Multi-Attribute Decision-Making Model for Security Perception in Smart Apartments from a User Experience Perspective

Abstract

With an aging population and the widespread adoption of smart technologies, elderly residents’ perceived safety in smart apartments has become a critical determinant of their quality of life and their acceptance of technology. However, much of the current research remains confined to either technical or psychological dimensions, with insufficient attention to the systematic interactions among multiple factors as experienced by elderly populations. This study aims to systematically evaluate and optimize the living environments of older adults, with the goal of enhancing their overall quality of life and subjective well-being. This study employs the DANP–mV model to empirically analyze the safety perception of older adults in smart apartments, integrating case-based investigation and evaluation to propose targeted optimization strategies and improvement pathways. Unlike traditional approaches that treat criteria as independent, this hybrid model reveals the interdependencies among factors and establishes a more realistic prioritization of improvement actions. The study found that, compared with merely reinforcing physical security measures, factors such as enhanced remote security support, a stronger sense of control and coping confidence, and higher satisfaction with the protective system exert a more fundamental influence on the overall safety perception. These results demonstrate that adopting a systems-thinking approach shifts the focus of decision-making from superficial safety risks to underlying causal drivers, thereby mitigating resource allocation imbalances and enhancing the effectiveness and sustainability of safety improvement measures.

1. Introduction

The pace of population aging has been accelerating globally since 2000, with the proportion of individuals aged 60 years and older rising in nearly all countries, and projections indicating that this group will surpass 2 billion worldwide by 2050 [1]. This structural demographic transition not only places increasing pressure on social security and public healthcare systems, but also generates growing demand for age-friendly housing models and smart living environments [2]. Driven by continuous advances in information and communication technologies (ICT) and smart grids, smart home technology (SHT) has gradually evolved into a key technological pathway for improving residential living conditions and enhancing quality of life [3]. However, the highly distributed architecture and massive device connectivity of IoT systems pose significant challenges to data security and system stability [4]. During system operation, the data management process encompasses data collection, processing, storage, and application, forming a highly sensitive information value chain. As each link is frequently exposed to open network environments, the resulting broad attack surface can easily become a target for malicious intrusion [5]. At the same time, with the rapid popularization of smart home systems within residential environments, their role in enhancing personal safety has attracted increasing scholarly and practical attention. Although smart door locks, wireless sensors, and remote monitoring systems have been widely adopted to improve physical security, studies have shown that key vulnerabilities persist within these systems. However, studies have revealed that these key security devices still exhibit significant vulnerabilities to wireless communication attacks, thereby exposing the underlying structural weaknesses of smart security systems in contemporary smart environments [6]. Notably, technological uncertainty tends to translate into heightened risk perception among older adults [7]. Ghorayeb et al. (2021) observed that any delay or abnormality in system feedback is often interpreted as a loss of control and an unpredictable environment, which in turn induces anxiety, heightened vigilance, and even technology resistance, thereby undermining users’ perceived security in smart living environments [8]. Therefore, developing a systematic understanding of the trust-building mechanisms among older adults within smart-apartment intelligent systems—from a user experience-oriented cognitive perspective—has become essential to uncovering the underlying mechanisms shaping perceived security in such environments.

As technology becomes increasingly embedded in everyday living environments, subjective security perception has emerged as a critical dimension of user experience in smart apartment contexts. Recent studies have identified it as one of the key determinants of successful smart living implementations [9,10]. Subjective security perception refers to an individual’s cognitive and emotional appraisal of the safety and reliability associated with a given technology or service [11]. For instance, in residential settings, subjective security perception manifests when users believe that their smart apartment provides effective intrusion prevention, emergency response, and privacy protection mechanisms—allowing them to perceive the environment as controllable, safe, and trustworthy, even when alone or at night [12]. However, because most users find it difficult to assess the system’s actual level of security, they tend to rely on interface cues and interaction experiences to construct their security perceptions [13]. Thus, for most users, security functions less as an objective condition and more as a subjective construct shaped by perception. Extensions of perceived risk theory and the technology acceptance model suggest that, under conditions of uncertainty and information asymmetry, users tend to recalibrate their perceived risk through subjective cognitive processes [14,15]. In this process, Studies have also shown that high levels of trust help users mitigate perceived uncertainty and potential losses, thereby increasing their willingness to participate in subsequent decision-making [14]. Therefore, subjective security perception plays a central role in users’ risk cognition and evaluation processes, while also serving as a key psychological foundation for trust development, service adoption, and continued use.

Existing research on security perception in smart apartments has predominantly examined isolated dimensions, thereby hindering a systematic understanding of how multiple interacting factors collectively shape users’ security perceptions [12,16]. Specifically, many studies have concentrated on individual aspects—such as improving technical security performance, examining user acceptance of smart home technologies, or assessing privacy risks—yet lack an integrative perspective that bridges technical and psychological dimensions [12,17,18]. For example, Yang et al. (2025) observed that the smart home field has long been dominated by an engineering-oriented paradigm, while research addressing the specific needs of older adults remains notably scarce [16]. While a few studies have examined older users’ perspectives, their focus has remained primarily on functional aspects such as physical health and safety monitoring, with limited exploration of older adults’ psychological needs and subjective experiences [8,19]. As a result, the actual concerns and behavioral patterns of older adults in smart living environments remain insufficiently understood, thereby hindering the development of targeted improvement strategies. More importantly, security perception is not determined solely by technological factors, but is better understood as the outcome of users’ cognitive construction of the system environment [4,20,21]. For example, Popoola et al. (2024) found that greater operational reliability can strengthen users’ sense of control and, in turn, enhance psychological safety, yet this positive effect can be undermined by privacy-related risk perceptions [4]. Therefore, when the synergistic or antagonistic effects among these factors are analyzed solely through static linear models, it becomes difficult to capture their actual interaction mechanisms within the cognitive domain [10].

Finally, existing research on users’ security perception in smart apartments has yet to systematically model or comprehensively evaluate the influencing factors shaping this perception. Smart apartment planning and operation are often challenged by practical constraints, including rapid technological advancements, diverse user needs, and limited financial resources. Without a theoretical framework and evaluative model grounded in systems thinking, design and decision-making processes can easily fall prey to local optimization, thereby overlooking the interdependencies among multiple factors and the overall perception pathways—ultimately resulting in imbalanced resource allocation, degraded user experience, and even a crisis of trust and subsequent technology rejection. More critically, such fragmented improvement strategies may fail to address the fundamental issue underlying users’ diminished sense of security and can perpetuate a reactive “fix-as-you-go” logic—thereby reinforcing a vicious cycle of perception breakdown, weakened trust, and reduced usage [22]. To overcome the aforementioned limitations, this study developed a multi-attribute decision-making (MADM) model by integrating the DANP–mV analytical framework. The model aims to identify the core external factors shaping users’ perceived security, elucidate the hierarchical structure and interaction mechanisms among these factors, and quantitatively assess their relative weights and overall effects in order to uncover the underlying psychological mechanisms and decision-making preferences. Accordingly, this study is guided by the following research questions:

• What are the key external factors that determine older adults’ perceived security within smart apartment contexts?
• Within a multi-attribute decision-making (MADM) framework, how can the perceived weight structure and interrelationships among security-related factors, as experienced by older adults, be identified?

