The Architectural Language of Biophilic Design After Architects Use Text-to-Image AI

Abstract

Biophilic design is an architectural concept that bridges the gap between modern buildings and the innate human longing for nature. In addition, it promotes physical and mental well-being while aligning with several Sustainable Development Goals. Recent research highlights that the architectural language used to describe the attributes of biophilic architecture remains unclear. Previous research has shown that text-to-image AI enhances architects’ ability to articulate their ideas more effectively. Therefore, this study aims to address the following research question: What are the architectural languages of biophilic design after architects use text-to-image AI? The initial step involves generating images of biophilic architecture by using three popular text-to-image AI tools: DALL-E 3, MidJourney, and Stable Diffusion. The 30 selected images were used to help architects develop the architectural language to describe the characteristics of biophilic design across 10 categories: Form, Space, Movement, Light, Color, Material, Object, View, Sound, and Weather. The terms obtained were analyzed using natural language processing (NLP) techniques, including word cloud analysis, frequency analysis, and topic modeling. The results indicate that the architectural language of biophilic design exhibits greater detail and clarity after architects utilize text-to-image AI. Nevertheless, in some instances, the language used to describe biophilic design is also constrained by the images generated by the text-to-image AI that the architects observe.

1. Introduction

Urban growth is spreading globally [1]. UN reports project that, by 2050, 68% of the global population will reside in urban areas [2], spending more than 90% of their time inside buildings [3]. Urbanization has significantly contributed to the gradual loss of green spaces [4]. While humans have migrated to urban areas, their connection to nature remains intact [5].

Driven by a growing research interest in biophilia, biophilic design has garnered increasing attention in the last two decades [6]. Biophilic architecture has emerged as a vital approach to bridging the gap between modern buildings and humanity’s innate connection to nature [7]. Consequently, this study’s key focus is architecture designed under the biophilic design concept.

A literature review reveals that biophilic design integrates nature into architectural design, enhancing human physical and mental well-being [8,9]. Additionally, biophilic architecture contributes to sustainability by reducing energy consumption, improving indoor air quality [10], promoting resource efficiency [9,11], and supporting biodiversity [12].

Furthermore, the elements of biophilic design align with several principles of the Sustainable Development Goals (SDGs) [12] and serve as the conceptual foundation for nature-based requirements in various architectural design standards, such as LBC, WELL, and LEED [13,14,15,16,17]. These observations indicate that biophilic architecture is playing an increasingly significant role.

While biophilic architecture is receiving considerable attention due to its benefits for human physical and mental health and global sustainability, recent research has nonetheless revealed a gap in academic study. Since biophilic design has only been applied in architecture for about 20 years, the architectural language used to describe biophilic design approaches remains unclear [18].

Although the terminology used to describe biophilic design remains ambiguous, previous research indicates that architects utilizing text-to-image AI to communicate architectural concepts can express their ideas more clearly [19,20,21,22]. Therefore, text-to-image AI is expected to play a crucial role in enabling architects to articulate the language for describing biophilic architecture more effectively.

This study was conducted to identify the architectural language of biophilic design as a guideline based on insights from architects utilizing text-to-image AI as a tool. This approach serves as a guideline for using text-to-image AI to assist architects in articulating their design concepts. The results will demonstrate the effectiveness of using text-to-image AI to assist architects in articulating architectural language.

The theoretical framework of this study is based on the idea that, when architects analyze biophilic architecture images generated by text-to-image AI, the output will be a more explicit architectural language for describing biophilic architecture. Thus, this study focuses on addressing the following research question: What are the architectural languages of biophilic design after architects use text-to-image AI?

In the next step, we conducted a literature review to examine how previous research has identified the architectural language used to describe the characteristics of biophilic design. The findings are summarized in the following section.

2. Literature Review

To address Research Question 1, we explored the architectural language used to describe the characteristics of biophilic design as identified in previous research, which can be summarized in Figure 1.

