整合 PIA 模型与三支决策理论的多粒度叙事云选择框架：面向建筑意义建构

Abstract

Architectural experience depends not only on perceptual fluency but also on whether users can successfully construct higher-order semantic meaning from complex built form. In the Pleasure-Interest Model of Aesthetic Liking (PIA model), the controlled-processing stage differentiates interest, confusion, and boredom. In conventional architectural cognition research, however, confusion is often treated as a negative indicator of communicative failure. This paper challenges that assumption and argues that confusion is especially important in architectural meaning-making because it identifies a population that still possesses interpretive motivation but has not yet achieved semantic resolution due to a mismatch between narrative granularity and cognitive accessibility. To address this problem, the paper proposes a conceptual and heuristic framework that integrates the PIA model, three-way decision theory, and granular computing. First, the three states in the controlled-processing stage of the PIA model are reinterpreted through a three-way decision structure, in which confusion is reconstructed as a strategic boundary region rather than as a purely negative endpoint. Second, a multi-granularity optimization mechanism and a utility-based selection function are introduced to identify narrative descriptions that balance novelty and comprehensibility for target perceivers. Third, a memorial-hall case and a computer-simulated illustrative statement-scoring exercise are used to illustrate the feasibility of the proposed procedure. Rather than claiming full empirical validation, the paper positions its contribution as a design-oriented framework for improving architectural communication, user-centered narrative intervention, and the cultural-social dimension of human-centered sustainability in the built environment.

1 Introduction

Recent scholarship in neuroaesthetics, architectural cognition, and human-building interaction has clarified that built environments are not merely visual objects but meaning-bearing settings that shape perception, affect, behavior, and interpretation through layered interactions among sensory, cognitive, and contextual processes (; ; ; ). At the same time, recent studies in architecture and the built environment increasingly emphasize visual preference, restorative response, public interpretation, virtual perception, and user-centered sustainable design as interconnected issues rather than isolated topics (; ; ; ; Wu et al., 2024; ; ). These developments suggest that architecture should be studied not only as form, but also as a medium through which meaning is organized, communicated, and experienced.

One useful framework for understanding this process is the Pleasure-Interest Model of Aesthetic Liking (PIA model) (). On the one hand, the PIA model captures the richness of aesthetic preference by distinguishing low-level automatic processing from high-level controlled processing. On the other hand, it explains why the same object may produce different aesthetic outcomes depending on whether the perceiver is willing and able to invest cognitive effort in resolving novelty, ambiguity, or complexity. In this sense, the PIA model reconciles the apparent conflict between fluency-based preference () and findings showing that complexity, novelty, and challenge may also support positive aesthetic evaluation when they can eventually be interpreted (; ; ; ; ).

The concept of fluency has long occupied a central place in aesthetic theory. Derived from experimental aesthetics and later developed in cognitive psychology, fluency has important implications for preference, familiarity, and judgments of meaning (Zajonc, 1980; Winkielman et al., 2006; ; Whittlesea, 1993). In the controlled-processing stage of the PIA model, three states are especially important: interest, confusion, and boredom. From the perspective of architectural meaning-making, confusion is particularly significant because it captures a condition in which perceivers are motivated to understand a design but remain unable to complete interpretation. In architecture, this condition is not trivial. Buildings are often appreciated for reasons that exceed pure beauty, including cultural symbolism, semantic associations, experiential narratives, and the social meaning embedded in space (; ).

This issue becomes more salient when architecture is approached as narrative. Narrative architecture may be understood as a design strategy through which buildings communicate stories, values, symbolic references, and semantic cues by means of form, spatial sequence, material expression, and atmosphere (; ). In this process, architecture is not only perceived; it is also read, inferred, and interpreted. However, different perceivers bring different cultural knowledge, professional backgrounds, and perceptual expectations. As a result, the same architectural intention may produce very different interpretive outcomes. If the selected narrative discourse is too fine-grained, highly abstract, or dependent on specialized background knowledge, users may remain in confusion rather than progressing toward meaningful engagement; if it is too coarse and overly explicit, the architectural narrative may become readable but lose semantic tension and experiential value.

To address these issues, this paper proposes a multi-granularity narrative cloud selection framework integrating the PIA model, three-way decision theory, and granular computing. The central claim is that confusion should not be treated as a purely negative result. Instead, it can be interpreted as a boundary region with strategic design value. Once confusion is modeled in this way, architectural narrative selection becomes a problem of choosing an appropriate granularity: a description that is sufficiently novel to remain meaningful while also sufficiently comprehensible to support successful interpretation. In other words, the architectural task is not to eliminate uncertainty at all costs, but to manage uncertainty so that perceivers remain within a productive interpretive range. This perspective provides a stronger basis for narrative architectural design and more directly supports communication between architects and perceivers. In public cultural buildings, such improved interpretive accessibility contributes to human-centered sustainability in a cultural-social sense by strengthening public legibility, emotional identification, and long-term civic engagement rather than by directly addressing environmental performance.

The next sections are arranged in the following: Section 2 introduces the relevant concepts of three-way decisions theory and granular computing and clarifies the conceptual bridge across psychology, computation, and design. Section 3 introduces the interpretation method for the aesthetic model from the perspective of three-way decision. Section 4 presents the multi-granularity narrative cloud selection framework. Section 5 offers a case-based demonstration and a computer-simulated illustrative evaluation of candidate narrative statements. Section 6 discusses comparative positioning and practical applicability. Ultimately, Section 7 concludes the paper and outlines future empirical validation.