Accordingly, the research design of this study proceeds as follows. First, a systematic literature review was conducted to identify and synthesize the key factors influencing users’ perceived security. Building on this foundation, a group of experts was invited to conduct pairwise comparisons to evaluate the interaction relationships among factors, thereby constructing the initial direct influence matrix. Subsequently, the matrix was normalized using the DEMATEL approach, and the comprehensive influence matrix was derived to determine the influencing degree (D) and influenced degree (R) of each factor [23]. In the second stage, a multi-attribute network model was constructed using the Analytical Network Process (ANP) to capture the dependencies and interrelationships among the factors. A key strength of the ANP lies in its ability to capture the interdependencies and feedback relationships among decision factors, making it particularly suitable for modeling complex structures in which multi-dimensional factors are interwoven within the security perceptions of smart apartment users [24]. In the final stage, the Modified VIKOR (mV) method was introduced as a comprehensive evaluation mechanism. By constructing positive and negative ideal solutions, this method identifies the optimal compromise solution, thereby enhancing the practical applicability and decision-making relevance of the evaluation results [25].

2. Literature Review

2.1. Perceived Security in Smart Apartments: A User Experience Perspective

From a user experience (UX) perspective, perceived security in smart apartments represents a multidimensional perceptual construct that integrates both technical features and situational experiences, rather than a single technical attribute [26]. In this context, research on aging has extensively examined the influence of housing environments on older adults’ health and well-being, with the home environment increasingly recognized as a key determinant of healthy aging [27]. For older adults, the home is not only a familiar and secure living environment, but also a cornerstone of their independence and autonomy [28]. Beyond its perceptual dimension, safety also reflects older adults’ recognition of security within their living environments—including buildings, communities, and homes—as shaped by their lived experiences [23]. By contrast, security perception differs fundamentally from conventional notions of technology acceptance or satisfaction. Whereas the latter centers on the convenience and enjoyment derived from functional utility and ease of use, security perception emphasizes the dynamic balance between controllability and trust under conditions of uncertainty [12]. Therefore, security perception may be conceptualized as a subjective experiential construct that evolves and is continuously refined through the process of technology use [29].

In the context of aging and later-life living, beyond concerns about physical health and domestic caregiving, it is equally important to attend to older adults’ physical and emotional comfort [30]. This sense of comfort encompasses both basic physical safety—such as the capacity of built infrastructure to withstand external threats—and psychological security arising from privacy protection and interactions with smart systems [8,31]. Building on this foundation, it is essential to incorporate the dimension of personal safety, which not only concerns individuals’ basic assurance of protection against external intrusion or accidental harm, but also underscores the capacity for timely detection and effective response in emergencies such as falls, sudden illness, or intrusion incidents [32,33].

2.2. Identification of Security Perception Dimensions in the Smart Home Domain

2.2.1. Physical Security

Physical security goes beyond the mere installation of protective devices and systems; it reflects residents’ holistic assessment of environmental controllability and predictability [34]. In the multi-layered device–app–cloud architecture, the availability and responsiveness of cameras, door locks, and alarm systems together shape residents’ intuitive judgments of safety and control [20]. At the same time, home security devices remain vulnerable to multi-vector attacks conducted via communication protocols such as 433 MHz radio, Bluetooth, and RFID, while firmware and application-layer flaws further exacerbate these systemic vulnerabilities [6]. Previous research has demonstrated structural vulnerabilities in the 433 MHz protocol used for residential security applications. With inexpensive software-defined radio devices such as the HackRF One, attackers can conduct capture-and-replay attacks on EV1527 and PT2262 signals, thereby simulating legitimate commands and rapidly bypassing fixed-code protections in garage-door systems and smart locks [35]. Empirical studies have also demonstrated that smart sensors and cameras employing Bluetooth Low Energy (BLE) protocols exhibit exploitable flaws in pairing and key negotiation processes—such as BLESA and KNOB attacks—while RFID access-control systems that continue to rely on unencrypted communication remain susceptible to eavesdropping and cloning [6]. These potential flaws are particularly consequential for older adults, who—due to limited technical literacy—often struggle to gauge the severity of system vulnerabilities. As a result, they tend to equate such technical issues with unpredictable risks, leading to a rapid decline in perceived security and, in some cases, heightened anxiety or outright technology rejection [36,37].

Conversely, visibility and physical barriers constitute a key perceptual foundation underpinning older adults’ sense of security. Himschoot et al. (2024) found in an environmental psychology study that enhancing nighttime illumination and adopting warm-white lighting significantly enhances users’ perceived sense of security [34]. Amid increasingly digitalized residential environments, tangible physical boundaries—such as doors, locks, windows, and gates—remain a primary source of perceived safety among older adults [38]. Perceived robustness of physical barriers helps mitigate residents’ concerns regarding the perceived fragility of digital systems [39].

2.2.2. Sense of Information Security

In the context of smart homes, information security extends beyond technical encryption mechanisms—it reflects residents’ subjective judgments about the visibility and controllability of personal data across the full data lifecycle, from collection and transmission to storage and sharing [40]. Compared with younger users, older adults are more likely to experience a loss of perceived control when navigating lengthy disclosure statements, fragmented permission interfaces, and specialized technical terminology [41]. Attacks targeting the radio frequency, Bluetooth, and protocol layers of home security devices (e.g., door locks, cameras, and gateways) have been repeatedly demonstrated in both laboratory and field settings. Although real-world incidents are rare, the mere awareness that such systems can be compromised fosters a persistent baseline of insecurity among older adults, discouraging continued use and limiting functional adoption [6].

Existing research often conflates technical defense efficacy with perceived security, while overlooking older adults’ core needs—particularly regarding understandability, operational load, and disengagement cost [42,43,44]. In this context, information security extends beyond preventing data breaches to encompass individuals’ agency and control over how personal data are circulated and used within their living environments [45].

