Measuring Personalized Learning in the Smart Classroom Learning Environment: Development and Validation of an Instrument

Abstract

Smart classrooms leverage intelligent and mobile technologies to create highly interactive, student-centered environments conducive to personalized learning. However, measuring students’ personalized learning experiences in these technologically advanced spaces remains a challenge. This study addresses the gap by developing and validating a Smart Classroom Environment–Personalized Learning Scale (SCE-PL). Drawing on a comprehensive literature review, content-expert feedback, and iterative item refinement, an initial pool of 48 items was reduced to 39 and subsequently to 34 following item-level analyses. Two datasets were collected from Chinese middle-school students across three provinces, capturing diverse socio-economic contexts and grade levels (7th, 8th, and 9th). EFA on the first dataset (n = 424) revealed a nine-factor structure collectively explaining 78.12% of the total variance. Confirmatory factor analysis (CFA) on the second dataset (n = 584) verified an excellent model fit. Internal consistency indices (Cronbach’s α > 0.87, composite reliability > 0.75) and strong convergent and discriminant validity evidence (based on AVE and inter-factor correlations) further support the scale’s psychometric soundness. The SCE-PL thus offers researchers, policymakers, and practitioners a robust, theory-driven instrument for assessing personalized learning experiences in smart classroom environments, paving the way for data-informed pedagogy, optimized learning spaces, and enhanced technological integration.

1. Introduction

Advances in intelligent and mobile technologies have given rise to smart classrooms (SCs)—technology-driven learning environments that enhance teaching quality, student engagement, and personalized learning (Gambo & Shakir, 2023; Zhan et al., 2021). SCs integrate information technology, the internet, and digital tools to create intelligent, personalized, and digitized educational spaces (Lycurgus et al., 2024; Ma et al., 2024; Mugruza-Vassallo, 2023; Shaw & Patra, 2022). However, fully harnessing digital technologies requires a shift toward a student-centered, personalized learning culture (C.-M. Chen et al., 2005; S. Y. Chen et al., 2016; Walkington & Bernacki, 2020). Here, defined as a technology-assisted closed environment, SCs support personalized learning and innovative pedagogy (Major et al., 2021; Zhang et al., 2020a, 2020b). In this study, we define personalized learning in smart classrooms as an approach that tailors learning experiences based on individual differences—including prior knowledge, cognitive and non-cognitive traits, interests, needs, and goals—through intelligent technologies and systematic instructional design (Hu et al., 2022; Z. Liu et al., 2022; Wang et al., 2023). Our aim is to develop a personalized learning measurement scale for smart classrooms, enabling the assessment of students’ experiences, optimization of teaching strategies, and advancement of personalized learning research.

1.1. Personalized Learning in Smart Classrooms

Personalized learning (PL), rooted in a learner-centered philosophy, respects individual differences by offering flexible methods that help students master knowledge and develop key skills at their own pace (C.-M. Chen et al., 2005; S. Y. Chen et al., 2016; Walkington & Bernacki, 2020). Historically, PL’s roots trace back to John Dewey’s early 20th-century learner-centered philosophy (Keefe & Jenkins, 2008; Redding, 2016). Characteristically, PL challenges the standardized, large-scale education model of the Industrial Revolution, driving modern educational reform (Basham et al., 2016). It acknowledges differences in students’ cognitive and non-cognitive characteristics (Sampson & Karagiannidis, 2002), where cognitive skills involve problem-solving and critical thinking, while non-cognitive traits influence personal growth, social engagement, and achievement (Farkas, 2003; Kell, 2018). S. Y. Chen and Macredie (2010) further identified factors like prior knowledge, cognitive styles, gender, and learning approaches that shape PL. To put it briefly, given its emphasis on individual differences, PL is central to student-centered instructional design.

Like many key concepts in the humanities and social sciences, PL lacks a unified definition. Principally, researchers viewed it through the lens of learner individuality, emphasizing a learner-centered philosophy that promotes active participation and choice in what, how, when, and where students learn (C.-M. Chen et al., 2005; S. Y. Chen et al., 2016; Walkington & Bernacki, 2020). Some define PL from a teaching perspective, considering individual differences (Patrick et al., 2013; U.S. Department of Education, 2010). Others emphasize the role of technology in PL, defining technology-enhanced personalized learning (TEL) as adaptive learning experiences based on real-time assessments (e.g., Bulger, 2016; Major et al., 2021; Zhang et al., 2022). All in all, PL is designed based on respecting students’ individual differences, focusing on their cognitive, thinking, emotional aspects, and maximizing the autonomy and subjectivity of their learning (Bulger, 2016; Major et al., 2021; Patrick et al., 2013; U.S. Department of Education, 2010; Zhang et al., 2022).

While teachers can address personalized needs through direct guidance (Bulger, 2016; Holmes et al., 2018), TEL provides scalable solutions for overburdened educators and overcrowded classrooms (Fake & Dabbagh, 2020). As a result, PL approaches and techniques have gained widespread attention in education (Major et al., 2021; Pane et al., 2015; Van Schoors et al., 2021; Xie et al., 2019). Researchers increasingly integrate intelligent technologies into SCs to develop PL systems that deliver tailored recommendations and individualized experiences (Niknam & Thulasiraman, 2020; Ouf et al., 2017; Sungkur & Maharaj, 2021; Troussas et al., 2020). Nevertheless, while technology has shown promise in enhancing PL, its impact in SC environments remains underexplored.