2 Preliminaries

2.1 Three-way decisions theory

Based on decision theory, the three-way decisions (3WD) theory was proposed by Yao (2010, 2011) from the perspective of decision risk. A domain is divided into three disjoint regions (positive region, negative region and boundary region) by computing the minimum decision risk in 3WD model. As shown in Figure 1, the rules generated from the positive region, negative region and boundary region represent the action of acceptance, rejection and deferment, respectively. In the view of cognitive science, the 3WD theory simulates the human thinking mechanism in solving problems. Recently, the concept of 3WD has attracted significant attentions in recent years, such as decision making (), three-way decisions spaces (), cognitive concept learning () social networks (), classification (), clustering () and recommendation (Ye and Liu, 2021).

Figure 1

3WD model brings new insight into the problem of parameter setting according to the minimum expected overall decision risk.

Definition 1. (Three-way decisions theory) (Yao, 2010) Given an decision system S = (U, C∪D, V, f), R⊆C and let X is a fuzzy set on U. The action set A = {a_P, a_B, a_N} represents three kinds of actions which are acceptation, rejection and deferred decisions. λ_PP, λ_BP, λ_NP denote the losses incurred for taking actionsa_P, a_B, a_N, respectively. When an object belongs to target set X, and λ_PN, λ_BN, λ_NN denote the losses incurred for taking these actions when an object does not belong to target set X. Thus, the expected losses associated with taking different actions with object x can be expressed as follows,

Where, denotes the membership degree of the equivalence class [x] induced by equivalence relation U/R belong to object concept X, and denotes the membership degree of the equivalence class [x] induced by equivalence relation U/R belong to target set X^C. From the perspective of probability and statistics, can be understood as the probability that a randomly selected object x∈[x] belong to target set X, and also can be understood as the probability that a randomly selected object x∈[x] not belong to target set X.

According to the Bayesian decision rule, the minimum-risk decision rules can be obtained as follows

(P) If R(a_P|[x]) ≤ R(a_N|[x])and R(a_P|[x]) ≤ R(a_B|[x]), decide x∈POS(X);

(B) If R(a_B|[x]) ≤ R(a_N|[x])and R(a_B|[x]) ≤ R(a_P|[x]), decide x∈BND(X);

(N) If R(a_N|[x]) ≤ R(a_B|[x])and R(a_N|[x]) ≤ R(a_P|[x]), decide x∈NEG(X).

(P) and

(B) and

(N) and

The decision rules can be re-expressed as follows:

(P) If and , decide x∈POS(X);

(B) If and , decide x∈BND(X);

(N) If and , decide x∈NEG(X).

Combining with the above Equations 1–3, we can obtain the three parameters α, β, γ, respectively, according to the minimum-risk decision rules (P‘)-(N‘). The three decision thresholds are defined explicitly in Equations 4–6.

For rule(P′)

For rule(B′)

For rule(N′)

In other words, one can systematically determine the required threshold values from loss functions. In the present paper, the losses are assumed to be non-negative and order-constrained, namely 0 ≤ λ_PP ≤ λ_BP ≤ λ_NP and 0 ≤ λ_NN ≤ λ_BN ≤ λ_PN, without restricting them to the unit interval. Under these assumptions, it still follows that 0 <α ≤ 1, 0 ≤ β ≤ 1, and 0 <γ ≤ 1. Furthermore, if , we have 0 ≤ β <γ <α ≤ 1 [25, 18]. The rules (P), (B), and (N) can be expressed as follows,

(P) if ,decide ;

(B) if , decide ;

(N) if , decide .

2.2 Granular computing

Granular computing (GrC) (Yao et al., 2013; Yao, 2020; Wang and Li, 2020; ) is an umbrella term for a category of models inspired by human cognitive processes. There are three main GrC models: fuzzy sets (Zadeh, 1965), quotient space (Zhao and Zhang, 2006), and rough sets (). Although each GrC model has its own method for solving problems, they have a common characteristic: a problem space will be divided into many subspaces that form a hierarchical structure before solving the problem. Granularity selection (Yang et al., 2019) is a prosperous issue in GrC, which aims to search an appropriate granularity to improve the efficiency of solving problems. At present, GrC has received much attention (; ; ; ; Wang et al., 2017; ). ; proposed the principle of justiable granularity by combining the construction and optimization method of granularity. presented the granularity selection for the target concept and feature selection from the perspective of uncertainty based on these characteristic functions. introduced the idea of granularity of attributes is into triadic contexts on the basis of the relationship between triadic concept analysis and formal concept analysis. Wang et al. (2017) summarized the work of GrC from three aspects: granularity optimization, granularity switching and multi-granularity computing. In this paper, based on GrC theory, we will construct a multi-granularity narrative cloud selection method to improve the interpretation rate for the confusion state of population. This model is able to search for an optimal level of narrative cloud, which contributes to better understand the meaning of architecture and provides a more valuable basis for the subsequent narrative architectural design for designers.