2.2.3. Psychological Safety

In the context of smart apartments, older adults’ psychological security is characterized by stable expectations regarding the explainability of automation, the controllability and reversibility of interactions, and the availability of social support [46]. Linkage services with high explainability and low false-positive rates can effectively alleviate technology-induced anxiety and foster long-term trust [47]. If the system integrates an intuitive one-click pause/withdrawal function and establishes a well-defined protocol for anomaly management—linking family members and property management—older adults are enabled to reframe previously unpredictable risks as controllable and manageable events [8,48]. Moreover, a privacy pre-consented closed-loop system for fall detection and timely rescue—coupled with remote care and companionship features—can help mitigate persistent nighttime anxiety associated with solitary living and social isolation, thereby improving older adults’ sleep quality and sense of security, while sustaining their autonomy and dignity [8,49,50].

2.2.4. Sense of Personal Safety

In the context of smart apartments, proactive intervention strategies should be implemented to evaluate and enhance the existing living environments of older adults. Such strategies are significant not merely for strengthening current risk prevention and safety protection capacities, but, more importantly, for fostering a sustainable and future-oriented living experience [51]. A well-designed outdoor environment within residential areas not only provides older adults with conditions that support safety and independence in daily activities, but also fosters continued social engagement and physical activity [52]. For residential buildings, it is essential to assess structural durability, as well as the strength and integrity of doors, windows, and related components, while also accounting for resilience against natural disasters such as typhoons and earthquakes [53,54]. For example, coastal regions such as Macau and the Pearl River Delta area have progressively developed building safety codes that emphasize resistance to wind pressure and structural impact in response to frequent typhoons [55]. Common causes of residential fires include the misuse of electrical appliances, electrical overloads, and damaged or leaking gas pipelines. For older adults, owing to age-related declines in perception and reaction capacity, such risks are more likely to escalate into severe injuries or even accidental fatalities [56]. Meanwhile, maintaining adequate indoor and outdoor lighting during both daytime and nighttime is directly linked to the environmental safety of older adults, while also fostering positive emotional and perceptual experiences that enhance their alertness and preventive awareness [57]. Finally, drawing upon the literature review, this study identifies and summarizes the evaluation indicators used in the analysis (see

Table 1. Evaluation metrics.

Primary Indicator	Secondary Indicators	Indicator Explanation	Source
Physical Security (A)	Equipment Operational Reliability (A1)	Do users believe that smart cameras, door locks, and alarm devices can effectively prevent intrusions?	[58]
	Equipment Vulnerability Concerns (A2)	Are users concerned that hackers could pose a physical threat by hacking smart door locks or surveillance systems?	[10]
	Monitoring Coverage Perception (A3)	Do users believe that key areas of the apartment (such as entrances, hallways, and balconies) are fully covered by security equipment with no obvious blind spots?	[4,17]
	Protection Structure Satisfaction (A4)	User satisfaction with physical security structures such as access control systems, stairwell barriers, and lock strength.	[4]
Sense of Information Security (B)	Privacy Leakage Concerns (B1)	Are users concerned that personal behavioral, health, or lifestyle information collected by smart devices may be used without authorization or leaked?	[59]
	Perceived Risk of Cyberattacks (B2)	Users are concerned that their devices or networks may be hacked, leading to risks of data breaches or manipulation.	[60]
	Trust in Data Usage Transparency (B3)	Do users believe that the device (or service provider) has a clear and transparent explanation mechanism regarding the purpose of data collection, storage, and permission controls?	[4,17]
	Perceived Implicit Data Inference (B4)	Are users aware that sensors may infer privacy-sensitive information such as their behavior and lifestyle patterns through non-visual data (e.g., temperature, motion)?	[4]
Psychological Safety (C)	Trust in System Automation (C1)	Do users trust that smart systems can automatically execute tasks with fault-tolerant mechanisms, minimizing operational errors to enhance psychological security?	[61]
	Sense of Control and Response Confidence (C2)	Do users feel they can effectively control/intervene in smart systems, respond to anomalies, and gain autonomy and security?	[62]
	Level of Peace of Mind and Comfort (C3)	Do smart facilities alleviate users’ daily anxieties, fostering mental relaxation and psychological comfort within living spaces?	[63,64]
	Perceived Social Trust and Support (C4)	Do users perceive that designers/manufacturers/property management assume responsibility for system safety and provide adequate technical/situational support, thereby enhancing trust and security?	[65]
Sense of Personal Safety (D)	Perceived Risk of Unexpected Events (D1)	Do users believe that smart systems (such as fire and gas sensors) can promptly identify accident risks and provide early warnings?	[66]
	Trust in Medical Emergency Response (D2)	Do users perceive that the smart home system can automatically alert emergency services and notify responders during health emergencies such as cardiac arrest or sudden illness?	[67]
	Remote Security Support (D3)	Do users believe that the smart system’s features such as remote viewing and emergency assistance enhance the convenience and practicality of home safety?	[12]

).

3. Methods and Data

3.1. Research Design and Methods

The key value of the multi-attribute decision-making (MADM) model lies in its ability to efficiently reveal the interdependencies and uncertainties among multiple factors within complex systems under resource constraints and to derive improvement strategies for real-world applications through the evaluation of empirical cases [23,68,69]. This technical approach has been widely applied across diverse disciplines, particularly in cutting-edge studies on energy development and environmental management [70,71]; in tourism management and destination marketing [72,73]; in urban planning and governance [74,75]. Therefore, this study employed a hybrid multi-attribute decision-making (MADM) model—integrating the DEMATEL-based ANP (DANP) and Modified VIKOR (mVIKOR) methods, commonly referred to in previous studies as the DANP–mV model to address the complex, multidimensional, uncertain, and conflicting contexts of safety perception evaluation in smart apartments from the perspective of older adults.

The DANP–mV model integrates three analytical techniques: Decision-Making Trial and Evaluation Laboratory (DEMATEL), DEMATEL-based Analytic Network Process (DANP), and the Modified VIKOR(See Figure 1). In recent years, this decision-making model has attracted growing academic interest among researchers across related disciplines due to its methodological advantages. DEMATEL can explicitly identify the direct and indirect effects among evaluation criteria, thereby overcoming the traditional analytical assumption of mutual independence. Even when criteria conflict or interact with one another, a robust determination of priorities and selection of solutions can be achieved through network-based weighting and compromise-ranking procedures [76]. By employing ideal and anti-ideal benchmarks instead of relative comparisons, the model avoids identifying merely a “relatively optimal” alternative when the overall system performance is low. By integrating causal structure analysis with gap measurement, the model can systematically identify key driving factors and thereby propose phased, actionable improvement strategies [23].

