Augmented Reality’s Impact on Student Creativity in Design and Technology: An Immersive Learning Study

Abstract

This quasi-experimental study examined the effectiveness of Augmented Reality (AR)-enhanced instruction on creativity development in Malaysian Design and Technology education. Forty-six, fifteen-year-old female students were assigned to AR-enhanced (n = 23) or traditional instruction (n = 23) groups for a four-week Mechatronic Design unit. Creativity was assessed using an adapted Torrance Tests of Creative Thinking-Figural (TTCT-F) instrument with expert validation and independent scoring by three raters. Bootstrapped ANCOVA (5000 iterations) controlling for pretest differences revealed significant improvements across all Guilford creativity components in the AR group: Elaboration (F = 27.093, p < 0.001, η² = 0.387), Originality (F = 20.445, p < 0.001, η² = 0.322), Fluency (F = 17.896, p < 0.001, η² = 0.294), and Flexibility (F = 7.593, p = 0.008, η² = 0.150). The differential effect pattern suggests AR operates through multiple mechanisms, primarily socio-constructivist collaborative scaffolding, followed by motivational enhancement and cognitive load reduction. These findings demonstrate AR’s substantial potential for creativity development in Design and Technology education, particularly for collaborative elaboration and generative ideation. However, single gender sampling, brief intervention duration, and quasi-experimental design limit generalizability, warranting future research with diverse populations and extended interventions.

1. Introduction

Creativity is widely recognized as a core 21st century competency and essential for innovation and problem-solving in contemporary societies [1]. In Design and Technology education, creativity is particularly critical as students are required to generate, refine, and evaluate design solutions through iterative processes that integrate technical knowledge and user-centered thinking [2]. However, research indicates that secondary school students often struggle to produce original and well-elaborated ideas, particularly in tasks involving spatial reasoning and system integration [3,4]. One contributing factor is the limited capacity of conventional instructional approaches to effectively support creative design processes. Traditional teaching methods frequently rely on static two-dimensional structures [5], while the integration of technology enables students to explore alternative configurations and engage in sustained collaborative refinement [6,7]. These constraints may hinder student’s ability to externalize ideas, iterate designs, and build upon peer feedback, which are processes central to creative thinking in design contexts.

Augmented Reality (AR), defined as the real-time overlay of digital onto physical environments [8], has emerged as a promising educational technology for addressing these challenges. By enabling interactive manipulation of three-dimensional models within authentic learning spaces, AR may reduce cognitive demands associated with spatial visualization and support exploratory learning [9,10]. In addition, AR environments allow rapid, low-risk iteration of design ideas and provide shared visual artifacts that can facilitate peer discussion and collaborative sense-making [11,12]. Meta-analytic evidence indicates that AR produces moderate to large effects on spatial ability and task performance [13]. However, findings related to creativity outcomes remain mixed, with reported effects ranging from negligible to substantial [14], suggesting that AR’s impact on creativity may depend on specific learning mechanisms and contextual conditions.

To explain these inconsistent, this study adopts a componential view of creativity grounded in Guilford’s Structure of intellect model [15], which conceptualizes creativity as comprising four divergent thinking components (Fluency, Flexibility, Originality, and Elaboration). Building on this framework, there are three complementary theoretical pathways proposed to explain how AR may differentially influence these components. From a cognitive-load perspective [16], AR may support Fluency and Originality by reducing extraneous cognitive demands, thereby freeing cognitive resources for idea generation. From a motivational perspective, Amabile’s Componential Theory [17] suggests that AR’s low-stakes, manipulability environments may enhance intrinsic motivation and encourage exploration of novel ideas, particularly supporting Originality. Finally, from a socio-constructivist perspective informed by Vygotsky [18], AR’s shared and persistent digital artifacts may function as boundary objects that scaffold collaborative dialogue, peer feedback, and iterative refinement, potentially strengthening Elaboration and Flexibility.

Based on these theoretical considerations and the identified empirical inconsistencies, this study investigates the effectiveness of AR-enhanced instruction in Design and Technology education by examining not only whether AR improves student’s creativity, but also how its effects vary across creativity components. Specifically, the study addresses two research questions:

• (RQ1) Does AR-enhanced instruction improve students’ creativity (Fluency, Flexibility, Originality, Elaboration) compared to traditional instruction?
• (RQ2) Do AR effects differ across creativity components in patterns consistent with cognitive, motivational, or social theoretical mechanisms?

By integrating theory-driven hypotheses with component-level creativity assessment, this study contributes to educational technology research in three ways. Theoretically, it advances understanding of the mechanisms through which AR may support different dimensions of creativity. Methodologically, it demonstrates the application of curriculum-aligned adaptation of the Torrance Tests of Creative Thinking-Figural (TTCT-F) [19]. Practically, the findings insights are for educators’ insight designers wondering which aspects of creativity benefit most from targeted AR integration in design-based learning.