2.3 The relationship among GrC, 3WD and cognitive science

A study of three-way decision and granular computing, in the context of results from cognitive science, is by itself an example of thinking in threes. Figure 2 shows the relationship among GrC, 3WD and cognitive science. For example, cognitive science provides a basis for three-way decision and the philosophy and methodology of three-way decision, as thinking in threes, can be used to study granular computing.

Figure 2

2.4 Conceptual bridge across psychology, computation, and design

To reduce conceptual jumps across disciplines, Table 1 summarizes how the main constructs used in this paper correspond to one another. The intention is not to claim that psychological states, decision-theoretic regions, and design operations are identical. Rather, the table clarifies how they can be linked in a design-support framework. In particular, confusion is treated as an interpretable uncertainty state: the user remains motivated, partial cues are available, and design intervention is still meaningful.

Table 1

Spatial element–narrative–perception correspondence in the memorial-hall example.

3 Interpretation of PIA model in architecture based on 3WD theory

In the PIA model, the perceptual results of automatic processing are classified into positive emotion and negative emotion according to the availability of fluency, while the perceptual results of controlled processing are classified into disfluency reduction (Yes DR), no disfluency reduction (No R), and no disfluency reduction (No D). Among these outcomes, Yes DR is associated with a positive aesthetic response, whereas the other two outcomes are usually associated with negative responses. Confusion is especially relevant here because the perceiver may still be motivated to invest cognitive effort, even though the stimulus has not yet produced sufficient progress in meaning construction or fluency increment.

In this paper, based on the idea of three-way decision (3WD) theory, we interpret the PIA model in architecture for people in confused state of controlled processing. To make the discussion easier to follow from a design perspective, the three states are described here through typical audience situations. The interest state refers to users who can connect spatial cues with intended meaning after a manageable amount of cognitive effort; for example, a design-trained visitor may understand that an intentionally broken circulation sequence symbolizes historical rupture and therefore experience interpretive reward. The confusion state refers to users who sense that the building is meaningful but cannot yet stabilize an interpretation; for example, a general visitor may notice the unusual path, tilted wall, and abrupt light transition, but may still be unsure whether these cues represent memory, disaster, resistance, or some other narrative. The negative or rejection state refers to users for whom the semantic distance remains too large; in that situation, the architectural message is not decoded and the visitor is more likely to abandon interpretation or reduce the building to a purely formal impression.

Example 1. As shown in Table 2, suppose that a memorial-type building contains 500 units of architectural information perceived by five representative audience groups. These groups may be understood as architecture students, art enthusiasts, local residents, the lead architect, and general tourists. The amounts of information perceived by the five groups are 50, 50, 100, 150 and 150, respectively. Let C denote the subset of 285, that is, the total amount of information that is successfully interpreted is 285.

Table 2

Conceptual bridge among PIA constructs, 3WD regions, observable indicators, and design interventions.

Where, C∩E denotes the number in the corresponding set (the amount of information successfully interpreted). C^C∩E denotes the amount of information that was not successfully interpreted. pr(C|E)and pr(C^C|E)denotes the conditional probability of interpreting the amount of successful information in the corresponding set, respectively. This design-oriented grouping is useful because it translates the abstract equivalence classes of 3WD into recognizable audience states. In practice, the architect is expected to occupy the most interpretable position, students and trained viewers often remain near the interest state, local residents and culturally engaged visitors frequently enter the confusion state, and general tourists are more likely to fall into rejection when the narrative remains too implicit.

The following loss values are used as illustrative parameters for demonstrating the decision logic of the framework. They are not intended as empirically calibrated constants. Rather, they express an asymmetric design situation in which falsely assuming successful interpretation, prematurely rejecting potentially interpretable users, and delaying decision for additional narrative support have different consequences.

Set the loss function as follows:

λ_PP = 0, λ_PN = 17, λ_BP = 9, λ_BN = 2, λ_NP = 15, λ_NN = 0,

Then,

We have α = 0.63, β = 0.25, γ = 0.53

The positive region, boundary region and negative region are calculated as follows:

Pawlak model:

Three-way model:

par Two-way model:

It can be seen that the boundary of two-way model is always the empty set.

From a design interpretation perspective, the three-way result is the most meaningful. E₁ and E₄ fall into the positive region because these groups already possess enough disciplinary knowledge or direct authorial intention to transform challenging spatial cues into interest. E₂ and E₃ fall into the boundary region because they do not fully fail; rather, they partially understand the architectural message and therefore represent the most valuable target population for narrative assistance. In practical terms, these users are the ones who would benefit most from a better calibrated narrative layer, for example, a more accessible exhibition text, a clearer interpretive sequence, or a more legible material-symbolic cue. E₅ falls into the negative region because the semantic distance is too large and the design, without additional support, is unlikely to produce stable meaning acquisition. From Table 3, for the equivalence class E₁: the Pawlak model makes delayed decisions with a misclassification cost as follows:

Table 3

Information table.

In contrast, both the two-way and three-way models made acceptance decisions. The associated misclassification cost are given by the following equation:

Therefore, the acceptance decision suggested by the three-way model has a low cost. Using similar calculations, Table 4 summarizes the decisions and associated costs for all equivalence classes.

Table 4

The comparison of the three decision models.

Obviously, the misclassification cost of the three-way model is always the same as the lower cost of the Pawlak and two-way models. For the Pawlak model the overall decision data are as follows.

