An AIoT Product Development Process with Integrated Sustainability and Universal Design

Abstract

The rapid development of contemporary artificial intelligence and Internet of Things (IoT) technologies has given rise to the emerging paradigm of the AIoT (Artificial Intelligence of Things), which is profoundly impacting human life and driving the digital transformation of industries and society. The AIoT not only enhances product functionality and convenience but also accelerates the achievement of the United Nations Sustainable Development Goals (SDGs). However, the widespread adoption of these technologies still poses challenges related to social inclusivity, particularly regarding insufficient accessibility for elderly users, which may exacerbate the digital divide and social inequality, contradicting SDG 10 (reducing inequality). This study integrates AIoT product development processes based on sustainability and universal design principles using methods such as Quality Function Deployment, the Analytic Hierarchy Process, the Scenario Method, the Entropy Weight Method, and Fuzzy Comprehensive Evaluation. The features of this process include ease of operation and high flexibility, making it suitable for cross-departmental product development while prioritizing the needs of diverse age groups throughout the development process. The research findings indicate that the AIoT product concepts proposed can meet the needs of diverse users, contributing to the realization of age-friendly products. This study provides a reference point for future AIoT product development, promoting the inclusive and sustainable development of smart technology.

1. Introduction

The rapid advancement of contemporary artificial intelligence (AI) and Internet of Things (IoT) technologies is profoundly reshaping patterns of human daily life and giving rise to the emerging technological paradigm of the Artificial Intelligence of Things (AIoT) [1]. This convergence not only drives technological innovation but also opens new avenues for industrial development and social progress. The emergence of the AIoT signifies the arrival of the intelligent era, integrating data collection, analysis, and decision-making capabilities into interconnected devices, bringing unprecedented opportunities and challenges to various industries [2]. As AIoT technology continues to mature, its widespread application in smart cities, industrial automation, healthcare, and other fields is expected to further drive societal digital transformation and intelligent upgrading. According to a report by Grandview Research (2024), the global smart home market had a value of USD 79.16 billion in 2022 and is projected to grow at a compound annual growth rate (CAGR) of 27.07% from 2023 to 2030 [3,4].

The widespread adoption and application of AIoT technology not only significantly enhances product functionality and convenience but also accelerates the achievement of the United Nations Sustainable Development Goals (SDGs), creating unprecedented opportunities for innovation in global industries and society [5]. By integrating the advantages of artificial intelligence and the Internet of Things, AIoT technology has demonstrated immense potential across multiple fields, such as healthcare [6], smart home appliances, smart buildings, sustainable urban development, and climate action [7]. It not only injects new growth momentum into global industries but also provides innovative solutions to social issues, becoming a key driver of sustainable social development [8].

However, despite its revolutionary potential, current AIoT technology faces multiple challenges that need to be addressed. A 2022 United Nations survey report suggested that AI technology may not be user-friendly for elderly users in practical applications. This finding highlights the social inclusivity challenges posed by AIoT technology during its widespread adoption, particularly with regard to ensuring the technology’s usability and acceptability for all age groups [9,10,11]. Currently, research and development data for smart products primarily target younger and middle-aged populations, failing to adequately consider the needs and user experiences of the elderly. This may exacerbate age discrimination issues [12,13] and contradict SDG 10, Reduce Inequality, particularly in the context of rapid global population aging, making resolving this issue all the more urgent [14,15,16]. This trend may not only widen the technological divide but also exacerbate social inequality. Therefore, in the design and development of smart products, it is necessary to adopt more inclusive approaches to ensure that the needs of the elderly are adequately considered. This approach not only aligns with the requirements of the SDGs but also promotes social fairness and harmonious development. Current practices for preventing AI from promoting age discrimination primarily include the following: first, involving the elderly in the design of AI technology and collaborating with them [11,17]; second, establishing age-diverse data science teams [11,18,19]; third, collecting data that include all age groups [11,15]. However, these measures are challenging to implement and costly, making widespread adoption and application difficult. As a result, AIoT product development often focuses primarily on AI and IoT technology aspects, while the importance of the physical product layer is often overlooked. Specifically, such technological efforts usually concentrate on the following: (1) artificial intelligence development, including data science, AI engineering, real-time analytics, and AI/ML operations; (2) IoT-related infrastructures, such as enterprise and cloud integration, edge computing, OTA (over-the-air) updates, and DevOps for system operation and maintenance. For example, the gray area in Figure 1 represents frequently neglected components, including product design, manufacturing, user research, user experience, and user interfaces [20].

Given the current trends in AIoT technology development, this study proposes a product development process for Artificial Intelligence of Things (AIoT) products based on sustainability [21,22] and universal design principles. This product development process prioritizes the usage needs, acceptance, and recognition of different age groups, requiring ease of operation and high flexibility, and is suitable for cross-departmental collaboration in product development and management processes. We hope that this study will provide empirical references for scholars and businesses engaged in AIoT innovation research or development, thereby promoting the development of more inclusive and sustainable smart technology. The application of this AIoT product development process is expected to drive the practical implementation of AIoT technology in fields such as age-friendly environments, aligning with the SDGs and international human rights conventions and laying the foundation for realizing universal design principles [23,24].

2. Literature Review

2.1. AIoT Technology and Market Development

The Artificial Intelligence of Things (AIoT) is the product of the convergence of artificial intelligence (AI) and the Internet of Things (IoT) [25], integrating advanced data analysis, machine learning algorithms, and ubiquitous connectivity technologies into a unified technological paradigm [25,26]. The IoT collects and transmits real-time data through sensors and connected devices, while AI grants the system the ability to interpret, learn, and act autonomously [27]. The synergistic effect of these two technologies enables smart systems to perform intelligent decision-making, predictive maintenance, context-aware services, and adaptive control, further expanding the application boundaries of interconnected devices across various industries [28]. Areas application of the AIoT are extensive, including smart cities, industrial automation, healthcare, environmental monitoring, smart transportation, and smart homes [29,30]. In smart city development, the AIoT leverages predictive analytics to optimize resource allocation, improve traffic management, and enhance public safety [31,32]; in the industrial sector, the AIoT accelerates the implementation of Industry 4.0, boosting production efficiency, reducing downtime, and supporting automated operations [33,34]. In healthcare, the AIoT can be applied to remote patient monitoring, early disease diagnosis, and personalized treatment plan development [35]. Additionally, in environmental and energy management, the AIoT can facilitate real-time pollution level monitoring, optimize energy consumption, and promote sustainable management practices [36]. In recent years, with the advancement of 5G communication, edge computing, and sensor technology, the global AIoT market has shown a trend of rapid growth, being estimated to grow from USD 18.37 billion in 2024 to USD 79.13 billion in 2030, with a compound annual growth rate (CAGR) of approximately 27.6% during the forecast period [37].

2.2. The Relationship Between the AIoT and the United Nations Sustainable Development Goals

The emergence of the Artificial Intelligence of Things (AIoT) has provided unprecedented technological support for global sustainable development. By integrating capabilities such as sensing, connectivity, computing, and intelligent decision-making, the AIoT enables cross-domain data flow and knowledge transformation, thereby optimizing resource allocation, enhancing system efficiency, and promoting synergistic development across social, economic, and environmental dimensions [38,39].

