The Development of the Two-Dimensional Model of Emotion Based on Both Architectural Emotion Words and Design Elements in China

Abstract

The purpose of this paper is to construct and validate a model centered on architectural emotion, to explore the role of emotion in architectural design, and to provide theoretical support for emotion-oriented design. This study collected 614 terms related to architectural emotion, screened 30 core terms, constructed a two-dimensional architectural emotion model, and verified the scientific and practicality of the model through three measurement methods. First, the one-dimensional scale analysis identified two dimensions of pleasure and charm, which portrayed the range of word variation; second, the Principal Component Analysis confirmed the periodic ordering pattern of words, which revealed its systematic relationship; and, lastly, the Multidimensional Scaling Analysis demonstrated the distribution of emotion words based on cognitive similarity in the multidimensional space. Based on this model, this paper proposes a three-layer circular model of “architectural emotion-architectural cognition-architectural elements”, which constructs the correspondence between architectural emotion and design elements, as well as how architectural cognition and architectural elements can synergize to create a spatial experience that triggers specific emotions. The model provides theoretical support for emotion-oriented architectural design and evaluation, and helps designers to better understand the relationship between emotion and space so as to create more valuable architectural works.

1. Introduction

The design of architectural space significantly affects the emotional experience of users, and architectural design elements such as space, form, color, light, and material can stimulate people’s emotional response, which, in turn, affects users’ cognitive evaluation and behavioral decision making. From Vitruvius’ three principles of “strong, applicable, and beautiful” [1] to the four principles of “practical, economic, green, and beautiful” proposed by China’s architectural industry [2], architectural design mainly focuses on the functional, aesthetic, technological, and energy-saving aspects of the building, but not on the emotional experience of users. The main focus of architectural design is on the functional, aesthetic, technical, and energy-saving aspects of buildings, whereas there is a lack of quantitative scientific research on the emotional and psychological impacts on users. Emotions and psychology are subjective and difficult to be systematically quantified and modeled, requiring interdisciplinary cooperation between architects and experts in the fields of psychology, cognitive science, neuroscience, and environmental design (biophilic design, fractal geometry). With the advancement of technologies and experimental methods such as brain science, sign monitoring, affective computing, and artificial intelligence, it has become possible to accurately measure and analyze the users’ emotional responses.

The core issue of this paper is the emotional experience of architecture, i.e., the spatial expression of emotional cognition. Among them, the emotional words expresses the emotional state through language, which is the tool and carrier of emotional cognition, and the architectural design elements are the compositional basis of emotional experience. In the research framework, multiple disciplines, techniques and methods mentioned above are involved. Through the following research review, the interdisciplinary and comprehensive framework of emotional experience is further organized.

Cognitive architecture combines cognitive science, psychology, neuroscience, and architecture to study how humans perceive and respond to the built environment through their senses (e.g., sight, sound, touch). Ann Sussman and Justin B. Hollander proposed the four principles of Edge Importance, Pattern Importance, Shape Importance, and Narrative Importance for guiding emotionally driven design [3]. Juhani Pallasmaa emphasizes the multidimensional impact of architecture on the senses and emotions [4], Peter Zumthor discusses the shaping of emotional experience by architectural ambience, and Charles Jencks puts forward the concept of “symbolic architecture” to highlight the emotional and narrative aspects of design [5].

Neuroarchitecture is based on neuroscience, which investigates how the built environment regulates emotion, cognition, and behavior by influencing neural activity in the brain, a concept first proposed by Jonas Salk and Francis Crick in the 1980s [6], and the establishment of the American Neuroarchitecture Faculty (ANFA) in 2003 marked the development of the field as an independent field. In recent years, neuroscience techniques (e.g., fMRI, EEG) have been widely used to study the link between architectural design and emotional responses in the brain. For example, Navid Khaleghimoghaddam’s study showed that pleasant and unpleasant spaces can significantly alter blood flow patterns in emotion-related brain regions [7].

Fractal geometry resonates with the human psyche through self-similarity and visual complexity, and has been applied to nature simulation and architectural design. Richard Taylor investigated the positive effects of fractal patterns on the psyche [8], and Michael Mehaffy and Nikolaus Salingaros proposed design principles based on fractal theory [9]. Donald Ruggles explores the potential of combining fractals with neuroscience in architecture, emphasizing the impact of design elements on emotion and health [10], while Alexandros A. Lavdas emphasizes combining visual complexity with psychological mechanisms to create emotionally resonant spaces [11].

Biophilic design proposes to enhance the emotional experience through the closeness of natural elements to humans, and Edward O. Wilson proposed the Biophilic Hypothesis in 1986, arguing that this connection to nature is essential for health [12]. Stephen R. Kellert systematized the theory, and Terrapin Bright Green’s team proposed 14 design patterns, including direct and indirect nature experiences in 2014 [13]. Dalay (2020) indicates that biophilic design elements, such as natural light, plants, and natural materials, significantly enhance the atmospheric perception of interior spaces and improve users' emotions and overall experience [14]. In 2024, Peter Gloor introduced artificial intelligence techniques to validate the positive effects of biophilic design on mood through facial emotion recognition (FER) and emotion detection [15].

There are five commonly used emotion models in the study of the application of emotion models in enhancing users’ emotional experience, among which, Norman’s Emotional Design Theory, Kano Model [16], and Osgood’s Semantic Differential Model are mainly applied to the research in the field of product design and product service enhancement. The assessment of emotional experience in architectural space mainly uses the following two-dimensional emotional model, Russell’s two-dimensional emotional ring model (pleasantness, arousal) proposed in 1980, and Russell and Mehrabian added dominant dimensions to the original two-dimensional model in 1974 to form the PAD three-dimensional model. Among them, pleasantness describes whether people’s subjective feelings about the space are positive or negative, and the degree to which the space makes people feel comfortable and pleasant to depressed and dissatisfied. Arousal describes the degree of emotional arousal of the space to people, i.e., whether the space is able to stimulate people’s emotional response. Dominance describes the degree of emotional control and sense of dominance that the space has over the user, i.e., whether the user feels that the space is oppressive to them, and whether they are able to use and control the space freely. Examples include hospitals, prisons, theaters, and affective research in large public and authority buildings [17,18,19,20]. These three dimensions do not exist in isolation; they influence and interact with each other. It is worth noting that pleasure and arousal are more commonly used in the study of emotions in buildings and spaces due to their universal applicability, ease of measurement, and central role. In contrast, dominance is more difficult to measure accurately as it relates to an individual’s sense of control, sense of power in a space, or autonomy, aspects that are not necessarily apparent or subjectively realized in a given space.