2. Literature Review

Creativity is a central learning outcome in Design and Technology education, emphasizing novel solutions to authentic design problems [20]. Educational researchers distinguish between domain-general creativity (assessed through divergent thinking tests) and domain-specific creativity (assessed through design product evaluations) [17]. While both are relevant, domain-general creativity, Guilford’s four divergent thinking components (Fluency, Flexibility, Originality, Elaboration) provides standardized metrics enabling cross-study comparisons [15]. Empirical studies demonstrate that Design and Technology curricula emphasizing iterative design processes, prototyping, and peer critique produce moderate improvements in divergent thinking [21]. However, effect sizes vary substantially based on instructional factors, including scaffolding quality, feedback specificity, and revision opportunities, suggesting that pedagogical methods critically influence creativity development [22].

AR has been extensively investigated in STEM education, with meta-analyses documenting moderate to large effects on spatial ability (d = 0.68) [12], conceptual understanding (d = 0.56) [23], and procedural skills (d = 0.72) [24]. These outcomes are theoretically attributed to three cognitive mechanisms: dual coding, which integrates visual and verbal channels to enhance memory retrieval; cognitive load reduction, which minimizes working memory demands by externalizing spatial information [16]; and embodied cognition, where gesture-based interactions ground abstract concepts in sensorimotor experiences. However, these mechanisms have been primarily examined in relation to knowledge acquisition and spatial reasoning; their applicability to creativity development which involves generative rather than reproductive cognitive processes remain theoretically underspecified.

Empirical evidence regarding AR’s effects on creativity is mixed. Chen et al. [14] found six weeks of AR-based design instruction substantially boosted Originality (d = 1.24) and Fluency (d = 0.89) in 64 Chinese middle school students, attributing originality gains to psychological safety offered by digital prototyping. Martinez et al. [13] observed large effects (d = 1.15) on product creativity and moderate effects (d = 0.67) on divergent thinking in 48 Spanish university students. Chen and Wang [25] reported moderate improvements in creative problem-solving (d = 0.54) across Fluency, Flexibility, and Originality for 92 Taiwanese elementary students following four weeks of AR-supported science inquiry. This variability suggests AR’s influence on creativity depends on subject, duration, and educational context.

Three factors may explain inconsistent findings. First, task alignment is critical with studies showing significant effects (e.g., Chen et al. [14], Martinez [13]) when utilizing design tasks where AR affordances (3D visualization, rapid prototyping) directly supported creative processes, whereas studies showing null effects used AR primarily for content delivery where spatial manipulation was peripheral. Second, intervention duration matters, as longer programs (≥6 weeks) yield larger effects (mean d = 0.89) than shorter ones (≤3 weeks; mean d = 0.31), suggesting learners require extended time to overcome interface learning curves. Third, measurement specificity influences outcomes; domain-specific assessments aligned with curriculum content often show larger effects than domain-general tests like TTCT-F, indicating that while AR enhances applied creativity, transfer to abstract divergent thinking may require structured scaffolding.

Critically, most AR creativity studies report aggregate scores or focus on single components (typically Originality or Fluency), precluding analysis of differential effects across Guilford’s four components. Only three studies provide complete component-level data (

Table 1. Previous studies report differential effects of AR on creativity components.

Study	Sample	Duration	Fluency (d)	Flexibility (d)	Originality (d)	Elaboration (d)	Pattern
Chen et al. [14]	n = 64 Chinese middle school	6 weeks	0.89	0.5	1.24	0.08	Originality > Others
Chen and Wang [25]	n = 92 Taiwanese elementary	4 weeks	0.56	0.51	1.24	0.08	Uniform (no differentiation)
Martínez-Beneito et al. [13]	n = 54 Spanish secondary	8 weeks	0.94	0.38	0.87	0.62	Fluency ≈ Originality > Others

).

These patterns suggest AR effects are not uniform across components, but existing studies lack theoretical frameworks predicting differential effects. The current study addresses this gap by operationalizing three competing mechanisms (cognitive load, intrinsic motivation, collaborative scaffolding) with explicit component-level predictions.

Bibliometric studies show that research on augmented reality (AR) in education has been steadily increasing. Singh et al. [26] analyzed over 1700 articles and reported that AR research has expanded across multiple regions and educational levels, with most studies coming from technologically advanced countries. Luthfi et al. [27] examined Scopus data from 2020–2024 and found a consistent rise in AR publications, with countries such as the United States, China, Indonesia, Malaysia, and Germany contributing the most. However, Southeast Asian nations, including Malaysia, are still underrepresented in empirical AR studies. Local educational settings characterized by teacher directed teaching, uneven access to technology, and high linguistic diversity, may affect how AR supports learning. By exploring AR in Malaysian secondary schools, this study adds valuable geographic and cultural perspective to the existing evidence.

