AI-Driven Biophilic Façade Design for Senior Multi-Family Housing Using LoRA and Stable Diffusion

Abstract

South Korea is rapidly transitioning into an aging society, resulting in a growing demand for senior multi-family housing. Nevertheless, current façade designs remain limited in diversity and fail to adequately address the visual needs and preferences of the elderly population. This study presents a biophilic façade design approach for senior housing, utilizing Stable Diffusion (SD) fine-tuned with low-rank adaptation (LoRA) to support the implementation of differentiated biophilic design (BD) strategies. Prompts were derived from an analysis of Korean and worldwide cases, reflecting the perceptual and cognitive characteristics of older adults. A dataset focusing on key BD attributes—specifically color and shapes/forms—was constructed and used to train the LoRA model. To enhance accuracy and contextual relevance in image generation, ControlNet was applied. The validity of the dataset was evaluated through expert assessments using Likert-scale analysis, while model reliability was examined using loss function trends and Frechet Inception Distance (FID) scores. Our findings indicate that the proposed approach enables more precise and scalable applications of biophilic design in senior housing façades. This approach highlights the potential of AI-assisted design workflows in promoting age-inclusive and biophilic urban environments.

1. Introduction

The transition of South Korea into a super-aged society has accelerated the demand for senior multi-family housing [1,2]. Statistical data indicate that the number of senior multi-family housing units in South Korea has steadily increased by approximately 4.73% annually from 2008 to 2023 [3]. Recently, the development of senior multi-family housing has evolved beyond basic residential spaces to incorporate integrated medical, welfare, and care services, with community-based housing models being introduced to support successful aging for seniors [4]. These changes necessitate design improvements in the exterior and living configurations of senior multi-family housing to reflect the unique characteristics of the elderly population. However, currently, most façade designs for senior multi-family housing in South Korea display high-density residential typologies characterized by monotony and uniformity [5]. The lack of diversity and differentiation in façade designs results primarily from prioritizing the economic efficiency associated with mass production [6].

In architecture, façade design extends beyond mere aesthetic elements to determine the visual identity of a building and serves as an important mediator connecting indoor and outdoor spaces [7,8,9,10]. Additionally, façade design plays a key role as the backdrop of urban spaces and local life while enhancing a sense of place [11], and it contributes to shaping the symbolic nature of buildings [12]. However, traditional façade design processes can be subjective depending on the architect’s prior knowledge and capabilities, and they often show an inverse relationship between time and efficiency. Therefore, innovative tools are needed to achieve more objective results while simultaneously improving time and efficiency. Recently, Generative Adversarial Networks (GANs) proposed by Goodfellow et al. [13] and Denoising Diffusion Probabilistic Models (DDPM) advanced by Ho et al. [14] have driven rapid advancements in Text-to-Image models, bringing transformative changes to AI-based image generation technologies. In particular, generative AI technologies can quickly generate and visualize diverse façade designs by training on large-scale datasets, thus enhancing creativity and efficiency simultaneously [15]. A growing number of studies have applied these generative models to façade design. For instance, several works using Stable Diffusion have proposed façade visualizations that reflect regional characteristics [11], incorporate various architectural styles [16], or balance historical preservation with commercial application through the integration of LoRA models [17]. Additionally, StyleGAN-based systems such as iFACADE have been developed to streamline the early design process through simple user inputs [18], while Pix2pix and other models have been used to train façade generation from traditional architectural datasets [19]. Other approaches have focused on producing detailed, high-quality façade renderings [20], or suggested methods to visualize single-family housing exteriors in the early stages of design [21]. These generative AI techniques, especially when combined with prompt engineering and low-rank adaptation methods like LoRA [16], can effectively reflect the specific perceptual and emotional needs of older adults in façade design.

These generative AI techniques, employing prompts and LoRA (low-rank adaptation) [22] models, can effectively reflect the specific needs and characteristics of older adults in façade design.

Older adults, typically defined as individuals aged 65 and above, commonly exhibit diminished visual information processing capabilities, along with decreased spatial awareness and impaired wayfinding abilities [23], thereby necessitating intuitive and legible façade designs in age-friendly environments. In this context, the façade design process should incorporate the concept of Aging in Place (AIP), which emphasizes enabling older adults to live independently and securely within familiar surroundings [24]. Applying clear color contrasts and easily recognizable exterior forms can help accommodate age-related visual impairments and facilitate intuitive navigation [25,26]. Moreover, the use of natural materials such as wood, brick, and vegetation in façade design contributes to psychological well-being by evoking a sense of familiarity and comfort through natural textures and colors [27].

Previous studies related to multi-family housing for older adults have primarily focused on indoor spatial planning [24,26,28], as well as medical and welfare perspectives [29,30,31]. Research addressing the façades of multi-family housing has generally been conducted from the viewpoint of energy efficiency and sustainability [32,33]. However, differentiated façade design research specifically considering the visual characteristics of older adults remains insufficient. Façade design refers to the architectural process of configuring a building’s external envelope. Beyond serving as a medium of aesthetic expression, it constitutes a complex design domain that synthesizes architectural identity, environmental responsiveness, and user-centered experience. Knaack et al. [34] characterize the façade as a “multifunctional interface” integrating form, materiality, structure, and environmental control, thereby functioning both as a connector between interior and exterior spaces and as a modulator of environmental conditions. Herzog & de Meuron [35] emphasize its role in shaping a building’s visual identity and symbolic presence within the urban fabric, while Schittich [36] and Yeang [37] underscore its contribution to sustainability and contextual integration. Furthermore, Browning et al. [38] highlight the psychological and emotional impact of façade elements, particularly in relation to users’ sense of environmental comfort and well-being. In light of these perspectives, façade design may be understood as a multidimensional architectural practice in which technical, functional, aesthetic, and psychological considerations converge. However, the present study delineates its scope more narrowly by focusing on the visual properties of façade design that directly respond to the sensory needs and spatial cognition of older adults. Specifically, this research limits its investigation to the color and form of the building envelope, given their critical role in promoting visual legibility, intuitive orientation, and psychological reassurance within age-friendly residential environments. Therefore, this study aims to explore the content and methodologies of façade design for older adults by specifically focusing on the visual and spatial qualities of building façades that support age-friendly living environments.