The research is approached from different perspectives, including the analysis of environmental psychology in architecture and urban spaces, the assessment of emotional experience in public spaces and interior design, the role of design elements (e.g., light, color, materials, etc.) on emotional states, the study of outdoor natural environments and soundscapes, and the measurement of emotion in virtual spaces with digital technology. For example, tall spaces are associated with a sense of openness and freedom and contribute to a positive emotional experience (efficient valence). Small and low spaces, on the other hand, may lead to a sense of oppression, lowering the valence and enhancing negative emotions. Soft colors and natural materials can enhance the pleasantness of a space, making people feel comfortable and relaxed. Cool colors and overly industrial materials (such as metal and concrete) can lead to feelings of emotional distance or indifference. Smoothly designed spatial movement lines can increase the sense of pleasure and freedom and reduce anxiety (efficient price), while complex and confusing movement line designs may lead to feelings of dissatisfaction and frustration, among others. Through the above emotional evaluation, mainly through the two dimensions of “arousal” and “pleasure” in the model, the researcher was able to quantify and analyze how the architectural design affects the users’ emotional responses (Table 1).

Table 1. Summary of the literature related to the application of Russell’s two-dimensional affective model in architecture.

Field	Reference	Research Content
Architectural Space	[21,22,23]	Banaei et al. (2020): This study shows that assessing people’s emotional responses to the form of building interiors through virtual reality reveals how personality differences affect people’s emotional interactions with space, supporting the application of the 2D affective model to architectural design. Robert and John (1982): This paper explored how store ambience affects customer emotions and behavior, showing that spatial design can elicit a range of emotional responses consistent with the two-dimensional affective model. Zhang et al. (2023): This study explored the emotional responses of museum visitors in a digital technology environment, emphasizing that the design of exhibition spaces can elicit different emotions, consistent with the valence and arousal dimensions of the two-dimensional affective framework.
Design Elements	[24,25,26,27,28,29]	Pereira M. (2018): this article explores how color affects people’s spatial experience through visual effects and emotional stimuli in architectural design, showing how color affects emotional responses through two-dimensional affective models such as pleasantness and activation. Zhang C et al. (2024): This study used affect theory to analyze the impact of color on the emotional responses of older adults in community outdoor spaces using a color–emotion loop diagram based on a two-dimensional affective model. Custers et al. (2010): This study looked at how lighting in retail environments affects emotional perception, suggesting that spatial design elements (e.g., lighting) can manipulate emotion on the pleasure-activation dimension, fitting a two-dimensional affective model. Rankel (2014): This study examines the impact of urban nighttime lighting design on urban appearance and emotion through a case study, showing how lighting as a design element can influence pleasure and activation in a two-dimensional emotion model. Kim and Mansfield (2021): This paper proposes a method to create a positive atmosphere in an office through lighting design, showing that lighting can significantly influence the emotional experience in office environments, fitting the pleasure and arousal dimensions of the two-dimensional affective model. Lee and Lee (2022): This study examines the effects of colored lighting on pleasure and the activation of emotions in different cultural contexts, demonstrating how design elements can influence emotional responses through color in a two-dimensional model of emotion.
Outdoor Environment, Soundscape Design	[30,31,32,33,34,35]	Yin et al. (2024): This study explored soundscape design in urban green spaces using Russell’s Affective Circumplex Model, demonstrating how soundscape design elements can influence human emotional experience, fitting the pleasure and activation dimensions of the two-dimensional affective model. Zhang et al. (2024): This study compares the difference between perceived and actual emotions in urban open spaces, demonstrating the role of sound design elements in emotional evaluation, fitting the validity–arousal framework of the two-dimensional emotion model. Jin et al. (2024): This paper explores the effects of visual and auditory design elements on human behavioral and emotional responses in urban parks, showing how these elements influence pleasure and activation in a two-dimensional emotion model. Ren et al. (2024): Using eye-tracking analysis, this paper assesses perceptions of rural public spaces, demonstrating how spatial design elements influence affective responses, fitting the framework of a two-dimensional affective model. Ren et al. (2024): This EEG-based study assesses the role of virtual reality technology in rural streetscape design, revealing how streetscape design elements affect human relaxation and activation responses, fitting the principles of a two-dimensional affective model. Malekinezhad et al. (2013): This paper compares the emotional responses of designers and non-designers to urban sculptures, using an emotional grid to show the impact of sculpture design elements on emotion, in line with the pleasure and arousal dimensions of the two-dimensional emotion model.

Although important progress has been made in architectural emotion research [36], current research mostly focuses on a single study of certain specific emotions (e.g., pleasure, relaxation, arousal), and a unified framework for evaluating architectural emotion is still lacking. The comprehensive correspondence between emotion types (e.g., awe, sense of belonging, concentration) and specific architectural elements has not yet been clarified, and it is difficult to systematically reveal the impact of architectural design on multidimensional emotional experience. In addition, the mechanisms of dynamic changes in emotion are understudied, especially the transition and transmission of emotion in spatial sequences. For example, Jeroen Nawijn (2024) pointed out, in his study of emotional clusters in concentration camp memorials, that visitors’ emotions experienced emotional transitions from pain and despair before the visit to sympathy and reflection during the visit, and then to thanksgiving, hope, and cherishing after the end [37]. Weber (2024) discusses how spatial changes and sequences profoundly affect emotional experience from the perspective of psychology [38]. This suggests that architectural emotional experience does not occur in a discrete manner, but has a spatial dimensional relationship, and is the result of the comprehensive experience of the overall senses of the environment and the long-term interaction of multiple senses.

This paper aims to construct a systematic classification and multidimensional model of architectural emotions, and explore the comprehensive correspondence between emotion types and architectural design elements. Specific objectives include generating a two-dimensional emotion model first based on psychological research, semantic analysis, and questionnaire surveys; core words related to architectural emotions are screened out and these emotion words are mapped to the corresponding positions in the two-dimensional emotion model through experimental data so as to generate a two-dimensional emotion model exclusive to architectural emotions. Second, a three-layer cyclic relationship model is constructed by combining architectural emotion, architectural cognition, and architectural elements; the association between emotion, cognition, and design is systematically presented to provide a theoretical basis for emotion-driven architectural design. Through these studies, the understanding of architectural emotional experience will be further deepened to provide more scientific and comprehensive theoretical support for emotionally driven architectural design. Therefore, the following hypothesis is proposed in this paper: The architectural emotion model can define two-dimensional emotions using architectural words.

2. Methods and Materials

2.1. Method

This study aims to explore in depth the mechanism of architectural emotional word induction. Firstly, referring to Nowlis’ [39] emotional word collection method, 614 emotional words closely related to architecture were collected through Visual Material Analysis, subjective experience, and a literature review. According to Osgood [40], the collected emotional words related to architecture were finely classified and screened using similarity reduction, participant scoring, and high-frequency word screening, and 30 representative emotional words for architecture were refined; an expert model was used to evaluate and revise the 30 emotional words.

The primary objective of this study was to elucidate the relationship between architectural elements and emotions, as well as to aid architects in intuitively analyzing emotional responses to architecture prior to the design process. To achieve this, participants were exclusively recruited from students and faculty with backgrounds in architectural design. A total of 15 participants were selected through convenience sampling methods, including word-of-mouth and email announcements. All participants were affiliated with the School of Architecture at Hebei University of Engineering in China; 12 were undergraduate and graduate students, while the remaining 3 were faculty members. Among the participants, 7 were female and 8 were male. Their ages ranged from 19 to 45 years, with a mean age of 26.5 years. All participants were confirmed to be free of cognitive impairments, and each provided written consent before engaging in the study activities. Each participant carefully examined the emotional aspects of the study prior to scoring.