From a macro-perspective, the AIoT aligns highly with the United Nations Sustainable Development Goals (SDGs) [38]. The SDGs focus not only on improvements in individual sectors but also on comprehensive, cross-system transformation and progress, aligning with the technical characteristics of the AIoT and applied AIoT models [38]. For example, the AIoT’s capacity for multi-level data integration can promote scientific and real-time decision-making, supporting evidence-based public policy formulation, while its intelligent operational mechanisms can help to enhance infrastructure resilience and service coverage, narrowing the gaps in technology and resource access between different regions and groups [2]. Additionally, if AIoT systems are designed and applied with sustainability principles embedded, they can deliver long-term benefits in terms of reducing energy waste, enhancing resource recycling, and improving quality of life [31]. However, the relationship between the AIoT and sustainable development does not automatically produce positive outcomes. Rapid technological expansion without consideration of inclusivity and ethical norms may lead to new digital divides and social inequalities, and even create cybersecurity and privacy risks [40,41]. In this study, ‘inclusivity’ is defined as ensuring that AIoT products and services are accessible and user-friendly for diverse groups, including the elderly, people with disabilities, children, and individuals from different socio-economic and cultural backgrounds, thereby reducing the risk of exclusion due to differences in age, physical conditions, or digital literacy. Meanwhile, ‘ethical norms’ encompass responsibility principles such as safeguarding data privacy and security, ensuring transparency and fairness in algorithmic decision-making, clarifying responsibility attribution in automated systems, and mitigating bias or discrimination in AI algorithms, ensuring that technological advances align with human values and the goals of sustainable development. Therefore, integrating the principles of fairness, inclusivity, and responsibility advocated by the SDGs into the development and governance framework of the AIoT is a critical factor in ensuring its long-term value. Only by balancing technological innovation with social responsibility can the AIoT truly become a core driver of global sustainable development rather than a potential source of risk.

2.3. The Challenge of Inclusion in AIoT Technology Adoption

Despite the rapid development of AIoT technology and its expanding application areas, ensuring that all user groups can benefit equally remains an important issue that needs to be addressed [41]. Among these, the digital divide and differences between age groups constitute significant obstacles in the adoption process [42,43]. Compared to younger generations who are digitally native, older users often lag behind in terms of digital literacy, information processing speed, sensory abilities, and familiarity with emerging technology ecosystems, which can lead to technological anxiety and affect their willingness to adopt and use smart products [43]. For the elderly, adopting AIoT products (such as smart home devices, smart medical monitoring systems, and smart mobility aids) often presents multiple usage barriers, including overly complex interface designs, reliance on unfamiliar interaction modes (such as voice assistants or gesture control), and the need to go through multi-step setup procedures [44]. Additionally, while AI-driven adaptive features can enhance system efficiency, users often find that their operational logic and response mechanisms lack transparency, making it difficult for older adults to predict, understand, or trust the system’s behavior [45].

Current solutions, such as simplified user interfaces, larger font sizes, or voice-guided tutorials, have improved usability to some extent. However, these solutions typically provide mostly superficial interface optimizations rather than a fundamental rethinking of interaction modes for accommodating different cognitive and physical abilities [46]. Additionally, the lack of unified design standards between AIoT platforms and devices forces older users to repeatedly learn to use and adapt to different products, further reducing their willingness to adopt such technologies [42,43]. Therefore, addressing inclusivity challenges in AIoT technology adoption requires an integrated user-centered design strategy. This should involve not only making localized adjustments to product interfaces but also incorporating universal design principles throughout the entire product development process to ensure that the system can accommodate diverse user capabilities while maintaining high transparency, thereby fostering trust [47]. Only by doing so can we prevent the digital divide from widening further and ensure that AIoT technology truly becomes an inclusive innovation that promotes social equity and enhances quality of life [48].

2.4. Advantages and Limitations of Applying Traditional Universal Design Principles to AIoT Product Development

Universal design was proposed in the 1970s by Ronald L. Mace, Director of the Center for Universal Design. Its core philosophy is to proactively consider the needs of users of different ages, abilities, and backgrounds during the planning and design of products, environments, and services, enabling them to be used safely, conveniently, and fairly by as many people as possible without the need for subsequent modifications or special designs [49]. When discussing the practical application and promotion of universal design, Story, Mueller, and Mace (1998) proposed three supplementary principles [50]. Therefore, the most commonly used universal design approach is based on seven core principles and three supplementary principles, comprising thirty-seven evaluation criteria [49], aiming to maximize accessibility and fairness for diverse users in the design processes of products, environments, and services. The seven principles encompass equitable use, flexibility of use, simplicity and clarity, perceptible information, tolerance for error, low physical effort, and appropriate spatial arrangement. These principles emphasize ensuring safe, convenient, and consistent user experiences across varying abilities, ages, and cultural backgrounds [51,52]. The three supplementary provisions further elaborate on practical design considerations, including adaptability to varying environmental conditions, optimizing the balance between cost and benefit, and integrating aesthetics with functionality [50], to ensure that universal design is functional, is feasible, and that its promotion and implementation provides long-term value. It provides an inclusive framework for product development, helping to address the insufficient consideration of diverse user needs in conventional product development processes. This is particularly important in the development of smart electronic products, as such products need to serve users of different ages, abilities, and technical proficiency levels. However, traditional universal design principles and evaluation indicators have limitations when addressing the characteristics of emerging smart products. Bai et al. (2016) and Yao et al. (2023) noted that traditional universal design principles fail to adequately consider the interconnectedness, adaptability, and artificial intelligence characteristics of AIoT products [53,54]. Even when following generalized interface optimization (e.g., font enlargement, step simplification), there is still a significant gap when applied to dynamic, data-driven smart scenarios [55]. Therefore, it is necessary to refine and expand these principles.

In light of this issue, this study proposes a set of improved universal design principles and product development processes, aiming to provide a more suitable guidance framework for AIoT product development. This new product development process not only retains the core concepts of traditional universal design but also incorporates AIoT characteristics, such as data privacy, adaptability, system transparency, and human–machine collaboration, among other evaluation criteria.

3. Methodology

This study focuses on theoretical exploration, process construction, and empirical validation to develop an AIoT product development process that integrates sustainability and universal design concepts. The detailed implementation steps of this process are defined in this section and shown in Figure 2.

3.1. Phase 1: Conceptual Framework and Requirements Definition

Step 1

This study is based on prior literature analysis, combining the principles of sustainability and universal design to establish the overall values and design evaluation criteria required for AIoT product development. Given the highly intelligent, adaptive, human–machine interactive, and data-driven characteristics of current AIoT products, traditional universal design indicators are no longer fully aligned with the actual usage scenarios of modern products [56]. Therefore, this study revisits and adjusts the original universal design achievement indicator (Product Performance Program (PPP)) evaluation criteria (as shown in

Table 1. Traditional universal design achievement PPP evaluation indicators.

Seven Principles and Three Supplementary Provisions	Thirty-Seven Universal Design Evaluation Indicators
Principle 1: Fair Use	UD 01 Equal use
	UD 02 Eliminate feelings of difference
	UD 03 Offering choices
	UD 04 Eliminate anxiety
Principle 2: Flexible Use	UD 05 Freedom of use
	UD 06 Accepting left-handed and right-handed people
	UD 07 Proper use in emergency situations
	UD 08 Usability under changing conditions
Principle 3: Simple and easy to understand	UD 09 Not overly complicated
	UD 10 Intuitive to use
	UD 11 Easy to use and understand
	UD 12 Operating tips and feedback
	UD 13 Easy to understand structure
Principle 4: Easy to identify information	UD 14 Provide multiple means of information transmission
	UD 15 Processed obstacle operation information
Principle 5: Allow for mistakes	UD 16 Considerations for preventing danger
	UD 17 Preventing accidents
	UD 18 Ensuring safety even when used incorrectly
	UD 19 Even if you fail, you can return to the status quo
Principle 6: Reduce physical strain	UD 20 Can be used in a natural posture
	UD 21 actions that exclude meaning
	UD 22 Low physical load
	UD 23 No fatigue even after prolonged use
Principle 7: Accessibility and operating space	UD 24 Maintain easy-to-use space and size
	UD 25 Suitable for users of all body types
	UD 26 Caregivers can use it together
	UD 27 Easy to transport and convenient to store
Supplementary provision 1: Economical for long-term use	UD 28 Consider using durability
	UD 29 Reasonable price
	UD 30 Economy in continuous use
	UD 31 Easy to maintain and repair
Supplementary provision 2: High quality and aesthetically pleasing	UD 32 Comfortable and attractive to use
	UD 33 Satisfactory quality
	UD 34 Effective use of materials
Supplementary provision 3: Harmless to humans and the environment	UD 35 Harmless to humans
	UD 36 Harmless to the natural environment
	UD 37 Promoting recycling and reuse

) [49,50], including the 7 principles, 3 supplementary provisions, and 37 universal design evaluation indicators, to reconstruct an evaluation framework that better aligns with contemporary needs.

