Sustainable design is growing exponentially in architecture as socially and environmentally aware structures become necessary. The field battles with ever-changing design criteria, design bias, and environmental uncertainties, introducing a new challenge of constant change to sustainable design. Exploring these challenges may open up a technological pathway in Artificial Intelligence for architects struggling to adapt to these changes, creating a more efficient solution to the problem. This study will investigate these technological opportunities by reviewing 54 sources analyzed by research methods such as quantitative analysis to investigate existing data, qualitative analysis to study the theoretical aspects of the issue, and CRAAP analysis to examine their results' relationship between the field, users, and machine learning models. Each source is filtered and analyzed through the themes that are highlighted within each individual source but also propose the pillars of machine learning implementation into the field. The sources are defined through the following themes: supervised machine learning affected by design bias in artificial intelligence, and supervised machine learning as a tool for problem solving. By pushing the boundaries of the relationship between technology and architecture, how might we use supervised machine learning as a tool for architects to find immediate sustainable solutions for their commercial and residential projects, even with design bias and design criteria constantly changing?

In Design Integration: Global Technological Advancement and Local Culture, the authors state that optimization is a concern for designers and scholars implementing sustainable design and manufacturing techniques in their work.[5] This topic expands on the issue of unsustainable practice in architecture as a reaction to ever-changing design criteria. Developing sustainable design methods with artificial intelligence (AI) can improve how architects adapt it to their work efficiently.

Sustainability, as defined architecturally by the Brundtland Report, refers to “Development that meets the needs of the present without compromising the ability of future generations to meet their own needs.”[4] Sustainability considers social, environmental, and economic factors, making it the future of design focus. It is important to realize that artificial intelligence is not a replacement for human intelligence; instead, it is a machine whose algorithms are influenced by humans’ passive or active consumption and production decisions.[6]

AI platforms, specifically supervised machine learning, can empower designers to design sustainably by quickly reviewing and synthesizing sustainable building methods affected by varying design criteria that can offer insights for their projects.

  Machine learning consists of models learned from data that are either supervised, unsupervised, semi-supervised, or reinforced. A supervised machine learning model responds with precise results based on the labeled, positively biased data fed into the algorithm. It consists of a multi-step model, its steps include: gathering input data, evaluating input data, selecting algorithm model, processing input data, training the data through the algorithm, validating the model, monitoring the output, evaluating output data, reworking model parameters, reevaluating output data, and lastly receiving accurate output.

   Others may ask, is machine learning needed to be successful in this? While architects may be able to create sustainable responses to problems in architecture, they cannot store all this information, patterns, and algorithms that AI does, therefore slowing the process and forcing them to work at the small-scale level instead of a more extensive overall solution.

Keywords

Artificial Intelligence

Machine Learning

Sustainable Architecture

Technology

Energy

Fabrication

Resources

Algorithms

Sustainable Design

Bias

Supervised Machine Learning

Consciousness

Problem-Solving

building methods

Materiality

Building Codes

Design

Criteria

Methodology

 Through the research, it becomes clear that there are recurring themes. Many sources investigate the relationships between the user and machine learning models. These relationships explore the negative versus positive impacts, the idea that more flexibility in the algorithms leads to more of a voice for architects or a more biased influence, and they explore the differences between biased and unbiased uses and responses of the tools. Others solely explore the existing uses of machine learning in the field of sustainable architecture as problem solving method; some of which include monitoring and processing a building's energy output and surveying out to reduce the increased outputs.

Supervised Machine Learning Affected by Design Bias:

Bias in Machine Learning is attributed to the human influence on its code, algorithm, or responses. An example, as stated by Amira Al-Khulaidy Stine and Hamdi Kavak, is “Two individuals having identical attributes, with the exception that one individual is male and the other female, or where two individuals have the exact attributes with only “race” being the differing attribute.”[1] AI users rely on the tool to respond with an accurate solution to the problem rather than allowing the tool to aid a step towards finding a solution. The future-generated responses will remain biased once a biased prompt is fed to a supervised model.

  The sources in this section refer to language as a driver of bias. By simplifying prompts and information input into algorithms to the most active voice and least stylistic prompt, there is a lesser margin of error for the limit of bias. Data bias is affected by this issue and is also impacted by representation, measurement, and intention issues.

  Algorithmic bias, as defined in Risk and the Future of AI: Algorithmic Bias, Data Colonialism, and Marginalization is the “Phenomenon by which an algorithm may perform particularly poorly on a population subgroup if it was not exposed to that subgroup's data during algorithm development and training.”[2] This bias underrepresents certain social groups, which must be addressed in sustainable design.