A key challenge is the lack of effective tools to measure PL support, limiting both theoretical research and practical implementation. Although some researchers have developed personalized learning support instruments (PLSIs) based on the universal design for learning (UDL) framework (Zhang et al., 2022), these tools are designed for traditional classrooms and are unsuitable for smart learning environments (Chatti et al., 2010; Cheung et al., 2021; Kloos et al., 2020; Price, 2015; Sharma et al., 2018; Toivonen et al., 2018). Additionally, from our comprehensive literature review, we found that many researchers have explored technology-driven solutions to enhance PL. Wang et al. (2023) describe SCs as an advanced form of online and multimedia-based learning, emphasizing that technology is the key differentiator between smart and traditional classrooms. Hu et al. (2022) introduced a learning analytics dashboard (LAD) in an AI-supported smart learning environment (SLE), allowing students to track and optimize their learning behaviors, leading to significant academic improvement. Kaur and Bhatia (2022) further highlight the role of information and communication technology (ICT) components such as the Internet of Things (IoT), cloud computing, and edge computing in realizing this vision. Niknam and Thulasiraman (2020) created a learning path recommendation system using an ant colony optimization algorithm, significantly improving knowledge retention. Yu et al. (2024) developed an intelligent Q&A system for entrepreneurship education, utilizing natural language processing (NLP) to enhance knowledge retrieval and comprehension. Overall, most current research results often center on the application of specific systems or algorithms, and their effectiveness and universality in different environments and broader populations remain unclear (Hu et al., 2022; Kaur & Bhatia, 2022; Niknam & Thulasiraman, 2020; Wang et al., 2023; Yu et al., 2024). Thus, understanding students’ experiences with these innovative solutions becomes essential. Therefore, we adopted a constructivist theoretical perspective for understanding student experiences (Radcliffe, 2009).

1.2. Constructivism

Constructivism suggests that individuals actively construct their own understanding of the world through their experiences and interactions with the environment, and that this process of construction is ongoing throughout the lifespan (Charmaz, 2021; Glaser & Strauss, 1964). Social constructionism/constructivism is, “a theoretical perspective that assumes that people create social reality(ies) through individual and collective actions” (Charmaz, 2006, p. 189). This perspective focuses on how people in a particular time and place create their own perceptions of reality. Instead of assuming that reality exists independently of us, social constructionists “view that social properties are constructed through interactions between people, rather than having a separate existence” (Robson & McCartan, 2016, p. 24). Hence, Radcliffe (2009) proposed the pedagogy–space–technology theoretical framework to foster an understanding of student-centered constructivist pedagogies in education.

1.3. Pedagogy–Space–Technology Framework

David Radcliffe, after launching the “next generation learning spaces” project, proposed a framework for designing and evaluating learning spaces (Radcliffe, 2009). This framework suggests that the composition of a learning space includes three elements: pedagogy, learning space, and technology. It classifies stimuli based on their constituent elements, considering natural objective objects (technology), artificial objective objects (learning space), subjective objects (pedagogy), and other factors. According to the three-dimensional model for learning environment design and evaluation, students engaging in PL connect with teaching methods, the effective integration of technology, and the design of learning spaces (Reushle, 2012). We will examine the rationale for developing this scale in terms of PST dimensions.

1.3.1. Pedagogy

What kind of teaching and learning do we aim to promote, and for what reasons? (Radcliffe, 2009). This question forms the basis of the “pedagogy” dimension. The constantly evolving nature of education and the (so-called) new generation of students places considerable emphasis on the adaptive teaching and characteristics of teachers (Göncz, 2017). According to Huber and Seidel (2018), our understanding of the interplay between teachers and students has recently begun to improve. Compared to teacher-centered pedagogies, which overly emphasize the teacher’s dominant role and neglect students’ individual differences and active learning needs, student-centered pedagogies foster creativity and critical thinking by promoting student autonomy (Bulger, 2016; Major et al., 2021; Patrick et al., 2013; U.S. Department of Education, 2010; Zhang et al., 2022). Furthermore, Basham et al. (2016) found that designing PL experiences requires a restructured educational model that emphasizes individual growth in various areas. Currently, Cheng and Carolyn Yang (2023) summarize three types of student-centered pedagogies—team-based learning, problem-based learning, and project-based learning—and further clarify the differences among them. In addition, precision teaching (Lindsley, 1992) and flipped classrooms are also considered effective student-centered teaching methods (Fan, 2024; C. Liu & Wang, 2018). Since SCs and student-centered pedagogies are interdependent (Koh et al., 2020; Malekigorji & Hatahet, 2020), we recognized the pedagogy dimension as the first rationale.

1.3.2. Learning Spaces

The key question for this dimension is “Which elements of space design and the arrangement of furniture and fixtures influence pedagogical approaches?” (Radcliffe, 2009). A learning space naturally influences how people interact within it, shaped by interactions between students and their environments (Blumer, 1966; Bryant & Charmaz, 2007; Charmaz, 2021; Dai & Chen, 2013; Dai et al., 2011; Lemert, 1974; Mead, 1934/1972; Morse et al., 2021; Robson & McCartan, 2016; Sternberg, 2022; Wirthwein et al., 2019; Ziegler & Bicakci, 2023; Ziegler & Stoeger, 2012; Ziegler et al., 2012). A smart learning space is essentially an intelligent physical space equipped with interactive whiteboards, projectors, automatic assessment/feedback tools, cameras for recording and storing lectures, and sensors that control temperature, humidity, air quality, and acoustics (Saini & Goel, 2019). This is distinct from the commonly discussed “learning environment”, which solely refers to the social, psychological, or conceptual environment (Reushle, 2012; Zhan et al., 2021).