While existing research confirms AR can enhance creativity in design education, reported effect sizes vary significantly (d = 0.08–1.24), and underlying mechanisms remain underspecified. Three critical gaps limit advancement: (1) a theoretical design lacking specific predictions regarding creativity components; (2) measurement inconsistency across diverse assessments hindering cross-study synthesis; and (3) contextual concentration leaving Southeast Asian settings underrepresented. This study addresses these gaps by operationalizing three theoretical frameworks (Guilford, Amabile, Vygotsky) to generate component-level predictions and examining AR effects within Malaysian secondary schools, contributing essential contextual diversity to the field.

3. Materials and Methods

3.1. Research Design

This study employed a quasi-experimental pretest–posttest comparison group design to examine the effectiveness of AR-enhanced instruction on student creativity [28,29]. Randomized controlled trials (RCTs are gold standard for causal inference) and randomized assignment were not feasible in this educational context due to administrative, ethical and ecological constraints. In Malaysian secondary schools, students are assigned to intact classes based on administrative criteria and random assignment. It could disrupt established classroom dynamics and reduce ecological validity. In addition, school administrators expressed ethical concerns regarding withholding potentially beneficial instructional approaches from randomly selected students.

To address these constraints while maintaining methodological rigor, intact classes were assigned to experimental and comparison conditions, and Analysis of Covariance (ANCOVA) was used to statistically control pre-existing group differences through pretest covariates. This approach is widely accepted as an appropriate approximation of experimental control in classroom-based intervention research when randomization is impractical [28,29,30].

Potential threats to internal validity, including selection bias, maturation, testing effects, and instrumentation, were addressed through design and analytical strategies such as short intervention duration, parallel test forms, blinded scoring procedures, and covariance adjustment. A summary of identified threats and mitigation strategies is provided in

Table 2. Internal validity threats and mitigation strategies.

Threat	Description	Mitigation Strategy	Residual Risk
Selection Bias	Pre-existing differences between groups	ANCOVA with pretest covariates; propensity score sensitivity analysis	Low-Moderate
Maturation	Developmental changes during 4-week intervention	Brief intervention duration; comparison group controls for time effects	Low
Testing Effects	Pretest exposure influencing posttest performance	4-week interval between tests; parallel test forms (different prompts)	Low
Instrumentation	Scorer drift or inconsistency	Dual independent scoring; inter-rater reliability monitoring (ICC ≥ 0.80); periodic calibration	Low

.

3.2. Participants

The participants were 46 Form 3 female students (mean age = 15.2) enrolled in Design and Technology classes at a government funded secondary school in Kuala Lumpur, Malaysia. Students were drawn from two intact classes assigned by school administrators prior to recruitment based on balanced academic performance. One class (n = 23) was designated as the AR-enhanced instruction group while the other served as traditional instruction comparison group.

All participants completed both pretest and posttest assessments, resulting in zero attrition. Although two students in the AR group missed one instructional session due to illness, sensitivity analyses excluding these cases produced substantively similar results, indicating minimal impact on study outcomes.

3.3. Instruments

Creativity was measured using an adapted version of the TTCT-F. This test is widely used and has proven reliability. Acar et al. [31] reported that TTCT-F scores are consistently reliable across studies (ω ≈ 0.81). In Malaysia, adapted versions also showed strong inter-rater agreement, with ICC values of 0.90 or higher [32]. These results support using the adapted TTCT-F to assess creativity in secondary school students. To enhance ecological validity and curricular relevance, the instrument was adapted to align with Malaysian Design and Technology curriculum, specifically within the context of mechatronic product design.

Prior research demonstrates that curriculum-aligned creativity assessments produce larger effect sizes and stronger predictive validity than decontextualized measures, supporting the decision to adapt TTCT-F to Design and Technology content.

TTCT-Figural Adaptation Process

The process included the following:

1:

Rationale for Instrument Adaptation

The Torrance Tests of Creative Thinking-Figural (TTCT-F) was chosen as the primary creativity assessment because it is grounded in Guilford’s Structure of Intellect model [15] and has a long history of validation in educational research [19]. However, the standard TTCT-F tasks are domain-general and abstract, which may limit their ecological validity when applied to discipline specific contexts such as Design and Technology education [19].

To address this limitation, the instrument was systematically adapted to align with the Malaysian Form 3 Design and Technology curriculum, specifically the Mechatronic Design unit. The adaptation aimed to preserve the core constructs of divergent thinking, fluency, flexibility, originality, and elaboration while embedding them within an authentic, curriculum-relevant design task. This approach is consistent with prior research indicating that curriculum-aligned creativity assessments yield stronger validity and more meaningful educational interpretations [30,33].

2:

Curriculum Analysis and Task Selection

A curriculum task analysis was conducted on the Form 3 Design and Technology syllabus to identify design activities that (a) required integration of mechanical, electrical and ergonomic considerations, (b) involved spatial reasoning and iterative refinement, and (c) allowed sufficient conceptual openness for divergent thinking. The adapted scoring rubrics were operationalized to reflect design-based responses in a Design and Technology context.

Table 3. Criteria for design task selection.