  In some cases, Bias can be seen as a positive intention, specifically through Design Bias.

Supervised Machine Learning as a Tool for Problem Solving:

Individuals using supervised machine learning or those implementing it in the working environment must be informed about AI and its capabilities. AI literacy, as explored by existing research, “Investigates the extent to which digital divide, cognitive absorption, and computational thinking skills affect...AI literacy.”[3] This knowledge leads to a user understanding of how machine learning can be used for problem-solving.

Problem-solving involves expertise and a vast knowledge of different topics. The source, Opportunities of artificial intelligence for supporting complex problem-solving: Findings from a scoping review, investigates how the experience and expertise of human intelligence work together with the infinite data artificial intelligence holds can be used as a process for problem-solving. The separate roles the users and the AI take in the combined method, or what leads to successful complex problem-solving. The existing research highlights the components of complex problem-solving that have been positively affected by machine learning.

There is a gap in the current research on how problem-solving using AI is implemented into sustainable design, while also considering design bias. Combining the two themes will lead to possibilities for AI implementation in architecture.

Supervised Machine Learning Model:

When applying the previously discussed findings it is important to consider how may Design Bias be positively harnessed into a supervised machine learning model. The data must be selected and validated, while taking into account the subgroup's perspective. This data must abide by the sustainable architecture standards while incorporating the subgroup's sustainable intent. The figure below highlights the multi-step process of a supervised machine learning model.

Through the extent of this study, it is apparent that there is no existing overlap or combination between the study of bias in supervised machine learning, the integration of machine learning in the sustainable architecture field, and the use of machine learning as a problem-solving tool. It has been implemented in the field without regard for its relationship with its user and how the user or architect in the field will have a say over what the model is generating. By continuing to analyze existing studies, patterns will emerge, creating a line between the two components of successful integration of supervised machine learning into the field, which still interacts with the user but adapts to changes in criteria while doing so.

Each source touches upon the user's relationship with the machine learning model or Artificial Intelligence's relationship with the field in which it is being used. Some studies found that using generative AI will usually lead to biased outcomes, while using deductive AI will have a more factual and direct response. Other studies focused on the positive and negative impacts of supervised machine learning in the field of architecture by analyzing the different uses in smart buildings, spatial layout, and energy efficiency. By exploring these two topics, the study of the integration of supervised machine learning as a tool in the field will be solidified. Moving forward, a case study conducted through an ethnographic method or cyber ethnographic method may be the most successful way to address this problem because it will consider the users’ behavioral issues in relation to the chosen supervised machine learning model; while also considering the technological implications the machine learning models may have been studied over time.

References

[1] Al-Khulaidy Stine, Amira, and Hamdi Kavak. 2023. “4 - Bias, Fairness, and Assurance in AI: Overview and Synthesis.” In AI Assurance, edited by Feras A. Batarseh and Laura J. Freeman, 125–51. Academic Press. https://doi.org/10.1016/B978-0-32-391919-7.00016-0.

[2] Arora, A., M. Barrett, E. Lee, E. Oborn, and K. Prince. 2023. “Risk and the Future of AI: Algorithmic Bias, Data Colonialism, and Marginalization.” Information and Organization 33 (3): 100478. https://doi.org/10.1016/j.infoandorg.2023.100478.

[3] Celik, Ismail. 2023. “Exploring the Determinants of Artificial Intelligence (AI) Literacy: Digital Divide, Computational Thinking, Cognitive Absorption.” Telematics and Informatics 83 (September): 102026. https://doi.org/10.1016/j.tele.2023.102026.

[4] Cocklin, Chris, and Katie Moon. 2020. “Environmental Policy.” In International Encyclopedia of Human Geography (Second Edition), edited by Audrey Kobayashi, 227–33. Oxford: Elsevier. https://doi.org/10.1016/B978-0-08-102295-5.10788-7.

[5] Markopoulou, Areti, and Philip F. Yuan. 2020. “Design Integration: Global Technological Advancement and Local Culture.” In Design Transactions, edited by Bob Sheil, Mette Ramsgaard Thomsen, Martin Tamke, and Sean Hanna, 92–97. Rethinking Information Modelling for a New Material Age. UCL Press. https://doi.org/10.2307/j.ctv13xprf6.18.

[6] Weller, Amanda J. 2019. “Design Thinking for a User-Centered Approach to Artificial Intelligence.” She Ji: The Journal of Design, Economics, and Innovation 5 (4): 394–96. https://doi.org/10.1016/j.sheji.2019.11.015.

Gallery