This research direction is particularly relevant in the context of rapid urbanization, which has led to living environments increasingly disconnected from nature [39,40,41]. Such disconnection has been associated with various negative outcomes, including the spread of disease, reduced social interaction, heightened environmental stress, and psychological anxiety [42]. These challenges are especially detrimental to older adults, who are more vulnerable to environmental stressors and their associated physical and emotional impacts [43]. Recently, biophilic design (BD) has gained attention as an architectural and environmental planning alternative to address the challenges of rapid urbanization [44]. Biophilic design is defined as a design approach that seeks psychological stability and visual satisfaction through direct and indirect natural planning elements integrated within architectural and environmental contexts [45]. Particularly for older adults, prominent natural features or imitations can aid their cognition and memory, reinforce the meaning of place, enhance life satisfaction, and provide motivation [46,47]. Additionally, the benefits of nature include improved immune function and increased physiological [48] and psychological resilience [49,50]. In other words, integrating natural elements into senior housing can activate physical and cognitive functions in older adults while supporting familiarity with spaces to better realize the concept of Aging in Place (AIP) [51]. Therefore, the objective of this research is to propose façade design strategies for multi-family housing for older adults that incorporate biophilic design attributes by employing generative AI technology.

In the process of applying biophilic design to senior housing façades, the following research questions (RQs) were formulated:

(i)

Identify the Problem: What are the current limitations of façade design in senior multi-family housing, and how do these shortcomings affect the spatial experience and well-being of older adults?

(ii)

Define Theoretical Foundation: Which principles and attributes of biophilic design are applicable to façade design for older adults, and how can these be effectively structured for implementation?

(iii)

Design the Framework: How can generative AI techniques—specifically Stable Diffusion with LoRA—be leveraged to integrate biophilic design attributes into senior housing façades?

(iv)

Apply the Methodology: How can the façade generation pipeline—including prompt engineering, dataset construction, and LoRA hyperparameter optimization—be structured and adjusted to generate façade designs that reflect the perceptual and visual needs of older adults?

(v)

Validate the Methodology: What evaluation methods can be used to verify the generation quality, representational accuracy, and training stability of the proposed AI-based façade design approach?

2. Materials and Methods

The methods and scope of this study are as follows (see Figure 1). First, the concepts of “generative AI and façade design” and “older adults and biophilic design” are explored through literature reviews and prior studies. Second, façade “colors” and “shapes and forms” are analyzed by selecting case studies of senior multi-family housing from South Korea and abroad. In particular, this study aims to explore the content and methodologies of façade design for older adults by specifically focusing on the visual and spatial qualities of building façades—as perceived from the exterior—that support age-friendly living environments. Case studies are selected based on criteria specifying housing designed for older adults aged 60 or older, capable of independent living and self-care. The reason for selecting senior multi-family housing façades in this study is that façade planning in Korean multi-family housing is typically monotonous and standardized due to design practices prioritizing economic efficiency through mass production [6]. Given the growing demand for senior multi-family housing in South Korea, driven by its transition into a super-aged society, differentiated approaches to façade design are deemed necessary. Third, biophilic façade design elements and associated text prompts are derived from analyses of Korean and worldwide case studies. Fourth, a generative AI-based biophilic façade design model is developed for application to senior multi-family housing. This process includes constructing an image and text dataset, setting hyperparameters, and training a biophilic façade design LoRA model. For the quantitative validation of the developed LoRA model, a loss function analysis and a Frechet Inception Distance (FID) evaluation are conducted to objectively verify the model’s performance and determine optimal hyperparameter and LoRA weight values. Finally, biophilic façade design images for senior multi-family housing are generated and analyzed. In this process, façade images are created by combining prompts, LoRA, and ControlNet. The generated façades are then analyzed based on color, shapes and forms, or combinations of both, followed by discussions on future developments and potential applications.

3. Theoretical Insight and Literature Review

3.1. Older Adults and Biophilic Façade Design

Humanity has evolved over centuries through interactions with the natural environment, relying more on nature than on urban or artificial environments [52]. In particular, spending time in nature has been shown to promote social interaction, improve memory, reduce stress, and activate physical activity for the older adults [53]. Furthermore, Peters & Verderber [54] argued that simply viewing natural landscapes can improve the quality of life for older adults. Therefore, addressing natural elements that are vital to the well-being and overall quality of life of older adults is essential. Biophilic design, a theoretical approach supporting this view, integrates natural elements into architecture, promoting interaction between humans and nature and thereby fostering positive experiences [45,55].

Table 1. Experiences and attributes of biophilic design by Stephen Kellert [45].

Experiences	Attributes
Direct Experience of Nature	Light/Air/Water/Plants/Animals/Landscape/Weather/ Views/Fire
Indirect Experience of Nature	Images/Materials/Texture/Color/Shapes and Form/Information Richness/Change, Age, and the Patina of Time/Natural Geometries/Simulated Natural Light and Air/Biomimicry
Experience of Space and Place	Prospect and Refuge/Organized Complexity/Mobility/Transitional Spaces/Place/Integrating Parts to Create Wholes

shows three experiences and attributes [45] for biophilic design application and utilization in the built environment. Direct experience with nature is the most basic concept in the practice of biophilic design and means actual contact with nature. Indirect experience of nature involves methods of imitation or representation rather than direct exposure, symbolically or metaphorically communicating nature through visual scenes or portrayals. The experience of space and place involves exposure to constructed environments characterized by attributes inspired by natural settings. At this time, the characteristics of the natural environment induce the involuntary attention of humans [56], and nature is conveyed realistically or expressed symbolically and metaphorically according to intention, thereby arousing human curiosity and interest.

This study reviewed previous research on façade design strategies in Korean and worldwide multi-family housing, summarized in

Table 2. Analysis of façade design elements in multi-family housing.