Participants rated 30 word items for similarity, and the mean similarity matrix was calculated. To conduct a comprehensive analysis of the sentiment words related to architecture, three methods were employed: Principal Component Analysis (PCA), Multidimensional Scaling Analysis (MDS), and Unidimensional Scaling Analysis (UDS). These methods were used to process and analyze the data from the average similarity matrix, ultimately leading to the development and validation of a two-dimensional coordinate annular model of architectural sentiment words. Finally, the collected words related to architectural elements, architectural perception, and emotions were structured to generate a three-layered circular model.

Principal Component Analysis (PCA) is a multivariate analysis technique aimed at extracting significant information from multiple interrelated data sets of quantitative dependent variables. It represents this information as a new set of orthogonal variables known as principal components, allowing for the visualization of patterns of similarity among observations and variables as points on a map. The data for this study consisted of similarity scores of words, with the objective of analyzing and visually representing the interrelationships between these words. Ultimately, the goal was to construct a two-dimensional model; therefore, PCA was initially selected to analyze and reduce the dimensionality of the variables comprising multiple word similarity scores.

Multidimensional Scaling Analysis (MDS) aims to transform multiple data matrices that represent similarities into a two-dimensional space that reflects the original multidimensional space. This transformation utilizes Euclidean distances to identify correlations among the data. In this study, the data were employed to calculate the interrelationships between words based on the similarity of word pairs. The ultimate goal was to create a two-dimensional model that validates a previously established two-dimensional model of building-related emotional words obtained through Principal Component Analysis. Consequently, MDS was selected to compute the similarity matrix and illustrate the positional relationships of words within a two-dimensional space.

The purpose of using Unidimensional Scaling Analysis (UDS) is to convert two-dimensional data into one-dimensional data, identify the maximum and minimum values of the two axes in the two-dimensional model, and provide a reference for defining the X- and Y-axes of the model.

The following (Figure 1) is the construction process of the two-dimensional sentiment model and the three-layer circular model of the architectural sentiment words.

Figure 1. Research steps.

While architectural functions are clear and concrete, emotional needs are subjective, diverse, and harder to quantify. Accurately understanding and quantifying these needs helps architects create designs that resonate emotionally, known as emotional architecture [22]. Emotions are influenced by architectural elements, user psychology, and cultural contexts. Based on Russell’s emotion cycle model, emotional words are collected and analyzed through similarity principles and participant ratings, clarifying the relationships between emotional words, cognitive phrases, and design elements. For instance, materials like natural stone, brick, and wood evoke feelings of naturalness, intimacy, and timelessness due to their historical and tactile qualities. Visual features such as soft colors and curved shapes also promote calm and relaxation. A two-dimensional emotional model links emotion, perception, and design elements, enabling a more precise emotional design. Incorporating emotion in architecture goes beyond aesthetics, aiming to enhance user experience. Emotion modeling allows designers to predict and improve emotional responses, creating effective user-centered spaces and supporting emotional evaluations, which provides valuable insights for future design innovations.

The following (Figure 2) is the one-to-one relationship of the three levels of the three-layer circular model.

Figure 2. Hierarchical relationship diagram of emotion model.

2.2. Research Steps

2.2.1. Data Collection

Word type: Collect emotion-related words and phrases that occur when individuals interact with buildings or living environments.

Number of words: The objective was to gather 600 words; ultimately, a total of 614 words were collected.

Word sources:

• Visual Material Analysis: This study prepared three sets of materials: texts, images, and videos. The texts were derived from the architectural design needs questionnaire and the architectural evaluation document. The images included photographs of buildings that represented various styles, functions, and cultural backgrounds. The videos consisted of film clips and documentaries focused on architectural themes. Members of the research team reviewed the materials and documented the emotional responses they elicited, ultimately generating words related to architectural emotions.
• This document provides a systematic compilation of emotional words related to architecture from the perspective of the architectural profession. It specifically addresses the physical and design elements of architecture, including form, structure, volume, space, function, technology, materials, color, light, sound, and thermal engineering. The objective is to comprehensively capture the emotion responses of users that may arise from various architectural elements.
• This study also referenced the literature related to architectural design [41,42,43] and architectural psychology [44]. It incorporated Russell’s [45] emotion model to enhance the understanding of the emotions evoked by buildings, considering both positive and negative emotion intensities, as well as varying levels of arousal.

During the word collection process, it was found that some words can be used as both architectural adjectives and emotional words, such as the word “innovation”. Some words belong to the words of architectural element perception, such as rhythm, sense of belonging, sense of creativity, sense of eeriness, and so on, which have a strong representation of architectural emotions. As an architectural cognition word, “eeriness” is composed of two parts, “eerie” and “sense”, which can trigger emotions such as fear, anxiety, and depression. Therefore, it is retained in the word database at the initial stage, which provides the word database for establishing the model of architectural elements, architectural cognition, and architectural emotion.

2.2.2. Similarity Reduction of 614 Words

• Reducing the words to 300 words through similarity analysis: Firstly, based on Tomkins’ [46] theory of grounded emotion, the collected words were ordered from left to right based on negative, neutral, and positive association, and then from top to bottom based on the arousal of the words. In the process of words sorting, the similarity assessment method of Osgood [40] was borrowed for word reduction and categorization. Chinese words usually consist of two to three characters, each of which has an independent meaning and is combined together to form word meanings. In the reduction process, we first considered words with the same characters in the words, which usually have a high degree of similarity, such as “appreciation”, “admiration”, “exclamation”, and so on. Words with weak architectural relevance were deleted. In addition, certain words that express the same architectural attribute or purpose, such as “curiosity”, “desire to explore”, “mystery”, etc., were grouped together for similarity reduction, and retained as words that were more instructive for architectural evaluation and design. The words that were more instructive for architectural evaluation and design were retained.
• Grouping words to 100 words: In order to facilitate the reduction in words again, the collected words were grouped from the process of emotion generation (see Table 2). People initially obtain information from the environment through visual, auditory, and other sensory modalities, which are combined with historical and cultural perceptions and transformed into emotions in the limbic system (amygdala) of the brain [47]. Therefore, the words were grouped in terms of physical perception, cognition, and emotion. People produce visual stimuli through viewing architecture combined with individual experience to produce cognition, such as magnificence, art, glamour, etc., which is eventually transformed into emotion in the brain, such as excitement, shock, and warmth. After grouping, words with little correlation with architectural emotions were deleted from the words with similar cognition and emotions, and words that were more responsive to architectural emotions among the similar words were retained.
  Table 2. A list of 100 emotional words.