The intelligent systems of smart classrooms are considered as another element in the learning environment with the following interaction types (a) teacher–student, (b) student–student, (c) teacher–media, and (d) student–media (Zhan et al., 2021). Hence, compared to traditional multimedia classroom (TMC) environments, the infrastructural resources of SCs have the potential to significantly influence individuals’ educational pathways and experiences (c.f., Ziegler, 2005; Ziegler & Stoeger, 2017; Ziegler et al., 2012). Bringing it all together, students’ experiences are shaped by their individual perceptions of the learning space (Ketscher et al., 2025; Merton, 1995; Thomas & Thomas, 1928). Eventually, we noted learning spaces as the second rationale.

1.3.3. Technology

In what ways can technology be incorporated into space design to reinforce and improve the desired teaching and learning approaches? (Radcliffe, 2009). Technology supports learning, particularly in promoting PL for students, with a positive and beneficial impact. Some researchers have found that artificial intelligence (AI) technology can not only provide personalized feedback and expert guidance to learners (Yu et al., 2024), but also enhance engagement, offer PL experiences, and ultimately increase motivation and foster autonomous learning (Santhosh et al., 2023). That is to say, the technological infrastructure of SCs can lead to distinct pedagogical improvements compared to traditional and/or technology-integrated classrooms (Chatti et al., 2010; Cheung et al., 2021; Kloos et al., 2020; Price, 2015; Sharma et al., 2018; Toivonen et al., 2018). How learners use smart devices (such as smart tablets, smartphones, etc.) flexibly and effectively in a smart classroom is a key factor in promoting their PL. Thus, we have concluded the third rationale.

1.4. Summary of Research Rationale

The (so-called) new generation of students perceives technology as an integral part of learning and may struggle in technology-limited environments (Tapscott, 2009). Nevertheless, while Basham et al. (2016) find that PL environments demand more than technology, smart classrooms offer this “more” (Yang & Huang, 2015). SCs enhance adaptability and interactivity through advanced technology and AI, offering a more dynamic learning environment than TMCs, which rely on fixed resources and conventional software (Zhan et al., 2021). “Smart classroom is one direction in smart learning environment research”(Yang & Huang, 2015, p. 157). Hence, in this research body, our aim is to develop a scale to measure PL experiences in SC environments.

2. Method

2.1. Development Process of Smart Classroom Personalized Learning Scale (SCE-PL)

This study adopted a deductive approach to scale development, beginning with a comprehensive literature review to define the target construct of PL for middle-school students in SC settings (Hinkin, 1995). Existing scales and questionnaires were evaluated to identify gaps and areas for refinement, focusing particularly on five key instruments: Fraser (1998)’s What is happening in this class? (WIHIC), Zhang et al. (2022)’s personalized learning experience support instrument (PLSI), EDUCASE Learning Initiative (2021)’s learning space rating system, Timothy et al. (2010)’s self-directed learning with technology scale (SDLTS), and Yang and Huang (2015)’s classroom environment evaluation scale (CEES). From these sources, an initial pool of 48 items was generated, covering topics such as teaching methods and learning spaces. Subsequently, two Chinese information technology experts conducted three rounds of discussions to ensure clarity, exclusivity, and alignment with the scale’s objectives (DeVellis, 1991; Hambleton, 2005; Hinkin, 1995). Through this process, four items were removed for failing to meet relevance or clarity requirements (e.g., The view outside the classroom window relaxes me and enhances my learning atmosphere). This phase concluded with 44 initial items.

Content validity—how effectively an instrument’s items capture the intended construct (Polit & Beck, 2006)—was assessed by 11 experts (8 university professors and 3 secondary school teachers) in information technology (IT). After two rounds of expert feedback, 5 additional items were eliminated due to issues of redundancy or ambiguity (DeVellis, 1991; Hambleton, 2005), leaving 39 items in the initial scale (Appendix A,

Table A1. The 39 items on SCE-PLE developed in the first stage.

