The persistent shortage of affordable housing in the United States, coupled with aging infrastructure and rising energy costs, disproportionately impacts low-income households, particularly in historically disinvested communities. Addressing this challenge requires innovative, scalable solutions that balance affordability, energy efficiency, and climate resilience. This study introduces the BUILT2AFFORD dashboard, an integrated tool leveraging machine learning (ML) and Google Street View (GSV) imagery to pre-identify low-cost passive retrofit strategies for preserving and improving affordable housing. The dashboard assesses existing building conditions, evaluates energy retrofit potential, and mitigates heat-related health risks. The research involved on-site audits of single-family homes and Section 8 apartments, combined with continuous monitoring of indoor environmental conditions such as temperature, humidity, and CO2 levels. Preliminary results reveal significant opportunities for energy savings and improved thermal comfort, with scenarios like infiltration reduction and insulation upgrades achieving substantial reductions in energy use. The dashboard’s validation through eight testbeds demonstrates its potential to address housing challenges in South Bend, Indiana while advancing carbon-neutral and equity goals. This paper highlights how integrating advanced technologies into retrofit planning can enhance housing quality, reduce displacement risk, and foster climate resilience in vulnerable communities.

Integrating generative Artificial Intelligence into architectural design has opened new possibilities for creative exploration, particularly in the early stages of spatial research and aesthetic development. Since 2022, the increasing accessibility of AI platforms has altered how designers work, enabling the translation of text prompts into images, videos, and even three-dimensional forms. While current AI tools generate fragmented visions with limited spatial coherence, these constraints offer productive opportunities for innovative ways to engage AI as a creative collaborator. “Fabricated Combine: Animated Surface”—design research establishing a computational workflow bridging data formats and dimensions between two- and three-dimensional realms—expands creative outcomes through an AI-informed approach. In this context, AI serves as an input mechanism, using text prompts to shape the creative direction. These prompts do not generate final designs but evolving possibilities that blend aesthetic architectural language, motifs, patterns, and social narratives. This approach facilitates the rapid generation of conceptual designs, minimizing the mediation between AI-driven sketching and imaginative translation into spatial visions. The process begins with AI-generated images transforming into 3D models through a series of steps utilizing projective and depth maps. These steps produce “combined surfaces,” hybrid composites refined through iterative loops as animated point clouds, eventually converted into operable meshes for optimization and materialization through digital fabrication. Animated point clouds derived from 2D sources—whether from GenAI-generated images or two-dimensional representations of physical models—facilitate a seamless generative workflow, allowing design iterations to occur at any stage. As Artificial Intelligence excels in accelerating data variations, iterations, and optimizations, the designer’s role remains fundamental in refining details and establishing contextual connections. Integrating generative AI with procedural modeling tools, “Fabricated Combine: Animated Surface” showcases a cross-dimensional workflow that deepens design exploration. Ultimately, this project highlights AI’s transformative potential in architectural practice while reaffirming the designer’s essential role in shaping and grounding AI-generated outcomes.

Exposure to biophilic design has many documented health benefits. Even digital exposure to biophilic design can reduce physiological stress. This restorative process has tremendous potential for cognitive and neural health benefits. However, the cognitive-affective mechanisms that underlie these benefits are still being explored. A key method in the emerging field of neuroscience research on the built environment (including biophilia) is electroencephalography (EEG). EEG provides widely validated and efficient (low-cost, easily implemented) measures of cognitive-affective processing and looks to be well-positioned to serve as a central approach in this emerging area. The present project reviews the literature of environmental design studies seeking EEG responses to biophilic design stimuli. A total of twelve papers were found and reviewed from PubMed, Web of Science, and Scopus following PRISMA. The research highlights the EEG, as well as cognitive measures, implicated in biophilic design research. Understanding how well biophilic design targets attention restoration and stress reduction can move us closer to understanding and confidently designing truly restorative spaces. Results from this systematic review show that current studies target neural markers associated with attention restoration and stress reduction but that ties to larger mechanisms are missing.

The precise prediction of dioxide (CO_₂) emissions from buildings is essential for improving sustainable energy practices and mitigating environmental damage. This paper introduces a machine-learning prediction model, an AI subset, for calculating CO₂ emissions in 140 buildings on the campus of the University of Maryland, College Park. The model employs a synthesis of actual building data and simulated solar radiation data. The merging dataset has attributes including the year of building, building area, energy use intensity (EUI), total energy consumption, CO₂ emissions, and seasonal solar radiation values for each building. Data preprocessing included mean imputation for missing values, feature normalization, and the creation of interaction terms to elucidate complex interactions, such as the influence of solar radiation on energy use. Furthermore, we considered feature engineering in order to improve the prediction model's accuracy. Three machine learning models, including Random Forest (RF), Support Vector Regression (SVR), and K-Nearest Neighbors (KNN), were utilized to predict CO₂ emissions, with performance assessed by R-squared, mean squared error (MSE), and mean absolute error (MAE). The results demonstrate that RF surpassed the other models. Attain a high accuracy with an R-squared value of 0.966 on the test set. Conversely, SVR and KNN exhibited inferior performance, with R-squared values of 0.600 and 0.698, respectively. The findings indicate that machine learning algorithms, especially RF, can accurately predict CO₂ emissions in university buildings, providing a significant perspective for energy management and sustainable efforts.