The strategy based on the arithmetic available 3WD is more advantageous in dealing with the three states of the PIA model. Based on the above results, this paper emphasizes the importance of the confusion boundary state in architectural design. Due to the unique advantages of narrative architecture, the aim is to guide the perception of users who remain motivated and possess partial interpretive ability, but still require better calibrated semantic cues. The scientific design of narratives in architecture is therefore the central concern of this paper.

4 A multi-granularity narrative cloud selection integrating aesthetic model

An obvious characteristic of human intelligence is to deal with problems hierarchically. In other words, complex problems are usually analyzed at different levels of granularity rather than on a single descriptive layer. Granularity reflects the degree of detail used to represent information, and when the problem is complex, coarse and fine descriptions should be coordinated to satisfy practical constraints of interpretation and decision making. In architectural aesthetics, although many studies have explored the communication distance between designers and perceivers, it remains difficult to identify a narrative form that is simultaneously understandable, meaningful, and sufficiently novel for most users. Prior cognitive and neuroaesthetic studies have shown that informational reward is associated with pleasurable perception, especially when the stimulus offers both learning potential and interpretable structure (; Yue et al., 2007). Recent built-environment research likewise suggests that preference, attention, restorative response, and perceived environmental quality depend on how environmental information is organized and interpreted by users (; ; ; ; Wu et al., 2024; ; ). Therefore, based on GrC theory, we propose a multi-granularity narrative cloud selection method to identify descriptions that are both novel and understandable for people in the confusion state. In this paper, novelty (NO) and comprehensibility (EU) are treated as normalized evaluation variables in the range [0, 1]. In the illustrative case, these values are generated through a computer-simulated scoring procedure to demonstrate the workflow of the framework. In practical applications, however, the same variables can be obtained through reproducible user ratings, expert assessments, semantic differential scales, Likert-type questionnaires, structured interviews, or controlled perception experiments. For example, target users or domain experts may rate each candidate narrative statement in terms of perceived novelty and comprehensibility; the raw scores can then be averaged and normalized to [0, 1] before being entered into the selection model. The following heuristic utility function is established to search for the balance between novelty and comprehensibility. The utility of each granular layer is calculated using Equation 7.

Where, U_NCi denotes the heuristic utility of granular layer NC_i, and θ∈[0, 1] reflects the user's relative preference for novelty and comprehensibility. A larger value of U_NCi indicates a more desirable balance between cognitive accessibility and semantic tension. In light of the inverted-U relationship discussed in neuroaesthetics (; Yue et al., 2007), the narrative cloud with the highest comprehensibility alone is not necessarily the best narrative cloud. In practical architectural communication, the preferred description should retain enough novelty to sustain interpretive interest while remaining understandable to the target audience. The parameter θ can be adjusted according to design goals and user groups. A memorial or museum may assign a higher weight to novelty when reflective engagement is desired, whereas healthcare, transportation, or other high-stress public settings may assign a higher weight to comprehensibility in order to reduce cognitive burden. In applied projects, θ should ideally be set through expert discussion, pilot rating, or iterative calibration with target users.

If the perceivers' requirements for novelty and comprehensibility are denoted by NO_user and EU_user respectively, the narrative cloud optimization aims to search a narrative cloud NC_i that meets both conditions NO_NCi≥NO_user and EU_NCi≥EU_user, and then express the description using this narrative cloud. As shown in Figure 3, this is the multi-granularity optimization mechanism for the narrative cloud. Herein, narrative cloud NC_r meets the novelty requirement but fails to meet the comprehensibility requirement. NC_s meets the comprehensibility requirement but fails to meet the novelty requirement. NC_t meets the requirements for both novelty and comprehensibility. This optimization mechanism can be expressed as the following formula. The resulting constrained optimization problem is stated in Equation 8.

Figure 3

Figure 4 shows the architectural design process, which is defined in detail as follows: (1) the architect brainstorms a narrative cloud about the theme (the architect should expand the scope of the narrative and also involve different architects or target perceivers of the building); (2) the statements in the narrative cloud are classified (the classification is based on the theme to determine the classification criteria); (3) researching the classified statements for the target audience of the building (based on novelty and comprehensibility); (4) determining the best combination of statements in each category to form the narrative cloud, and (5) designing the building according to the narrative cloud. Throughout the process, the narrative cloud is categorized in order to avoid possible errors in the research under different categories and to be able to describe the building more comprehensively.

Figure 4

5 Case-based demonstration and illustrative evaluation workflow

To acquire the optimal granularity for statement selection under the above constraints, we present an explicit search procedure for identifying the most suitable candidate in each narrative dimension. The overall procedure should be understood as a two-stage operationalization of the theoretical framework. In the first stage, the PIA model and three-way decision theory identify the interpretive state of the target audience. The thresholds α, β, and γ define whether a user group is located in the positive region, boundary region, or negative region. The boundary region corresponds to confusion-state users who retain interpretive motivation but have not yet achieved stable meaning-making. In the second stage, Algorithm 1 is applied to this boundary-region audience to select narrative statements by balancing novelty and comprehensibility. Thus, α, β, and γ do not function as direct utility weights in Algorithm 1; rather, they determine the decision context and the target population for which narrative-cloud optimization is required. The procedure is framed as a heuristic utility-maximization problem rather than as a fully validated prediction model. It therefore serves as an illustrative design-support algorithm that can be implemented and empirically refined in future work.