This study employed the modified Delphi method [57] as the primary initial research strategy, integrating the characteristics of expert questionnaire surveys and consensus judgments from a core expert panel. This method is suitable for research objectives such as adjusting and simplifying evaluation indicators and constructing conceptual frameworks. First, the research team designed an online questionnaire and invited 10 domain experts with backgrounds in the AIoT, product design, and sustainable design to assess the necessity and appropriateness of 37 general design evaluation indicators, as well as to provide suggestions for additions, deletions, mergers, or revisions. These experts were selected based on two key criteria: (1) professional training or practical experience in the AIoT, product design, or sustainable/universal design; (2) a minimum of five years of research or practical experience in related fields to ensure the validity and reliability of the evaluation results. The questionnaire also included open-ended fields to collect specific revision suggestions and potential new indicator recommendations. The collected data were preliminarily organized and categorized by the research team to form the basis of this analysis.

Furthermore, three experts with over 10 years of experience in product design were invited to form an “Expert Panel” [58] to conduct in-depth interpretation and discussion of the questionnaire results, clarify differences in expert opinions, and revise or semantically reconstruct indicators that were repetitive, ambiguous, or no longer aligned with current circumstances. This stage applied the Expert Consensus Method, emphasizing high-level integration based on expert criteria to ensure that indicator content balanced theoretical foundations with practical applicability. The original 37 universal design evaluation indicators were reorganized, and an enhanced universal design achievement evaluation framework (Enhanced Product Performance Program for AIoT (ePPP-AIoT)) was designed, making it more focused on the practical needs of AIoT product characteristics. This result also served as the basis for the subsequent implementation of the QFD (Quality Function Deployment) method for design requirement conversion and weighted analysis, further supporting the practice of universal design-oriented AIoT product development processes.

3.2. Phase Two: Design Concept and Decision Analysis

3.2.1. Step 2

This study adopted the Quality Function Deployment (QFD) method as its development process framework, incorporating the aforementioned five principles as user design requirements on the left wall of the House of Quality (HOQ) in QFD. When using QFD, to address the relative importance of design requirement indicators, quantitative tools are often used in conjunction. The Analytic Hierarchy Process (AHP) is an effective expert decision analysis tool [59] that can be used for calculating the importance of each indicator [60]. This method is suitable for calculating the weights of design factors and establishing clear design guidelines based on these weights [61]. It systematizes complex problems by decomposing them into hierarchical levels from different perspectives, using quantitative calculations to identify patterns and establish the importance weights of the five design requirement criteria. In this study, a method known as Aggregating Individual Priorities (AIPs) was applied in the AHP procedure. By using the arithmetic mean method, the individual weight vectors obtained from each expert were aggregated into a group weight, ensuring that the independent judgments of all experts were equally respected and integrated into the final results. The application of the AHP method involved the following five steps:

• Questionnaire development and distribution:

The five design requirement criteria were developed into an AHP questionnaire, and an expert panel was invited to complete the questionnaire. Through the questionnaire, the weightings of the importance of the five design requirement criteria could be determined.

• Converting Questionnaire Results:

We used Equation (1) to convert the questionnaire results into a pairwise comparison matrix. In the below equation, $a_{ij}$ refers to the comparison scale between the $i$-th design requirement and the $j$-th design requirement:

$$\mathrm{A}={\left[a_{ij}\right]}_{5 \times 5}= \left[\begin{array}{cccc}1& a_{12}& \cdots & a_{15}\\ {{\displaystyle \frac{1}{a}}}_{12}& 1& \cdots & a_{25}\\ ⋮& ⋮& 1& ⋮\\ {{\displaystyle \frac{1}{a}}}_{15}& {{\displaystyle \frac{1}{a}}}_{25}& \cdots & 1\end{array}\right]$$

(1)

• Calculate the weights of the design requirement criteria:

We used the normalization of the geometric mean of the rows (Equation (2)) to calculate the weights of the design requirement criteria:

$$w_i={\displaystyle \frac{{\left(\prod _{j=1}^5{a}_{ij}\right)}^{\frac{1}{5}}}{\sum _{i=1}^5{\left(\prod _{j=1}^5{a}_{ij}\right)}^{\frac{1}{5}}}} , i=1,2,\cdots ,5$$

(2)

• Consistency Test:

To confirm the validity of the expert questionnaire, a consistency test was conducted on the questionnaire results. First, the maximum eigenvalue ${\lambda }_{max}$ was calculated using Equation (3). Next, the consistency index (C.I.) was calculated using Equation (4) $(n=5)$. Finally, the consistency ratio (C.R.) was calculated using Equation (5). According to

Table 2. Random index (R.I.).

n	1	2	3	4	5	6	7	8	9	10
R.I.	0.00	0.00	0.58	0.90	1.12	1.24	1.32	1.41	1.45	1.49

, the random index (R.I.) was set to 1.12$(n=5)$. When C.R. < 0.1, this indicated that the questionnaire passed the test [60].

$${\lambda }_{max}={\displaystyle \frac{1}{5}}\left[\left({\displaystyle \frac{w_1^{\prime }}{w_1}}\right)+\left({\displaystyle \frac{w_2^{\prime }}{w_2}}\right)+\cdots +\left({\displaystyle \frac{w_5^{\prime }}{w_5}}\right)\right]$$

(3)

$$\mathrm{C}.\mathrm{I}.={\displaystyle \frac{{\lambda }_{max}-n}{n-1}}$$

(4)

$$\mathrm{C}.\mathrm{R}.={\displaystyle \frac{C.I.}{R.I.}}$$

(5)

• Weight integration:

Finally, we used Equation (6) to integrate the weights of the $k$ experts. In the below equation, ${\overline{w}}_i$ refers to the importance weight of the $i$-th design requirement criterion after integrating the opinions of the $k$ experts, while $w_{ik}$ refers to the weight value calculated based on the opinion of the $k$ expert in the $i$-th design requirement:

$${\overline{w}}_i=\left({\displaystyle \frac{w_{i1}+w_{i2}+\cdots +w_{ik}}{k}}\right),i=1,2,\cdots ,5$$

(6)

3.2.2. Step 3

In the field of technical measures in Quality Function Deployment (QFD), this study combined brainstorming with scenarios for expert panel discussions. The overall process was conducted in three rounds, with each limited to 30 min: Round 1 (Idea Generation)—The research team and experts individually brainstormed and proposed diverse story drafts, using an anonymous approach to avoid influence or dominance by individual members. Round 2 (Consolidation and Refinement)—Guided by a neutral host, the expert panel consolidated, categorized, and refined the preliminary drafts to derive candidate scenarios. Round 3 (Consensus and Convergence)—The experts engaged in in-depth discussions regarding the candidate scenarios. The representative scenarios were determined when no new significant ideas emerged and consensus was reached on their representativeness. This process produced a final set of representative scenarios.