Ref.	Contents	Façade Elements
[61]	Examined diverse approaches to apartment façade design through pattern analysis.	Pattern, Material, Color, Form
[62]	Analyzed the influence of design plans and regional design images on the residential preferences for façades.	Form, Material, Color
[63]	Investigated optimal strategies for shaping urban façade imagery by evaluating the façade color characteristics of apartment complexes in Seoul.	Form, Color
[58]	Conducted a consumer preference analysis for façade designs in apartment complexes.	Pattern, Material, Color, Form
[6]	Proposed guidelines for façade designs tailored to apartment complexes through data collection and wind analysis.	Pattern, Material, Color, Form
[59]	Reviewed façade design guidelines for sustainable apartment complexes and proposed practical improvement measures.	Form, Material, Color
[60]	Reviewed multi-objective optimization methods for high-performance building façade design and identified key algorithmic strengths, limitations, and future research opportunities.	Color, Material, Pattern
[57]	Analyzed the importance and preferences of apartment façade design elements from a user perspective to inform user-centered design.	Form, Material, Color
[5]	Analyzed how the exterior design of high-rise apartments in South Korea has changed over the past 40 years and presented the main factors that influenced the changes.	Form, Color, Material, Pattern, Roof
[64]	An analysis of the impact of aesthetic design elements on resident satisfaction in apartment complexes in Seoul.	Roof, Material, Light, Color, Form
[65]	Studied the history of the design of the exterior of apartment buildings in Siberia, Russia, and analyzed how various architectural and historical elements contributed to the formation of the exterior of residential complexes.	Form, Mass, Material, Color, Ground level

. Based on this review, core elements of biophilic design were derived to apply biophilic attributes to senior multi-family housing. According to the analysis of previous studies, research based on empirical data, such as importance and preference evaluations [57,58], as well as studies related to architectural regulations and urban aesthetics [6,59,60], identified form, material, color, and pattern as key elements of façade design. In addition, studies that analyzed the morphological and aesthetic characteristics of façades [61,62,63,64], and those that examined historical transitions in façade design [5,65], derived further elements such as mass, ground level, and roof. Among these, form and color were consistently presented as core elements of façade design across studies. This aligns with findings that human visual perception is primarily composed of two components: ‘forms’ and ‘colors’ [23,66]. In a survey evaluating preferences for façade elements of Korean multi-family housing, ‘shapes and forms’ and ‘colors’ were ranked as the most important factors [57]. Given that visual perception accounts for approximately 87% of information recognition, visual elements have significant implications, especially for older adults [67]. Therefore, this study aims to focus on planning façade designs by emphasizing design elements related to ‘color’ and ‘shapes and forms’ among the attributes of biophilic design.

3.2. Generative AI and Façade Design

3.2.1. AI and Architectural Design Automation

The architectural design field has continuously changed with the advancement of technology, and although there is initial resistance when a new technology is introduced, users tend to gradually accept the technology [68]. Traditionally, architectural design has transitioned from manual methods to Computer-Aided Design (CAD) and Building Information Modeling (BIM), with these technological advances significantly enhancing the efficiency and precision of the design process. Recently, artificial intelligence (AI) technology has gained attention as a tool for further improving automation, reliability, and efficiency in architectural design. In particular, generative AI can serve as a supportive tool for architects in the design process, aiding the exploration of novel ideas and automating repetitive design tasks [69]. Architectural design is closely linked to creativity, and finding solutions that optimize creativity in a practical manner is crucial [70]. Therefore, as AI technology continues to develop, some architects will utilize generative AI as a complementary or alternative tool to conventional methods of exploring design inspiration [71].

The use of AI in architectural design is not a new concept, and design optimization techniques using machine learning (ML) and deep learning (DL) have been studied for a long time. Recently, advances in deep generative models such as Generative Adversarial Networks (GANs) and Stable Diffusion (SD) models have accelerated discussions on AI-based architectural design automation [17]. GANs, artificial intelligence models that generate new data through adversarial competition between two neural networks [13], have been further developed by researchers in diverse directions [17]. Architectural design research using GAN has mainly been dealt with in building façade generation [18,19], and automatic spatial layout generation [72]. GAN has proven its usefulness in the early stages of design by reducing the need for prior knowledge of architects in the early stages of the design process and generating images that fit a specific architectural style [73]. The diffusion model shows more advantageous results than GAN in image generation quality [74]. In particular, the Stable Diffusion model (SD) provides a user-friendly interface [75] and is open source, so anyone can use it for free. Additionally, continuous updates to the SD model have consistently improved image quality. Cao et al. [68] analyzed the impact of SD on architectural design workflow and architectural education, proving that SD is a tool that can improve the architectural design process, and Li & Li [76] proposed automation of architectural design considering lighting using SD. Ma & Zheng [77] proposed a method to generate façade design by fine-tuning SD using LoRA training. As demonstrated, diffusion models not only save time but also enable architects to rapidly experiment with diverse ideas and concepts [17].

3.2.2. Generative AI and Façade Design Visualization

Generative AI creates new content based on learned data and can be applied to various fields such as image processing, data generation, and prediction [78,79,80]. In the architectural field, façade design methods using generative AI are actively being discussed.

Table 3. A study on façade design visualization using generative AI.

Ref.	Platform	Contents
[11]	Stable Diffusion	Propose building façade designs that reflect regional characteristics.
[16]	Stable Diffusion	Present visualizations of façade designs incorporating architectural styles.
[17]	Stable Diffusion	Utilize Stable Diffusion and LoRA models to transform traditional architectural façades into modern commercial styles, proposing a balance between historical preservation and commercial application.
[18]	StyleGAN	Develop a StyleGAN-based iFAÇADE system that generates new façade images with simple inputs, streamlining communication between architects and users during the early design stages.
[20]	Stable Diffusion	Generate detailed, high-quality façade images using generative AI models, proposing a creative and efficient design approach.
[19]	Pix2pix	Train and generate façade images using existing datasets and traditional Chinese architectural datasets to enable rapid recognition and design reconstruction.
[21]	Stable Diffusion	Create visualizations of single-family house exteriors reflecting architectural styles and propose a methodology to streamline the visualization process during the early design phase by building additional training models.

summarizes previous studies that proposed façade designs based on generative AI. Previous studies primarily emphasize that generative AI enables creative and efficient façade design proposals [18,20]. Furthermore, earlier studies proposed façade designs that reflect regional characteristics [11,17,19] and implemented façade designs using LoRA models that capture architects’ unique design styles [16,21]. Generative AI automates repetitive architectural design tasks and supports optimal design through data analysis. Recently, the use of generative AI has gained significant traction, with platforms such as Midjourney, DALL-E 2, Stable Diffusion, and Adobe Firefly drawing considerable attention. Stable Diffusion (SD) has the scalability to additionally learn data that existing AI models have not learned, and research is actively being conducted in the architectural field to utilize it as a visualization tool. In particular, SD shows high utility in the architectural design and visualization process by providing the advantage of being able to customize and apply specific design concepts and properties.

4. Case Study and Biophilic Façade Design Planning Elements

4.1. South Korea and Worldwide Senior Multifamily Housing

This study analyzed cases of senior multi-family housing in South Korea and internationally, with detailed information presented in

Table 4. Analysis of façade ‘color’ and ‘shapes and forms’ of senior multi-family housing complexes in South Korea and internationally.