3.1.1. DEMATEL

First, the direct impact matrix E was derived from the expert questionnaire results to capture the perceived causal relationships among evaluation criteria. The questionnaire adopted a 0–4 rating scale, where 0 indicated no influence and 4 indicated a very strong influence. Subsequently, the average direct influence matrix A was derived by averaging the direct influence matrices provided by all experts. Before proceeding to the next stage of analysis, a consistency check was performed to evaluate the reliability of the expert questionnaire results. A 5% threshold was adopted as the criterion for the gap value. When the calculated result falls below this threshold, it indicates that the reliability of the expert questionnaire exceeds 95%. This step helps to ensure the overall robustness and reliability of subsequent analyses. If the test results suggest instability, the reliability of the collected data and the adequacy of the expert sample size should be further examined. After the consistency test, the average influence matrix A was normalized to derive the standardized matrix D. The purpose of normalization is not only to eliminate dimensional differences, but also to convert all values into a comparable scale between 0 and 1. On this basis, the total impact matrix T is constructed using Equation (5), which is the basis for drawing the impact network relationship diagram (INRM) in this study. Equations (6) and (7) were then applied to sum the values in each row and column of the matrix, respectively. Consequently, the centrality degree (D + R) and the influence degree (D − R) of each evaluation criterion were derived.

$$E=\left[\begin{array}{ccccc}{e}_{11}& \dots & e_{1j}& \dots & e_{1n}\\ ⋮& & ⋮& & ⋮\\ e_{i1}& \dots & e_{ij}& \dots & e_{in}\\ ⋮& & ⋮& & ⋮\\ e_{n1}& \dots & e_{nj}& \dots & e_{nn}\end{array}\right]$$

(1)

$$A=\left[\begin{array}{ccccc}{a}_{11}& \dots & a_{1j}& \dots & a_{1n}\\ ⋮& & ⋮& & ⋮\\ a_{i1}& \dots & a_{ij}& \dots & a_{in}\\ ⋮& & ⋮& & ⋮\\ a_{n1}& \dots & a_{nj}& \dots & a_{nn}\end{array}\right]$$

(2)

$$\mathrm{Average} \mathrm{gap} - \mathrm{ratio} \mathrm{in} \mathrm{consensus} (\%)={\displaystyle \frac{1}{n\left(n-1\right)}}\sum _{i=1}^n \sum _{j=1}^n \left({\displaystyle \frac{\left|a_{ij}^H-a_{ij}^{H-1}\right|}{a_{ij}^H}}\right)\times 100\%$$

(3)

$$D=b\cdot Ab=min\left\{{\displaystyle \frac{1}{\underset{1\le i\le n}{max} \sum _{j=1}^n a_{ij}}},{\displaystyle \frac{1}{\underset{1\le j\le n}{max} \sum _{i=1}^n a_{ij}}}\right\}$$

(4)

$$\begin{gathered} T=D+D^2+\dots +D^q=D\left(I+D+D^2+\dots +D^{q-1}\right){\displaystyle \\ }=D(I+D+D^2+...+D^{q-1})(I-D)(I-D{)}^{-1}=D(I-D{)}^{-1},\mathrm{w}\mathrm{h}\mathrm{e}\mathrm{n} lim{D}^q=[0{]}_{n\times n} \end{gathered}$$

(5)

$$o=(o_i{)}_{n\times 1}={\left[\sum _{j=1}^n t_{ij}\right]}_{n\times 1}=(o_1,\dots ,o_i,\dots ,o_n)$$

(6)

$$r=(r_i{)}_{n\times 1}={\left(r_j\right)}_{1\times n}^{\prime }={\left[\sum _{i=1}^n t_{ij}\right]}_{1\times n}^{\prime }={\left(r_1,\dots ,r_j,\dots ,r_n\right)}^{\prime }$$

(7)

3.1.2. DANP

Within the DANP–mV framework, DANP is employed to derive the influence weights (IWs) of all criteria in the system. First, the total influence matrix $T$ was partitioned vertically based on the predefined dimensions, and the sum of the criteria under each dimension was then calculated. Next, each element in the partitioned matrix was normalized by dividing it by the total of its corresponding dimension, thereby obtaining the normalized matrix ${\mathit{T}}_c^a$. Following the logic of the AHP pairwise comparison process, the normalized matrix ${\mathit{T}}_c^a$ was then transformed into the unweighted supermatrix. The dimensional weight matrix ${\mathit{W}}^a$ was obtained using the same normalization procedure, in which the total influence matrix among the dimensions was normalized to derive the matrix ${\mathit{T}}_D^a$. Multiplying the matrix ${\mathit{W}}^a$ by the matrix ${\mathit{T}}_D^a$ yields the weighted supermatrix ($\mathit{W}$ as shown in Equation (12)). Finally, the limit supermatrix was obtained by raising the weighted supermatrix $\mathit{W}$ to successive powers until it converged to a stable state. After the limit supermatrix ${\mathit{W}}^g$ was obtained, the value in each column represented the global weight of the corresponding criterion, referred to as the influence weight (IWs).

$${\mathit{T}}_c^{\alpha }=\begin{array}{cc}& \begin{array}{c}\begin{array}{ccccc}{\mathit{D}}_1& & {\mathit{D}}_j& & {\mathit{D}}_m\\ c_{11}\dots c_{1m_1}& \dots & c_{j1}\dots c_{j{m}_j}& \dots & c_{n1\dots }{c}_{mmm}\end{array}\end{array}\\ \begin{array}{cc}{\mathit{D}}_1& c_{11}\\ & c_{12}\\ & ⋮\\ ⋮& c_{1m_1}\\ & ⋮\\ & ci1\\ {\mathit{D}}_i& c_{i2}\\ & ⋮\\ & c_{i{m}_i}\\ ⋮& ⋮\\ & c_{m1}\\ & c_{m2}\\ & ⋮\\ {\mathit{D}}_m& c_{m{m}_m}\end{array}& {\left[\begin{array}{ccccc}{\mathit{T}}_c^{\alpha 11}& \dots & {\mathit{T}}_c^{\alpha 1j}& \dots & {\mathit{T}}_c^{\alpha 1m}\\ \\ \\ ⋮& & ⋮& & ⋮\\ \\ \\ \\ \\ & & & & \\ {\mathit{T}}_c^{\alpha i1}& \dots & {\mathit{T}}_c^{\alpha ij}& \dots & {\mathit{T}}_c^{\alpha im}\\ \\ \\ ⋮& & ⋮& & ⋮\\ \\ {\mathit{T}}_c^{\alpha m1}& \dots & {\mathit{T}}_c^{\alpha mj}& \dots & {\mathit{T}}_c^{\alpha mm}\end{array}\right]}_{n\times n\mid m<n,{\sum }_{j=1}^m m_j=n}\end{array}$$