  Awake	High	Desire	Majestic	Twisted	Infectious	Embarrassment	Peace	Excited
  Brisk	care	Urgent	Worried	Depressed	Prosperous	Convenient	Art	Apathetic
  Happy	Calm	Accept	Agitated	Respect	Comfortable	Affectionate	Anxiety	Sleepiness
  Shock	Cure	Bright	Pleasure	Relaxing	Stimulation	Satisfaction	Inspire	Loneliness
  Rules	Joy	Curious	Disgust	Arrogant	Liberation	Coordination	Anger
  Sober	Fair	Sadness	Release	Intimate	Respectful	Magnificent	Hate
  Safe	Avant	Smooth	Cherish	Gratitude	Dreariness	Affirmation	Shameful
  Rigor	Dark	Freedom	Hopeful	Cultural	Diversity	Perfunctory	Abstract
  Sinful	Ugly	Gloomy	Delicate	Confused	Restraint	Entanglement	Obsession
  Pride	Cozy	Fragile	Remorse	Depressed	Burdensome	Nervousness	Rough
  Strange	Sharp	Clever	Cautious	Desperate	Attachment	Challenging	Pain
  Unity	Fear	Beauty	Creative	Favorite	Incentive	Interesting	Boredom

• Redefining the emotional words and reducing the words to 30 (see Table 3): The emotional and cognitive words of the 100 words were combined into one group, which served as the base material for the construction of the three-layer ED model of architectural emotion language. Redefining the emotional words as per Ekman’s theory of emotion [48]: Emotions are rapid and spontaneous mental states, usually triggered by specific stimuli and accompanied by facial expressions, body gestures, voice changes, and physiological responses. Fifteen architecture students and faculty members were recruited to screen 30 words representative of architectural emotions from 100 words, respectively. Before screening the words, participants learnt about the concept of emotional words and the process of constructing emotional words. They were then asked to select non-pure emotional words from the 100 words and transform them into emotional words, e.g., “beauty” could be transformed into “happiness” and “satisfaction” to form a new emotional words. Through word frequency statistics, the frequency of each emotion word was calculated and ranked and, based on the principle of architectural representativeness and word similarity, 30 strictly architectural emotional words were filtered out.
  Table 3. The final 30 words.

  Excite	Curious	Shocking	Joy	Favorite	Pride	Hopeful	Hate
  Respectful	Pleasure	Happy	Majestic	Satisfaction	Cozy	Comfort	Boredom
  Relaxation	Safe	Calm	Apathetic	Sad	Gloomy	Burdensome
  Desperate	Painful	Surprise	Fear	Anxiety	Cautious	Anger

  The word frequency statistics formula is as follows:

  $$\mathrm{f}(w)={\displaystyle \frac{cout(w)}{\mathrm{n}}}$$

  (1)

  where f(w) is the frequency of the word, count(w) is the number of times the word w appeared in the glossary, and n is the total number of all words in the glossary.
  The participants’ respective 30 emotional words were summarized and, still according to the word frequency statistics method, the words were frequency-calculated and sorted; the words with a higher frequency of occurrence were screened and the final 30 emotional words were selected by combining the principles of whether or not they have strong architectural representation and similarity.
• The explanation of the 30 architectural emotional words, including the concepts corresponding to the 30 emotional words, the architectural elements that may correspond to them, the expressions of architectural domains, and the corresponding expressions of architectural perception and cognition words with corresponding Chinese and English translations in the document. Architecture experts were invited to determine the final 30 most architecturally evoked representative emotional words through an expert discussion model.
• The 30 emotional words were rated by the participants for similarity with similarity scores ranging from 0.1 to 1, with the highest similarity score being 1 and the lowest similarity score being 0.1, creating a similarity scoring matrix of the 30 words. The mean of the participants’ 30 word similarity scores was calculated to create an average similarity matrix. Finally, the matrices were analyzed using three methods: Principal Component Analysis (PCA), Multidimensional Scaling Analysis (MDS), and Unidimensional Scaling Analysis (UDS), respectively.

3. Data Analysis

The data of 30 words with architectural emotions were processed by combining every 2 words, since 30 words were used. Their number can be calculated by [(30 × 29)/2] = 435. Whenever two words are assigned together, each of the 435 pairs is given a similarity score. Similarity scores ranged from 0.1 to 1, with higher similarities resulting in higher scores. Each participant would score each word pair three times, and the average score would be used as the final score for the word pair.

In the three sets of experiments, the highest similarity score was 1 and the lowest similarity score was 0.1, and the data from the three sets of participants were combined and normalized (see Table 4).

Table 4. Normalized list of mean matrices of similarity scores of architectural emotion words.

Data Normalization
	Average Value	Standard Deviation	Variance
Excite	0.429	0.224	0.050
curious	0.480	0.248	0.062
Shocking	0.407	0.275	0.076
Joy	0.362	0.274	0.075
Favorite	0.433	0.273	0.074
Pride	0.487	0.214	0.046
Hopeful	0.421	0.270	0.073
Respectful	0.414	0.274	0.075
Pleasure	0.453	0.258	0.066
Happy	0.441	0.208	0.043
Majestic	0.477	0.246	0.061
Satisfaction	0.437	0.257	0.066
Cozy	0.413	0.251	0.063
Comfort	0.449	0.244	0.060
Relaxation	0.423	0.258	0.067
Safe	0.412	0.277	0.077
Calm	0.414	0.282	0.080
Apathetic	0.449	0.245	0.060
Sad	0.414	0.239	0.057
Gloomy	0.442	0.251	0.063
Burdensome	0.420	0.231	0.053
Desperate	0.371	0.260	0.067
Painful	0.396	0.220	0.048
Surprise	0.389	0.200	0.040
Fear	0.387	0.243	0.059
Anxiety	0.373	0.237	0.056
Cautious	0.351	0.245	0.060
Anger	0.354	0.254	0.065
Hate	0.334	0.248	0.061
Boredom	0.432	0.293	0.086

3.1. Principal Component Analysis (PCA)

The scores for the similarity of the 435 word pairs were transformed into a word matrix of 30 words × 30 words for each participant. The word matrices were computed using Principal Component Analysis (PCA) in SPSS version 26 to obtain the potential relationships between each word factor. In order to control for acquiescent differences, calculations were performed by subtracting the median (centered value) from each participant’s score prior to analysis; this was performed to obtain a better two-dimensional feature structure that would represent the data relationships for the architectural affective words. The fit measure (‘dispersion’ ranging from 0 to 1) was 0.95. In order to remove the collinearity problem, we used the covariance analysis (ANOVA) model to process the original architectural sentiment word data, merged them into independent indicators, and extracted two groups of indicators that could fully reflect the overall information for subsequent analysis. After KMO and Bartlett’s test, the KMO values of the first three groups were 0.51, 0.85, and 0.91, respectively. The KMO values higher than 0.6 were taken as the final eigenvalues. At the same time, using Bartlett’s Test, p was less than 0.01, indicating that there was sufficient correlation between the variables and it was statistically significant (Table 5).

Table 5. Eigenvalues by PCA.