Heuristic selection of narrative statements based on GrC.

In the following, the procedure is illustrated through a design-oriented memorial-hall example together with a computer-simulated illustrative scoring exercise.

Example 2. Suppose that a city plans to build a memorial hall with the theme of “memory and rebirth” on a post-war ruin site. The project adopts fragmented geometry, non-linear circulation, rough material transitions, and abrupt changes in light in order to metaphorize rupture, resilience, and temporal transformation. Here, the memorial hall should be understood as an illustrative design-oriented case rather than as a completed empirical validation. A set of narrative cloud S = (NC₁, NC₂, NC₃) is generated for describing the building, where NC₁ = (R₁₁, R₁₂, R₁₃), NC₂ = (R₂₁, R₂₂), and NC₃ = (R₃₁, R₃₂, R₃₃, R₃₄). To make the computational procedure explicit, novelty and comprehensibility values are treated as computer-simulated illustrative scores and then normalized to the interval [0, 1]. These simulated scores are used to demonstrate how the proposed framework operates under novelty and comprehensibility constraints; they should not be interpreted as empirical user-rating data. We assume that θ = 0.5, which means that the user's preference for novelty and comprehensibility is balanced. In addition, the user's requirements are set as NO_user = 0.2 and EU_user = 0.1. In this application, the task is to search for the optimal statement in each narrative dimension and then combine these statements into a new narrative cloud, denoted by NC_opt.

Table 1 links the narrative statements to spatial elements, perceptible cues, and likely audience interpretations, thereby clarifying that the framework is not limited to explanatory text alone. Rather, the narrative cloud may include concept narrative, spatial narrative, material narrative, path narrative, symbolic narrative, exhibition narrative, guide-text narrative, signage, wayfinding support, digital guidance, and post-design public interpretation. Table 5 shows the illustrative results obtained by the proposed heuristic procedure using computer-simulated illustrative scores. Furthermore, Figures 5, 6 are presented to describe the results more intuitively. Figure 5 shows the novelty, comprehensibility, and utility of each statement. Figure 6 shows utility under different novelty and comprehensibility combinations.

Table 5

Multi-granularity evaluation of narrative statements in the memorial-hall case.

Bold values indicate the utility scores of the statements selected for the optimal narrative cloud after application of the novelty and comprehensibility thresholds within each narrative dimension.

Figure 5

Figure 6

From these results, it can be seen that the algorithm does not simply prefer the most understandable or the most novel description. Rather, it filters out statements that are too ordinary to sustain interpretive interest and also excludes statements that are too obscure to support comprehension. The final optimal narrative cloud is NC_opt = {R₁₂, R₂₁, R₃₃}.

This result can be interpreted from a design perspective. R₁₁ is easy to process but lacks semantic tension, and therefore it fails to activate sufficient interpretive curiosity. R₂₂, although relatively understandable, does not satisfy the novelty threshold and is therefore excluded under the current settings. R₃₄ is highly novel but falls below the comprehensibility threshold; if imposed on a confusion-state audience, it is likely to push that audience toward frustration rather than toward meaningful engagement. By contrast, R₁₂, R₂₁, and R₃₃ all preserve a clear phenomenological anchor in the physical environment—walls, stairs, and light-space transition—while still leaving enough poetic openness for users to construct meaning. In this sense, the algorithm identifies not merely an optimal statement set, but an optimal design scaffold for people located in the confusion boundary region.

6 Comparative positioning and practical applicability

The proposed framework is not intended to replace architectural judgment, cultural interpretation, or participatory design. Its purpose is to provide a structured decision-support procedure for organizing candidate narrative statements and identifying a suitable balance between novelty and comprehensibility for users who remain in the confusion boundary region. Table 6 summarizes the qualitative difference between the proposed framework and several common approaches to architectural narrative selection.

Table 6

Qualitative comparison with common narrative selection approaches.

In practical terms, the framework can be adapted to several architectural contexts. In museums and exhibition buildings, it may support the selection of interpretive routes, display texts, symbolic cues, and spatial sequences. In educational buildings, it may help align spatial narratives with institutional memory, learning identity, and collective belonging. In healthcare facilities, the framework should be used more cautiously, with greater weight given to comprehensibility, emotional comfort, and reduced cognitive load. In public spaces, it may support wayfinding, civic interpretation, cultural memory, and inclusive communication for users with different levels of background knowledge. Across these contexts, the values of NO, EU, NO_user, EU_user, and θ should not be treated as universal constants. They should be selected according to the building type, target audience, narrative goal, and design stage, and should preferably be calibrated through expert discussion, small pilot evaluation, or user-centered semantic assessment.

7 Conclusion

In this paper, we reinterpret the three perceived outcomes of the controlled-processing phase of the PIA model from the perspective of three-way decision theory and show the design value of the confusion state. Rather than treating confusion as a purely negative endpoint, the proposed framework models it as a boundary region in which design intervention remains meaningful and potentially valuable. From the viewpoint of architectural communication, this is important because many perceivers possess sufficient motivation and partial interpretive ability, but still encounter barriers between the architect's intended meaning and the meaning that can actually be acquired from the built form.