Items	1	2	3	4	5
1. The teacher sets personalized learning goals for me based on my learning data.
2. I clearly understand the learning objectives of each smart classroom lesson.
3. The smart platform or smart devices (such as tablets, laptops, etc.) recommend suitable learning tasks to help me achieve the learning goals.
4. The teacher provides feedback on my progress toward achieving learning goals through the smart platform.
5. The smart classroom offers learning content and resources that I am interested in.
6. The teacher provides multiple learning content options through the smart platform, and I can choose the most suitable content based on my interests and needs.
7. The teacher uses smart tools (such as online resource libraries and learning path recommendation systems) to recommend diverse learning materials (such as videos, articles, exercises, etc.) for me.
8. I can participate in the course in different ways (such as online discussions, virtual experiments, group collaboration, etc.) through the smart platform or smart devices.
9. The teacher offers various ways to showcase learning outcomes (such as text, charts, videos, etc.) via the smart platform, allowing me to choose the most appropriate method.
10. The smart platform or smart devices provide abundant learning resources and materials, allowing me to flexibly choose the most suitable learning approach.
11. The teacher monitors my learning progress in real time through the smart platform and adjusts the teaching content and methods accordingly.
12. When I make mistakes in class, the teacher provides more accurate guidance based on the diagnostic feedback from the smart platform.
13. The teacher checks my learning progress in real time through the smart platform and provides timely feedback.
14. The teacher regularly checks my progress on completing learning tasks through the smart platform.
15. The teacher designs personalized learning tasks based on my learning data to help me better understand the course topics.
16. The teacher provides me with challenging tasks suited to my learning level through the smart platform.
17. The teacher uses smart tools (such as group functions) to help me collaborate with classmates to meet my personalized learning needs.
18. The teacher adjusts teaching strategies based on learning analysis reports from the smart platform to ensure that each student learns in the most suitable way.
19. The classroom has sufficient natural light, allowing me to see the blackboard and textbooks more clearly, which enhances my learning experience.
20. I can adjust the temperature of the smart classroom according to my needs, ensuring comfort without distractions.
21. I can adjust the brightness of the classroom lighting as needed, making the learning environment more comfortable.
22. The classroom’s sound system ensures that I can clearly hear the teacher’s explanations.
23. I find the seating in the classroom very comfortable, which helps me focus on learning.
24. The seating arrangement in the classroom is flexible, allowing me to move in and out easily and interact with others.
25. I can easily adjust the position of smart devices to participate in different types of learning activities.
26. The desk size is just right, providing enough space for textbooks, smart devices, and other materials.
27. The layout of the classroom’s central control system, interactive whiteboard, and smart projectors is well-organized, suitable for both teaching and learning.
28. I can easily use the smart devices provided by the school for learning.
29. The smart devices provided by the school are high-quality and capable of supporting a variety of learning tasks.
30. I can continue learning at home using the smart devices or smart platform provided by the school.
31. The school regularly updates and maintains the smart devices to ensure they are functioning properly.
32. I can easily access the learning resources recommended by the teacher (such as videos, articles, exercises, etc.) through the smart platform or smart devices.
33. I can access supplementary learning resources related to the course content through the smart platform or smart devices.
34. I can share learning resources with other classmates through the smart platform or smart devices.
35. The smart platform generates a personalized list of learning resources to help me complete learning tasks more effectively.
36. The smart platform or smart devices automatically analyze my learning weaknesses based on my learning data (such as quiz results, study time, etc.).
37. When I encounter learning difficulties, the smart platform or smart devices recommend suitable learning resources or solutions to assist me.
38. The smart platform or smart devices generate a personalized learning report to help me track my learning progress.
39. After completing learning tasks, the smart platform or smart devices provide detailed error analysis and improvement suggestions.

). The questionnaire comprised two sections: (1) demographic information about the respondents, and (2) the 39-item SCE-PL, rated on a 5-point Likert scale (1 = “strongly disagree”, 5 = “strongly agree”).

2.2. Sampling Frame and Participants

A combination of stratified and convenience sampling was employed to capture diverse educational contexts. Three provinces in China (Guangdong, Shaanxi, and Inner Mongolia) were selected to reflect variations in economic development and educational practices. Within each province, one representative school was purposively chosen, and students from multiple grade levels (7th, 8th, 9th) were included to enhance generalizability. All participating schools had adopted technology-enhanced PL strategies, providing a suitable environment for investigating the proposed scale.

The research targeted Chinese middle-school students who had experience in SC settings. A SC was defined as a learning space equipped with interactive whiteboards, recording systems, VR/AR devices, and student terminal devices, e.g., tablets and laptops (Lycurgus et al., 2024; Ma et al., 2024; Mugruza-Vassallo, 2023; Shaw & Patra, 2022). Prior to data collection, data collection approval was obtained from local school administrators, superintendents or principals, and class teachers. Furthermore, in order to avoid an ethical dilemma, we prepared our research under the guidance of people like Kropotkin (1922/1924) or Resnik (1998), and ethical codes such as the American Educational Research Association (AERA, 2011), American Psychological Association (APA, 2017), “The Nuremberg Code (1947)” (The Nuremberg Code (1947), 1996), and Multidisciplinary Digital Publishing Institute’s research and publication ethics (MDPI, n.d.).

2.3. Data Collection

The questionnaire was distributed online through Wenjuanxing (https://www.wjx.cn/), with the assistance of IT teachers. Data were collected in two waves. The first dataset (n = 440 initially, 424 valid after screening) was designated for exploratory factor analysis (EFA), while the second dataset (n = 598 initially, 584 valid after screening) was reserved for confirmatory factor analysis (CFA) to verify the structural validity of the SCE-PL. The demographic composition of the participants is shown in

Table 1. The demographic composition of the samples.

Participant Demographics	Sample 1 n = 440 *	Sample 2 n = 584 *
Grade level	n (%)	n (%)
Seventh grade	159 (36.14%)	225 (38.53%)
Eighth grade	135 (30.68%)	173 (29.62%)
Ninth grade	146 (33.18%)	186 (31.85%)
Gender
Male	253 (57.5%)	281 (48.12%)
Female	187 (42.5%)	303 (51.88%)

* Before data purification.

.

2.4. Data Preparation

Two researchers jointly managed the statistical analyses to ensure consistency and accuracy. SPSS 26.0 was employed to conduct item analysis using the critical ratio (CR) method and reliability tests (Hair et al., 2010). The decision rules were (a) retain items with significant CR values (p < 0.05) and t > 3; and (b) ensure Cronbach’s α remains above 0.70, item-total correlation (CITC) exceeds 0.35, and that deleting an item does not elevate the dimension’s α (Yuan et al., 2021; Zhao et al., 2023). Based on these criteria, five items (CG3, FMM4, SLV3, LD2, and DA3) were eliminated, leaving 34 items. Further reliability checks confirmed that all remaining items had α coefficients above 0.70, that removing them would lower the dimension’s reliability, and that item-total correlations exceeded 0.40 (Tabachnick & Fidell, 2007).