Criterion	Description	Justification
Curriculum alignment	Covered throughout the intervention period	Ensures instructional validity
Cognitive demand	Requires spatial reasoning and system integration, crucial for creative thinking	Aligns with creativity constructs
Design openness	Allows multiple solution pathways	Supports divergent thinking
Assessment feasibility	Can be completed within a 60-minute time frame, practical for classroom implementation	Practical for classroom use

below summarizes the criteria for design task selection. Several potential design contexts were evaluated, including drone innovations, small household appliances, and washing machine modifications. The washing machine design task was selected as the primary assessment context based on its alignment with instructional content during the intervention period, consistency of cognitive demands and adequacy of design space for generating diverse creative responses.

3:

Development of Parallel Assessment Tasks

Two parallel task prompts were developed for pretest and posttest administration to minimize testing effects while maintaining equivalent levels of difficulty and cognitive demand. Both prompts required students to design an improved washing machine for household users, incorporating specified components such as the drum mechanism, motor system, and control interface.

Students were instructed to produce annotated sketches and brief written explanations within a 60-min time limit. Both task versions were designed to elicit responses scorable across the four TTCT creativity components without privileging specific technical solutions.

4:

Scoring Rubric Adaptation

The standard TTCT-F scoring rubrics (

Table 4. Adapted TTCT-Figural scoring rubrics.

Component	Score	Operational Definition	Example
1. Fluency	0–n	Number of distinct ideas	Touch screen, voice control
2. Flexibility	0–5	Conceptual categories	Mechanical, ergonomic
3. Originality	0–2	Statistical infrequency	<2% responses
4. Elaboration	0–3	Level of detail	Multi-view technical drawing

) were adapted to reflect Design and Technology design outputs. Operational definitions were contextualized to ensure that scores reflected meaningful design features rather than superficial embellishments. This norm-referenced approach ensured that originality judgments were empirically grounded rather than subjective.

5:

Expert Validation of the Adapted Instrument

Content validity was established through expert review by a three-member panel comprising a creativity researcher and two instructional design specialists as shown in

Table 5. Expert panel validation results.

Validation Criterion	Expert 1	Expert 2	Expert 3	Agreement
Appropriateness	Yes	Yes	Yes	100%
Language Clarity	Yes	Yes	Yes	100%
Item Clarity	Yes	Yes	No	66.7%
Scoring Clarity	No	Yes	Yes	66.7%
Objective Alignment	Yes	Yes	Yes	100%
Overall	4/5 (80%)	5/5 (100%)	4/5 (80%)	86.7%

Provided qualitative feedback for improvement.

. Experts independently evaluated the adapted instrument using a binary (Yes/No) protocol across five validation criteria [34,35,36].

Inter-Rater Agreement (IRA) among experts was assessed using Fleiss’ kappa (k = 0.74), indicating substantial agreement. Minor revisions were made following expert feedback, primary to clarify scoring descriptors and task wording. Student responses were evaluated by three raters: the principal investigator and two experienced Design and Technology teachers. Raters underwent calibration training using sample responses representing varying creativity levels.

All responses were initially scoring independently, followed by consensus discussions to resolve discrepancies. Raters were blinded to group assignment throughout the scoring process to reduce bias. This consensus-based approach prioritized scoring reliability while remaining feasible in an authentic school context.

Representatives’ student responses were selected to illustrate low, moderate and high-performance levels across four creativity components. Exemplars include annotated sketches and excerpts from written explanation, demonstrating variation in idea quantity (Fluency), category diversity (Flexibility), novelty (Originality) and design detail (Elaboration). These exemplars are provided to support transparency in scoring interpretation and rubric application [37,38,39,40].

3.4. Intervention

The group using AR worked with an app called MekaAR. This application allowed students to view 3D digital models of machine parts overlaid onto their real world workspace, visible through the device’s camera. Students worked in pairs, sharing a single Android. This setup was designed to make them collaborate, discuss ideas, and solve problems together. The comparison group received instruction on identical content pedagogy with traditional tools (textbooks, diagrams, paper-based sketching). The duration for this experiment was 4 weeks, with 2 sessions per week and 60 min per session (8 sessions, 480 min of instruction).

Table 6. The comparison between the experimental group and the control group activity.

Week	AR-Enhanced Instruction Group	Traditional Instruction Comparison Group
1	Session 1: AR Technology Orientation Introduction to MekaAR application interface and navigation Session 2: Design Analysis Through AR Visualization Analyzing existing washing machine designs using interactive 3D AR models.	Session 1: Design Analysis Through Static Media. Analyzing washing machine diagrams and cross-sectional illustrations from textbooks Session 2: Mechanical Principles and Initial Ideation Generate initial design concepts grounded in mechanical understanding
2	Session 3: Ideation and Concept Visualization: Visualizing conceptual designs using AR 3D models and experimenting with different component configurations in AR environment Session 4: Iterative Design with Peer Feedback Collaborative design refinement using shared AR di splay	Session 3: Iterative Design Development Iterative design refinement using paper sketching and colored pencils Session 4: Peer Feedback and Revision Structured peer feedback through verbal discussion and written comment cards.
3	Session 5: Mechanical Principles Exploration Investigating mechanical principles through interactive AR simulations. Session 6: Collaborative Problem-solving Applying mechanical principles to washing machine design challenge.	Session 5: Mechanical Principles Study Studying mechanical principles through textbooks. Session 6: Collaborative Problem-solving Applying mechanical principles to washing machine design challenges.
4	Session 7: Final Design Production Through the AR learning experience, students produce a final sketch of improvements for the washing machine product Session 8: Peer Evaluation and Reflection Formal peer presentations (5 min per student)	Session 7: Final Design Production Students produce a final sketch of improvements for the washing machine product Session 8: Peer Evaluation and Reflection Formal peer presentations (5 min per dyad)