The study by Weijie Zhong et al., 2022 provided a detailed description of the architectural language of biophilic design [18]. It also highlighted that several other researchers have proposed definitions, including Kellert in 2008 and 2018 [9,23], Abdelaal in 2019 [24], Xue et al. in 2019 [25], Browning and Ryan in 2020 [26], Wijesooriya et al. in 2021 [27], and Indre Grazuleviciute-Vileniske et al. in 2022 [28]. This architectural language of biophilic design can be summarized by categorizing the topics based on the components of architectural atmosphere as follows:

The form of biophilic design often features natural shapes, natural forms, and natural patterns while resisting straight lines and right angles, inspired by nature. It employs approaches like biophilic design and biomorphic design, which imitate natural contours such as shells, spirals, eggs, ovals, tubular forms, arches, vaults, domes, and motifs of organisms in building forms and structural systems. Additionally, biomimicry design focuses on learning from other species to address functional needs. The form is often defined by landscape features and organized using hierarchically structured ratios and scales, as well as presented through the concept of biophilic ratios, including intermediate ratios (1:1.3–1.75), the Fibonacci series (0, 1, 1, 2, 3, 5, 8, 13, 21, 34 …), and the Golden Ratio (1:1.618).

Space often embodies characteristics such as space as shape and form, along with exposure to nature and connection to nature. It also incorporates geographical, ecological, historical, and cultural connections to place and designs for transitional spaces like porches, patios, balconies, courtyards, pavilions, gardens, entry areas, foyers, atria, etc., to conceptualize interior–exterior connections and foster human–nature interactions in both interior and exterior settings. This approach provides opportunities for human–nature interactions through elements like cantilevers, infinity edges, transparent facades, pathways under/over water, and scenes defying gravity, which create a sense of ‘peril’. The layout also emphasizes incorporating nature, creating spatial harmony and variability, and applying the concept of prospect and refuge, such as providing open views/vistas (prospect) and being under shelters/safe environments (refuge). It further promotes connectivity and order, lifestyle well-being areas, shared spaces and facilities, and the experience of space and place. Additionally, it includes indirect experiences of nature, the spirit of a place, a sense of spaciousness, mastery, and control in utilizing spaces, and the complexity of organization in its design.

Movement often exhibits mobility in spaces, including entrances, exits, corridors, stairs, and high-glass elevators. Additionally, it emphasizes transportation connectivity and the development of a connected pedestrian network. The space also supports accessibility by including cycling lanes and parking facilities. At the same time, it focuses on creating attraction and gathering spaces. Furthermore, it incorporates wayfinding visual information systems and navigation collaboration.

Light often has characteristics of natural light, daylight or sunlight, as well as spectral and ambient qualities of natural light, utilizing multiple low-glare electric light sources, ambient diffused lighting on walls or ceilings, and daylight-preserving window treatments. Additionally, light incorporates nature, light and shadow, and light as shape and form, including lighting control refuge effects, along with light ray and line, filtered and diffused light, and dynamic and diffused light, along with reflected light, including light contrasts where high-contrast lights draw attention and evoke a sense of sacredness, as well as light pools, white light, warm light, spotlights, and light from outside.

Color is often characterized by natural colors, including color rays, hues, and complementary contrasts such as red and green, yellow and purple, orange and blue, and green and magenta. It also includes a color change from age, change, and the patina of time, growth, and efflorescence.

The materials of biophilic design often include wood, clay, and concrete, along with natural materials, a connection with nature, and translucence and indigeneity. It also incorporates material textures, light, color, and sound beyond material texture. It also includes natural patterns and processes that evolved human–nature relationships, such as biomorphic patterns, botanical motifs, and animal motifs, mainly vertebrates. Furthermore, it features permeable surfaces and arranges materials with specific textures and colors with the environment, creating dynamic balance and tension.

Objects include environmental features, as well as nature incorporation, such as natural scenes, plants, animals, various species, water, landscapes, geological features, or human–nature survival experiences depicted through paintings, photographs, videos, and fabrics. Additionally, they feature word reminders of nature, sign reminders of nature, and observable artworks and biomorphic elements from botanical or animals, with the number of objects for 5% of common areas.