2.5. Data Analysis

EFA is routinely employed to assess the construct validity of an instrument. Prior to factor extraction, the data’s suitability was evaluated using the Kaiser–Meyer–Olkin (KMO) test and Bartlett’s test of sphericity (Huck et al., 1974; Zhao et al., 2023). The KMO value was 0.889 (>0.70), indicating strong inter-item correlations, and Bartlett’s test was highly significant (χ² = 10,258.843, df = 561, p < 0.001), confirming that common factors were present (

Table 2. KMO test and Bartlett’s test of sphericity.

KMO	Bartlett’s Test of Sphericity
	χ2	df	Sig.
0.889	10,258.843	561	0.000

). Principal component analysis (PCA) with varimax rotation then extracted nine factors with eigenvalues greater than 1 (DeVellis, 1991).

CFA was conducted using AMOS 23.0 on the second dataset (n = 584) to verify whether the model derived from EFA aligned with theoretical expectations. Model fit was assessed via multiple indices: χ², RMSEA, CFI, TLI, RMR, GFI, AGFI, NFI, and IFI. Common acceptance criteria include χ²/df < 5, RMR ≤ 0.05, RMSEA < 0.08, GFI and AGFI > 0.80, and CFI, NFI, IFI, and TLI > 0.90. After confirming model fit, reliability and validity tests (including convergent and discriminant validity) were performed to finalize the scale structure.

3. Results

EFA was conducted on the first dataset (n = 424) to identify the latent structure of the initial scale. The KMO value (0.889) and Bartlett’s test of sphericity (χ² = 10,258.843, df = 561, p < 0.001) confirmed the adequacy of the data for factor analysis. PCA with varimax rotation extracted nine factors with eigenvalues exceeding 1. Together, these factors accounted for 78.12% of the total variance, surpassing typical benchmarks in educational and psychological research (Gruijters, 2019; Härdle & Simar, 2015; Tabachnick & Fidell, 2007). The scree plot (Figure 1) shows a sharp drop in eigenvalues over the first nine factors, followed by a distinct “elbow”, after which the remaining factors level out with eigenvalues below 1. Factor loadings exceeded 0.60 for all retained items, and no additional deletions were necessary, thus yielding a robust nine-factor structure for subsequent validation.

The second dataset (n = 584), with a ratio of approximately 17:1 respondents per item, was used to verify the nine-factor structure. AMOS 24.0 estimated a model comprising nine first-order factors—Timely Progress Monitoring (TPM), Intelligent Diagnosis and Services (IDS), Layout and Display (LD), Flexible Instructional Methods and Materials (FMM), Clear and Relevant Goals (CRG), Environmental Quality (EQ), Access to Resources (AR), Supporting Learner Variability (SLV), and Device Access (DA)—organized under three broader dimensions (pedagogy, space, and technology). Model fit was excellent (χ²/df = 1.515, RMSEA = 0.030, GFI = 0.929, AGFI = 0.918, NFI = 0.946, TLI = 0.979, and CFI = 0.981), and all factor loadings were significant (p < 0.001). No item showed cross-loadings, verifying that the 34 retained items accurately measured their intended constructs (Huck et al., 1974). Figure 2 illustrates the final model, while further analyses confirmed that variables were correctly assigned to the nine factors, each aligning with the theoretical dimensions established through expert review.

The rotated component matrix (

Table 3. Rotated component matrix.