Component Matrix
	Component
	1	2
Majestic	0.932	0.142
Desperate	−0.921	0.137
Pleasure	0.919	−0.184
Painful	−0.896	−0.156
Fear	−0.892	−0.251
Anxiety	−0.888	−0.195
Shocking	0.884	−0.366
Satisfaction	0.882	0.353
Cozy	0.862	0.306
Respectful	0.848	−0.437
Favorite	0.821	−0.421
Hopeful	0.804	−0.467
Happy	0.777	−0.121
Hate	−0.776	−0.471
Joy	0.757	−0.554
Anger	−0.747	−0.522
Surprise	−0.745	−0.393
Burdensome	−0.740	0.439
Cautious	−0.699	−0.517
Gloomy	−0.654	0.626
Calm	−0.035	0.959
Curious	0.130	−0.954
Excite	−0.039	−0.931
Sad	−0.180	0.922
Safe	0.356	0.900
Boredom	0.366	0.888
Relaxation	0.500	0.813
Apathetic	−0.597	0.713
Comfort	0.569	0.703
Pride	0.623	−0.694

As shown in Table 6, in order for the data to be free of covariance, a covariance matrix-style weighted summation was used to extract each of the main components from the data, combining them into a new type of combination with the information of the original data indicators. As shown in Figure 3, all the extracted eigenvalues have a trend from steep to smooth, and after the fourth eigenvalue, the line segment becomes gentle. However, it is not easy to explain the feature relationship in high-dimensional dimensions, so the two most prominent eigenvalues are selected here to draw the spatial coordinate system.

Table 6. Feature extraction.

Total Variance Explained
Component	Initial Eigenvalues			Extraction Sums of Squared Loadings
	Total	% of Variance	Cumulative %	Total	% of Variance	Cumulative %
1	15.277	50.923	50.923	15.277	50.923	50.923
2	10.250	34.166	85.089	10.250	34.166	85.089

Figure 3. Gravel analysis.

We found the value with the largest degree of discretization to be the most X-axis of the new coordinate system and plot the new coordinate system. As shown in Figure 4, the variable projections on the X-axis carry most of the original amount of information and, therefore, become the first dimension, while the Y-axis, which carries the secondary amount of information, is the second dimension.

Figure 4. Fitted scatterplot.

According to the newly plotted coordinate system in Figure 5, “happiness” and “anger” are opposites of each other, and “happiness” and “depression” are opposites of each other. “Happiness” and “Anger” are in opposition to each other, and “Happiness” and “Depression” are in opposition to each other, clearly indicating a potency relationship in dimension 1. Positive words such as “love”, “adore”, “happiness”, and “joy” on the right side were in contrast to the left side’s “apathy”, “sadness”, “despair”, and “anxiety”, becoming opposites and forming negative emotions, and this dimension distinguishes positive and negative emotions into a back-and-forth row. From the “Y” axis of the two-dimensional circular axis in Figure 5, it can be seen that “Excite”, “Curious”, “Anger”, and “Hate” are in the positive direction, and “Calm”, “Gloomy”, and “Apathetic” are in the negative direction, showing a very obvious relationship between the arousal levels. Because the high-arousal “Excite” and “Curious” and the low-arousal “Calm” and “Gloomy” all appear in the positive and negative directions of the “Y” axis, showing an obvious “Arousal” relationship.

Figure 5. Two-dimensional cyclic model of PCA.

As shown in Figure 6, because, in the PCA scaling model in the first dimension, many words are arranged vertically and equally in order to better distinguish the order of words from each other, the maximum value of “Majestic” is used as the dependent variable in the positive direction, the linear regression model yields R = 0.919, F < 0.01, and the residual value is 4.521, which shows that, in this dimension, the larger the value of “Majestic” is, the higher the building emotion predicted by the regression model [49]. The larger the value of “Majestic”, the higher the architectural emotion predicted by the regression model. When the intensity of “Pleasure” and “Majestic” decreases, the architectural emotion words in the first dimension are the same as that of “Majestic”. “Majestic” is positively correlated the more negative emotions the model presents. Therefore, the “majesty” of architectural emotion is the greatest value.

Figure 6. Unidimensional Scaling model of PCA algorithm—first eigenvalue.

As shown in Figure 7, in dimension two, words that represent positive qualities such as “comfort”, “safety”, “relaxation”, etc., have a positive correlation with “Activation”. Words with very low values of “Activation” are on the right side of the second dimension with words with high values of “Activation”, such as “Excite”, “Curious”, etc., which represent negative emotions. “Despair” and “Fear”, which represent negative emotions, are in the middle of the second dimension, and are associated with “Pleasure”, “Happiness”, “Joy”, and “Reverence”. Therefore, the characteristics of this dimension are related to the degree of “activation” of the emotion, not to the potency of the emotion, and it also shows that the second dimension is related to the independent self. the second dimension is related to the experience of an independent sense of self (seeing the self as embedded in social relationships with others) [50,51]. As the second dimension carries a secondary amount of information, “Curious” and “Excite” appear as the negative direction of the one-dimensional map and “Relaxation”, “Safe”, and “Calm” appear as the positive direction of the one-dimensional graph [52]. From Figure 7, it is observed that “Curious” and “Excite” are very close to each other and need to be used as “Curious”, with “Excite” as the dependent variable, and utilize the regression model to calculate them separately.

Figure 7. Unidimensional Scaling model of PCA algorithm—second eigenvalue.

In summary, using “Excite” as the dependent variable, the linear regression model can obtain R = 0.920, F < 0.01, and the residual value is 3.05. Using “Curious” as the dependent variable, the linear regression model can obtain R = 0.932, F < 0.01, and the residual value is 3.05, 0.932, F < 0.01, residual value 3.682. “Curious” is more prominent than “Excit”, so when the value of “Curious” is higher, the value of “Curious” is more prominent than “Excite”. The higher the value of “Curious”, the higher the building emotion predicted by the regression model. On the other hand, when the value of “Curious” is lower, the realized emotions of “Relaxation”, “Safe”, and “Calm” are higher. From the architect's perspective, when “Curious” and “Excite” appear, the higher the participants' interest in the building, the higher the “Curious” value of the building itself. “Curious” is very influential in this dimension, so the second dimension is defined as “Curious”.

3.2. Multidimensional Scaling Analyses

The pre-preparation is similar to the Principal Component Analysis, where the word matrix of 30 words × 30 words is given to the Multidimensional Scaling Analysis to show the relative spacing of the 30 architectural emotion words and their location in the multidimensional space in which they are located. Based on the Multidimensional Scaling Analysis method mentioned by Russell (1980), the similarities and differences between the data were weighed equally to enable the results to respond to the conceptual distances between the various architectural affective words with similar affective mappings.

In the analysis, the dependent variable was treated as an ordered scale, so the neighboring variables of the model were set to be ordered. Since the data consisted of ratings of 435 word pairs, to avoid bias, the “unbind bound observations” function was used to solve this problem. A weighted Euclidean model was used as a standard model to solve the problem of individual differences in ratings within participants and the problem of not being able to rotate, using the importance of each dimension for each participant as the dimension weights. The model is a lower triangular matrix where the approximations are analyzed using similarity analysis. The initial setting for the model’s multidimensional scaling was Torgerson. Stress values were computed iteratively by adding one dimension at a time until their stress values approached 0.01. As the dimensions increase, the stress value decreases until a dimension no longer decreases significantly, and the maximum number of iterations is set to 100. The fit of the model is 0.20 for poor, 0.10 for fair, and 0.05 and below for good [53]. In the present results, the stress value keeps decreasing and the iteration stops when the calculation appears to be in the third dimension and the stress value is less than 0.001. At this point, all obtained RSQ values are 0.986 and stress values are 0.05. These values are in accordance with the criteria for a good fitting model. The 2D model of MDS obtained by calculating all the above is shown in Figure 8.