Views are characterized by a visual connection with nature, including views of natural landscapes such as constructed wetlands, grasslands, prairies, forests, and other habitats. They also include views of waterscapes or water-covered areas such as fountains, constructed wetlands, ponds, water walls, rainwater spouts, aquaria, etc. Additionally, views of weather changes and seasonal changes in plants are included, along with views of prominent landmarks and geological forms, inside and outside experiences through window views, and transitional spaces, along with the stimulation of natural features, landscape orientation, and landscape ecology. Animal-friendly living areas are also designed to attract animals, such as nest boxes, gardens, green roofs/walls, and urban farming. Plants are incorporated into buildings through green roofs, walls, facades, large atria with park-like settings, green pockets, etc. Accessible sky gardens, sky terraces, internal courtyards, and rooftop gardens are provided as accessible areas, along with tree and columnar supports. Outdoor biophilia (25% of the site area with landscaped grounds or rooftop gardens and 70% plantings including tree canopies) and indoor biophilia (potted plants or planted beds > 1% of the floor area per floor, and wall areas covering ≥ 2% of the floor area) are also included. Interior landscapes in common areas of ≥5% are added, along with greenery indoors, such as potted plants and green walls. Water features include the presence of water, water within the building, low-maintenance water features, and indirect sensory experiences through water.

The sound elements of biophilic design include amplified sound, natural sound design, and imperceptible sound sources.

Weather management includes natural ventilation and airflow variability, climate with place-based relationships, and temperature, thermal, humidity, and barometric pressure. It also encompasses time and seasonal changes, including exposure to weather through openings and open spaces such as operable windows, porches, balconies, terraces, courtyards, etc. The natural air and ventilation simulation is achieved through vents, airshafts, operable windows, porches, clerestories, HVAC systems, etc. Additionally, meteorological condition awareness is enhanced through transparent roofs, rainwater collectors, spouts, etc.

In addition to the ten elements mentioned above, it is also noted that biophilic design identifies other functions, including biomimicry, as well as functional needs, and regular and provisional programs, as well as management and maintenance, factors related to geography, and enhancing people’s physical and mental health, productivity, and wellbeing. Additionally, it involves the integration of culture and ecology, as well as security and protection. Regarding users’ emotions, it includes natural sensory experiences, sensory variability, and avoiding placelessness while ensuring no significant adverse effects on main health aspects. These emotions also encompass mystery, risk/peril, fear and awe, curiosity and enticement, change and metamorphosis, affection and attachment, attraction and beauty, information and cognition, and reverence and spirituality.

The literature review reveals that the architectural language of biophilic design is relatively detailed in the categories of space, light, material, view, and weather. However, the terminology used to describe form, movement, color, and objects remains limited, with sound being remarkably underexplored.

The next section of the study will explain the methodology used to explore how the architectural language used to describe biophilic design evolves after architects analyze images generated by text-to-image AI.

3. Methodology

In order to address the research question, “What are the architectural languages of biophilic design after architects use text-to-image AI?”, we have designed the research methodology with three steps.

The first step is to generate images of biophilic design architecture to serve as a database for architects to develop the architectural language to describe its characteristics. The second step is to collect the architectural language that architects use to describe biophilic design after viewing images generated by text-to-image AI. The final step is to analyze the data to summarize the architectural language of biophilic design after architects use text-to-image AI. These three steps are illustrated in Figure 2 address Research Question 2; we have designed the research methodology with three steps, as seen in Figure 2.

3.1. Database

Previous research found that text-to-image AI helps architects better explain architectural concepts [21]. The first step is to generate biophilic design architecture images using text-to-image AI, which will serve as material for architects to develop descriptions of the characteristics of biophilic design. These biophilic architectural images were created using the three most popular tools: DALL-E 3, MidJourney, and Stable Diffusion [29,30,31].

This image creation process was carried out by an architect with expertise in using text-to-image AI to generate architectural design concepts. To minimize bias, this architect had no prior involvement with our research team. To ensure transparency, this architect will not be involved in any other stages of the research process.

The prompts for creating images with DALL-E 3 are constantly modified as the architect inputting the commands finds that using the same prompt results in similar photos. In contrast, MidJourney and Stable Diffusion can generate new images with the same prompt, so the architect uses the same prompt every time to create images. The prompts used to create images with all three models are shown in

Table 1. This table shows the prompts for generating biophilic architecture images using text-to-image AI.

Text-to-Image AI	Prompts
DALL-E 3	Biophilic Design Architecture, Biophilic Design Architecture, Exterior Perspective & re-roll picture, change the location to the environment that surrounds nature, can you use biomimicry design, try again with 1-story architecture, Biophilic Design Architecture Exterior Perspective Sustainable material nature surround environment natural light picture style taken by Canon Mark V Camera, please make it more curvy style, re-generate again, make it different from the same pic
Midjourney	Biophilic Design Architecture Exterior Perspective
Stable Diffusion	Biophilic Design Architecture Exterior Perspective

.