Items	Component
	1	2	3	4	5	6	7	8	9
TPM3	0.845	0.013	0.079	0.145	0.173	0.048	0.065	0.159	0.054
TPM2	0.839	0.034	0.103	0.236	0.215	0.007	0.019	0.172	0.027
TPM4	0.819	0.064	0.028	0.221	0.124	−0.008	0.008	0.122	0.025
TPM1	0.808	0.005	0.042	0.206	0.218	0.039	0.002	0.149	−0.010
IDS3	0.028	0.892	0.054	0.033	0.075	0.057	0.161	0.060	0.167
IDS2	−0.009	0.861	0.099	0.079	0.049	0.066	0.194	0.045	0.162
IDS1	0.066	0.853	0.109	0.058	0.061	0.064	0.174	0.014	0.167
IDS4	0.034	0.814	0.088	0.087	0.067	0.070	0.212	0.005	0.160
LD1	0.052	0.064	0.868	0.055	0.060	0.209	0.075	0.032	0.036
LD5	0.074	0.097	0.855	0.059	0.134	0.264	0.100	0.095	0.067
LD3	0.066	0.095	0.835	0.139	0.054	0.202	0.049	0.059	0.116
LD4	0.070	0.116	0.831	0.136	0.081	0.281	0.136	0.098	0.078
FMM5	0.257	0.041	0.100	0.848	0.166	0.058	0.007	0.153	0.036
FMM2	0.182	0.092	0.102	0.824	0.154	0.072	0.034	0.176	0.072
FMM1	0.204	0.150	0.122	0.794	0.246	0.117	0.025	0.180	0.066
FMM3	0.300	0.010	0.096	0.696	0.217	0.122	0.109	0.184	−0.011
CRG5	0.123	0.022	0.093	0.111	0.839	0.027	0.016	0.107	0.123
CRG4	0.130	0.097	0.068	0.198	0.790	0.053	0.106	0.141	0.015
CRG2	0.268	0.105	0.058	0.192	0.790	0.126	0.019	0.161	−0.021
CRG1	0.304	0.052	0.108	0.235	0.774	0.057	0.062	0.207	−0.004
EQ4	−0.018	0.137	0.210	0.025	0.059	0.825	0.067	0.036	0.048
EQ2	0.017	0.017	0.153	0.116	0.083	0.822	0.037	−0.008	0.040
EQ3	0.044	0.045	0.226	0.034	0.014	0.820	0.083	0.096	0.062
EQ1	0.047	0.047	0.263	0.117	0.063	0.806	0.097	0.025	0.016
AR2	0.029	0.195	0.071	0.034	0.001	0.097	0.830	0.043	0.133
AR1	0.066	0.160	0.083	0.067	0.116	0.017	0.819	−0.019	0.140
AR4	−0.003	0.174	0.108	0.047	0.087	0.033	0.814	−0.091	0.165
AR3	−0.002	0.175	0.058	−0.013	−0.024	0.138	0.763	0.092	0.153
SLV2	0.281	0.065	0.053	0.224	0.150	0.060	−0.024	0.816	0.051
SLV4	0.190	−0.006	0.131	0.203	0.230	0.009	0.029	0.807	0.032
SLV1	0.153	0.059	0.077	0.192	0.190	0.081	0.016	0.807	−0.010
DA1	0.047	0.231	0.066	0.056	0.032	0.076	0.198	−0.002	0.811
DA4	0.024	0.203	0.064	0.055	0.056	0.058	0.179	0.029	0.808
DA2	0.011	0.184	0.122	0.019	0.022	0.025	0.203	0.034	0.794
Eigenvalue	3.344	3.310	3.261	3.058	3.048	3.042	2.950	2.342	2.205
% of Variance	9.835	9.734	9.591	8.994	8.965	8.947	8.677	6.888	6.484
Cumulative %	9.835	19.569	29.160	38.155	47.120	56.066	64.743	71.631	78.116

) shows how each survey item clusters onto nine distinct factors, with each column (Components 1–9) corresponding to one factor. For example, items TPM1–TPM4 load strongly on Component 1 (all above 0.80), indicating they measure a single underlying construct (Timely Progress Monitoring). Similarly, items IDS1–IDS4 load on Component 2 (Intelligent Diagnosis and Services), LD1–LD5 on Component 3 (Layout and Display), etc. Consequently, this table supports the conclusion that the final 34-item questionnaire reliably measures nine distinct constructs, each capturing a unique aspect of PL in a SC environment.

Cronbach’s alpha was used to assess internal consistency. The overall alpha coefficient for the scale was 0.938, while the subscales ranged from 0.887 to 0.924, all well as being above the commonly accepted threshold of 0.70. Table 3 presents the means, standard deviations, and correlation coefficients among the nine latent constructs (CRG, FMM, TPM, SLV, EQ, LD, DA, AR, and IDS). The bold-faced diagonal entries (e.g., 0.820, 0.839) are the square roots of the AVE for each factor.

The Fornell–Larcker criterion supported discriminant validity: the square roots of the AVE for each factor surpassed the correlations with other factors, indicating that the nine constructs are empirically distinct. As shown in

Table 4. Discriminant validity.

	Mean	SD	CRG	FMM	TPM	SLV	EQ	LD	DA	AR	IDS
CRG	3.59	1.06	0.820
FMM	3.73	0.96	0.505 **	0.839
TPM	3.55	1.02	0.491 **	0.524 **	0.839
SLV	3.63	1.05	0.489 **	0.477 **	0.497 **	0.830
EQ	3.60	0.98	0.241 **	0.258 **	0.163 **	0.234 **	0.816
LD	3.67	1.07	0.374 **	0.335 **	0.274 **	0.321 **	0.540 **	0.886
DA	3.69	1.01	0.275 **	0.225 **	0.247 **	0.226 **	0.235 **	0.326 **	0.802
AR	3.79	0.88	0.249 **	0.238 **	0.268 **	0.219 **	0.136 **	0.271 **	0.454 **	0.806
IDS	3.72	0.93	0.221 **	0.258 **	0.235 **	0.224 **	0.207 **	0.287 **	0.504 **	0.431 **	0.873

** Statistically significant at the p < 0.01 level.

, the EFA and CFA findings, alongside favorable reliability and validity metrics, affirm that the nine-factor, 34-item SCE-PL is both psychometrically sound and theoretically coherent.

Table 5. Convergent validity.