shows a comparison of the activities conducted in the experimental group, which utilized an AR application as an instructional tool, and the control group, which maintained traditional teaching and learning methods.

4. Results

Table 7. Descriptive statistics of creativity components by group and time point.

Component	Time	Experiment Group (n = 23)		Control Group (n = 23)
		Mean	SD	Mean	SD
Fluency	Pre	1.57	0.66	2.61	0.66
	Post	5.78	1.62	4.22	0.93
Flexibility	Pre	1.57	0.59	2.13	0.82
	Post	3.83	1.77	2.78	0.85
Originality	Pre	1.09	0.42	1.43	0.59
	Post	2.57	0.99	1.57	0.66
Elaboration	Pre	1.96	0.88	1.30	0.93
	Post	6.35	2.81	2.26	1.57

Note: Experiment = AR-enhanced instruction group; Control= Traditional instruction group.

summarizes the descriptive statistics for all creativity components pretest and posttest. Baseline differences between groups were evident, with the control group scoring higher in Fluency, Flexibility, and Originality, while the experimental group showed higher initial Elaboration. At posttest, both groups improved across all components; however, gains were consistently larger for the experimental group, particularly for Elaboration and Fluency. Greater posttest variability in the experimental group suggests a wider range of creative responses following AR-enhanced instruction.

Specifically, the control group scored higher on Fluency (M = 2.61 vs. 1.57) and Flexibility (M = 2.13 vs. 1.57), while the experiment group had a higher score on Elaboration (M = 1.96 vs. 1.30). Secondly, both groups showed posttest improvements across all components compared to their pretest scores. Crucially, however, the experiment group demonstrated substantially larger gains, particularly for Elaboration (an increase of 4.39 points) and Fluency (an increase of 4.21 points). Finally, an analysis of the variability patterns indicates that the experiment group exhibited greater posttest variability, with larger standard deviations for both Elaboration (SD = 2.81) and Flexibility (SD = 1.77), suggesting a wider diversity of creative responses following the AR instruction.

Independent samples t-test (

Table 8. Pretest equivalence tests.

Component	Experiment M (SD)	Control M (SD)	t(44)	p	Cohen’s d
Fluency	1.57 (0.66)	2.61 (0.66)	−5.45	<0.001	−1.58
Flexibility	1.57 (0.59)	2.13 (0.82)	−2.70	0.010	−0.78
Originality	1.09 (0.42)	1.43 (0.59)	−2.31	0.026	−0.67
Elaboration	1.96 (0.88)	1.30 (0.93)	2.56	0.014	0.74

) revealed significant pretest differences across all components (p < 0.05), with moderate to large effect sizes. These findings justify the use of ANCOVA with pretest scores as covariates to adjust for baseline imbalance.

Shapiro–Wilk and Levene’s tests indicated violations of normality and homogeneity, of variance across all components (all p > 0.05;

Table 9. Shapiro–Wilk normality test results.

Component	Time Point	Group	Statistic	p-Value	Normality Assumption
Fluency	Pre	Experiment	0.878	<0.001	Not met
Flexibility	Pre	Experiment	0.850	<0.001	Not met
Originality	Pre	Experiment	0.703	<0.001	Not met
Elaboration	Pre	Experiment	0.888	<0.001	Not met
Fluency	Post	Control	0.884	<0.001	Not met
Flexibility	Post	Control	0.831	<0.001	Not met
Originality	Post	Control	0.872	<0.001	Not met
Elaboration	Post	Control	0.905	0.001	Not met

and

Table 10. Levene’s test for homogeneity of variances.

Component	Time Point	F-Value	p-Value	Equal Variance Assumption
Fluency	Pre	28.80	<0.001	Not Met
Fluency	Post	16.79	<0.001	Not Met
Flexibility	Pre	7.26	0.010	Not Met
Flexibility	Post	6.47	0.015	Not Met
Originality	Pre	5.33	0.026	Not Met
Originality	Post	16.16	<0.001	Not Met
Elaboration	Pre	6.01	0.018	Not Met
Elaboration	Post	37.13	<0.001	Not Met

). According, bootstrapped ANCOVA procedures (5000 respondents) were employed to obtain robust estimates.