Finally, the architect creates images with each model 10 times to use as representative images in the next step. However, since MidJourney generates four images at a time, the architect selects one image from each set of four to proceed with the next step. The 30 images generated in this step thus serve as prototype images of biophilic architecture, which will be used as material to assist architects in developing the architectural language needed to describe biophilic design.

3.2. Data Collection

The second step involves collecting the architectural language used to describe biophilic design from architects after reviewing the images generated in the previous step.

To facilitate data collection, all responses were obtained through open-ended questionnaires. The questionnaire included all 30 biophilic architecture images as attachments, which was expected to help architects provide more detailed architectural language descriptions due to viewing the AI-generated images [19,20,21,22]. The questions were designed based on the conclusions from the literature review to facilitate easy comparison with the results [32]. The literature review summarized the architectural language used to describe biophilic architecture from previous research into 10 categories: Form, Space, Movement, Light, Color, Material, Object, View, Sound, and Weather. Therefore, the questionnaire consisted of 10 questions, each asking how biophilic design is characterized when considering Form, Space, Movement, Light, Color, Material, Object, View, Sound, and Weather.

To ensure validity and fairness, this study employed triangulation using multiple research methods conducted by an investigator [33,34]. Hence, the three research assistants were trained in data access, code definitions, and recording methods [33]. The first coder, Bhumin Bhutdhakomut, is a faculty member at the School of Architecture and holds a master’s degree. The other two coders, Pornteera Chunhajinda and Prima Phaibulputhipong, have been research assistants for over three years. One is pursuing a master’s degree, while the other has already earned a master’s degree in architecture. All three individuals hold professional architecture licenses in Thailand.

Furthermore, none of the three architects were involved in the literature review or the image creation process, ensuring that the terms used to describe biophilic design in this step were based solely on the images generated by text-to-image AI as material for developing descriptions. The three architects independently carried out this step using a triangulation approach to ensure the accuracy and reliability of the research results [34]. They decoded all images using the content analysis method, which involves defining, analyzing, and interpreting the meaning within the pictures of verbal form [35].

Since all three architects are Thai and primarily use Thai, the architectural language used to describe biophilic design in this step will be collected and translated from Thai to English for the next step.

3.3. Data Analysis

The third step is to use various natural language processing (NLP) techniques to identify meaningful patterns from the keywords from the previous step. NLP is a branch of artificial intelligence (AI) focused on the interaction between computers and human languages, enabling machines to understand, interpret, and generate human language effectively [36]. The main tools for this work are the Natural Language Toolkit package and the Scikit-learn package in Python 3 [36,37,38]. The dataset was preprocessed to standardize and clean the architects’ keyword responses by removing newline characters, converting text to lowercase, eliminating numeric values, and stripping punctuation. This ensured consistency and focused the analysis on relevant architectural descriptors, facilitating more accurate results in subsequent analyses like word cloud, frequency, and topic modeling analysis.

This step was carried out by one data science expert who was not involved in any previous steps to ensure no bias from viewing the images or formulating descriptions. All three steps were designed to align with the data obtained from the earlier stages, to suit the answers required by the research question, and to be compatible with the data interpreter, an architecture expert familiar with simple numerical data and images.

Word cloud analysis was used to visually represent the frequency of keywords related to biophilic architectural design, highlighting key descriptors by adjusting word size based on frequency [39]. This method provided an intuitive overview of the most prevalent terms, helping identify central themes in architects’ language without in-depth statistical analysis [39]. All text data were combined to allow the WordCloud function to analyze the entire dataset as a unified text, ensuring the representation of keywords from all aspects. A set of stopwords, including common and custom terms, was defined to filter out irrelevant words, focusing the visualization on meaningful, specific descriptors. This word cloud analysis was created using the matplotlib package, with a 10 × 5-inch figure size and axis ticks removed for clarity [40].