Variable	Item	Factor Loading	Composite Reliability (CR)	Average Variance Extracted (AVE)
Smart Classroom Personalized Learning Scale (SC-PLS)	Pedagogy	0.719	0.752	0.503
	Space	0.720
	Technology	0.688
Pedagogy	CRG	0.766	0.836	0.560
	FMM	0.741
	TPM	0.750
	SLV	0.735
Space	EQ	0.800	0.747	0.597
	LD	0.744
Technology	DA	0.798	0.767	0.525
	AR	0.662
	IDS	0.707
Clear and Relevant Goals (CRG)	CRG1	0.883	0.891	0.672
	CRG2	0.864
	CRG4	0.765
	CRG5	0.758
Flexible Instructional Methods and Materials (FMM)	FMM1	0.871	0.905	0.704
	FMM2	0.832
	FMM3	0.750
	FMM5	0.896
Timely Progress Monitoring (TPM)	TPM1	0.811	0.905	0.704
	TPM2	0.895
	TPM3	0.841
	TPM4	0.805
Supporting Learner Variability (SLV)	SLV1	0.772	0.868	0.688
	SLV2	0.883
	SLV4	0.830
Environmental Quality (EQ)	EQ1	0.857	0.888	0.665
	EQ2	0.766
	EQ3	0.837
	EQ4	0.799
Layout and Display (LD)	LD1	0.833	0.936	0.785
	LD3	0.855
	LD4	0.923
	LD5	0.929
Device Access (DA)	DA1	0.842	0.844	0.644
	DA2	0.778
	DA4	0.785
Access to Resources (AR)	AR1	0.810	0.881	0.650
	AR2	0.835
	AR3	0.754
	AR4	0.824
Intelligent Diagnosis and Services (IDS)	IDS1	0.864	0.928	0.762
	IDS2	0.879
	IDS3	0.910
	IDS4	0.837

offers a detailed overview of the convergent validity indices for the SCE-PL. Convergent validity was established by factor loadings (>0.50), composite reliability (>0.70), and average variance extracted (>0.50) for each factor, confirming that items within each subscale converge on a single underlying dimension.

4. Discussion

This study developed and validated the smart classroom environment–personalized learning scale (SCE-PL), a measurement instrument designed to capture middle-school students’ PL experiences in SC environments. Drawing on literature reviews, expert consultations, and iterative item refinement, we generated a final instrument comprising 34 items and nine factors aligned with three overarching dimensions: pedagogy, learning space, and technology (Radcliffe, 2009). EFA indicated that these nine factors accounted for a substantial portion of the total variance, while confirmatory factor analysis confirmed strong model fit and factor loadings. Reliability tests (Cronbach’s alpha and composite reliability) met or exceeded standard criteria, and convergent and discriminant validity were supported through factor loadings, AVE, and the Fornell–Larcker criterion. These findings underscore that the nine-factor structure and strong psychometric properties of the SCE-PL scale corroborate the claim that PL in SCs emerges from a dynamic interplay of student-centered pedagogy, thoughtfully designed learning spaces, and well-integrated technology (Basham et al., 2016; Bulger, 2016; Chatti et al., 2010; S. Y. Chen et al., 2016; Fraser, 1998; Gambo & Shakir, 2023; Z. Liu et al., 2022; Mead, 1934/1972; Radcliffe, 2009; Yang & Huang, 2015; Zhan et al., 2021; Ziegler, 2005; Ziegler & Stoeger, 2017).

The results enrich the theoretical dialogue surrounding PL and SC research by systematically bridging pedagogy, space, and technology factors under a single measurement framework (Ouf et al., 2017; Pane et al., 2015; Price, 2015; Radcliffe, 2009; Reushle, 2012; Saini & Goel, 2019; Van Schoors et al., 2021; Walkington & Bernacki, 2020; Xie et al., 2019; Yang & Huang, 2015; Zhan et al., 2021; Zhang et al., 2020a). Earlier research often centered on specific technologies (e.g., learning analytics dashboards and algorithm-based recommendation systems) or conceptual models lacking robust validation in varied contexts (e.g., Hu et al., 2022; Kaur & Bhatia, 2022; Niknam & Thulasiraman, 2020; Wang et al., 2023; Yu et al., 2024). By integrating Radcliffe’s (2009) PST framework with the conceptual underpinnings of PL, this study contributes a nuanced perspective that acknowledges both the instructional design (teacher strategies, student autonomy, and flexible methods) and environmental design (physical layout and device access) critical for personalizing students’ learning processes. Consequently, the SCE-PL may serve as a foundational tool for future empirical inquiries into how these dimensions interact to shape adaptive learning experiences.

From a constructivist perspective, learning is a process in which students actively construct meaning based on their experiences, interactions, and reflections. This study emphasizes the situated nature of PL by confirming that each of the nine subfactors is distinct yet correlated. Students are not merely passive recipients of knowledge; they engage with instructional materials, classroom technologies, peers, and teachers in ways profoundly shaped by individual differences and contextual factors. Such a view aligns with social constructionism, wherein reality—and, by extension, learning experiences (Charmaz, 2021; Glaser & Strauss, 1964; Robson & McCartan, 2016)—are co-constructed in specific social and technological contexts. The robust loadings of subfactors related to pedagogy further highlight that PL is not only about customizing content or pacing; it also depends on the interactive, learner-centered strategies that spark motivation and autonomy (Fraser, 1998; Tapscott, 2009; Yang & Huang, 2015; Zhan et al., 2021). The consistent internal consistency of these constructs suggests that educators who adopt learner-oriented pedagogies guided by constructivist principles significantly influence students’ PL experiences.

The findings validate prior arguments that student-centered instructional methods (e.g., team-based learning and flipped classrooms) are essential for personalizing the learning experience (Basham et al., 2016; Bulger, 2016; Cheng & Carolyn Yang, 2023). That Timely Progress Monitoring (TPM) loaded strongly under pedagogy underscores the importance of continuous, data-driven feedback loops, aligning with constructivist theories that posit real-time interaction as a core facilitator of deeper learning. A smart learning space includes elements like interactive whiteboards, flexible seating, and environmental controls (Saini & Goel, 2019). Hence, the significant loadings for Environmental Quality (EQ) and Layout and Display (LD) confirm the critical role of physical and infrastructural design in creating learning spaces (Blumer, 1966; Reushle, 2012). The data suggests that these factors are not merely aesthetic or comfort-related; they tangibly shape students’ perceptions of autonomy, engagement, and collaboration.