The statistical analysis revealed significant heterogeneity of variances (meaning the spread of scores was significantly different between the groups) for all components measured, both at the pretest and posttest (all p < 0.05). These violations were particularly severe for Elaboration at the posttest (F = 37.13, p < 0.001). These findings provide additional strong justification for using the bootstrapping approach in the primary ANCOVA analysis, as this method is more robust and reliable when the assumption of equal variances is violated.

Bootstrapped ANCOVA results are presented in

Table 11. Bootstrapped ANCOVA results: AR effects on creativity components.

Creativity Component	Group	Adjusted Mean	Bootstrapped 95% CI	F (1,43)	p-Value	Partial η2
Fluency	Experiment	6.022	[5.41, 6.64]	17.896	<0.001	0.294
	Control	3.978	[3.36, 4.60]
Flexibility	Experiment	3.914	[3.31, 4.52]	7.593	0.008	0.150
	Control	2.695	[2.09, 3.30]
Originality	Experiment	2.644	[2.29, 3.00]	20.445	<0.001	0.322
	Control	1.486	[1.13, 1.84]
Elaboration	Experiment	6.097	[5.15, 7.05]	27.093	<0.001	0.387
	Control	2.511	[1.56, 3.46]

Note: Adjusted means control for pretest scores. All confidence intervals are bias-corrected and accelerated (BCa) bootstrap estimates based on 5000 iterations.

. After controlling pretest scores, significant group effects were observed for all creativity components (p < 0.008). Effect sizes ranged from moderate to large, with the largest effects observed for Elaboration, followed by Originality, Fluency and Flexibility.

AR-enhanced instruction was associated with significantly higher posttest creativity score across all components after controlling for baseline differences. The strongest effects were observed for Elaboration and Originality, with comparatively smaller but still significant effects for Fluency and Flexibility. This pattern suggests that AR may particularly support design detail, refinement, and novelty, although interpretation of underlying mechanisms remains tentative.

5. Discussion

This study examined the effects of AR-enhanced instruction on students’ creativity in Design and Technology education, focusing on Fluency, Flexibility, Originality, and Elaboration. Overall, the findings indicated that AR-enhanced instruction was associated with significantly higher posttest creativity scores across all components after controlling for baseline differences using bootstrapped ANCOVA [41,42].

Importantly, the effects of AR were not uniform across creativity components. Elaboration showed the largest effect, followed by originality, fluency and flexibility. This pattern suggests, rather than conclusively demonstrates, that AR-supported instruction in this context is consistent with social-constructivist learning mechanisms, in which shared digital artefacts facilitate collaborative dialogue and iterative refinement of ideas [18,43].

The strong effect observed for Elaboration may reflect the role of AR as a persistent and manipulable shared representation that supports peer explanation, critique and design refinement. In line with Vygotsky perspectives [18], AR appears to externalize thinking processes and scaffold collaborative meaning making. However, this interpretation remains theoretical, as social interaction processes were not directly measured.

Originality and Fluency also demonstrated substantial, though comparatively smaller, effects. These findings are consistent with the view that AR can reduce visual spatial cognitive load and performance anxiety by enabling low-risk digital prototyping, thereby supporting idea generation and exploration [16,44]. Their relative ranking suggests that cognitive and motivational pathways may have operated alongside but secondary to social mechanisms in this instructional context. Flexibility exhibited the smallest, albeit statistically significant, effect. This outcome may indicate that students’ domain knowledge constraints and tendencies toward early group convergence limited exploration across diverse conceptual flexibility without explicit pedagogical strategies that structure divergent category exploration [15,17].

Interpretations involving Malaysian collectivist cultural norms should be treated with caution. Although collectivist orientations may plausibly support collaborative learning and shared ownership of ideas, no cultural measures were collected in this study [43,45]. Accordingly, cultural explanations are offered as theoretical interpretations rather than empirically substantiated claims.

The findings underscore that the impact of AR on creativity is context-dependent, shaped by task design, instructional structure, and patterns of learner interaction. AR should therefore be understood not as a uniform intervention, but as a mediating tool whose educational value depends on how it is pedagogically integrated.

Comparison with Prior Research

This study’s effects (Elaboration d = 1.33, Originality d = 1.20, Fluency d = 1.13, Flexibility d = 0.81) substantially exceed meta-analytic averages for AR interventions, revealing important insights about when and why AR enhances creativity.

Martinez-Beneito et al.’s [13] meta-analysis of 61 studies reported an overall AR effect on learning outcomes of d = 0.68, with larger effects for procedural knowledge (d = 0.76) than conceptual knowledge (d = 0.54). Even this study’s smallest effect (Flexibility d = 0.81) exceeds the meta-analytic mean (d = 0.68), and the largest effect (Elaboration d = 1.33) is nearly double. This suggests AR may be particularly effective for creativity outcomes when instruction is structured around collaborative elaboration configuration underrepresented in Martinez-Beneito et al.’s [13] sample, which primarily examined knowledge acquisition (recall, comprehension) rather than creative production.