Name (City, Year)	Image	Façade Color	RGB	H V/C *	Analysis
					Color	Shapes and Forms
A. Seoul Seniors Gayang Tower (Seoul, 2007)			177 203 226	5PB 7/6	Overall use of low-chroma colors	Rectangular, vertical, and simple shapes
			137 118 103	5YR 5/4
			91 98 90	10GY 4/2
			128 152 156	5BG 6/2
B. The Heritage (Gyeonggi-do, 2017)			200 200 200	10PB 8/1	Use of moderate and low-saturation colors	Symmetrical structure and straight-line shape
			69 74 55	10Y 3/2
			175 141 106	10YR 6/4
			150 114 84	10YR 5/4
C. Gangbyeon Neulpureun Town (Andong, 2017)			227 227 227	N 9/0	Creating a strong brightness contrast by increasing the difference in lightness	Horizontal and stepped shape
			173 173 173	N 6.5/0
			209 166 104	2.5Y 7/6
			96 96 96	N 4/0
D. Dongbaek Spring County Xi (Gyeonggi-do, 2019)			217 230 237	5PB 8/2	Shapes of straight lines and right angles	Use of high brightness white and medium saturation orange
			170 174 178	N 7/0
			139 117 106	10YR 5/4
			158 175 199	5PB 7/4
E. Jangseong Nuri Town (Jeolla-do, 2019)			55 75 84	10BG 3/2	Providing a color contrast using a combination of white and yellow tones	Straight lines and symmetrical structure/flower-like patterned design
			218 230 216	10GY 8/2
			191 175 62	2.5Y 6/6
			162 177 171	5BG 7/2
F. Tiger Island (Morgan, 2023)			163 176 176	5BG 7/2	Low saturation and medium brightness color/grayish blue color	Symmetrical two-gable roof form combining triangles and straight lines
			91 67 33	10YR 3/4
			192 201 210	5PB 8/2
			194 201 204	5PB 8/1
G. Armstrong Place Senior (San Francisco, 2011)			206 79 52	10R 5/6	Use warm colors with high brightness and saturation/high brightness and saturation contrast	Cube=shaped protruding window/rectangular shape
			81 68 70	5RP 3/2
			58 46 45	10R 2/2
			238 229 167	5Y 9/4
H. Morangis Retirement Home (Paris, 2013)			230 193 158	10YR 8/4	Highly saturated orange and yellow with high chroma and high brightness	Vertical and horizontal shapes of rectangles and squares
			245 188 140	10YR 8/6
			240 154 31	5YR 7/10
			230 193 158	10YR 8/4
I. Belle Vue Senior Residence (London, 2019)			82 68 66	5RP 3/2	Use of colors with high brightness and low saturation	Modular, vertical, and horizontal forms
			180 138 116	10R 6/4
			238 164 144	7.5R 7/4
			154 112 94	10R 5/4
J. Brazos Towers (Houston, 2015)			216 196 181	2.5Y 8/2	Medium brightness, low saturation beige, sand, and brown tones	Featuring large glass windows and open balconies in a straight, vertical form
			64 72 87	10PB 4/2
			110 122 148	7.5PB 6/4
			115 122 134	10PB 6/2

* Note. H V/C: Hue, Value, Chroma/Image credits: A—©SEOUL SENIORS TOWER; B—©NEWSIS; C—©DailyGNews; D—©Author; E—©Jangseong Weekly Newspaper; F—©Leonid Furmansky; G—©Brian Rose; H—©Florent MICHEL/11 h 45; I—©Jack Hobhouse; J—©Brazos Towers at Bayou Manor.

. To analyze façade designs of Korean and worldwide senior multi-family housing, the following selection criteria were established: First, the study focused exclusively on residential facilities for older adults aged 60 years or above. Second, housing types were limited to independent senior multi-family housing, which refers to housing intended for daily living rather than care-oriented or medical-centered facilities. Third, the cases selected were those completed after the year 2000, with priority given to examples around 2010 to reflect recent design trends. Fourth, to avoid excessively small buildings and to ensure the inclusion of diverse scales, the cases selected had a total floor area of at least 3000 m². Based on these criteria, five Korean cases (A–E) and five worldwide cases (F–J) were analyzed. In addition, color analysis was performed using the Korean Standard Color Analysis (KSCA) [81] program, and the top four major colors were extracted and displayed.

The analysis results indicated the following characteristics in façade designs of senior multi-family housing in South Korea and internationally. Regarding colors, Korean housing primarily utilized mid-to-low brightness and low-to-mid saturation colors, creating visual stability in buildings. Korean senior multi-family housing generally followed color schemes featuring neutral tones with mid-to-low saturation. Color combinations tended to be somewhat limited, often using neutral shades such as beige, gray, and brown, with relatively low brightness contrast. Overall, Korean cases showed stable and uniform color planning strategies. In contrast, worldwide cases frequently demonstrated noticeable color contrasts by incorporating high-brightness and high-saturation colors. Worldwide examples showed diverse and experimental approaches to color use, often emphasizing individuality through strong accent colors. However, Korean cases also demonstrated a gradual trend toward stronger color contrasts and brighter accent colors in more recent examples compared to those from the early 2000s. This shift highlights the importance of aesthetic satisfaction and visual accessibility in residential environments for older adults. Regarding shapes and forms, Korean housing façades commonly displayed simple, box-like shapes characterized by symmetrical structures and linear designs. In contrast, worldwide cases typically exhibited more three-dimensional façade designs. Although predominantly characterized by linear and horizontal forms, worldwide examples displayed distinctive three-dimensional features through elements such as windows and balconies. In the case of Case F, a unique sense of space was created through a triangular roof and an irregular façade configuration.

In other words, in Korean senior multi-family housing (A–E), the design maintains stability and functionality; however, it exhibits relatively insufficient color contrast and morphological diversity compared to worldwide cases. Therefore, this study aims to derive and apply façade design elements related to ‘color’ and ‘shapes and forms’ attributes in biophilic design while considering the visual characteristics of older adults.