Strong factor loadings for Access to Resources (AR), Device Access (DA), and Intelligent Diagnosis and Services (IDS) reinforce the claim that smart technology is integral to PL, notably when it offers adaptive feedback, varied resources, and real-time assessments (Hu et al., 2022; Niknam & Thulasiraman, 2020; Yu et al., 2024). Such findings echo emerging stances on how technology mediates learning by providing interactive experiences that allow students to co-construct knowledge rather than passively receive it. Moreover, the evidence that these technological subfactors are empirically distinct yet interrelated with pedagogy and space underscores Radcliffe’s (2009) assertion that technology must be purposefully embedded within pedagogical strategies and physical infrastructure to maximize its educational impact. Furthermore, rather than seeing technology as a standalone enabler, the data illustrate that technology’s efficacy in personalizing education is contingent upon compatible pedagogical methods (such as immediate feedback and flexible pacing) and well-designed spaces (supportive seating, adjustable lighting, and interactive displays). Consequently, this study provides a more granular, empirically grounded articulation of how the synergy among pedagogy, space, and technology contributes to student-centered, adaptive learning cultures.

Finally, the validated structure of the SCE-PL enriches ongoing scholarly discourse on PL by merging classic learner-centered theories (John Dewey, Piaget, Vygotsky) with modern, technology-integrated approaches (Bulger, 2016; Major et al., 2021; Zhang et al., 2022). The high internal consistency of subscales like Flexible Instructional Methods and Materials (FMM) or Supporting Learner Variability (SLV) highlights that personalizing education is not a single-pronged strategy but a holistic endeavor. Each dimension—be it real-time progress monitoring, physical layout, or AI-driven diagnostics—contributes uniquely to the learner’s constructivist process of active knowledge building (S. Y. Chen & Macredie, 2010). Moreover, mapping these subscales onto the pedagogy–space–technology matrix illustrates that personalization extends beyond simply providing digital resources or customizing tasks; it requires cohesive pedagogical design, flexible spatial arrangements, and sophisticated technological infrastructure. This integrated view resonates with the broader notion of “smart learning environments”, wherein learning transcends mere digitization to become an adaptive, interactive, and profoundly learner-centered experience (Ma et al., 2024; Mugruza-Vassallo, 2023; Yuan et al., 2021; Zhan et al., 2021).

Practitioners and policymakers can leverage the SCE-PL to diagnose and optimize PL initiatives within SCs. By pinpointing strengths and weaknesses across nine distinct factors—such as Timely Progress Monitoring (TPM), Device Access (DA), or Layout and Display (LD)—teachers and administrators gain actionable insights for targeted improvements. For instance, if students rate “Layout and Display” lower, schools might invest in rearranging seating or upgrading interactive displays to boost engagement. Likewise, higher scores in “Intelligent Diagnosis and Services” (IDS) but moderate scores in “Flexible Instructional Methods and Materials” (FMM) could imply that while real-time data analytics are robust, the range of pedagogical materials needs expansion. Overall, these diagnostic applications can foster continuous quality enhancement in SCs, aligning with broader global shifts toward data-informed educational practice (Kaur & Bhatia, 2022; C. Liu & Wang, 2018; Radcliffe, 2009; Reushle, 2012; Sharma et al., 2018; Zhan et al., 2021).

Despite its rigorous methodology, this study has several limitations. First, generalizability may be constrained by the sampling strategy, which, although stratified across three Chinese provinces, remains situated in specific cultural and educational contexts. Future validations in diverse regions—especially outside East Asia—are encouraged to ascertain the scale’s broader applicability. Second, self-reported data can be susceptible to social desirability bias, as students may overstate or understate their experiences (Diefes-Dux, 2019; Zhang et al., 2022). Incorporating observational or performance-based measures could yield a more holistic evaluation of PL (Sampson & Karagiannidis, 2002). Third, while the SCE-PL captures key dimensions aligned with the PST framework, other latent variables—such as motivation, self-efficacy, or teacher autonomy support—may also influence PL experiences (Fraser, 1998; Yang & Huang, 2015). Integrating these constructs could offer richer insights.

Building on the SCE-PL, researchers might explore longitudinal designs to assess how students’ PL experiences evolve over time, particularly as SC infrastructure and pedagogical innovations mature. Cross-cultural replication could illuminate whether the nine-factor structure holds under varying educational policies, socio-economic conditions, or technological advancements. Additionally, experimental or quasi-experimental designs could investigate causal relationships between improved scores on specific subscales and measurable outcomes such as academic performance, engagement, or self-regulated learning skills. Finally, further integration with learning analytics—for instance, correlating SCE-PL scores with real-time data from learning management systems—could refine our understanding of how different aspects of the scale map onto students’ actual learning behaviors.

To conclude, by uniting pedagogy, space, and technology elements into a validated nine-factor instrument, this study addresses a critical gap in measuring PL experiences within SCs. The SCE-PL offers a theoretically grounded and empirically substantiated tool for both researchers and practitioners, enabling more precise identification of strengths and challenges in technology-enriched instructional settings. Doing so lays the groundwork for data-driven refinements to PL strategies, ultimately advancing the broader goal of placing learner individuality at the heart of educational innovation.