Subsequently, a word frequency analysis was conducted to quantify the occurrence of each keyword across the dataset. This quantitative approach provided a structured view of which terms appeared most often, allowing the study to identify prominent descriptors and recurring themes across the architectural aspects beyond mere visual observation [41]. The stopwords utilized in word cloud analysis were applied to highlight more meaningful and specific keywords. The CountVectorizer function from the sklearn package [37] was used to process text into a matrix of token counts to count word occurrences. The matrix was converted into an array and summed along the rows to obtain the total count for each word. All word occurrences were sorted in descending order, and the top 10 most frequent words were selected. This information is then visualized through the matplotlib package [40].

To analyze the data deeper, topic modeling was conducted to uncover latent themes within the architects’ descriptions of biophilic design elements. Topic modeling is an NLP technique that identifies related terms or multiple topics, helping researchers understand overarching themes without manually coding or labeling text [42]. This study used the Latent Dirichlet Allocation (LDA) model, a widely used algorithm [43]. It is a probabilistic model that assumes a set of keywords belongs to multiple topics, each characterized by a distribution of words [44]. The custom stopwords were combined with standard English stopwords to exclude general or uninformative words from the topic modeling process. Two vectorization methods were implemented to represent the text data as a matrix of term frequencies: term frequency (TF) to identify terms that appear frequently and term frequency–inverse document frequency (TF-IDF) to down-weight common words across documents [45]. The LDA algorithm was applied separately to the TF and TF-IDF document-term matrices; this was carried out to identify one cohesive theme or cluster within the corpus and examine if the weighting would yield different insights by emphasizing unique terms in the dataset. The LDA model was instantiated with a single topic (n_components = 1), as the goal was to identify one cohesive theme or cluster among the analyzed words. For each topic generated by the LDA model on the TF matrix, the top 5 words most associated with that topic were retrieved.

4. Result

The results from the procedures outlined in Section 3 indicate that, after examining a biophilic architecture generated by text-to-image AI, the architects described its architectural characteristics as follows:

The form of biophilic design often manifests as a box shape, combining horizontal and vertical lines in a sequential arrangement. These forms are predominantly characterized by sharp angles, although slight curves occasionally appear. Additionally, the design demonstrates a systematic repetition of forms, utilizing a modular structure assembled from individual components, commonly referred to as a frame structure. This results in a modern overall aesthetic. Therefore, it can be concluded that the architectural language related to the form of biophilic design is typically associated with terms such as straight, vertical, line, box, and horizontal (Figure 3).

The space of biophilic design is often associated with exterior areas that may be connected to interior spaces, such as balconies that can link to other levels or areas. The resulting open spaces are typically characterized by openness, transparency, and easy air circulation or by enclosed spaces that provide a sense of enclosure. Variations in elevation within the same space may also be present. Therefore, it can be concluded that the architectural language related to the space of biophilic design is commonly associated with terms such as ‘balcony’, ‘airy’, ‘connect’, ‘inside’, and ‘outside’ (Figure 4).

The movement in biophilic design is prominently expressed through a sense of swinging or undulating motion with clearly defined pathways. These paths are typically characterized by predominantly horizontal straight lines, although curving wave-like forms may occasionally appear, although less frequently. The use of trees in various parts of the design further enhances the building’s appeal, creating an eye-catching and inviting atmosphere that encourages exploration. Therefore, it can be concluded that the architectural language related to the movement of biophilic design is often associated with terms such as ‘clear’, ‘straight’, ‘path’, ‘horizontal’, and ‘swaying’ (Figure 5).

The lighting in biophilic design is primarily characterized by natural sunlight, with artificial lighting being less prominent. The light that appears is often soft and diffused, sometimes filtering through openings to create natural shadow patterns within a space. This creates an atmosphere that feels gentle and harmonious. Therefore, it can be concluded that the architectural language related to light in biophilic design is commonly associated with terms such as ‘soft’, ‘natural’, ‘bright’, ‘artificial’, and ‘sunlight’ (Figure 6).

The color palette in biophilic design typically features earthy tones such as deep green, brown, and warm hues, which are commonly found in wood and natural materials. These colors evoke a sense of warmth and a natural ambiance. Additionally, other colors like translucent brown, gray, and black may also appear, but they are generally used in more muted or opaque tones. Therefore, it can be concluded that the architectural language related to color in biophilic design is often associated with terms such as ‘green’, ‘warm’, ‘natural’, ‘wood’, and ‘brown’ (Figure 7).