4.2. Biophilic Façade Design Planning Elements and Prompts for Older Adults

Older adults tend to prefer primary colors over achromatic tones due to diminished color perception ability [82]. This is closely related to changes in color perception caused by age-related vision decline. Therefore, in color design for older adults, it is effective to use primary colors as accent colors and plan base and secondary colors through appropriate color combinations. In addition, considering the decreased and deteriorated vision of older adults, colors should be applied to a wide area [83], and a color scheme with a clear contrast should be divided to increase visual convenience [26]. In particular, colors found in nature have a positive effect on the visual experience of older adults [26,45], and blue and green, which are representative colors of nature, reduce stress and provide a sense of stability [46,84]. Therefore, it is important to consider primary shades of green and blue and to use color schemes resembling natural landscapes and scenery to maximize the effects of biophilic design.

Natural forms and shapes are one of the representative ways to integrate nature into the built environment [45], and applying curves to architectural design can strengthen the sense of connection with nature [85,86]. Nature favors organic curves over straight lines, and humans instinctively perceive comfort and softness through curves [85,87]. Therefore, façade designs that incorporate curves rather than relying solely on straight lines are necessary. Accordingly, façade design planning elements and prompts based on selected attributes of biophilic design were derived as shown in

Table 5. Biophilic façade design planning elements and prompt derivation.

BDA *	Biophilic Planning Elements		Prompt
Color	1	Utilize natural colors such as sky blue and green.	Natural tones: sky blue or green or brown, broad areas for visibility, high contrast for readability
	2	Apply colors to large areas to aid seniors with reduced vision.
	3	Use high-contrast colors for better visibility.
Shapes and Forms	1	Apply smooth curves.	Smooth curves for stability and comfort that minimize angular forms and reflect natural shapes and structures
	2	Minimize angular shapes.
	3	Reflect natural forms and structures.

Note. BDA *: Biophilic design attributes.

.

5. Building a Senior Multi-Family Housing Façade Model Based on Biophilic Design

5.1. Building a Learning Model Dataset

This study utilized the low-rank adaptation (LoRA) training model to develop biophilic façade designs for senior multi-family housing. LoRA is an efficient training method that reduces memory and computational costs by adding low-rank matrices while keeping the pre-trained model’s weights fixed [22]. A dataset is required to construct the LoRA model. The dataset consists of combined image and text data. When constructing the image dataset, the color attributes were referenced from a previous study [88], which developed a biophilic design palette derived from natural landscape images. The biophilic design color palette is categorized into eight attributes, including sunlight, sky, water, and plants. Accordingly, this study converted the color codes (RGB) of plants and the sky into images and learned them, taking into account the psychological and mental effects for older adults. For shape and form attributes, the dataset was constructed through the process outlined in Figure 2. The selection process for the image dataset is as follows. Keywords related to shapes and forms were extracted through a literature review and the use of ChatGPT-4o and Interrogate Clip functionality. Based on the extracted keywords, 40 images of building façades incorporating curves were collected. The dataset images were screened through expert evaluation by five architectural design professionals with over 10 years of experience in biophilic design research. The evaluation was conducted using a Likert scale, with the criteria outlined in

Table 6. Evaluation metrics for biophilic façade training dataset.

Metric	Details
Domain Fidelity	Does the dataset imagery faithfully reflect the principles of biophilic design?
Visual Clarity and Quality	Does the dataset maintain a resolution of 512 × 512 or higher with appropriate lighting and contrast for accurate AI learning of shapes and structures?
Diversity and Balance	Does the dataset encompass diverse architectural types without regional or stylistic bias while incorporating various curved forms, including gentle, geometric, and organic curves?
Form Expressiveness	Does the dataset provide sufficient and clearly defined curved elements for the AI model to learn and naturally apply?
Training Suitability	Does the dataset enable the AI model to flexibly apply curves while preserving seed image characteristics and ensuring natural curve representation in generated outputs?

. Based on the Likert scale evaluation, 20 images with the highest scores were selected from the initial set of 40. The selected 20 images were preprocessed to a resolution of 512 × 512 for training, and a dataset combining image and text data was constructed using BLIP Captioning.

Table 7. Dataset for building LoRA model.

Attributes	Color and Image Data
Color	Elements	Colors	RGB
	Plant		71 42 18
			39 81 33
			160 147 96
	Sky		205 232 235
			55 95 163
			153 207 241
Shapes and Forms

summarizes the image data used for training in the construction of the LoRA model.

5.2. Learning a Biophilic Façade Design LoRA Model

In this study, seed images for image generation were selected from one case of senior multi-family housing in South Korea and another case of senior multi-family housing abroad. The Korean case selected was Gayang Seniors Tower. This building is located in Seoul, the capital of South Korea, and represents the regional characteristics of Korean senior multi-family housing. Additionally, it is the case among the Korean examples (A–E) in Table 4 that employs the lowest saturation in its color scheme. The low contrast in its colors suggested a need for visual enhancement. Brazos Towers was selected as a worldwide case. Brazos Towers represents senior housing in the southern United States and is characterized by a subdued color palette and a traditional linear design. With its low color contrast and standardized façade design, it was deemed an appropriate case for analyzing visual changes through the application of biophilic design.

Image generation utilized the extended model ControlNet. ControlNet preserves the basic structure of the seed image while enabling flexible concept modification [89,90]. Thus, it allows both flexible concept alteration and spatial control [90]. Specifically, images are generated using the Canny model, one of ControlNet’s control methods. In addition, image generation values and quality prompts are essential to obtain high-quality image generation results. The image generation parameters and quality prompts are detailed in

Table 8. Image quality generation values.

Check Point	Image Generation Values
	Image Size	Sampler	Steps	CFG Scale	Weight
SD 1.5	1152 × 712	DPM++ 2M SDE Exponential	25	9	0.7

and

Table 9. Prompts for enhancing image quality.

Common Prompt	Positive Prompt	Negative Prompt
façade, biophilic design (considering the proportion and perspective of the façade)	high quality, highly detailed, attention to detail, photorealistic 8k, UHD, HDR, professional photograph, realistic, precise	human, error, bad quality, tiling, drawing, sketch, ugly, pixelated, low resolution, high contrast, split image, distortion, text, watermark, name, signature

. Hyperparameter configuration, which critically influences the LoRA model’s quality, is also essential. Hyperparameters are parameters that control the learning process of a model. The main hyperparameter settings are as follows. First, batch size refers to the number of samples in the training dataset. Second, epoch refers to the number of times the entire dataset is passed through the model during training. Third, repeat refers to the number of times each training sample is used. Fourth, the learning rate is the value used to adjust the model’s weights during updates, controlling the sensitivity of the learning process. Lastly, the optimizer is a hyperparameter that selects the optimization algorithm for the deep learning model. The Lion optimizer selected for this study provides faster convergence and a more stable learning process compared to the traditional Adam optimizer [91]. The Lion optimizer maintains stable performance even with fewer epochs, supporting fast and reliable weight updates during the model convergence process [92]. This plays a crucial role in improving the learning speed and preventing overfitting. Thus, the Lion optimizer used in this study allowed for rapid convergence on the dataset while maintaining the quality of the learning process. The hyperparameter settings and descriptions are outlined in

Table 10. Hyperparameter configuration for LoRA training.