Following the determination of scenarios, the experts further identified design pain points for each role across the scenarios and mapped them to the universal design evaluation indicators. On this basis, specific creative solutions and design strategies were proposed. The final results were represented as technical measures at the ceiling of the House of Quality (HOQ), ensuring traceability between design requirements and design concepts (as shown in Figure 3).

Using Quality Function Deployment (QFD), the weights of the technical measures could be derived as follows:

Firstly, we constructed the relation matrix (R): Through expert discussions, the relationships between five design requirements and n technical measures were assessed to build the relation matrix R, as shown in Equation (7). The degree of correlation $r_{ij}$ was rated as 9 (very strong), 3 (moderate), 1 (weak), or 0 (none).

$$R={\left[r_{ij}\right]}_{5\times n}=\left[\begin{array}{ccc}{r}_{11}& \cdots & r_{1n}\\ ⋮& \ddots & ⋮\\ r_{51}& \cdots & r_{5n}\end{array}\right]$$

(7)

Next, we calculated the weights of technical measures: The design requirement weights ${\overline{w}}_i$ (obtained from Equation (6)) were substituted into Equation (8) to obtain the weight $t_j$ of each technical measure. A higher $t_j$ value indicated that the technical measure $j$ better fulfilled users’ needs and should, therefore, be prioritized in the design strategy.

$$t_j=\sum _{i=1}^5\left[{\overline{w}}_i\times t_{ij}\right],\left(j=1,2,\cdots ,n\right)$$

(8)

3.2.3. Step 4 to 6

We entered the requirements and technical items into the QFD matrix. Through correlation analysis, we evaluated the extent to which each technical item met the general design requirements, derived a comprehensive weighted score, and ranked the priority order of implementation for each technical concept. This will serve as a reference point for the design strategy for subsequent AIoT product development.

3.3. Phase Three: Design and Validation

3.3.1. Step 7

Three experts with product development backgrounds used this product development process to develop a smart home appliance.

3.3.2. Step 8

The purpose of this step was to assess the quality of the design outcomes, requiring an effective quantitative tool for analysis. The Entropy Weight Method (EWM) combined with the Fuzzy Comprehensive Evaluation (FCE) method can be applied to evaluate complex issues with multiple dimensions [62], demonstrating reliable assessment results and scientific rigor [63]. Therefore, this study adopted this method as the assessment tool for design outcomes. This study invited participants with expertise in product design and smart technology to complete an evaluation questionnaire regarding the development outcomes. The evaluation criteria for the development outcomes were the AIoT product development design criteria proposed by this study, with the following evaluation scale: very poor, poor, average, good, very good. Then, the evaluation results were established as a single-factor evaluation matrix $\tilde{D}$. Next, the Entropy Weight Method (EWM) was applied to the matrix to calculate the weights of the evaluation factors. Finally, the Fuzzy Comprehensive Evaluation (FCE) method was used to evaluate the design outcomes, thereby verifying the effectiveness of the proposed product development process. The evaluation procedure consisted of the following four steps:

• Establish evaluation factors and evaluation ratings:

This study summarized $h$ evaluation indicators as evaluation factors for assessing design results. The factor set was $B=\left\{b_1,b_2,\cdots ,b_h\right\}$. The rating level was expressed as the set $C=\left\{c_1,c_2,\cdots ,c_5\right\}=\left\{Very bad, bad, average, good, very good\right\}$.

• Convert the questionnaire results into an evaluation matrix:

First, we converted the evaluation questionnaire results into a single-factor evaluation matrix $\tilde{D}$ (Equation (9)). In the below equation, the value $d_{ij}$ in the matrix $\tilde{D}$ refers to the percentage of experts’ evaluation ratings for the $j$-th evaluation factor in response to the $i$-th evaluation factor, which serves as the membership function of FCE [64,65]:

$$\tilde{D}={\left[d_{ij}\right]}_{h\times 5}=\left[\begin{array}{ccc}{d}_{11}& \cdots & d_{15}\\ ⋮& \ddots & ⋮\\ d_{h, 1}& \cdots & d_{h, 5}\end{array}\right]$$

(9)

• Calculate entropy weight:

First, we used Equation (10) to calculate the entropy of the h-term evaluation factor $E_i(i=h)$, and when $d_{ij}=0$, $d_{ij}\times lnln \left(d_{ij}\right) =0$. Next, we used Equation (11) to calculate the entropy weight $w_i^{\left(\epsilon \right)}(i=h)$:

$$E_i=-{\displaystyle \frac{1}{\ln \left(5\right)}}\sum _{j=1}^5{d}_{ij}\times \ln \left(d_{ij}\right), i=1,2,\cdots ,h$$

(10)

$$w_i^{\left(\epsilon \right)}={\displaystyle \frac{1-E_i}{\sum _{i=1}^h\left(1-E_i\right)}} ,i=1,2, \cdots ,h$$

(11)

• Calculate the comprehensive evaluation:
  First, we calculated the entropy weight obtained in the previous step, with $w_i^{\left(\epsilon \right)}$ defined as the weighting for comprehensive evaluation. Next, fuzzy operations were performed on $\widetilde{D }$ using Equations (12) and (13) to obtain the Fuzzy Comprehensive Evaluation $\widetilde{F }$ of the design results. Specifically, Equation (13) refers to the calculation method used for composition operations in fuzzy mathematics. When calculating $f_j$, the minimum value was taken sequentially between $w_1^{\left(\epsilon \right)}$ and $d_{1j}$, then between $w_2^{\left(\epsilon \right)}$ and $d_{2j}$, continuing until $w_h^{\left(\epsilon \right)}$ and $d_{hj}$. Finally, among these selected values, the maximum was chosen:

  $$\widetilde{F }=W^{\left(\epsilon \right)}. \widetilde{D }=\left(w_1^{\left(\epsilon \right)}, w_2^{\left(\epsilon \right)},\dots , w_h^{\left(\epsilon \right)}\right). \left[\begin{array}{ccc}{d}_{11}& \cdots & d_{15}\\ ⋮& \ddots & ⋮\\ d_{h, 1}& \cdots & d_{h, 5}\end{array}\right]=\left(f_1, f_2,\dots , f_5\right)$$

  (12)

  $$f_j=\underset{i=1}{\overset{h}{\bigvee }}(w_i^{\left(\epsilon \right)}\wedge d_{ij}), (\mathrm{j}=1,2,\dots ,5)$$

  (13)

4. Results and Discussion

4.1. Phase 1 Experimental Results

Design Requirements

This study systematically reorganized the 7 principles, 3 supplementary provisions, and 37 universal design evaluation indicators included in the traditional PPP based on a modified Delphi method and a consensus judgment provided by an expert integration group. The research findings indicate that while the existing framework retains theoretical value in the realm of universal design, traditional indicators are insufficient for fully capturing the actual application scenarios of AIoT products, which exhibit highly intelligent, adaptive, human–machine interactive, and data-driven characteristics. After evaluation and integration by the expert group, redundant or overly traditional items were eliminated, overlapping content was consolidated and expanded, and new indicators necessary for addressing AIoT characteristics were added. Recent studies (e.g., Zallio et al., 2025, Kronlid, 2024) also emphasize that inclusive AIoT design must integrate equity, accessibility, and safety considerations, reinforcing the need to introduce new indicators that address diverse user needs and sociotechnical challenges [66,67]. Ultimately, this study proposes an enhanced universal design achievement evaluation framework (ePPP-AIoT), which encompasses five major design requirements and thirty evaluation indicators (as shown in

Table 3. Enhanced Product Performance Program for AIoT (ePPP-AIoT).