(8)

$$t_i^{14}={\sum }_{j=1}^{m_4} t_{ij}^{14},i=\mathrm{1,2},\dots ,m_1$$

(9)

$${\mathit{T}}_c^{\alpha 14}=\begin{array}{c}\\ c_{11}\\ ⋮\\ c_{1i}\\ ⋮\\ c_{1m_1}\end{array}\left[\begin{array}{ccccc}{c}_{41}& \dots & c_{4j}& \dots & c_{4m_4}\\ t_{11}^{14}/t_1^{14}& \dots & t_{1j}^{14}/t_1^{14}& \dots & t_{1m_4}^{14}/t_1^{14}\\ ⋮& & ⋮& & ⋮\\ t_{i1}^{14}/t_i^{14}& \dots & t_{ij}^{14}/t_i^{14}& \dots & t_{i{m}_4}^{14}/t_i^{14}\\ ⋮& & ⋮& & ⋮\\ t_{m_{1}1}^{14}/t_{m_1}^{14}& \dots & t_{m_{1}j}^{14}/t_{m_1}^{14}& \dots & t_{m_1{m}_4}^{14}/t_{m_1}^{14}\end{array}\right]=\left[\begin{array}{ccccc}{t}_{11}^{\alpha 14}& \dots & t_{1j}^{\alpha 14}& \dots & t_{1m_4}^{\alpha 14}\\ ⋮& & ⋮& & ⋮\\ t_{i1}^{\alpha 14}& \dots & t_{ij}^{\alpha 14}& \dots & t_{i{m}_4}^{\alpha 14}\\ ⋮& & ⋮& & ⋮\\ t_{m_{1}1}^{\alpha 14}& \dots & t_{m_{1}j}^{\alpha 14}& \dots & t_{m_1{m}_4}^{\alpha 14}\end{array}\right]$$

(10)

$${\mathit{W}}^{\alpha }={\left({\mathit{T}}_C^{\alpha }\right)}^{\mathrm{\prime }}=\begin{array}{cc}& \begin{array}{c}\begin{array}{rrrrr}{D}_1& & D_i& & D_m\\ c_{11}\dots c_{1m_1}& \dots & c_{i1}\dots c_{i{m}_i}& \dots & c_{m1}\dots c_{mmm}\end{array}\end{array}\\ \begin{array}{cc}{D}_1& c_{11}\\ & c_{12}\\ ⋮& ⋮\\ & c_{1m_1}\\ & ⋮\\ & c_{j1}\\ D_j& c_{j2}\\ & ⋮\\ & c_{j{m}_j}\\ ⋮& ⋮\\ & c_{m1}\\ & c_{m2}\\ & ⋮\\ D_m& c_{m{m}_m}\end{array}& {\left[\begin{array}{ccccc}{\mathit{W}}^{11}& \dots & {\mathit{W}}^{i1}& \dots & {\mathit{W}}^{m1}\\ \\ \\ ⋮& & ⋮& & ⋮\\ \\ \\ {\mathit{W}}^{1j}& \dots & {\mathit{W}}^{ij}& \dots & {\mathit{W}}^{mj}\\ \\ \\ ⋮& & ⋮& & ⋮\\ & & & & \\ \\ \\ \\ \\ {\mathit{W}}^{1m,}& \dots & {\mathit{W}}^{im}& \dots & {\mathit{W}}^{mm}\end{array}\right]}_{n\times n\mid m<n,{\sum }_{j=1}^m m_j=n}\end{array}$$

(11)

$$\mathit{W}={\mathit{T}}_D^{\alpha }{\mathit{W}}^{\alpha }=\begin{array}{cc}\begin{array}{cc}{D}_1& c_{11}\end{array}& \begin{array}{c}\begin{array}{rrrrr}{D}_1& & D_i& & D_m\\ c_{11}\dots c_{1m_1}& \dots & c_{i1}\dots c_{i{m}_i}& \dots & c_{m1}\dots c_{mmm}\end{array}\end{array}\\ \begin{array}{cc}& c_{12}\\ ⋮& ⋮\\ & c_{1m_1}\\ & ⋮\\ & c_{j1}\\ D_j& c_{j2}\\ & ⋮\\ & c_{j{m}_j}\\ ⋮& ⋮\\ & c_{m1}\\ & c_{m2}\\ & ⋮\\ D_m& c_{m{m}_m}\end{array}& \left[\begin{array}{ccccc}{t}_{11}^{\alpha D}\times W^{11}& \dots & t_{i1}^{\alpha D}\times W^{i1}& \dots & t_{m1}^{\alpha D}\times W^{m1}\\ \\ \\ ⋮& ⋮& & & ⋮\\ \\ t_{1j}^{\alpha D}\times W^{1j}& \dots & t_{ij}^{\alpha D}\times W^{ij}& \dots & t_{mj}^{\alpha D}\times W^{mj}\\ \\ \\ ⋮& & ⋮& & ⋮\\ & & & & \\ \\ \\ \\ t_{1m}^{\alpha D}\times W^{1m}& \dots & t_{im}^{\alpha D}\times W^{im}& \dots & t_{mm}^{\alpha D}\times W^{mm}\end{array}\right]\end{array}$$

(12)

$${\mathit{W}}^{\mathrm{g}}={\mathrm{l}\mathrm{i}\mathrm{m}}_{z\to \mathrm{\infty }} (\mathit{W}{)}^z$$

(13)

3.1.3. Modified VIKOR

This study employs a Modified VIKOR method to classify evaluation alternatives in specific cases into positive and negative ideal solutions. Subsequently, the relative closeness of each performance outcome to the positive ideal solution is calculated to derive the overall priority ranking of the evaluation criteria. To address the inherent limitations of the traditional VIKOR approach—such as passive selection of the best alternative under generally low performance levels and ranking instability caused by the absence of ideal solutions or zero improvement margins—Xiong et al. (2017) introduced the concept of an “ideal criterion” to replace the conventional positive ideal solution [68]. This conceptual shift prevents the method from merely identifying the “relatively best” option among suboptimal alternatives, thereby aligning it more closely with the non-ideal and uncertain conditions encountered in real-world performance evaluations. Equations (14) and (15) was applied to compute the total gap and overall performance values in the empirical case, serving as the analytical foundation for formulating targeted improvement strategies.