The materials used in biophilic design often feature smooth, seamless surfaces that are transparent, allowing light to pass through and creating a sense of openness. Common materials include glass and translucent wood. Opaque materials, such as solid wood, concrete, and steel, are also frequently used in structural applications. These materials contribute to a natural aesthetic while maintaining durability and functionality. Therefore, it can be concluded that the architectural language related to materials in biophilic design is commonly associated with terms such as ‘transparent wood’, ‘steel’, ‘concrete’, ‘wood’, and ‘glass’ (Figure 8).

The objects commonly found in biophilic design include walls, planes, and overhead coverings such as roofs rather than rooftops, often integrated with green spaces featuring trees. The elements tend to have a rough texture and significant weight, with wood being the predominant material. Additionally, materials that can reflect light, such as glass, are also present. Therefore, it can be concluded that the architectural language related to objects in biophilic design is often associated with terms such as ‘wooden’, ‘plane’, ‘roof’, ‘green’, and ‘wall’ (Figure 9).

The views in biophilic design often feature abundant tree coverage throughout the area, with clear visibility of the sky. Additionally, there are scenic landscapes of gardens and forests organized in rhythmic arrangements. These views provide a sense of vitality and calm simultaneously, creating an inviting atmosphere. Therefore, it can be concluded that the architectural language related to views in biophilic design is commonly associated with terms such as ‘covered’, ‘calm’, ‘lively’, ‘sky’, and ‘trees’ (Figure 10).

The sound in biophilic design often consists of natural, soothing vibrations, primarily from the rustling of leaves, the sound of wind blowing, and the flowing of water. These sounds are complemented by the calls of birds and the chirping of various insects that inhabit the space. Therefore, it can be concluded that the architectural language related to sound in biophilic design is commonly associated with terms such as ‘birds’, ‘singing’, ‘water’, ‘wind’, and ‘leaves’ (Figure 11).

The weather in biophilic design is often characterized by a high humidity and a warm, mild atmosphere. The skies may be either overcast or clear, with soft, gentle sunlight that is not too intense. Occasionally, there may be cooler, more comfortable conditions, similar to being in the shade. Therefore, it can be concluded that the architectural language related to weather in biophilic design is commonly associated with terms such as ‘humid clear’, ‘sunny hot’, ‘cool cool’, ‘mild hot’, and ‘humid’ (Figure 12).

5. Discussion

After obtaining the results of the architectural language of biophilic design after architects use text-to-image AI in the previous section, the differences from the architectural language identified through the literature review became evident, as follows:

The first element is form. This aspect shows both similarities and differences. The similarities are that, while previous research discussed shape, form, and pattern, architects also mentioned terms such as straight lines, curves, shapes, and forms. However, these descriptions do not convey a biophilic architecture style, as they are general elements of architecture. The notable difference is that previous research often referenced natural and bio terms to explain biophilic design forms, whereas architects focused more on the forms of modern architecture. The difference here indicates that the descriptions of biophilic design from the literature are relatively more straightforward. Therefore, using text-to-image AI to help architects visualize architecture more clearly may not always be accurate.

Regarding space, the architectural language used to describe biophilic design from the literature review and the architects’ responses after considering AI-generated images are similar. Both sources mention interior and exterior spaces, connections, and open areas. Previous research discusses connections to nature, interaction with nature, shared experiences within a space, and transitional spaces. Architects also mention balconies, an architectural element that provides a similar experience. The interesting aspect of this topic is that architects provide clear architectural elements, which is more specific than the description of the spatial experience found in the literature review.

In terms of movement, the literature review discusses the connections of transportation, movement within spaces, and navigation systems of pathways. However, most architects’ descriptions still focus on lines, shapes, and forms, similar to the discussion on form, which differs from the literature review. However, when comparing this topic to the previous two, it is evident that the descriptions of biophilic design from the literature tend to be more abstract and detailed, while the descriptions from architects are often more concrete and rigid.

Previous studies’ and architects’ descriptions of light in biophilic design align with natural light, with architects identifying it as sunlight. However, architects also refer to artificial light, which may seem to contrast with natural light. Yet, when considering other terms associated with artificial light, such as ‘soft’, ‘dim’, and ‘natural’, it can be inferred that biophilic design light, from the architects’ perspective, may also involve designing artificial light to resemble natural light.