Design Requirements	Evaluation Indicators
A1 Fairness	b01 Accepting users of different age groups
	b02 Accepting different gender groups
	b03 Accepting people with different physical conditions
	b04 Providing personalized choices and operational flexibility
	b05 Eliminate anxiety
	b06 Eliminate feelings of difference
A2 Ease of use	b07 Easy to use and understand, not overly complicated
	b08 Intuitive to use
	b09 Can automatically learn and optimize
	b10 Operating tips and feedback
	b11 Provide multiple means of information transmission
	b12 Organized and categorized operational information
A3 Safety	b13 Preventing dangerous situations from occurring
	b14 Focus on network security to prevent data leaks
	b15 Even if you fail, you can return to the status quo
	b16 Correct usage in emergency situations
	b17 Reduce repetitive movements
	b18 Reduce operational fatigue
A4 Compatibility	b19 Different users and caregivers can jointly manage/use smart products
	b20 The system’s operating methods and interface design are consistent with the user’s cognitive model
	b21 The shape, size, and layout of the product should be accessible to the user’s body structure and range of motion
	b22 The ability to exchange information with other systems or devices for data integration
	b23 The operating direction of the control device is consistent with the natural movement direction of the user
	b24 The response speed of the system or equipment matches the user’s perception of time
A5 Sustainability	b25 Collect and recommend information appropriately based on user preferences
	b26 Economy during continuous use (including durability and maintenance considerations)
	b27 Comfortable and attractive to use
	b28 Satisfactory quality
	b29 Harmless to humans and the environment
	b30 Promoting recycling and reuse

).

The five key design requirements are fairness, usability, security, compatibility, and sustainability. Among these, “fairness” has been further refined in this study, building on traditional concepts such as “equal access”, “eliminating feelings of discrimination”, and “reducing anxiety”. Additionally, new indicators have been introduced to assess the acceptability of AIoT technology for users of different ages, genders, and physical conditions, addressing the challenges of digital divides and inclusive use across diverse populations regarding AIoT technology. Additionally, the traditional notion of “providing choices” has been expanded to “personalized operations and flexibility”, emphasizing the need for AI systems to offer differentiated service capabilities. The “usability” requirement consolidates overlapping items such as “simple and understandable”, “intuitive operation”, and “structural understanding”, distilling them into items necessary for ensuring that system operations are simple, are intuitive, and do not impose excessive cognitive burdens. Additionally, the new framework introduces the requirement for “automatic learning and optimization”, reflecting the adaptive nature of AI systems, while retaining elements such as operation prompts, real-time feedback, diverse information communication, and message categorization and organization, ensuring that the product remains clear and usable in highly intelligent scenarios. In terms of “safety”, traditional PPP focuses on hazard prevention, fault tolerance, and error recovery mechanisms. The new architecture retains these foundational elements while adding “emphasizing network security and preventing data leakage”, addressing the high reliance on data exchange in AIoT products. This requirement also integrates indicators such as “correct use in emergency situations” and “reducing repetitive actions and operational fatigue”, ensuring that users not only receive physical safety guarantees but also establish trust and protection in the digital realm. Compatibility is a key breakthrough in this architectural adjustment. Traditional PPP focuses on product spatial and physical compatibility, as well as the possibility of shared use between caregivers and users. ePPP-AIoT expands on this foundation to include cross-system and cross-platform compatibility. For example, the requirement that “different users and caregivers can jointly manage/use the system” has been retained and strengthened, but new items such as “the system can interoperate with other devices and integrate data” and “the response speed must align with user time perception” have been added to address complex interaction needs in smart home and multi-device networking environments. The “sustainability” requirement integrates traditional supplementary content, reorganizing factors such as “durability”, “economy”, “maintainability”, “comfort and aesthetics”, and “quality” while removing the overlapping “appropriate price” indicator and adding the “collecting and recommending information based on user preferences” indicator to address the data-driven needs of AI systems during long-term use. This requirement retains sustainability-related indicators such as “harmless to humans and the environment” and “promoting recycling and reuse”, enabling the new framework to balance the social and environmental value of the product lifecycle. This study refines the traditional 37 PPP indicators into 30 final indicators, and detailed reconstruction mapping of these indicators is presented in Figure 4 and Appendix C.

Compared to traditional PPP, ePPP-AIoT constructs a set of evaluation criteria more aligned with contemporary smart products and sustainability requirements by eliminating redundant items, consolidating overlapping concepts, and adding new indicators related to AIoT characteristics. This served as the input for the left-hand wall design requirements in the next phase of the development process.

4.2. Results of the Second-Phase Experiment

4.2.1. Calculation Results of Design Requirement Weights

This study invited 10 experts to complete a questionnaire, which was analyzed using the Analytic Hierarchy Process (AHP) to calculate the weights of the five design requirements. First, the questionnaire results were converted into a pairwise comparison matrix, as shown in Appendix A. Next, Equation (2) was used to calculate the weights of the design requirements, as shown in

Table 4. Weighting of design requirements.

Expert No.	1	2	3	4	5	6	7	8	9	10
A1	0.098	0.067	0.192	0.293	0.263	0.217	0.102	0.253	0.254	0.075
A2	0.256	0.182	0.275	0.099	0.242	0.249	0.393	0.142	0.256	0.293
A3	0.470	0.377	0.253	0.280	0.278	0.270	0.327	0.430	0.366	0.464
A4	0.086	0.174	0.146	0.068	0.105	0.100	0.060	0.099	0.056	0.086
A5	0.091	0.200	0.134	0.259	0.112	0.164	0.118	0.075	0.069	0.081

.

Next, consistency tests were conducted using Equations (3)–(5), as shown in

Table 5. Consistency test results.

Expert No.	1	2	3	4	5	6	7	8	9	10
λmax	5.240	5.305	5.318	5.124	5.438	5.173	5.193	5.422	5.435	5.039
C.I.	0.060	0.076	0.080	0.031	0.110	0.043	0.048	0.106	0.109	0.010
C.R.	0.053	0.068	0.071	0.028	0.098	0.039	0.043	0.094	0.097	0.009

. The analysis confirmed that all 10 expert questionnaires were valid, and all consistency ratios (C.R.) were below the acceptable threshold of 0.1. Whenever values approached this limit, the corresponding expert judgments were carefully reviewed.

Finally, Equation (6) was used to integrate the weights of the 10 experts, and the results are shown in

Table 6. The weightings of the integrated design requirements.

Design Requirements	A1	A2	A3	A4	A5
Integrated weighting	0.181	0.239	0.352	0.098	0.130

. From the data results, the weight values of design requirements A3 and A2 are 0.352 and 0.239, respectively, making them the two most important design development indicators.

4.2.2. Converting Design Requirements into Specific Technical Measures

To further validate the feasibility of the enhanced universal design achievement evaluation framework (ePPP-AIoT) for use in practical applications, three experts in the field of product design were invited to jointly discuss and establish five everyday life scenarios (as shown in

Table 7. Setting typical scenario stories.