Figure 8. MDS 2D toroidal model.

As shown in Table 7, the positive maximum value of the first eigenvalue calculated by the MDS model is 0.675, which corresponds to the word “Shocking”, and the positive words in descending order are “Pleasure“, “Excite”, “Satisfaction”, etc. The architectural emotion words show an increasing trend of joyfulness, “Pleasure”, “Excite”, “Satisfaction”, and other positive words, and the building emotion words show an increasing trend of joyfulness. The minimum value is −0.757, which corresponds to the word “Anxiety”, and the values are “Anger” and “Hate” in descending order, “Fear”, and other negative words, and the building emotion words show a negative trend in terms of joy. This shows a clear indication of a potency relationship in dimension 1. This one conclusion is the same as all the results obtained by Principal Component Analysis (PCA), where the “Shocking” of architectural emotion is the greatest value.

Table 7. Eigenvalues extracted by MDS.

Highest Score				Minimum Score
X		Y		X		Y
Shocking	0.675	Excite	0.646	Anxiety	−0.757	Calm	−0.723
Pleasure	0.646	Curious	0.594	Desperate	−0.744	Boredom	−0.72
Excite	0.069	Shocking	0.592	Fear	−0.714	Safe	−0.685
Satisfaction	0.599	Joy	0.498	Hate	−0.675	Relaxation	−0.605

As shown in Table 7,the maximum value of the second eigenvalue calculated by the MDS model is 0.646, which corresponds to the word “Excite” and, according to the values in descending order, “Curious”, “Joy”, “Shocking”, “Negative” and other words with high “Activation” values. The smallest value is −0.723, which corresponds to the word “Calm”, followed by “Boredom” and “Safe” in descending order, “Boredom”, “Safe”, “Relaxation”, and other words with high “Activation” values, which are the same as the findings of PCA; this dimension is defined as “Curious”. When the sense of “Curious” is higher, the model predicts higher architectural emotions and vice versa; the model predicts that the architectural emotions become more related to “Relaxation” and more reflective of “Boredom”, “Safe”, and “Relaxation”.

4. Results

The experimental results and model construction study reveal the interrelationship between architectural emotion and architectural element. Through the analysis, the emotional words present circularity and continuity, and verify that the architectural emotion model can define two-dimensional emotions using architectural words. The three-layer circular model proposed by this study integrates emotional needs, architectural perception, and design element selection, providing a systematic framework and theoretical basis for emotional architectural design, which is of great significance as a practical guide.

4.1. Experimental Results

The experimental results show that both Principal Component Analysis (PCA) and Multidimensional Scaling Analysis (MDS) support that architecture-induced emotions have a similar cyclic structure to the Russel emotion model. The distribution of architectural affective words in the two-dimensional space shows a continuous cyclic pattern, and the similarity relationship between affective words is clear, showing that emotions are not independently dispersed, but are interrelated. The two-dimensional model of architectural emotion words further reveals that the emotion words are uniformly distributed in the four quadrants, with higher emotional pleasantness in quadrants one and four, and lower pleasantness in quadrants two and three. Meanwhile, the word curiosity was higher in quadrants one and two and lower in quadrants three and four. The results of the PCA and MDS analyses were similar, validating the reliability of the architectural affective loop model.

In addition, by combining the results of PCA, MDS, and UDS, we analyzed the outcomes of PCA, MDS, and UDS, as the score of “pleasure” is higher on the X-axis of both methods. The X-axis ranges from negative “despair” and “anxiety” to positive “pleasure” and “majesty”, which reflects the increase in architectural pleasantness; thus, the X-axis is defined as pleasantness. “Curiosity” scores were high on the Y-axis for both methods, ranging from “calm” and “disgust” at the low end of the Y-axis to “excite” and “curiosity”, reflecting the increase in the building’s charisma, thereby defining the Y-axis as “charm”. The higher the degree of pleasure and charisma, the stronger the emotional experience of architecture, which is often related to the perception of self-consciousness; conversely, the lower the degree of pleasure and charisma, the more basic the physiological reaction based on the emotion of architecture. Together, these results demonstrate that an architectural emotion model can define two-dimensional emotions using architectural words.

4.2. Architectural Emotion, Architectural Cognition, and Architectural Elements—Affective Modeling Map

In affective architectural design, a causal transformational interaction is formed between architectural emotion, cognition, and elements. First is the causal relationship: designers need to clarify the emotional needs of the target user (such as relaxation or reverence), and select architectural elements that match the emotional goals through emotion-oriented design. Second is the translation relationship: through the architectural cognitive layer, the abstract emotional needs (e.g., reverence) are translated into concrete architectural qualities (e.g., sacred) to form an operable design language. Architectural emotion is the subjective response of people when interacting with space, and architectural cognition is the rational perception of architectural qualities by the senses, where perception is the direct experience of architectural design elements through the sensory system (e.g., visual, auditory, tactile). Perception is an intuitive and immediate response that is the basis of emotional experience. Cognition is the rational interpretation and understanding of perceived information, such as the perception of tall space as related to “sacredness” or “ritual”. Cognition, through the brain’s information processing, elevates the perceptual experience to a meaningful cognitive evaluation, which, in turn, influences the emotional response, and together they shape the emotional experience.

The three-layer circular model (Figure 9, Table 8) proposed in this paper provides a systematic framework for emotional architectural design. The third layer is the architectural emotion layer, which contains 30 emotion words such as “excitement”, “comfort”, “shock”, etc., reflecting people’s emotional experience after interacting with the building. The second layer is the architectural cognitive layer, which consists of 60 words describing architectural qualities (e.g., sense of rhythm, sense of sacredness), translating emotional needs into cognitive expressions. The first layer is the architectural element layer, covering design elements such as spatial layout, color, and material, which trigger the user’s perception and emotion through morphological expression. The three layers are interlocked, building a logical path from emotional needs to design implementation.

Figure 9. The three-layer circular model.

Table 8. Architectural affective words, architectural perception, and architectural elements.