On this basis, a multi-granularity narrative cloud selection method is constructed to support architectural interpretation among confusion-state users while preserving the novelty required for architectural meaning-making. By utilizing granular computing theory, the proposed method searches for an appropriate level of narrative description that balances comprehensibility and novelty and therefore provides a more rigorous basis for subsequent narrative architectural design. More broadly, the study contributes to current research on neuroaesthetics, architectural cognition, and human-centered building design by clarifying how meaning-oriented design decisions may be linked to interpretable uncertainty. The memorial-hall example and computer-simulated scoring exercise illustrate the feasibility of this framework, but they do not constitute full empirical validation. A small pilot evaluation with target users or domain experts would be a necessary next step for testing whether the selected narrative statements are perceived as intended. Future work should therefore further test and refine the framework through real-case applications, user studies, expert assessment, eye-tracking, immersive environments, physiological measures, and user-based semantic evaluation.

Statements

Author contributions

QL: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing.

Funding

The author(s) declared that financial support was received for this work and/or its publication. This research was supported by the Open Research Fund of the Watershed Vernacular Built Environment Shanxi Provincial Key Laboratory under the project ‘Research on the Spatial Morphological Characteristics of Shanxi Merchant Vernacular Dwellings from the Perspective of Evolutionary Aesthetics' (Grant No. WaBEL2024-05), and by the 2025 Shanxi Province Art Science Planning Project under the project ‘Research on the Spatial Gene Characteristics and Value Activation of Ming and Qing Shanxi Merchant Forts Based on Human Factors Technology' (Grant No. 25G126).

Acknowledgments

The author thanks the reviewers and colleagues who provided comments on the theoretical development and manuscript preparation.

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that Generative AI was used in the creation of this manuscript. Generative AI was used in a very limited manner for minor language and formatting assistance only. It was not used for the development of the core ideas, methods, results, or conclusions. The author takes full responsibility for the entire content of the manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

References

• 1

  AbdarM.SamamiM.MahmoodabadS. D.DoanT.MazoureB.HashemifesharakiR.et al. (2021). Uncertainty quantification in skin cancer classification using three-way decision-based bayesian deep learning. Comput. Biol. Med. 135:104418. doi: 10.1016/j.compbiomed.2021.104418

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 2

  AchtenH. (2019). Interaction narratives for responsive architecture. Buildings9:66. doi: 10.3390/buildings9030066

  • CrossRef
  • Google Scholar
  • View reference in article

• 3

  AlaviH. S.ChurchillE. F.WibergM.LalanneD.DalsgaardP.Fatah gen SchieckA.et al. (2019). Introduction to human-building interaction (HBI): interfacing HCI with architecture and urban design. ACM Trans. Comput. Hum. Interact. 26, 6, 1-10. doi: 10.1145/3309714

  • CrossRef
  • Google Scholar
  • View reference in article

• 4

  BiedermanI.VesselE. A. (2006). Perceptual pleasure and the brain: a novel theory explains why the brain craves information and seeks it through the senses. Am. Sci. 94, 247-253. doi: 10.1511/2006.59.247

  • CrossRef
  • Google Scholar
  • View reference in article

• 5

  BunyarittikitS.WongvorachanT.PetlaiT.SomngamP.RatanapongN.ThampanichwatC.et al. (2026). Exterior biophilic design attributes supporting urban mental well-being: a case study of Singapore's architecture. Sustainability18:1469. doi: 10.3390/su18031469

  • CrossRef
  • Google Scholar
  • View reference in article

• 6

  ChatterjeeA.VartanianO. (2014). Neuroaesthetics. Trends Cogn. Sci. 18, 370-375. doi: 10.1016/j.tics.2014.03.003

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 7

  CoburnA.VartanianO.ChatterjeeA. (2017). Buildings, beauty, and the brain: a neuroscience of architectural experience. J. Cogn. Neurosci. 29, 1521-1531. doi: 10.1162/jocn_a_01146

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 8

  DengM.JinW.GuoH.ChenX.WangY.XuL.et al. (2026). The physiological and psychological effects of the built environment: research progress and implications. Buildings16:1144. doi: 10.3390/buildings16061144

  • CrossRef
  • Google Scholar
  • View reference in article

• 9

  FuS.WangG. Y.XiaS. Y.LiuL. (2020). Deep multi-granularity graph embedding for user identity linkage across social networks. Knowl. Based Syst. 193:105301. doi: 10.1016/j.knosys.2019.105301

  • CrossRef
  • Google Scholar
  • View reference in article

• 10

  GrafL. K. M.LandwehrJ. R. (2015). A dual-process perspective on fluency-based aesthetics: the pleasure-interest model of aesthetic liking. Pers. Soc. Psychol. Rev. 19, 395-410. doi: 10.1177/1088868315574978

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 11

  GüçlütürkY.JacobsR. H.LierR. v. (2016). Liking versus complexity: decomposing the inverted U-curve. Front. Hum. Neurosci. 10:112. doi: 10.3389/fnhum.2016.00112

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 12

  HassanA.BarberS. J. (2021). The effects of repetition frequency on the illusory truth effect. Cogn. Res. Princ. Implic. 6, 1-12. doi: 10.1186/s41235-021-00301-5

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 13

  HuB. Q. (2014). Three-way decisions space and three-way decisions. Inf. Sci. 281, 21-52. doi: 10.1016/j.ins.2014.05.015