No.	Theme	Application Scenario	Design Pain Points and Scenario Diagram
S1	Mom’s health management assistant	Mrs. Chen is a full-time mother who is busy taking care of her family and her children’s diet and health. She uses this smart refrigerator to track the freshness of ingredients and plan daily menus. Mrs. Chen opened the touch panel on the refrigerator door and found that the application showed that the chicken label inside the refrigerator had turned yellow, indicating that it was about to expire. The refrigerator immediately sends a notification reminding her to use this batch of chicken today. She quickly checks the freshness status of other vegetables in the system to ensure she can pair them with healthy side dishes. She is grateful that this refrigerator helps her to manage ingredients effortlessly, avoid waste, and ensure that her family enjoys fresh, healthy meals.	Traditional solutions cannot meet user needs. s1-1 automatically detects the shelf life of ingredients and provides timely reminders to avoid waste caused by forgetting. s1-2 provides ingredient pairing suggestions to help users quickly plan healthy meal plans.
S2	Father’s remote collaboration	Mr. Zhang frequently travels for work, but he still wants to help his wife and children keep track of the contents of the refrigerator, especially the children’s snacks and daily healthy beverages. While working away from home, Mr. Zhang used a mobile app to remotely check the contents of the refrigerator and discovered that his children’s favorite yogurt was almost gone. He marked the yogurt with a reminder label and set up a notification to remind himself to buy new yogurt when he had time. On the other hand, his wife could also ask him to restock the fridge with new ingredients. This refrigerator allows Mr. Zhang to stay informed about the situation at home even when he is away, enabling him to interact and collaborate with his family.	Traditional solutions cannot meet user needs. s2-1 supports remote real-time viewing of refrigerator contents, solving the problem of not being able to check the status of food ingredients when away from home. s2-2 sets tags and notifications to help family members to collaborate and reduce shopping omissions.
S3	My daughter’s culinary creativity	Xiao Mei, the daughter of the Zhang family, is a high school student who loves cooking. She often creates new dishes based on the ingredients available at home. Xiao Mei gently taps on the refrigerator door, uses the scanning function to check what ingredients are left inside, and searches for remaining ingredients using the keyword search function on the touchscreen panel. She discovers some green peppers and steak, immediately clicks on the smart function to look up recipes, and begins creating a new dish. This refrigerator allows Xiao Mei’s culinary creativity to no longer be limited by whether she remembers what ingredients are in the refrigerator, enabling her to improvise based on the current availability of ingredients at any time.	Traditional solutions cannot meet user needs. s3-1’s scanning and search functions make ingredient lists transparent, reducing the burden of memory and searching. s3-2 is paired with a recipe recommendation system to assist in improvisational cooking and enhance the cooking experience.
S4	Grandparents who are not familiar with technology can use it without stress.	The Zhang family’s grandparents live together, but they are not familiar with smart devices and worry that a smart refrigerator might be difficult for them to operate. Before opening the refrigerator, when the grandmother touches the door handle, the refrigerator door displays information about the contents inside, allowing her to see at a glance which foods are still fresh. Since the interface is simple, the grandmother only needs to touch the panel to check the expiration dates of the ingredients and use or discard them based on color-coded labels. The smart refrigerator requires no complicated operations, and even if she ignores this information, it does not affect the operation. It does not change her usage habits, making her feel stress-free when using it, and also helps her reduce food waste.	Traditional solutions cannot meet user needs. s4-1 features a simple and intuitive interface design, using color coding to help determine the freshness of ingredients. s4-2 employs touchscreen technology to reduce complex operations and lower barriers to use for the elderly.
S5	Children’s breakfast reminder	Xiao Ming is an elementary school student who often forgets to eat breakfast in the morning due to his busy schedule, and his mother is concerned about his nutritional balance. His mother set up reminder tags for his breakfast foods (such as milk and bread) in the app. When Xiao Ming wakes up in the morning, the refrigerator automatically sends a notification to remind him, and the touchscreen clearly displays the location of the foods he needs for breakfast. This smart refrigerator’s thoughtful feature ensures that Xiao Ming can enjoy his mother’s carefully prepared breakfast every day before school.	Traditional solutions cannot meet user needs. s5-1 automatically sends reminders to ensure that children do not skip breakfast. s5-2’s clear positioning on the touch panel allows children to quickly find the locations of ingredients.

) as the foundation for design thinking. These scenarios encompassed home cooking, use of smart appliances by the elderly, parent–child shared devices, multi-user cross-device interaction, and sustainable consumption and recycling, thereby reflecting the complex usage contexts that AIoT products may encounter in real-world operations. To illustrate the convergence process of the three rounds of expert discussions, this study systematically organized the sixteen initial drafts (R1-S), which were consolidated and refined into six categories of candidate scenarios (R2-S), from which five representative scenarios (S) were ultimately established. Detailed mapping of this convergence process is provided in Appendix D. These representative scenarios not only highlight the complexity and diversity of AIoT product applications but also serve as the foundation for subsequent design requirement conversion and the development of technical measures. Through this method, the experts re-examined the five design requirements represented by the left wall of the Quality House—fairness, usability, safety, compatibility, and sustainability—within specific real-life contexts. These abstract design requirements were then gradually transformed into actionable creative concepts (technical measures). For example, in the “elderly using smart appliances” scenario, the fairness requirement was transformed into larger font sizes and voice prompts in the interface. This aligns with empirical evidence (Zhou et al., 2022), which shows that elderly users significantly benefit from larger fonts, voice feedback, and simplified interfaces, thereby validating our design choice [68]. In the “parent–child shared device” scenario, safety is concretized as user permission levels and error recovery mechanisms. This process helped the expert group to clarify the connection between design requirements and technical implementation and extends the construction of the Quality House from a theoretical level to concrete design strategies that can be adopted by product teams. Creative ideas discussed and compiled by the expert team included the following: G1: Clearly displaying the expiration date of food items in text on the interface. G2: Using color labels (green, yellow, red) to intuitively indicate the expiration status of food items on the interface. G3: Automatic push notifications for food items nearing expiration are sent to the user’s phone. G4: Real-time synchronization of food item information between the refrigerator display and the mobile app. G5: Use RFID (Radio Frequency Identification) food clips to enable the refrigerator to automatically identify and record the types and quantities of ingredients. G6: The refrigerator automatically scans and updates internal photos every time the door is opened or closed. G7: The refrigerator display and mobile phone can synchronize ingredient inventory management in both directions. G8: The refrigerator is designed with a shallow depth for easy access to ingredients. G9: The refrigerator is designed with a deep storage capacity to increase ingredient storage volume. G10 The system can be connected to online food ordering platforms for convenient restocking. G11: Users can use smart features to view recommended recipes based on remaining ingredients. G12: The system can be connected to online cooking tutorial videos to assist users in cooking. G13: The transparent refrigerator door design allows users to directly view the contents. G14: Users can view the contents list or images via the refrigerator display or mobile app. G15: The refrigerator display interface shows no more than seven items of information at a time to maintain clarity. G16: The refrigerator offers smart voice control functions (such as queries or reminders). G17: The display interface automatically adjusts to a suitable height based on the user’s height. G18: All smart functions are designed without affecting the traditional refrigerator’s basic usage methods. G19: Digital memo labels are provided for reminders, making it easy to mark messages specific to family members. G20: Digital labels have intelligent tracking and recognition functionality, enabling them to follow items without confusion. G21: The system has user permission tiering functionality, restricting operation scope based on different family members. G22: A misoperation recovery mechanism is provided, enabling quick restoration to the previous step or default state.

4.2.3. Design Decisions

Due to the weighting calculation of the House of Quality, the product development requirements (left wall) and the technical measures (ceiling) were listed in the matrix. The correlations between the two were analyzed and evaluated by experts. Based on this process, the priority order of the technical measures was determined. In this study, the weight values served as the reference criterion, and through expert discussions, technical measures with higher weights (G22, G7, G20, G4, G6, G15, G17, G21, G2, and G11) were prioritized for further development. These measures were identified as the main development strategies for subsequent AIoT product design, as illustrated in Figure 5. Specifically, this study adopted a quartile-based criterion: technical measures within the first quartile (top 25% of cumulative weights) were prioritized as the primary set for development; those within the second quartile (25–50%) were considered as secondary options; and those within the third quartile (50–75%) were further discussed for their potential applicability in complementary or supportive roles.