The literature review on the architectural language used to describe color in biophilic architecture generally aligns with natural colors and the changing of colors over time. Meanwhile, architects often refer to colors of wood, brown, warm tones, green, and natural tones, which align with previous research. However, it can be observed that, after considering the images generated by text-to-image AI, architects can identify colors more clearly.

The architectural language architects use to describe materials is quite similar to the literature review, such as natural materials, wood, and concrete. Additionally, the term ‘glass’ was mentioned, which is a material that provides transparency, aligning with the literature. However, architects mention steel after considering the biophilic design images generated by text-to-image AI, which does not align with previous literature and remains unclear in its interpretation.

When considering the architectural language used to describe the object aspects of biophilic design, the literature and the architects’ responses were unaligned. Previous literature discusses natural environments, plant and animal species diversity, landscapes, and various phenomena in nature. Architects mention physical features such as walls, green spaces, roofs, planes, and objects made of wood and glass. In this category, it was found that, after viewing AI-generated images, architects discussed architectural objects more frequently.

The descriptions of view in the literature review are linked to natural landscapes, environmentally friendly landscape arrangements, and designs that integrate nature with buildings, focusing on the coexistence between humans and the environment. Meanwhile, architects describe the view using terms such as trees, sky, gardens, and forests, promoting a calm, lively, and welcoming feeling. It can be observed that both sources are pretty similar, and, after considering the AI-generated images, there is an increased emphasis on terms related to emotions.

The language used to describe sound is consistent between the literature and the architects’ responses. Previous research often discusses natural sounds, including echoes or sounds where the source is not identifiable. Meanwhile, when architects viewed biophilic architectural images generated by text-to-image AI, they described sounds occurring in nature, such as the sounds of leaves, wind, water, birds, and insects. It can be observed that architects clearly identified the types of sounds they expected to hear and also included descriptive terms such as ‘shaking’. This may suggest that viewing AI-generated images helps enhance the imagination of sounds.

The literature review’s architectural language of weather in biophilic design often discusses general factors such as weather conditions, ventilation, and outdoor air. However, the weather conditions architects expect to experience and feel from the overall images relate more to heat and humidity. While the descriptions in both sections may not be closely aligned, the terms architects use after viewing images expand the meaning of weather conditions more clearly.

Finally, we compiled the characteristics of biophilic design identified both in the architectural language from the literature review and from the architects’ questionnaire responses after viewing AI-generated images, as presented in Figure 13.

6. Conclusions

Even though biophilic architecture reconnects humans with nature [6,7], benefiting both physical and mental well-being and global sustainability [8,9,10,11,12], the architectural language used to describe biophilic design remains insufficient [18]. Given that previous studies have shown text-to-image AI can help architects articulate their ideas more effectively [19,20,21,22], this research explores how architects describe biophilic design after viewing AI-generated images of biophilic architecture.

We found that, in many elements of architectural atmosphere, architects could identify architectural language more clearly after viewing AI-generated biophilic design images. For example, this includes the identification of balconies in terms of space, the approach to adjusting artificial lighting, clear descriptions of color, the identification of architectural objects, the expression of feelings derived from viewing the landscape, the identification of types of sound, and the recognition of heat and humidity in weather conditions.

This study presents the architectural language of biophilic design after architects engage with text-to-image AI. An interesting finding is that text-to-image AI enables architects to describe biophilic design with greater detail and clarity. This section demonstrates that the results are consistent with previous studies [19,20,21,22]. This method could potentially be applied to help architects articulate other design concepts more effectively. However, the limitation is that architects use limited, concrete, and overly straightforward terminology, particularly when describing form and movement features.

Nonetheless, these results are derived from the coding experiment conducted by the research assistants using triangulation with multiple researchers [46,47]. This issue may stem from architectural myopia as their mental framework—shaped by symbols, metaphors, and geometric fundamentalism—dominates their perception [48]. Thus, the limitations should be further explored in future research, particularly by conducting repeat experiments with a larger group of architects or with the general public who are not influenced by the architect’s mindset.

Furthermore, the primary material used in this study may be a key limitation that caused the results to differ from expectations. Although AI can enhance human intelligence and creativity in design, it still has limitations in terms of data bias [49]. Previous research on similar topics has found that real-world experiences have a more significant impact than observing images displayed on a screen [50]. Given this limitation, future research should conduct experiments in real-world settings.