  • CrossRef
  • Google Scholar
  • View reference in article

• 14

  JiX.LiK. (2024). Bridging built environment attributes and perceived city images: exploring dual influences on resident satisfaction in revitalizing post-industrial neighborhoods. Sustainability16:7272. doi: 10.3390/su16177272

  • CrossRef
  • Google Scholar
  • View reference in article

• 15

  JiaX. Y.DengZ.MinF.LiuD. (2019). Three-way decisions based feature fusion for chinese irony detection. Int. J. Approx. Reason. 113, 324-335. doi: 10.1016/j.ijar.2019.07.010

  • CrossRef
  • Google Scholar
  • View reference in article

• 16

  JoS.LeeK. (2007). Architecture as narrative: on bernard franken's ruminations on characterization, integration, and imagination. J. Asian Archit. Build. Eng. 6, 213-220. doi: 10.3130/jaabe.6.213

  • CrossRef
  • Google Scholar
  • View reference in article

• 17

  KhanG. A.HuJ.LiT. R.DialloB.ZhaoY. M. (2022). Multi-view low rank sparse representation method for three-way clustering. Int. J. Mach. Learn. Cybern. 13, 233-253. doi: 10.1007/s13042-021-01394-6

  • CrossRef
  • Google Scholar
  • View reference in article

• 18

  LiJ. H.HuangC. C.QiJ. J.QianY. H.LiuW. Q. (2017). Three-way cognitive concept learning via multi-granularity. Inf. Sci. 378, 244-263. doi: 10.1016/j.ins.2016.04.051

  • CrossRef
  • Google Scholar
  • View reference in article

• 19

  LiS.YangJ.WangG. Y.ZhangQ. H.HuJ. (2022). Granularity selection for hierarchical classification based on uncertainty measure. IEEE Trans. Fuzzy Syst. 30, 4841-4855. doi: 10.1109/TFUZZ.2022.3161747

  • CrossRef
  • Google Scholar
  • View reference in article

• 20

  LiuJ. B.LiH. X.HuangB.LiuY.LiuD. (2021). Convex combination-based consensus analysis for intuitionistic fuzzy three-way group decision. Inf. Sci. 574, 542-566. doi: 10.1016/j.ins.2021.06.018

  • CrossRef
  • Google Scholar
  • View reference in article

• 21

  LiuY.GuoX. (2024). Online public feedback on mid- to high-rise biophilic buildings: a study of the Asia-pacific region over the past decade. Buildings14:2394. doi: 10.3390/buildings14082394

  • CrossRef
  • Google Scholar
  • View reference in article

• 22

  LoiaV.OrciuoliF.PedryczW. (2018). Towards a granular computing approach based on formal concept analysis for discovering periodicities in data. Knowl. Based Syst. 146, 1-11. doi: 10.1016/j.knosys.2018.01.032

  • CrossRef
  • Google Scholar
  • View reference in article

• 23

  MinissaleG. (2013). The Psychology of Contemporary Art. Cambridge: Cambridge University Press. doi: 10.1017/CBO9781139094313

  • CrossRef
  • Google Scholar
  • View reference in article

• 24

  PapadimitriouF. (2020). “Spatial complexity, visual complexity and aesthetics,” in Spatial Complexity: Theory, Mathematical Methods and Applications (Cham: Springer), 243-261. doi: 10.1007/978-3-030-59671-2_16

  • CrossRef
  • Google Scholar
  • View reference in article

• 25

  PawlakZ. (1984). Rough classification. Int. J. Man Mach. Stud. 20, 469-483. doi: 10.1016/S0020-7373(84)80022-X

  • CrossRef
  • Google Scholar
  • View reference in article

• 26

  PearceM. T.ZaidelD. W.VartanianO.SkovM.LederH.ChatterjeeA.et al. (2016). Neuroaesthetics: the cognitive neuroscience of aesthetic experience. Perspect. Psychol. Sci. 11, 265-279. doi: 10.1177/1745691615621274

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 27

  PedryczW.Al-HmouzR.MorfeqA.BalamashA. (2013). The design of free structure granular mappings: the use of the principle of justifiable granularity. IEEE Trans. Cybern. 43, 2105-2113. doi: 10.1109/TCYB.2013.2240384

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 28

  ReberR.SchwarzN.WinkielmanP. (2004). Processing fluency and aesthetic pleasure: is beauty in the perceiver's processing experience?Pers. Soc. Psychol. Rev. 8, 364-382. doi: 10.1207/s15327957pspr0804_3

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 29

  SedghikhanshirA.MontelliR. (2026). Biophilic design interventions and properties: a scoping review and decision-support framework for restorative and human-centered buildings. Buildings16:515. doi: 10.3390/buildings16030515

  • CrossRef
  • Google Scholar
  • View reference in article

• 30

  SonS.KimH. (2024). User preference maps: quantifying the built environment. Buildings14:3463. doi: 10.3390/buildings14113463

  • CrossRef
  • Google Scholar
  • View reference in article

• 31

  SongJ.KwakY.KimC.-Y. (2021). Familiarity and novelty in aesthetic preference: the effects of the properties of the artwork and the beholder. Front. Psychol. 12:694927. doi: 10.3389/fpsyg.2021.694927