Hyperparameters	Description	Value
Batch size	Number of samples processed per training step	8
Epoch	Number of times the entire dataset is iterated through during training	30
Repeat	Number of times each image is used during training	50
Optimizer	Optimization algorithm used for weight updates	Lion
Learning rate	Controls the step size for weight updates during training	0.00005
Max resolution	Maximum image size used for training	512 × 512

, while the application of the seed image and Canny model is illustrated in Figure 3. LoRA training was conducted using the Stable Diffusion (SD) platform on a PC equipped with an RTX 4090 GPU and 64 GB RAM, with a training duration of approximately one hour.

5.3. LoRA Model Training Loss Analysis and Evaluation

5.3.1. LoRA Model Learning Process and Loss Function Analysis

Figure 4 illustrates the variation in the loss function during the training process of the LoRA model. Analysis of the loss function reveals a gradual decrease in loss values as training progresses, indicating that the model is effectively learning the data distribution [93]. In particular, temporary volatility is observed in the 10–20 value range in the loss graph of the epoch, but the model tends to converge after 30 epochs. These results suggest that the LoRA model identifies optimal weight values through training, progressively improving image generation performance. However, excessive reduction in loss values may indicate a risk of overfitting. In addition, the LoRA model does not simply adjust the overall weight but uses a method to optimize the model by learning a low-rank matrix added to a specific layer [22]. Specifically, LoRA adjusts model performance via trainable low-rank matrices while keeping the original model’s primary weight matrices fixed. Due to these characteristics, the quality of the generated image may differ depending on the weight setting of LoRA even for the same model [94]. The suitability of the LoRA model cannot be evaluated simply by reducing the loss function value, and a process is needed to find the most appropriate setting by analyzing the Frechet Inception Distance (FID) score according to the weight value of LoRA.

5.3.2. Quantitative Analysis

This study measured the Frechet Inception Distance (FID) score to evaluate the LoRA model’s weight configurations and the quality of generated images. FID is a standard metric for assessing the similarity between generated and real data, where lower values indicate more realistic outputs. FID has become an essential tool for evaluating performance in image generation tasks, and it is especially important when a direct evaluation of model performance using loss curves is difficult to obtain [17]. While lower FID scores signify better models, their acceptable range varies depending on the dataset and model architecture [95]. Figure 5 shows the FID score graph according to the LoRA model’s weight value. The initial weight (0.3–0.5) showed a somewhat high score in the range of 170–150, but as it increased to 0.7 or more, the FID score decreased to less than 100, confirming that the generated image became similar to the ground truth (GT). In addition, Figure 6 shows one of the selected dataset images, and shows the results of a visual comparison of the quality changes in the generated images based on the weight of GT and LoRA Model. As the weight value increases, the generated image has higher structural similarity with GT, and the lower the FID score, the more consistent the results are with GT. Therefore, in this study, it is judged appropriate to set the weight value in the range of 0.7 to 1.0 to secure the optimal performance of the LoRA model.

6. Generation of Biophilic Facade Images Based on Generative AI

6.1. Image Generation Results

This study generated façades for senior multi-family housing incorporating biophilic design using a LoRA model. For the color LoRA model, additional training was conducted for the six colors presented in Table 6, with prompts such as “Using blue (or green or brown) as the accent color” to implement biophilic colors as accents. Furthermore, based on FID score analysis, the weight was set to 1.0 to optimize color and form representation.

For the ‘color’ attribute, the sky element was intended to be rendered in blue, while the plant element was designed to incorporate brown and green into the façade design. For the shapes and forms attribute, the goal was to generate façade designs incorporating moderate curves.

Table 11. Prompts and LoRA training data for image generation.

Attributes	LoRA Training Data			Prompts
Color	Sky		71 42 18	Using blue as the accent color, (natural tones: sky blue), broad areas for visibility. High contrast for readability, <lora:sky1:1.0>, <lora:sky2:1.0>, <lora:sky3:1.0>
			39 81 33
			160 147 96
	Plants		205 232 235	Using green (or brown) as the accent color, (natural tones: green or brown), broad areas for visibility. High contrast for readability, <lora:plants1:1.0>, <lora:plants2:1.0>, <lora:plants3:1.0>
			55 95 163
			153 207 241
Shapes and forms				Smooth curves for stability and comfort minimize angular forms and reflect natural shapes and structures, <lora:forms and structures:1.0>
Color + shapes and forms	Sky			Using blue as the accent color, (natural tones: sky blue), broad areas for visibility. High contrast for readability, Smooth curves for stability and comfort minimize angular forms and reflect natural shapes and structures, <lora:sky1:0.7>, <lora:forms and structures:0.7>
	Plants			Using green (or brown) as the accent color, (natural tones: green or brown), broad areas for visibility. High contrast for readability, Smooth curves for stability and comfort minimize angular forms and reflect natural shapes and structures, <lora:plants1(or 2,3):0.7>, <lora:forms and structures:0.7>

summarizes the prompts and LoRA training data used for image generation.

Table 12. Biophilic design attribute LoRA models generated images.

			Case A	Case J
Seed image
Attributes	Color	Sky
		Plants
	Shapes and forms
	Color + shapes and forms	Sky
		Plants

presents the results of biophilic façade image generation using the biophilic attribute LoRA model. For the color attribute LoRA model, sky elements were emphasized with sky blue, while plant elements were highlighted using green and brown as accent colors. Compared to conventional senior multi-family housing, blue and green were applied more distinctly, demonstrating pronounced color contrast. Through this, it was confirmed that the color LoRA model can effectively reflect the colors of the biophilic design palette. The application of the shapes and forms LoRA attribute model generated biophilic façade images featuring organic curves, contrasting the linear and repetitive designs of conventional senior multi-family housing. Notably, natural forms were integrated into architectural elements such as roofs, columns, and windows, indicating that the generative AI accounts for both aesthetic and structural aspects of design. In addition, it was confirmed that curves were applied while maintaining the symmetry and structural features of the existing seed image. Finally, combining color with shape and form models produced façades with balanced proportions of curves and straight lines, reflecting biophilic design principles. In particular, a design featuring the harmonious application of blue and green was implemented. The simultaneous application of color and form attributes resulted in a more pronounced color contrast than that observed in existing senior multi-family housing, along with a well-balanced proportion of curved and straight lines. Nevertheless, in order to validate the visual enhancements observed in the generated façades, a quantitative evaluation of color contrast (e.g., ΔE) and curve ratio between the seed and generated images is required.