The Third Layer	The Second Layer	The First Layer
Architectural Emotional	Architectural Cognition	Architectural Elements (Color, Light and Lighting, Spatial Layout, Shape and Line, Material and Texture)
Excite	Sense of Drama	Strong contrasting colors, dark tones, and local bright colors; light and dark contrast, spotlight effects, and sharp shadows; large-scale changes and sharp switching between open and closed spaces; sharp angles, exaggerated shapes, and irregular designs; dark stone, smooth metal, and reflective glass.
	Sense of Stimulation	Bright colors and high saturation; dynamic light, flickering lighting, and light and shadow changes; fluidity, dynamic layout, and multi-level exploration; streamlined, curved, asymmetrical, and rotating elements; reflective metal, transparent glass, and smooth and modern materials.
Curious	Sense of Exploration	Bright vibrant colors; variable lighting, natural light penetration, and dynamic light-guided exploration; multi-level winding paths, hidden corners, and sudden open spaces; streamlined, curved lines, irregular geometry, and strong interactivity; transparent glass, reflective surfaces, and warm wood.
	Sense of Mystery	Dark tones and monochrome with local accents; dim light, shadows, strong contrast between light and dark, and ambient light that blurs details; hidden spaces, narrow corridors, fragmented layouts, and semi-hidden areas; sharp angles, fragmented forms, maze-like layouts, and closed or semi-closed spaces; opaque stone, concrete, dark metal, and hollow screens or frosted glass.
Shocking	Sense of Impact	Strong contrasting colors (black, red, silver); strong contrast between light and dark and dramatic changes in light and shadow; huge scale and dramatic changes in space; sharp, broken, and irregular forms; metal, reflective glass, and rough concrete.
	Sense of Amazement	Dark tones with bright colors (dark blue, purple, gold); high windows for natural light and soft focused light; grand atriums and towering open spaces; smooth curves, exaggerated geometry, and complex details; colored glass, marble, and exquisite metal decoration.
Joy	Sense of Delight	Bright and lively colors (yellow, orange, pink); bright natural light and soft diffuse light; open layout and highly interactive space; streamlined, curved, and interesting forms; wood, colored glass, and warm materials.
	Sense of Cheerfulness	Warm tones (light yellow, off-white, light green); plenty of natural light and warm artificial light; open space and sunny courtyard; simple lines, geometric shapes, and harmonious proportions; wood, ceramics, and soft textures.
Favorite	Sense of Attachment	Natural tones (brown, beige, dark green); soft light and warm diffuse light; warm small-scale comfortable private space; traditional form and classic design; wood, masonry and traditional materials.
	Sense of Affinity	Low-key tones (gray, light brown, soft white); combination of natural and artificial light and comfortable lighting; shared space and courtyard connecting space; simple functional form; local materials, stone, and concrete.
Pride	Sense of Honor	Highly saturated warm colors (gold, dark red); solemn spotlight and brilliant light; symmetrical open space; vertical solemn lines; marble, metal, and bronze decoration.
	Sense of Triumph	High-contrast warm and cold colors (gold, silver, dark red); strong focused light and reflected light; magnificent open rising space; sharp angles and rising lines; bronze and polished stone.
Hopeful	Sense of Expectation	Gradual cold and warm colors (light blue, light yellow, white); soft gradient light; progressive flow space; curves and open layout; glass and metal.
	Sense of Positivity	Bright warm colors (bright yellow, orange, green); plenty of natural light and vibrant light; dynamic open layout; streamlined rounded shape; wood and stained glass.
Respectful	Sense of Veneration	Low-saturation warm colors (dark brown, gray); soft indirect light and long shadows; tall and solemn space; vertical and symmetrical lines; dark stone and bronze metal.
	Sense of Solemnity	Cold tones (dark gray, dark blue); dim light and local spotlight; closed and calm space; vertical and square lines; rough stone and concrete.
Pleasure	Sense of Enjoyment	High-saturation warm colors (red, yellow, blue); bright and uniform light; open interactive space; smooth curves and dynamic shapes; stained glass and wood.
	Sense of Contentment	Soft warm colors (beige, light brown); warm and soft light; comfortable open layout; smooth curves and soft lines; soft package and warm wood.
Happy	Sense of Bliss	Soft warm colors (pink, lavender); diffuse light and natural light; open free space; rounded dreamy curves; lightweight textiles and wood.
	Sense of Elation	Highly saturated warm colors (bright yellow, orange); dynamic bright light; high open space; rising streamline; reflective metal and glass.
Majestic	Sense of Grandeur	Deep warm colors (gold, dark green); natural lighting and brilliant light; magnificent symmetrical space; complex lines and magnificent shapes; marble and bronze.
	Sense of Magnificence	Highly saturated gold and dark blue; alternating light and dark and focused light; gorgeous open layout; delicate decoration and complex lines; gold foil and carved stone.
Satisfaction	Sense of Fulfillment	Soft warm colors (off-white, light yellow); uniform soft light; balanced and harmonious space; symmetrical and smooth lines; wood and marble.
	Sense of Accomplishment	Highly saturated cold and warm colors (blue, silver); bright focused light; gradual open space; sharp rising lines; metal and polished glass.
Cozy	Sense of Warmth	Medium-saturated warm colors (beige, orange); warm and soft light; enclosed and intimate space; rounded and soft lines; wood and soft package materials.
	Sense of Intimacy	Soft warm tones (light brown, off-white); warm and soft light; small-scale private space; rounded and organic lines; soft package and warm wood.
Comfort	Sense of Ease	Medium- and low-saturation warm and cold tones (light blue, off-white); soft natural light; transparent and open layout; smooth and flowing lines; light wood, cotton, and linen.
	Sense of Serenity	Low-saturation cold tones (blue-gray, light green); quiet and soft light; quiet and open space; simple and straight lines; stone and frosted glass.
Relaxation	Sense of Tranquility	Soft cold tones (light blue, gray); diffuse soft light; quiet and stable space; smooth curves and simple shapes; frosted stone and wood.
	Sense of Repose	Warm tones (beige, light yellow); warm and soft light; comfortable and relaxing space; smooth and soft lines; soft materials and warm wood.
Safe	Sense of Security	Low-saturation warm colors (dark gray, brown); uniform soft light; enclosed protective space; heavy straight lines; stone, brick, and metal.
	Sense of Protection	Deep warm tones (dark green, dark brown); indirect soft light; closed and stable space; vertical lines and stable shapes; stone, bronze, and concrete.
Calm	Sense of Stillness	Low-saturation cool tones (gray, cool white); static uniform light; simple static space; vertical and horizontal lines; stone and concrete.
	Sense of Healing	Soft cool and warm tones (light green, light blue); diffuse soft light; healing open space; soft curves; natural wood and ceramics.
Apathetic	Sense of Isolation	Low-saturation cool tones (gray-blue, black); dim isolated light; closed and narrow space; sharp straight lines; metal and cool glass.
	Sense of Alienation	Cool tones (dark gray, blue); shadows and local light; isolated and divided space; sharp angles and broken lines; metal and concrete.
Sad	Sense of Melancholy	Low-saturation cool tones (gray-blue, purple); dim diffuse light; closed and oppressive space; drooping curves; rough stone and dark wood.
	Sense of Sorrow	Deep cool tones (dark gray, black); dim local light; closed and heavy space; vertical lines and collapsed shapes; rough concrete and dark metal.
Boredom	Sense of Dullness	Low-saturation achromatic tones (gray, brown); single diffuse light; simple closed space; straight lines and flat shapes; concrete and rough masonry.
	Sense of Tedium	Low-saturation dark yellow, gray-brown; uniform dull light; monotonous repetitive space; straight lines and rectangles; rough bricks and concrete.
Gloomy	Sense of Dreariness	Dark and cold tones (dark gray, blue); dim diffuse light; empty and oppressive space; vertical monotonous lines; stone and frosted glass.
	Sense of Despondency	Dark and cold tones (gray-blue, purple); low brightness light; closed and oppressive space; drooping broken lines; rough stone and concrete.
Burdensome	Sense of Oppression	Dark tones (black, dark gray); strong shadows; low and closed space; heavy vertical lines; concrete and dark metal.
	Sense of Heaviness	Warm dark tones (dark brown, gray); soft dark light; low and heavy space; thick straight lines; heavy stone and masonry.
Desperate	Sense of Despair	Cold and cold tones (black, dark blue); dim light and blurred shadows; closed space; twisted broken lines; rough concrete and metal.
	Sense of Helplessness	Cold tones (gray, light blue); dim diffuse light; isolated space; drooping curved lines; cold metal and glass.
Painful	Sense of Agony	High-contrast cold and warm colors (dark red, black); sharp spotlight; distorted and broken space; sharp lines; reflective metal and rough concrete.
	Sense of Torment	Dark red and gray-black contrast; local spotlight; oppressive and distorted space; broken and piercing lines; metal and rough stone.
Surprise	Sense of Wonder	Cold and warm gradient colors (gold, blue); focused light and reflected light; towering fantasy space; streamlined winding lines; colored glass and metal.
	Sense of Astonishment	Strong contrast colors (black, red, silver); strong contrast between light and dark; drastically changing space; sharp and broken lines; reflective metal and concrete.
Fear	Sense of Dread	Dark and cold tones (black, blue-gray); dim light and long shadows; closed and oppressive space; sharp angles and drooping lines; cold metal and concrete.
	Sense of Alarm	High-contrast warm colors (red, orange); warning spotlight; open and tense space; sharp angles and piercing lines; reflective metal and glass.
Anxiety	Sense of Unease	Low-saturation cold and warm colors (gray, dark green); blurred light; intricate space; curved irregular lines; frosted glass and metal.
	Sense of Tension	High-contrast cold and warm colors (dark red, gray); sharp spotlight; oppressive space; sharp lines and tight shapes; metal and concrete.
Cautious	Sense of Wariness	Low-saturation cold colors (gray, dark green); faint light and local spotlight; hidden and cautious space; fragmented sharp lines; metal and dark glass.
	Sense of Vigilance	High-contrast warm colors (orange, red); bright alert light; open warning space; sharp angles and dynamic lines; reflective metal and glass.
Anger	Sense of Fury	High-saturation warm colors (dark red, black); violent dynamic light; explosive space; distorted broken lines; metal and rough concrete.
	Sense of Resentment	Cold and warm contrast colors (dark red, gray); low light; oppressive space; distorted lines and closed shapes; metal and rough stone.
Hate	Sense of Loathing	Low-saturation cold and warm colors (gray green, purple); dark light and local reflection; distorted and split space; sharp angles and piercing lines; metal and glass.
	Sense of Disgust	Low-saturation warm colors (dark yellow, green); dark light; chaotic and distorted space; curved irregular lines; rough metal and plastic.

