-play or actor-critic that do not rely on human heuristic datasets.
Therefore, I propose Enactive Genesis, a novel environment to train generative architecture AI agents through reinforcement learning and human-centric evaluation metrics. The environment is composed of three open-source Software Development Kits (SDK), each comprising a novel and foundational infrastructure towards general and human-centric AI in generative architecture design:
1. BIMGen: An open-source SDK that researchers could leverage to create generic architectural designs using a universal BIM grammar.
2. Promenader: An open-source SDK game engine asset that uses phenomenology as an explicit evaluation criteria for architecture design. We implement accessibility and phenomenal transparency as numeric evaluation metrics in an embodied architecture simulation.
3. EnGen: An open-source SDK for training intelligent agents to generate and modify grammar in component 1 according to human-centric evaluation metrics in component 2. EnGen allows user to perform enactive learning, using the generation system from step1, and the human-centric evaluation metrics from step 2 as loss function for artificial neural networks. With EnGen, AI agents can gradually learn emergent strategy for generating an architectural space that has high accessibility and phenomenal transparency value.
Citable link to this pagehttps://nrs.harvard.edu/URN-3:HUL.INSTREPOS:37367874