Akçayır and Akçayır’s [12] systematic review identified increased student engagement and enhanced collaboration as AR’s most consistent benefits. This study demonstrates how enhanced collaboration translates into measurable creativity gains: collaborative elaboration manifested as the largest effect on Elaboration specifically (d = 1.33). While previous reviews noted AR enhances collaboration, they did not specify how collaborative processes differentially affect distinct creativity components. However, the same collaborative processes that enhanced Elaboration may have inadvertently constrained Flexibility (d = 0.81) through convergence effects, revealing a cognitive social tradeoff unacknowledged in prior reviews.

Chen et al. [14] found the reverse pattern (Originality d = 0.89 > Elaboration d = 0.62), which is attributed to three moderator differences: instructional design (individual vs. collaborative), possible cultural context (though unverified), and measurement sensitivity (aesthetic vs. functional rubrics). This divergence demonstrates that AR does not uniformly enhance all creativity components; rather, its effects are component-specific and context-dependent. The same AR technology produced Originality dominant effects in Chen et al.’s [14] individual-task, aesthetic-focused context but Elaboration dominant effects in this study’s collaborative-task, functionality-focused context. This simplistic challenge “AR enhances creativity” claims common in educational technology literature and supports Amabile’s componential view that different environmental factors (task structure, social environment, evaluation criteria) affect different creativity components.

This research advances AR creativity literature through three contributions. First, differential effects analysis reveals AR’s impacts are highly differentiated, with social affordances (supporting Elaboration, d = 1.33) exerting stronger influence than cognitive affordances (supporting Fluency, d = 1.13) within collaborative settings, a component-level specificity obscured in prior research using aggregate creativity measures. Second, theoretical mechanism testing through hypothesis-driven design demonstrates that social constructivist mechanisms (H3) superseded cognitive (H1) and motivational (H2) factors when instruction was culturally and pedagogically aligned with collaborative elaboration. The rank order, Elaboration, Originality, Fluency and Flexibility—provide empirical evidence for pathway interactions proposed in H4’s integrated framework while revealing contextual contingency not anticipated in the original hypotheses. Third, context-sensitive framework development provides an integrated model specifying four moderators’ cultural context, instructional design, learner characteristics, and task assessment alignment that determine the relative strength of cognitive, motivational, and social pathways. This framework explains divergent findings across prior studies (e.g., why Chen et al. [14] found Originality dominant while this study found Elaboration dominant) and enables more precise predictions for future research, moving beyond “does AR work?” questions toward “when, why, and for whom does AR enhance specific creativity components?”

6. Conclusions

This study found that AR-enhanced instruction produced large to very large effects on all four creativity components in Malaysian Design and Technology education, with a theoretically significant pattern.

The findings support an integrated, context-sensitive theoretical model where AR’s effects depend on cultural context: collectivist values (may amplify social pathways, though unverified in this study); instructional design (collaborative structures amplify elaboration), learner characteristics (novices benefit from external scaffolding), and assessment alignment (tasks rewarding refinement amplify elaboration). This model may help explain divergent findings through prior research and offers a framework for predicting when AR will enhance specific creativity components.

Practically, the results suggest that Malaysian Design and Technology educators may benefit from prioritizing collaborative AR implementation with explicit scaffolding for category switching and domain knowledge integration.

For Malaysian educators and policymakers, AR may represent a high-impact pedagogical tool for creativity development in design-focused Design and Technology units, but successful implementation likely requires the following: (1) explicit collaborative instructional design, (2) adequate teacher professional development, and (3) realistic expectations about technology constraints.

7. Limitations

This study has several limitations that affect the generalizability and causal interpretation of findings. First, the sample included only female students from one urban school in Malaysia, limiting generalizability across gender and geographic contexts. Gender differences in spatial reasoning and collaborative behavior are well-documented, suggesting that AR effects may differ in male or mixed-gender groups. Critically, although Malaysian collectivist cultural values are invoked as explanatory mechanisms, such as collective ownership reducing evaluation anxiety and cultural amplification of social pathways (comparison with Chen et al. [14]), no cultural orientation measures were collected in this study.

The interpretation that collectivist values [46] amplify social mechanisms remains theoretical and speculative without empirical assessment using established cultural instruments, such as Hofstede’s Cultural Dimensions framework or other individualism–collectivism scales. This limitation constrains causal claims regarding the cultural moderation of AR’s social affordances. Cross-cultural comparative studies incorporating explicit cultural measurements are therefore necessary to determine whether collectivist values moderate AR’s effects on specific creativity components. Additionally, participants were mechatronic novices—Form 3 students following the Malaysian Design and Technology curriculum with no prior mechatronic training. AR may be particularly beneficial for novices requiring cognitive support and social scaffolding, whereas experienced students might benefit more from cognitive or motivational pathways, a hypothesis requiring empirical testing through expertise moderation studies.