4.3. Design Outcomes

In the era of smart IoT, home appliances are evolving toward greater intelligence and humanization, redefining the role of refrigerators. However, most existing smart refrigerator designs are limited to the concept of expensive “built-in cameras”, which not only fail to address users’ daily needs from their perspective but also make the user experience complex and impractical. This design returns to the core value of refrigerators—assisting families in monitoring food conditions and extending food freshness—while incorporating the concepts of “equitable use” and “sustainability” to create a smart refrigerator that balances convenience, inclusivity, and environmental responsibility. Through smart scanning technology, users can instantly view inventory types and quantities without opening the door, and information is automatically updated when the door is opened, reducing the number of duplicate purchases and forgotten items. Unlike conventional smart refrigerators that rely on costly multi-camera or continuous video systems, this design employs a single-image capture method triggered each time the door is opened. This low-cost approach not only reduces hardware and computational expenses but also maintains real-time efficiency. The captured image provides a clear and intuitive snapshot of the refrigerator’s contents, displayed on both the external interface and the mobile application. Additionally, the system automatically compares new and previous images to detect added or removed items, ensuring that inventory data remain synchronized without requiring manual input. This allows users of all ages and digital literacy levels to use the system intuitively and effortlessly. The user-friendly interface features simple and clear information design, ensuring that all family members, regardless of gender, age, or physical condition, can equally benefit from the smart features. Furthermore, the “Food Expiration Date Guardian System” uses green, yellow, and red labels with real-time reminders to effectively reduce food waste and promote sustainable resource utilization. The dedicated mobile app also offers a “Note Label” feature, allowing family members to share and allocate ingredients, enhancing interaction and collaboration. This design integrates smart management, fair use, and environmental friendliness, not only addressing the practical needs of modern families but also embodying the future value of smart appliances that balance technological innovation, social inclusion, and sustainable development, as shown in Figure 6.

4.4. Validation of Design Outcomes

A total of 40 experts completed the evaluation questionnaire in this study. Among them, 31 (77.5%) specialized in industrial design (e.g., product development, 2D computer graphics/3D modeling software, materials and manufacturing processes, mechanisms, and ergonomics), while 9 (22.5%) specialized in smart technology (e.g., information technology, big data analysis, the Internet of Things (IoT), and the artificial intelligence industry). Basic information about the experts is shown in

Table 8. Basic information about the experts.

Gender	Male	22 (55%)
	Female	18 (45%)
Age	21–30 years old	11 (27.5%)
	31–40 years old	20 (50%)
	41–50 years old	7 (17.5%)
	51–60 years old	2 (5%)
Education Level	University/technical college	1 (2.5%)
	Master’s degree	15 (37.5%)
	Doctorate	21 (52.5%)
	Others	3(7.5%)
Specialization Time	Within 5 years	3 (7.5%)
	6–10 years	14 (35%)
	11–15 years	12 (30%)
	16–20 years	7 (17.5%)
	Over 21 years	4 (10%)

.

This study’s evaluation method for design results combined the Entropy Weight Method and Fuzzy Comprehensive Evaluation method. First, Equation (6) was used to convert the expert evaluation results into a single-factor evaluation matrix, as shown in Appendix B. From the data results, most of the expert evaluation opinions were “good” evaluations. It should be noted that the fuzzy comprehensive evaluation in this study did not employ triangular or trapezoidal fuzzy numbers. Instead, the analysis used deterministic values derived directly from entropy-based weights and expert voting percentages. Accordingly, the results were already expressed as crisp values without requiring an additional defuzzification step, a practice that has also been adopted in previous studies [65,69].

Next, we used Equation (9) to evaluate the entropy of the factors and Equation (10) to calculate the entropy weight $w_i^{\left(\epsilon \right)}$, and the results are shown in

Table 9. Entropy and Entropy Weight of Evaluation Factors.

$\mathit{i}$	${\mathit{E}}_{\mathit{i}}$	${\mathit{w}}_{\mathit{i}}^{\left(\mathit{\epsilon }\right)}$
1	0.576	0.032
2	0.729	0.020
3	0.677	0.024
4	0.528	0.035
5	0.576	0.032
6	0.506	0.037
7	0.541	0.034
8	0.524	0.035
9	0.546	0.034
10	0.420	0.043
11	0.663	0.025
12	0.503	0.037
13	0.653	0.026
14	0.660	0.025
15	0.589	0.031
16	0.633	0.027
17	0.551	0.033
18	0.466	0.040
19	0.512	0.036
20	0.433	0.042
21	0.565	0.032
22	0.512	0.036
23	0.501	0.037
24	0.418	0.043
25	0.637	0.027
26	0.433	0.042
27	0.526	0.035
28	0.365	0.047
29	0.685	0.023
30	0.648	0.026

.

Next, we used $w_i^{\left(\epsilon \right)}$ as the weight for Fuzzy Comprehensive Evaluation and Equations (11) and (12) to obtain an expert comprehensive evaluation of the design results. Finally, the comprehensive evaluation value $f_j$ was standardized, as shown in

Table 10. A comprehensive evaluation of the design results.

	c1	c2	c3	c4	c5
$f_j$	0.000	0.037	0.043	0.047	0.047
Standardization	0.000	0.212	0.247	0.270	0.270

. The data results indicate that the evaluation was 0% for “very poor”, 21.2% for “poor”, 24.7% for “average”, 27% for “good”, and 27% for “very good”. According to the Maximum Degree of Membership principle, this design outcome can be considered to represent an evaluation of ‘good’ or “very good”.

4.5. Discussion and Findings

Through literature review and expert discussion, this study found that current AIoT products still face hidden challenges in terms of “friendliness” and “fairness”. While traditional universal design methods centered on multi-dimensional fairness have made significant contributions in the past, they are no longer sufficient to fully address the highly intelligent, interconnected, and continuously evolving characteristics of AIoT products. Similar arguments are raised by Song et al. (2024) and Zallio et al. (2025), who highlight that inclusive AI design frameworks must dynamically adapt to diverse user contexts and sustainability challenges, supporting the inclusion of indicators such as “automatic learning and optimization” and “personalized operations” in the ePPP-AIoT framework [66,70]. Investigating quality aspects for the UX evaluation of IoT-based systems identifies qualities such as response speed [71], interface consistency, and multimodality in information transmission as crucial UX dimensions that older universal design frameworks often under-represent. Similarly, the study of HK Adli et al., 2023 highlights automatic learning, data integration, and adaptation to diverse user contexts as emerging requirements [71,72]. Based on this, this study adopted a modified Delphi method, integrating expert questionnaires and a core expert group, to re-examine and adjust the 7 principles, 3 supplementary provisions, and 37 universal design evaluation indicators originally covered by PPP, thereby constructing an enhanced universal design achievement evaluation framework (ePPP-AIoT) that better aligns with the needs of contemporary AIoT products, as shown in Figure 7. This framework encompasses five design requirements and thirty design criteria, laying the foundation for subsequent method integration.