$$f^{\mathrm{a}\mathrm{s}\mathrm{p}\mathrm{i}\mathrm{r}\mathrm{e}\mathrm{d}}=\left(f_1^{\mathrm{a}\mathrm{s}\mathrm{p}\mathrm{i}\mathrm{r}\mathrm{e}\mathrm{d}},\dots ,f_j^{\mathrm{a}\mathrm{s}\mathrm{p}\mathrm{i}\mathrm{r}\mathrm{e}\mathrm{d}},\dots ,f_n^{\mathrm{a}\mathrm{s}\mathrm{p}\mathrm{i}\mathrm{r}\mathrm{e}\mathrm{d}}\right)$$

(14)

$$f^{worst}=\left(f_1^{worst},...,f_j^{worst},...,f_n^{worst}\right)$$

(15)

$$s_k={\sum }_{j=1}^n{w}_j{r}_{kj}={\sum }_{j=1}^n{w}_j(\mid f_j^{aspired}-f_{kj}\mid )/(\mid f_j^{aspired}-f_j^{worst}\mid )$$

(16)

3.2. Empirical Cases and Geographic Descriptions

The empirical case is conducted at the Heyuan·Yiyang Health Care Center, located in Zhuhai’s Xiangzhou District. The project has been designated as a district-level priority elderly care institution in Zhuhai City. Scheduled to open in 2024, it is designed as a comprehensive senior care facility integrating medical services, long-term nursing, and rehabilitation functions. The center covers a total floor area of approximately 77,000 square meters and comprises two principal buildings—namely, a healthcare building and a medical building. It provides around 1200 beds in total, encompassing publicly funded beds as well as market-based senior apartments and nursing units.

At the architectural level, the Heyuan·Yiyang Health Care Center adopts a dual approach that combines aging-friendly and intelligent design principles. Indoors, spacious corridors and barrier-free circulation routes are implemented, while elevators, handrails, and non-slip flooring are installed to meet the daily mobility needs of older adults with limited physical capacity. The center is also equipped with a range of smart devices, including fall-detection radar, respiration and heart-rate monitoring systems, intelligent smoke sensors, pan–tilt cameras, and sleep-monitoring belts, enabling real-time monitoring of residents’ daily status and early detection of abnormalities.

In addition, the facility houses a secondary-level general hospital and four specialized medical centers focusing on health management, chronic disease management, geriatric medicine, and rehabilitation, ensuring continuous and timely medical support for residents. Compared with traditional apartment-style housing, this project features a more integrated system encompassing medical support, information management, and emergency response. However, it simultaneously introduces new challenges, such as data privacy risks, operational complexity, and issues of user trust in intelligent equipment.

3.3. Data Collection

First, during the expert evaluation stage, a DEMATEL-based questionnaire was administered to collect expert judgments and construct the causal relationships among the various criteria within the smart apartment security perception system. During expert selection, the study adhered to the principles of interdisciplinary diversity and professional expertise to ensure the rigor and reproducibility of the evaluation results. A total of 17 experts were invited to complete the questionnaire and contribute to the construction of the judgment matrix, representing three major disciplinary domains: urban planning and architecture, computer science and technology, and the industrial Internet of Things (IoT). Seven experts specialized in urban planning and architecture, including four professors and three associate professors. Their primary research areas focus on urban renewal, community space design, and the development of age-friendly environments. Six experts specialized in computer science and technology, including three professors and three associate professors. Their primary research areas include artificial intelligence, information security, and human–computer interaction. Four experts specialized in the field of Industrial Internet of Things (IoT), all of whom are senior product managers with extensive practical experience in smart apartment system architecture and security protection. In terms of professional experience, four experts (23.5%) had 4–5 years of experience, eight experts (47.1%) had 5–10 years, and five experts (29.4%) had more than 10 years (See

Table 2. Expert information.

Serial Number	Field	Occupation	Years in the Field
1	Urban Planning and Architecture	Professor	4 to 5 years
2	Urban Planning and Architecture	Professor	5 to 10 years
3	Industrial IoT	IoT Product Manager	5 to 10 years
4	Computer Science and Technology	Professor	4 to 5 years
5	Computer Science and Technology	Professor	More than 10 years
6	Computer Science and Technology	Professor	More than 10 years
7	Urban Planning and Architecture	Professor	More than 10 years
8	Industrial IoT	IoT Product Manager	5 to 10 years
9	Urban Planning and Architecture	Associate Professor	5 to 10 years
10	Urban Planning and Architecture	Professor	More than 10 years
11	Computer Science and Technology	Associate Professor	5 to 10 years
12	Urban Planning and Architecture	Associate Professor	5 to 10 years
13	Urban Planning and Architecture	Associate Professor	5 to 10 years
14	Industrial IoT	IoT Product Manager	4 to 5 years
15	Industrial IoT	IoT Product Manager	5 to 10 years
16	Computer Science and Technology	Associate Professor	5 to 10 years
17	Computer Science and Technology	Associate Professor	5 to 10 years

). The questionnaire consisted of 15 safety perception criteria, each evaluated on a 0–4 Likert-type scale, with 0 representing “no impact” and 4 representing “extremely high impact”.

Secondly, a Modified VIKOR questionnaire was distributed among elderly residents in the case study smart apartment to assess their satisfaction levels and to quantify the performance and gap values of each criterion under real living conditions. Respondents’ satisfaction with 15 safety perception criteria was assessed through face-to-face interviews using a 0–10 Likert-type scale, where 0 denoted “very poor” and 10 denoted “excellent”. To ensure privacy protection, respondents’ names and specific addresses were not collected. The study initially aimed to obtain 200 valid questionnaires; in total, 174 respondents were approached, yielding 161 valid and 13 invalid responses. Questionnaires were deemed invalid if the responses were excessively uniform or if key items were left unanswered. Among the valid samples, 79 were male (49%) and 82 were female (51%).

4. Results and Discussion

This study applies the DEMATEL method to analyze the interrelationships among the dimensions. Based on the analysis results, the INRM presented in Figure 2 clearly illustrates the structural relationships underlying security perception in smart apartments. First, the network relationships reveal the dominant impact pathways among the dimensions. The INRM analysis indicates that the order of influence follows the sequence B–D–A–C. This indicates that sense of information security (B) functions as the primary causal source within the network, exerting a direct influence on sense of personal safety (D) as the main driving dimension, while also indirectly affecting physical security (A) and psychological security (C). At the same time, Sense of Personal Safety (D) exerts a secondary driving influence on physical security (A) and psychological Safety (C). Together, dimensions B and D serve as the core causal drivers of overall security perception, whereas A and C function primarily as outcome dimensions influenced by the former (See Figure 2). These findings suggest that although physical and psychological safety represent the most immediate forms of perceived security for residents, relying solely on such surface-level improvements is insufficient to achieve substantive enhancement in overall security perception. To achieve the sustainable development of a secure community, efforts should prioritize enhancing physical and psychological security.