Second, the four-week intervention may have favored immediate social effects over long-term cognitive internalization. The large Elaboration effect (η² = 0.387) may reflect temporary performance enhancement through external social scaffolding (recursive elaboration cycles) rather than durable internalization of elaborative thinking strategies. Longer interventions might reveal different patterns as students internalize AR-supported strategies. The collaborative instructional design, aligned with the Malaysian curriculum’s emphasis on teamwork [2], effectively amplified social affordances but may have underutilized cognitive affordances. Alternative designs incorporating individual exploration might produce different creativity patterns. Additionally, different AR platforms with advanced features (gesture-based interfaces, haptic feedback) might produce different effects on creativity components [12,13], potentially mediated by cognitive load and multimedia learning principles [16,44].

Third, measurement limitations must be acknowledged. The adapted TTCT-F scoring rubric emphasized curriculum specific features, which may have increased sensitivity to Elaboration compared to generic TTCT scoring. While this adaptation was intentionally designed to align with domain relevant creativity criteria, it may partially explain why Elaboration showed the largest effect size (η² = 0.387) relative to other components. This measurement choice affects the comparison with prior studies using generic TTCT scoring as different rubrics may capture different aspects of creative output [15,17]. The single-post intervention assessment captured only immediate effects; no delayed post tests were administered to assess retention or transfer after AR support was removed. While inter-rater reliability was adequate, subjective scoring introduces measurement error [47,48,49]. Future research should supplement expert scoring with objective behavioral measures such as design iteration frequency and digital trace data.

Fourth, the quasi-experimental design with intact classroom groups prevented random assignment. Although ANCOVA controlled pretest differences [28,29], unmeasured confounding variables (teacher effects, classroom climate) may have influenced outcomes. These confounds are particularly concerning cultural moderation interpretation, as classroom climate differences could produce collaborative patterns that mimic collectivist cultural effects without reflecting individual cultural values. The control group received traditional instruction rather than comparable technology, making it difficult to isolate AR-specific effects from general technological effects. Comparison with alternative digital tools would provide clearer differentiation. Additionally, a single teacher implemented both conditions, introducing potential experimenter bias despite scripted protocols.

These limitations suggest critical directions for future research, each directly addressing specific methodological concerns identified above: (1) to address the cultural measurement limitation replication in diverse cultural contexts with validated cultural orientation measures to empirically test whether collectivist values moderate AR’s creativity effects); (2) to address sample limitations examination of gender diversity, extended interventions with delayed posttests, and expertise moderation; (3) to address duration and retention concerns longitudinal studies assessing whether effects persist after AR support is removed; (4) to address the technology comparison limitation comparative studies isolating AR-specific affordances from general digital-tool effects; (5) to address measurement concerns triangulation using multiple creativity assessments (standardized tests, expert evaluations, portfolios, behavioral metrics); and (6) to address the lack of process data mixed methods approaches incorporating discourse analysis and classroom observations to provide richer mechanistic insights. Addressing these limitations would advance theoretical understanding of how AR technologies can be strategically deployed to enhance creativity in design education, as emphasized by Vygotsky’s sociocultural theory [18] and Amabile’s Componential Theory [17].

8. Implications and Future Research

This study contributes to learning theory by extending and challenging existing frameworks. First, it extends Vygotskian socio-constructivist theory [18] by suggesting that digital artifacts with persistence, manipulability, and shared visibility may function as powerful scaffolds that amplify social dialogue, a mechanism instantiated through the recursive elaboration cycles. Second, it suggests that Cognitive Load Theory (CLT) [16] may underestimate the role of social processes in enhanced learning technology, such as Elaboration (social pathway, d = 1.33) exceeding Fluency (cognitive pathway, d = 1.13) despite effective cognitive load reduction, a finding detailed that challenges CLT’s individual-focused framework. Third, it supports Amabile’s Componential Theory [17] by demonstrating synergistic interaction between motivational and social pathways, with Originality enhanced by both low-stakes experimentation (motivational) and collective ownership (social). Finally, these insights support an integrated, context-sensitive framework where AR effects on creativity depend on cultural context, instructional design, learner characteristics, and task assessment alignment [12,13,14,44].

Future research should pursue five directions. First, cultural moderation studies should replicate this study in individualist cultures with validated cultural orientation measures to test whether social pathways weaken while cognitive or motivational pathways strengthen. Second, instructional design experiments using factorial designs should manipulate collaboration structure (individual vs. group) and task emphasis (quantity vs. quality) to isolate which features amplify specific pathways addressing the instructional design limitation. Third, longitudinal studies with extended interventions and delayed posttests should determine whether effects persist after AR support is removed (indicating internalization) or fade (indicating dependence on external support) addressing the retention concern. Fourth, mechanism studies using process measures (video analysis of peer dialogue, eye tracking) should directly test proposed mechanisms (sustained joint attention, externalized thinking) rather than inferring them from outcomes addressing the lack of systematic observational data. Fifth, boundary condition studies should investigate moderators (gender, prior spatial ability, domain expertise, cultural background) to refine the context-sensitive model developed and enable personalized AR design [44,50,51,52].