In terms of methodology, this study combines Quality Function Deployment (QFD), the Analytic Hierarchy Process (AHP), the Scenario Method, and Fuzzy Comprehensive Evaluation with Entropy Weighting (FCE-EW) to propose an AIoT product development process centered on sustainability and universal design principles. The Scenario Method, which is an analytical tool commonly used in product design to simulate real-life situations and explore authentic user needs [73,74], was introduced to complement QFD. Prior studies have demonstrated that scenarios can be effectively integrated with QFD as an industrial innovation decision-making tool [75], and further research has shown that scenario building can serve as an ergonomics method to explore user needs and support design development [73]. Additionally, QFD is often applied in combination with other methods to overcome its methodological limitations [76], and recent work has proposed quantifying user requirements through scenarios to further strengthen its application in product–service design [77]. Furthermore, the introduction of FCE-EW into the AIoT product development process provides a systematic way to evaluate and validate early-stage design concepts and outcomes. As shown in Liu et al. (2023), the method enables rapid, structured assessment by integrating expert opinions with entropy-based weighting, thereby improving both the efficiency and reliability of design verification [65]. This process specifically incorporates the Scenario Method and brainstorming techniques to address the limitations of traditional QFD in user experience design and integrates user scenario requirements into the “ceiling” of the House of Quality (HOQ), as shown in Figure 8, thereby aligning the design more closely with real-world usage contexts. Research results show that design teams following this process for concept proposals and validation achieved outcomes that not only aligned with expert recommendations but also met key design requirements, demonstrating excellent evaluation outcomes. The proposed AIoT product development process based on sustainability and universal design concepts not only provides an evaluation basis for design teams during the concept development stage but also facilitates early collaboration among cross-departmental experts, enabling effective communication and integration from the outset of product development. This advantage helps to ensure that design decisions simultaneously address usability, inclusivity, and sustainability, making the development process more forward-looking and comprehensive. Through such a mechanism, development teams can more efficiently filter design concepts, reduce redesign costs at later stages, and enhance practical market acceptance of the resulting products. Adli et al. (2023) pointed out that AIoT applications must incorporate adaptive learning and sustainable integration mechanisms to remain competitive in complex environments [72]. This aligns with our proposition that embedding sustainability considerations at the concept development stage can enhance long-term market viability. Furthermore, Choma and Zaina (2025) emphasized that multimodality, interface consistency, and responsiveness are critical UX qualities influencing acceptance across heterogeneous user groups [71]. This resonates with our argument that systematically integrating multimodal design into the early stages of product development helps with proactively addressing usability and inclusivity issues, rather than deferring them to late-stage adjustments. In addition, Kruhlov et al. (2025) provided empirical evidence on social inclusivity in smart city governance, underscoring that inclusivity for elderly and vulnerable groups must be treated as a structural design criterion [78]. Taken together, these studies reinforce our argument that an effective AIoT product development process should integrate sustainability, multimodality, and inclusivity from the earliest stages, ensuring both design efficiency and broader user acceptance in practice. However, the scope of validation of this study primarily focused on kitchen smart appliances, so its applicability across different contexts still needs to be expanded. It should be noted that the validation in this study relied primarily on expert evaluations within the domain of smart kitchen appliances, which limits external validity. While this approach enabled the efficient convergence of expert consensus, it may bias the results toward a specific application domain. Future studies should expand to include general users and multiple application contexts to validate the framework’s adaptability across diverse cultural and consumer environments. Future research is recommended to further explore three areas: (1) explore the application potential of this development process in other AIoT product categories (such as smart healthcare, sustainable living, wearable devices, etc.) and optimize the structural relationship between requirements and indicators through quantitative empirical validation; (2) expand the research subjects from design or technology experts with professional backgrounds to general users and validate the findings by integrating regional cultural and consumer behavior differences to enhance global adaptability and localization flexibility; (3) given the rapid development of generative AI, future considerations may include integrating AI design assistance tools into the development process to enhance efficiency, creative quality, and decision-making accuracy, thereby further expanding the innovation and foresight of AIoT product development processes. The process proposed in this study focuses on the front-end concept development stage of product design, aiming to assist development teams in establishing an initial design direction. It should be acknowledged that most experts involved in this study were from the design field, which may introduce disciplinary bias and thus affect external validity. Building upon this, future studies should further incorporate empirical validation with end-users (e.g., elderly participants, caregivers, and children) after developing functional prototypes. Both subjective and objective indicators should be applied: the System Usability Scale (SUS), a 10-item standardized and internationally validated questionnaire, should be used to measure perceived usability [79]; meanwhile, Task Completion Time should serve as an objective performance metric to evaluate operational efficiency and intuitiveness [80]. This dual approach will enhance external validity, strengthen inclusivity, and ensure that the framework’s applicability extends beyond expert-driven evaluation to real-world user contexts. It is recommended that future studies adopt a “two-layer review mechanism” to maintain the long-term effectiveness and forward-looking nature of the product development process. First, at the technology and product features level, a rapid review should be conducted every 2–3 years to respond to emerging technologies and user needs. Second, at the overall framework level, a comprehensive review is recommended every 4–5 years, referencing the revision cycles of ISO/IEC international standards and the review mechanisms of the United Nations Sustainable Development Goals (SDGs), to ensure continuous alignment with long-term industrial and societal development. It is recommended that future research consider adopting fixed numbers of experts across different subgroups (e.g., age, gender) to facilitate systematic comparisons and cross-group analysis. After developing a functional prototype, future research should establish a set of measurable validation indicators and conduct prototype testing to strengthen the external validity of the framework. For safety, suggested indicators include failure rates, abnormal event logs, and compliance with international standards (e.g., IEC 61508, ISO/IEC 62368-1). For privacy, indicators should involve GDPR/CCPA compliance, encryption and anonymization rates, and user perceptions of data security [81,82,83]. For sustainability, quantifiable measures such as energy consumption, carbon footprint, recycled material content, and lifecycle assessment (LCA) outcomes are recommended, directly aligned with the United Nations Sustainable Development Goals (SDGs 7, 12, and 13) [84]. Within the proposed product development process, the weighting calculation of the House of Quality allows for prioritizing technical measures. However, the actual number of technical measures to be adopted as development strategies does not follow a fixed standard. Instead, the process intentionally retains flexibility, leaving development teams to determine the final selection based on factors such as budget constraints, specific user requirements, and product positioning.

5. Conclusions

The AIoT product development process constructed in this study integrates sustainable development concepts, universal design principles, and systematic design tools, demonstrating high application potential and flexibility. By integrating the enhanced universal design achievement evaluation framework (ePPP-AIoT) reconstructed using the modified Delphi method, Quality Function Deployment (QFD), Analytic Hierarchy Process (AHP), Scenario Method, and Fuzzy Comprehensive Evaluation with Entropy Weighting (FCE-EW), this framework combines creative inspiration with decision-making precision, effectively assisting design teams in achieving balanced development across the concept generation, functional decision-making, and validation stages.

The research results show that this process can enhance design teams’ creative thinking and technical integration, validating its potential as a new-generation smart product design tool. This study particularly emphasizes the critical role of the five major design requirements in AIoT product development. These aspects not only affect users’ acceptance of the product but also directly impact its social value and cultural inclusivity. Therefore, incorporating diverse user perspectives and sustainability thinking from the early stages of design can effectively avoid the one-sided focus on functionality and technology present in traditional product development and help with achieving a balance between technological development and social responsibility. The main contribution of this study lies in establishing and preliminarily validating an AIoT product development process based on sustainability and universal design concepts. This process can be applied in the early stages of product development, and its practical and user-friendly characteristics make it particularly suitable for cross-departmental expert collaboration. In addition, the proposed AIoT product development process provides design teams with a tool that balances theoretical rigor and practical applicability. It not only addresses current challenges in smart product design related to inclusivity, sustainability, and user experience but also offers a concrete and feasible pathway for cross-disciplinary product development. Through this process, AIoT product development is no longer merely a technology-driven outcome but can more comprehensively embody social value, environmental responsibility, and human-centered care, thereby demonstrating the profound significance of design at the intersection of technology and humanity. Recent empirical studies (e.g., Adli et al., 2023; Choma & Zaina, 2025) likewise emphasize that automatic learning, interface consistency, user diversity, and sustainability are increasingly critical in AIoT systems, reinforcing the importance of embedding these considerations from the early design stages [71,72].