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 32

  ThampanichwatC.WongvorachanT.BunyarittikitS.ChunhajindaP.PhaibulputhipongP. (2024). The architectural design strategies that promote attention to foster mindfulness: a systematic review, content analysis and meta-analysis. Buildings14:2508. doi: 10.3390/buildings14082508

  • CrossRef
  • Google Scholar
  • View reference in article

• 33

  Van de CruysS.DamianoC.BoddezY.KrólM.GoetschalckxL.WagemansJ. (2021). Visual affects: linking curiosity, Aha-Erlebnis, and memory through information gain. Cognition212:104698. doi: 10.1016/j.cognition.2021.104698

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 34

  WanQ.LiJ. H.WeiL. (2021). Optimal granule combination selection based on multi-granularity triadic concept analysis. Cogn. Comput. 14, 1844-1858. doi: 10.1007/s12559-021-09934-6

  • CrossRef
  • Google Scholar
  • View reference in article

• 35

  WangG. Y.YangJ.XuJ. (2017). Granular computing: from granularity optimization to multi-granularity joint problem solving. Granul. Comput. 2, 105-120. doi: 10.1007/s41066-016-0032-3

  • CrossRef
  • Google Scholar

• 36

  WangX. Z.LiJ. H. (2020). New advances in three-way decision, granular computing and concept lattice. Int. J. Mach. Learn. Cybern. 11:945-6. doi: 10.1007/s13042-020-01117-3

  • CrossRef
  • Google Scholar

• 37

  WhittleseaB. W. (1993). Illusions of familiarity. J. Exp. Psychol. Learn. Mem. Cogn. 19:1235. doi: 10.1037/0278-7393.19.6.1235

  • CrossRef
  • Google Scholar

• 38

  WinkielmanP.HalberstadtJ.FazendeiroT.CattyS. (2006). Prototypes are attractive because they are easy on the mind. Psychol. Sci. 17, 799-806. doi: 10.1111/j.1467-9280.2006.01785.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 39

  WuY.WuY.PanY. (2024). Sustainability optimization method of built environment with integrated physical environment and virtual perception simulation: a case study of campus open space. Sustainability16:8936. doi: 10.3390/su16208936

  • CrossRef
  • Google Scholar

• 40

  YangJ.WangG. Y.ZhangQ. H.ChenY. H.XuT. H. (2019). Optimal granularity selection based on cost-sensitive sequential three-way decisions with rough fuzzy sets. Knowl. Based Syst. 163, 131-144. doi: 10.1016/j.knosys.2018.08.019

  • CrossRef
  • Google Scholar

• 41

  YaoJ. T.VasilakosA. V.PedryczW. (2013). Granular computing: perspectives and challenges. IEEE Trans. Cybern. 43, 1977-1989. doi: 10.1109/TSMCC.2012.2236648

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 42

  YaoY.-Y. (2010). Three-way decisions with probabilistic rough sets. Inf. Sci. 180, 341-353. doi: 10.1016/j.ins.2009.09.021

  • CrossRef
  • Google Scholar

• 43

  YaoY.-Y. (2011). The superiority of three-way decisions in probabilistic rough set models. Inf. Sci. 181, 1080-1096. doi: 10.1016/j.ins.2010.11.019

  • CrossRef
  • Google Scholar

• 44

  YaoY.-Y. (2020). Three-way granular computing, rough sets, and formal concept analysis. Int. J. Approx. Reason. 116, 106-125. doi: 10.1016/j.ijar.2019.11.002

  • CrossRef
  • Google Scholar

• 45

  YeX. Q.LiuD. (2021). An interpretable sequential three-way recommendation based on collaborative topic regression. Expert Syst. Appl.168:114454. doi: 10.1016/j.eswa.2020.114454

  • CrossRef
  • Google Scholar

• 46

  YueX.VesselE. A.BiedermanI. (2007). The neural basis of scene preferences. Neuroreport 18, 525-529. doi: 10.1097/WNR.0b013e328091c1f9

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 47

  ZadehL. A. (1965). Fuzzy sets. Inf. Control 8, 338-353. doi: 10.1016/S0019-9958(65)90241-X

  • CrossRef
  • Google Scholar

• 48

  ZajoncR. B. (1980). Feeling and thinking: preferences need no inferences. Am. Psychol. 35:151. doi: 10.1037/0003-066X.35.2.151

  • CrossRef
  • Google Scholar

• 49

  ZhaoL.-Q.ZhangL. (2006). “Advances in the quotient space theory and its applications,” in Rough Sets and Knowledge Technology, eds. G. Y. Wang, J. F. Peters, A. Skowron, and Y. Yao (Cham: Springer), 445-452. doi: 10.1007/11795131_52

  • CrossRef
  • Google Scholar

Keywords

architectural meaning-making, granular computing, human-centered sustainability, narrative cloud, neuroaesthetics, three-way decision

Citation

Li Q (2026) A multi-granularity narrative cloud selection framework for architectural meaning-making integrating the PIA model and three-way decision theory. Front. Psychol. 17:1869490. doi: 10.3389/fpsyg.2026.1869490

Received

30 April 2026

Revised

15 July 2026

Accepted

21 September 2026

Published

06 October 2026

Volume

17 - 2026

Edited by

Yile Chen, Macau University of Science and Technology, Macao SAR, China

Updates

Copyright

© 2026 Li.

This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Qi Li, liam@sxu.edu.cn

Disclaimer

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.