In practical application, the model realizes emotional goals through the three steps of “emotion recognition—cognitive translation—design implementation”. For example, in order to create the emotion of “respect”, the designer can enhance the sense of sacredness and ceremony through the combination of high-ceilinged space and soft natural light; coupled with natural materials and harmonious color tones, the user can experience the emotional response of reverence and solemnity. The three-layer circular model transforms abstract emotions into concrete design with clear cause-and-effect and translation logic, providing theoretical and practical support for emotion-driven architectural design and evaluation.

5. Discussion

In this paper, we collected 614 words and phrases related to architectural emotions through various methods. The words were categorized and refined for similarity by 15 participants, resulting in the identification of 30 architecture-related emotion words, 60 architecture-related cognitive phrases, and their corresponding architectural elements through high-frequency word analysis. Three researchers conducted a detailed analysis of the architectural emotion words by comparing similarity scores before and after the analysis.

During the analysis process, we employed Principal Component Analysis (PCA) to extract the three primary features from the emotion word similarity matrix. We utilized the first feature value to develop a model of the emotion words associated with architecture. Subsequently, we applied Multidimensional Scaling (MDS) to validate this model and integrated Unidimensional Scaling (UDS) to define the two axes—charisma and pleasantness—of the emotion model related to architecture.

Based on word grouping during lexical processing, a three-layer circular model encompassing architectural emotion, architectural perception, and architectural elements was established by integrating insights from architectural and psychological experts. This study developed two models that substantiate the hypothesis regarding architecture-related emotions and their interrelationships, as outlined in the introduction. The emotional experiences elicited by architecture are analyzed through a word framework, ultimately organized into a two-dimensional ring, with two axes emphasizing the primary characteristics of these emotional experiences. Furthermore, the process of grouping and refining the words revealed that the emotional words associated with architecture, the perception of architecture, and the elements of architecture are structured hierarchically. Most research on design emotions has been conducted through the lens of emotional semantics and physiological indicators, lacking a systematic framework for analyzing the processes underlying architectural emotions.

In the process of categorizing the 614 terms related to architectural emotions that we collected, we discovered that some terms pertain to emotions, some refer to architectural elements, and others represent cognitive words that bridge architectural elements and emotional words. The identification of cognitive words enables us to explore the intricate relationship between architectural elements and emotions. According to the theory of emotional evocation, when an individual interacts with architectural elements, their brain analyzes the mechanisms that generate emotions through a cognitive assessment of external physical attributes. This cognitive analysis induces emotions, thereby supporting the validity and scientific basis of the three-layer circular model. Specifically, the relationship between architectural elements and architectural emotions is not directly established through psychological analysis; rather, it must be interrelated at the cognitive level. Contributions of this study: first, the collection of information on architectural emotions through various means shows that our study covers almost all areas related to architectural emotions and forms a comprehensive word database related to architectural emotions. Second, in analyzing the words, grouping and reduction were used to find both the main categories of research on architectural emotions and to identify representative emotion categories through reduction. A model of architecture-related emotion words was scientifically constructed and validated using PCA, MDS, and UDS, and the relationship between architecture-related emotion words was clarified, providing a visual modeling basis for quantifying the emotional experience provoked by architecture. Third, the relationships among architectural elements, architectural cognition, and architectural elements are sorted out, and the three-layer circular model is established to provide a theoretical framework for architectural emotion research for later studies using physiological indicator measures and architectural generation with AI.

6. Conclusions

Architecture evokes human emotion and expresses it in the form of words, following the theory of Russell’s classical emotion model. The relationship between emotional words is downgraded to a two-dimensional model in a ring-like arrangement, and the pleasure and charm axes highlight the emotional characteristics of architecture. By analyzing the 614 words related to architectural emotion that have been collected, and establishing the architectural emotion model and the three-layer circular model, it is found that architectural emotion, architectural cognition, and architectural elements can be constructed in the form of a hierarchy. This provides theoretical support for the architectural emotion evaluation system based on physiological data and the language-based architectural generation of AI.

Since this study is a basic conceptual modeling study, only subjective ratings were used to obtain data, which, relatively, lacks objective physiological data for support. The emotion of architecture can be substantiated with more objective data by collecting physiological data combined with the affective computing in the current study. Future research can expand its application in the field of architectural design, evaluation, and generation by collecting data such as eye movements, posture, behavior, EEG, ECG, and the picoelectricity of a person in architecture, combined with affective computing, and further optimizing the model using deep learning algorithms.