Similarly, according to the INRM analysis, within the dimension of physical security (A), the perception of monitoring coverage perception (A3) emerged as the most influential indicator, followed by A1, A2, and A4. Within the dimension of sense of information security (B), privacy leakage concerns (B1) exert the strongest influence, followed by B2, B3, and B4. For psychological security (C), perceived social trust and support (C4) emerges as the key indicator, followed by C1, C2, and C3. In the case of Sense of personal safety (D), remote security support (D3) is identified as the most dominant indicator, followed by D1 and D2.

The influence weights derived from the DANP analysis (see

Table 3. Performance evaluation of Modified VIKOR case studies.

Heyuan·Yiyang Health Care Center
Dimensions/Criteria	Influential-Weights (IWs)	Performance	Unweighted (Gap)	Weighted (Gap)
(A) Physical Security	0.172	6.427		0.437
(A1)Equipment Operational Reliability	0.280	6.180	0.124	0.035
(A2)Equipment Vulnerability Concerns	0.252	6.006	0.547	0.138
(A3)Monitoring Coverage Perception	0.214	6.329	0.391	0.084
(A4)Protection Structure Satisfaction	0.254	7.199	0.714	0.181
(B) Sense of Information Security	0.352	6.744		0.466
(B1)Privacy Leakage Concerns	0.235	6.745	0.416	0.098
(B2)Perceived Risk of Cyberattacks	0.301	7.385	0.429	0.129
(B3)Trust in Data Usage Transparency	0.200	6.801	0.882	0.177
(B4)Perceived Implicit Data Inference	0.264	5.969	0.236	0.062
(C) Psychological Safety	0.155	6.872		0.596
(C1)Trust in System Automation	0.334	7.627	0.298	0.099
(C2)Sense of Control and Response Confidence	0.239	5.075	0.969	0.232
(C3)Level of Peace of Mind and Comfort	0.224	6.752	0.534	0.120
(C4)Perceived Social Trust and Support	0.203	7.882	0.714	0.145
(D) Sense of Personal Safety	0.321	7.600		0.438
(D1)Perceived Risk of Unexpected Events	0.323	7.286	0.317	0.102
(D2)Trust in Medical Emergency Response	0.359	8.056	0.211	0.076
(D3) Remote Security Support	0.317	7.404	0.820	0.260
Total Performance		6.984
Total Gap				0.472
Cronbach’s α	0.966

) indicate that the evaluation and optimization of the senior housing environment can be further examined through Modified VIKOR-based performance analysis. The total gap value is calculated as the weighted sum of the gap values for all indicators, representing the deviation from the overall ideal performance. In this study, the gap value for each indicator was assessed via a survey of residents in the case-study apartment (see Table 3). Reducing these gaps would contribute to enhancing the comprehensive performance of each dimension and its constituent indicators.

As shown in Table 3, the INRM model reveals that psychological safety (C) exhibits the largest gap value (0.596), indicating that it should be prioritized for improvement among all dimensions. This finding is highly consistent with the influence pathway of the causal network. Within dimension C, the gap value of “Sense of Control and Response Confidence” (C2) is the largest (0.232), indicating the most significant weakness in this dimension. Focusing on the optimization of indicator C2 can drive multi-dimensional collaborative improvements by enhancing users’ sense of control and adaptability, and ultimately achieve a systematic improvement in the security perception of smart apartments and an improvement in the quality of life of elderly residents.

Based on the INRM (Figure 2) and the performance evaluation results (Table 3), a comprehensive analysis of improvement priorities was conducted by considering the impact relationships and intensities among the framework indicators, and a continuous improvement strategy was subsequently proposed. The Gap value serves as the basis for determining the improvement priorities. Table 3 shows that elderly residents across the empirical cases are generally dissatisfied with their living environment and experience a lack of psychological security (C). Compared with the gap values of other dimensions, the overall satisfaction level is relatively low, indicating that elderly residents in the empirical cases express generally low satisfaction across all safety dimensions and are particularly concerned about the lack of psychological security in their living environment. Among the 15 indicators listed in Table 3, those with the largest gap values are D3 (0.260), C2 (0.232), A4 (0.181), B3 (0.177), and C4 (0.145). Accordingly, D3, with the highest gap value, should be prioritized for improvement. To address the fundamental issue of insufficient safety perception identified in the empirical case, a systematic strategy can be formulated based on the DANP model.

First, with respect to D3, the current help-seeking mechanism that depends on phone calls or family members presents structural deficiencies, including complex operation procedures and significant response delays among the elderly. Similarly, wearable alarm devices are often ineffective due to high false-alarm rates and poor user compliance [77]. Every minute of delay in emergency response markedly elevates the risk of disability or death among elderly individuals who experience falls. Therefore, remote support should not be regarded as merely an auxiliary safeguard, but rather as a critical determinant of the overall ceiling of safety performance [78,79]. Therefore, it is necessary to reconstruct the support system with IoT monitoring, one-button alarms, and emergency linkage as its core framework, while simultaneously simplifying the interaction design and enhancing ergonomic adaptation to increase user willingness and system accessibility [80].

Based on the interrelationships among the indicators shown in

Table A2. Total impact Matrix for Standard TC.

	A1	A2	A3	A4	B1	B2	B3	B4	C1	C2	C3	C4	D1	D2	D3
A1	0.097	0.089	0.077	0.132	0.172	0.366	0.249	0.277	0.136	0.109	0.103	0.092	0.327	0.336	0.210
A2	0.120	0.102	0.126	0.144	0.193	0.403	0.256	0.338	0.124	0.112	0.097	0.107	0.330	0.416	0.405
A3	0.154	0.141	0.086	0.102	0.291	0.343	0.272	0.326	0.153	0.103	0.115	0.102	0.350	0.379	0.340
A4	0.163	0.135	0.087	0.104	0.299	0.342	0.273	0.315	0.155	0.120	0.107	0.113	0.384	0.372	0.356
B1	0.127	0.112	0.110	0.124	0.168	0.252	0.191	0.224	0.146	0.099	0.098	0.090	0.362	0.314	0.298
B2	0.118	0.121	0.101	0.122	0.210	0.209	0.181	0.232	0.138	0.107	0.090	0.079	0.292	0.380	0.305
B3	0.108	0.094	0.089	0.097	0.186	0.219	0.119	0.186	0.121	0.095	0.088	0.073	0.275	0.296	0.263
B4	0.098	0.091	0.081	0.108	0.192	0.213	0.143	0.148	0.103	0.068	0.078	0.069	0.256	0.281	0.257
C1	0.096	0.096	0.093	0.118	0.161	0.316	0.257	0.296	0.106	0.126	0.128	0.119	0.211	0.349	0.335
C2	0.116	0.101	0.099	0.108	0.236	0.180	0.202	0.245	0.102	0.061	0.059	0.054	0.286	0.289	0.271
C3	0.138	0.127	0.110	0.130	0.272	0.327	0.258	0.288	0.168	0.126	0.077	0.070	0.319	0.357	0.330
C4	0.137	0.125	0.117	0.128	0.286	0.335	0.264	0.320	0.175	0.123	0.123	0.073	0.340	0.367	0.345
D1	0.163	0.059	0.065	0.097	0.264	0.160	0.111	0.130	0.076	0.066	0.056	0.054	0.154	0.173	0.146
D2	0.074	0.088	0.072	0.071	0.114	0.292	0.105	0.127	0.113	0.070	0.063	0.054	0.146	0.159	0.148
D3	0.078	0.108	0.066	0.065	0.122	0.165	0.116	0.285	0.120	0.061	0.063	0.059	0.152	0.171	0.144