6.2. Quantitative Evaluation of Generated Biophilic Façade Designs

To objectively evaluate the generated façade images, a quantitative analysis was conducted by comparing them with the corresponding seed images, focusing on two key attributes: color contrast and curve ratio. For the color attribute, the CIE76 color difference formula (ΔE*ab), based on the Euclidean distance in the CIELAB color space, is commonly used as a fundamental method for quantifying color differences in digital images [96]. In this study, the color contrast between each generated and seed façade image pair was measured using the ΔE*ab value, thereby enabling an objective assessment of color variation. According to Mokrzycki & Tatol [97], when ΔE*ab exceeds 2, the color difference becomes perceptible to an average observer, and values exceeding 3.5 are generally recognized as distinct differences. According to the analysis results presented in

Table 13. Comparison of color differences (ΔE*ab) between weed and generated images.

Attributes	Elements	Color ΔE*ab
		Case A	Case J
Color	Sky	+18.18	+9.40
	Plant	+2.63	+15.63
Color + shapes and forms	Sky	+17.77	+10.57
	Plant	+4.61	+14.63

, most of the generated images showed ΔE*ab values exceeding 3.5, indicating a clear improvement in color contrast compared to the seed images. However, the plant element under the color attribute in Case A recorded a ΔE*ab value of 2.63, which still represents a perceptible color difference for general observers. Therefore, it can be concluded that, overall, the generated images exhibit a distinct enhancement in color contrast relative to the seed images.

Regarding the ‘Shapes and Forms’ attribute, the extent to which curved elements were incorporated into the generated façades was quantitatively analyzed using ImageJ version 1.41, an image processing software. ImageJ is a publicly available, cross-platform tool widely used in biomedical and visual research for its flexible capabilities in structural analysis and quantitative measurement [98].

Table 14. Comparison of curve ratio (%) and improvement rate.

Attributes		Curve Ratio (%)
		Case A			Case J
		Seed Image	Generated Image	Improvement Rate	Seed Image	Generated Image	Improvement Rate
Shapes and forms		6.0	51.36	+45.36	0	61.53	+61.53
Color + shapes and forms	Sky	6.0	48.65	+42.65	0	38.19	+38.19
	Plant	6.0	40.48	+34.48	0	56.87	+56.87

shows that the curve ratio increased by at least 30% and up to over 60% in all cases, indicating a significant enhancement of curved elements compared to the seed images. These quantitative results support the visual observations discussed in Section 6.1 and indicate that the generative AI model effectively integrated biophilic design attributes—such as color variation and curved forms—into the façade designs.

7. Discussion

This study presents a biophilic façade design approach for senior multi-family housing through the use of generative AI. The findings highlight the potential of interdisciplinary research in bridging aging-friendly architectural design with AI-driven design solutions. In particular, previous studies have emphasized that biophilic design has a positive effect on the cognitive function and emotional stability of the elderly [45,87]. Building on these findings, this study explored foundational approaches to efficiently incorporate biophilic attributes into façade designs for senior multi-family housing using Stable Diffusion and LoRA training. In particular, the application of ControlNet confirmed that biophilic design elements can be effectively integrated while preserving the structural consistency of seed images. This can be suggested as a method to improve the limitations of existing generative AI-based architectural design [17,89].

An analysis of previous studies enabled the identification of façade design elements that align with the visual preferences and perceptual characteristics of older adults. Existing studies indicate that older adults experience reduced contrast sensitivity and color perception, necessitating façade designs that enhance visual readability and wayfinding functionality [99]. The differentiation of façade design is meaningful as it reflects the principles of Aging in Place (AIP) within architectural and environmental planning for older adults. This study analyzed the monotony and lack of sensory elements in existing senior multi-family housing façades and derived design prompts focused on color and shapes and forms to address these issues. The reason why we focused on color and shape in this study is because they affect the emotional stability and cognitive ability of older adults [54,100]. By employing the diffusion model, LoRA, and ControlNet, this study demonstrates that AI-based architectural design offers significant potential for enhancing time and resource efficiency during the early design phase [11,17,89].

The primary distinction of this study lies in addressing the monotony of existing Korean senior multi-family housing façade design and integrating the characteristics of older adults with biophilic design attributes. Biophilic elements, including natural textures, organic forms, and color adjustments, have been found to have a positive impact on the mental well-being and social interactions of older adults [45]. This study concretized these elements into color and shape and form attributes and applied them to senior multi-family housing façade design. Additionally, to validate the performance of the LoRA model, the loss function graph was analyzed and the Frechet Inception Distance (FID) score was evaluated according to weight changes. This aligns with previous studies demonstrating that the FID score is an effective metric for quality assessment in AI-based architectural design [17]. Through this analysis, the training stability and performance of the LoRA model were objectively assessed, confirming the practical feasibility of AI-based models in senior multi-family housing façade design.

However, this study should consider the following potential limitations. The biophilic design attributes for visualization focus primarily on specific façade elements such as color and shapes and forms. This prioritizes visual attributes that emphasize critical characteristics in biophilic façade design. Furthermore, the image dataset used for color training contained only RGB information, which imposed certain limitations in generating realistic color compositions and applications as seen in actual buildings. In particular, while the color LoRA model successfully emphasized individual color attributes, it demonstrated limitations in accurately recognizing specific RGB codes. Recently, studies such as ColorPeel [101] and Control Color [102] have advanced techniques for precise RGB-based color control within diffusion models. Integrating such methods with the LoRA model is expected to enable more accurate and controllable color representation. In addition, from a morphological perspective, the LoRA model exhibits structural limitations in generating complex forms, such as nonlinear or free-form surfaces. Therefore, future research should consider the possibility of integrating three-dimensional shape-aware generative models, such as 3D-aware diffusion [103], to address these limitations. In addition, it is necessary to expand the diversity of biophilic design attributes and façade elements by incorporating additional features such as texture into the LoRA training process. Older adults perceive differences in visual textures, as the diversity of materials influences their sensory experience [104]. For the color LoRA model, integrating diverse AI technologies is essential to achieve precise color implementation. Furthermore, quantitative analysis and validation are required to assess whether façade designs harmonize with the surrounding landscape.