, C2 (Sense of Control and Response Confidence) exerts an influence on the performance of D3 (Remote Security Support). Meanwhile, D3 has the second largest weighted gap value in the Modified VIKOR performance evaluation. In the empirical cases, the absence of C2 not only reduces user willingness but also triggers a self-reinforcing negative feedback loop [81]. Alarm terminals and sensor devices designed to enhance safety, when subjected to dual pressures from operational complexity and the risk of accidental triggering, can exacerbate elderly users’ doubts about their ability to cope effectively in emergencies [82]. Although the system was designed to enhance the sense of control, it paradoxically generated a deeper psychological sense of loss of control, which in turn undermined its activation rate and functional effectiveness—making the remote support network appear complete in form yet remain functionally ineffective over time. For the above reasons, emergency operations should be transformed—through cognitive reframing and micro-situational familiarization—from sporadic tasks into predictable, internalized coping habits. This transformation enables individuals to automatically execute appropriate responses based on procedural memory, even under high-pressure conditions, thereby reducing the perceived threat and uncertainty of sudden events and restoring a sense of control over risk situations [83].

In the INRM network, the physical security dimension (A) lies at the terminal end of the system’s causal chain, receiving upstream influence from Sense of Information Security (B) and Sense of personal safety (D). The core indicator of this dimension, A4 (Protection Structure Satisfaction), exhibits the third largest gap value (0.181) in the Modified VIKOR performance evaluation. This finding indicates that elderly residents generally lack confidence in the reliability of protective facilities. During the empirical investigation, we observed that although elderly residents interviewed were equipped with access control systems, protective railings, and other basic safety facilities, many still installed additional bolts at night, repeatedly checked windows and doors, or reduced nighttime outings. This behavior did not stem from actual danger but rather from the absence of a clear and tangible safety confirmation mechanism, leaving them in a persistent state of psychological vigilance toward “possible risks” [84]. To fill the psychological gap created by this “perceptual mismatch,” the core of the intervention lies not in adding more protective facilities but in reconstructing the confirmability experience of existing protective structures. By embedding lightweight and familiar interaction experiences into everyday situations, these facilities can be transformed from static environmental backgrounds into perceptible and reassuring safety elements [33]. At the same time, simple and intuitive inspection prompts can be employed to continuously provide residents with feedback on the integrity of protective facilities, thereby transforming their sense of safety from “speculation” to “visibility”.

5. Conclusions

5.1. Theoretical and Management Implications

This study aims to enhance the safety perception of elderly residents in smart apartments, thereby improving their living experience and quality of life. It systematically reviews the relevant literature and identifies the perception factors related to elderly safety as the primary analytical focus. Furthermore, this study integrates multiple dimensions to refine the theoretical framework across related fields, establish a systematic evaluation model and standards, and reveal the mutual influence pathways among the various components. A hybrid MCDM approach integrating DEMATEL and ANP with a Modified VIKOR method was employed to conduct a systematic evaluation and develop effective improvement strategies for sustainable development. This study’s assessment of safety perception in the empirical cases not only evaluates existing strengths and weaknesses but also reveals a deeper insight: although technical support systems have been fully implemented, elderly residents still fail to internalize them as reliable sources of safety because of their limited perceptibility and availability in daily life. Although the overall performance has reached a high level, the gap value of remote security support (D3) remains as high as 0.260 in terms of the expected-level gap, representing the most prominent shortcoming and indicating that hardware deployment has not translated into stable psychological confidence. Second, this study proposes a global strategy that moves beyond simple gap-filling and instead re-prioritizes improvement efforts according to the system’s influence structure. The analysis indicates that prioritizing the enhancement of remote safety support (D3) can drive a systematic transformation from perceptual discontinuity to overall confidence by rebuilding the sense of control in risk situations. This, in turn, fosters a synergistic improvement in both sense of control and response confidence (C2, 0.232) and Protection Structure Satisfaction (A4, 0.181). For the design of future smart apartments, this implies that improvements should no longer be understood as the mere accumulation of local technologies but should instead focus on structural reconstruction centered on psychological leverage points to achieve sustainable enhancement of elderly residents’ safety experience.

5.2. Limitations and Future Research

This study is subject to several limitations. First, although first-hand feedback from elderly users was incorporated into the data analysis, the sample size and coverage were insufficient to fully capture the diverse needs across different groups of elderly residents. Future research should integrate large-scale surveys with multi-channel data sources to improve the representativeness and robustness of the findings. Second, given the current research focus, evaluating the safety perception of smart apartments from an expert perspective helps ensure the consistency and validity of the assessments. Nevertheless, elderly residents—the primary users—may exhibit subjective perceptions and behavioral responses that differ from expert judgments. Therefore, future research should directly survey elderly residents to complement and validate the findings of this study. Third, the case studies were geographically concentrated in Zhuhai, which may limit the generalizability of the findings to broader cross-regional or cross-cultural contexts. Future studies could undertake comparative analyses across multiple cities and cultural contexts to enhance the generalizability and applicability of the findings. Third, the Modified VIKOR method used in this study operates as an additive aggregation model, which offers the advantages of computational simplicity and intuitive interpretation. However, this feature also limits its ability to capture nonlinear relationships and potential interaction effects among multiple factors, possibly underestimating the coupling between indicators in complex systems. Future research should consider adopting non-additive evaluation approaches to make performance assessments more reflective of real-world conditions. It should be noted that these limitations affect the applicability of the conclusions rather than the generalizability of the methodology itself. As a process-oriented analytical tool, the DANP-mV framework remains applicable for indicator screening across different regions when integrated with local characteristics, thereby enabling methodological transferability and localized application.