8. Conclusions

This study proposed a differentiated façade design approach that reflects the color palette and morphological characteristics of biophilic design by utilizing generative AI and LoRA models.

The primary conclusions addressing the research questions proposed in this study are as follows. First, existing façade designs for senior multi-family housing inadequately account for the reduced visual information processing speed and diminished spatial cognition abilities of older adults, while also lacking differentiation in color and shapes and forms or integration of natural elements. Furthermore, prior studies predominantly focus on interior spaces, welfare facilities, and multi-family housing façades, whereas research on façade design for senior multi-family housing remains insufficient. Therefore, it is important for the façade design of senior multi-family housing to align with the concept of Aging in Place (AIP) by incorporating clear color contrasts and intuitive exterior forms to support visual recognition and wayfinding. Furthermore, reflecting familiar architectural elements is essential to evoke a sense of recognition and psychological comfort. In addition, it is necessary to increase psychological stability by introducing natural materials such as wood, brick, and greenery using biophilic design, and to create an environment that supports physical and cognitive functions by reflecting the AIP concept.

Second, we analyzed cases of senior multi-family housing in South Korea and abroad to derive biophilic design attributes that can be linked to façade design. The color scheme of the façade in senior multi-family housing in South Korea primarily uses neutral tones with medium-to-low brightness to convey a sense of stability. However, the lack of color contrast results in reduced visibility and individuality. In contrast, worldwide cases tend to emphasize color contrast and uniqueness by using high-brightness, high-saturation colors. Therefore, leveraging primary colors as accents—particularly natural hues such as blue and green—is effective for inducing psychological stability, considering the diminished color perception of older adults. While senior multi-family housing in South Korea predominantly features simple box-shaped structures and linear designs, worldwide cases emphasize spatial richness through the use of three-dimensional façades, windows, and balconies. Incorporating organic curves observed in nature into façade design enhances psychological stability and strengthens connectivity with the natural environment. Integrated strategies for color and form are as follows. It is important to use primary colors as accent colors to create clear color contrasts, and to apply biophilic colors that can harmonize with the surrounding nature, such as green and blue, to induce psychological stability. In addition, it is important to introduce curved shapes that represent nature to strengthen the connection between the physical building and nature; to use windows, balconies, and three-dimensional façades to enhance the sense of three-dimensionality and space; and to design a façade that gives personality. Incorporating distinctive elements into façade design is regarded as a strategy to foster a sense of identity among older adults in their residential spaces.

Third, this study proposed a methodology to enhance the accuracy and applicability of façade design for senior multi-family housing through generative AI, LoRA, and ControlNet models. Existing senior multi-family housing façade designs often lack consistency and restrict experimental iterations due to reliance on manual processes. In contrast, the LoRA model enables quantitative learning of biophilic color and shapes and forms attributes, improving design feasibility, while ControlNet allows experimental application of diverse design variations while preserving structural integrity. Thus, the synergistic integration of LoRA and ControlNet maximizes design adaptability while maintaining consistency. In particular, it is effective in automatically generating a contrasting color scheme that considers the color perception characteristics of older adults and a three-dimensional façade that reflects natural curves. It demonstrates the potential for systematic scalability of biophilic design in senior multi-family housing, offering enhanced precision over conventional manual design methods.

Fourth, this study constructed an integrated generation pipeline for façade design that reflects the perceptual and visual needs of older adults, incorporating prompt engineering, dataset construction, and LoRA hyperparameter configuration. First, the prompts were designed based on the color perception characteristics of older adults and principles of biophilic design, emphasizing natural hues (such as blue and green), high-brightness contrast, and curved forms. The dataset was built through a multi-step process involving keyword-based image collection, expert evaluation for high-quality image selection, and image-text pairing via BLIP captioning, ensuring domain fidelity and learning suitability according to the criteria outlined in Table 6. In the LoRA model training stage, key hyperparameters were configured, and the Lion optimizer was applied to enable efficient and stable learning. Subsequently, ControlNet (Canny model) was utilized to preserve the structural integrity of the seed images while enabling flexible image generation with curved and color-enhanced attributes. This methodology effectively supports the generation of biophilic façade designs tailored to the sensory characteristics of older adults, demonstrating both the practical applicability and scalability of AI-based architectural design for senior housing environments.

Finally, loss function analysis and Frechet Inception Distance (FID) score evaluation were conducted to validate the training stability and generative quality of the LoRA model. The loss function analysis revealed a gradual decrease in loss values as training progressed, indicating that the model effectively learned the data distribution. After 30 epochs, the model exhibited convergence patterns, suggesting that the LoRA model progressively improved image generation performance by identifying optimal weight values through training. However, relying solely on loss reduction to assess model suitability has limitations, and additional quantitative evaluation is required to mitigate the risk of overfitting. Based on FID score analysis, setting the weight values within the range of 0.7–1.0 was determined optimal to ensure the LoRA model’s peak performance. The evaluation methodology presented in this study facilitates the precise application of biophilic designs tailored to the color perception of older adults, enhancing the visual quality of façades in senior multi-family housing.

This study examined the necessity of improving façade design for senior multi-family housing and proposed a novel approach to the façade design process. It also confirmed the practical feasibility of AI-based architectural design. Notably, the methodology developed in this study—integrating biophilic façade design through generative AI models and fine-tuning—represents a significant contribution to architectural design and related fields. In other words, the AI-driven biophilic façade design holds critical significance by providing a foundation for visualization potential during the initial architectural design phase. The image data and LoRA model developed in this study demonstrate practical applicability when integrated with immersive technologies such as virtual reality (VR), augmented reality (AR), and the metaverse. Furthermore, the proposed approach is expected to extend beyond senior multi-family housing and be applicable to various domains within architectural design.

While this study focused exclusively on the visual and morphological attributes of building façades, future research should expand its scope to explore the interrelationship between interior and exterior environments. Specifically, investigating how façade design influences residents’ interior experiences—such as window views, natural lighting, and psychological comfort—can further enrich the understanding of age-friendly architecture and support more holistic residential design strategies.