To ensure that the CoCreAL model development responded to a measurable, context-specific competency gap rather than a theoretical assumption, a systematic needs analysis was conducted prior to model development. Three diagnostic instruments were administered to the target population (30 third-semester PGMI students at UNIPDU Jombang) during the initial weeks of the 2024/2025 academic year: a Collaboration Skills Rubric adapted from ( Bach & Thiel, 2024), a Creative Thinking Assessment derived from ( Beghetto et al., 2015), and a self-report Digital Literacy Survey. Instrument validity was confirmed through expert review (CVI > 0.80 for all items) and internal consistency (Cronbach’s alpha: collaboration rubric α = 0.87; creative thinking assessment α = 0.84; digital literacy survey α = 0.81).

As presented in Table 1, the diagnostic assessment revealed that between 73.3% and 80.0% of participants scored below the institutional competency threshold across all three domains, with a composite baseline mean of 51.5 (SD = 8.4). These results confirm that the deficit targeted by CoCreAL is empirically present in the study population, not merely assumed from the general literature. The digital literacy deficit (80.0% below threshold) is particularly noteworthy given that AR integration constitutes a central component of the proposed model, as it indicated that AR scaffolding would need to be embedded from the outset rather than assumed as a pre-existing competency. These baseline findings directly informed the design priorities of CoCreAL’s five phases.

In response to the identified theoretical gap and empirically confirmed competency deficits, this study aims to: (1) design and develop the CoCreAL model with systematic theoretical grounding in constructivism and 21st-century skills frameworks; (2) validate the model and its assessment instruments using expert review and Content Validity Index analysis; (3) assess the model’s practicality and feasibility through student perception instruments; and (4) evaluate its effectiveness in improving collaboration and creative thinking skills through pre/post quasi-experimental design. The study contributes theoretically by extending constructivist pedagogy into AR-mediated learning contexts and practically by providing a scalable, validated instructional framework for Indonesian MI teacher education.

This study employed a Research and Development (R&D) approach following the Dick and Carey instructional design model, selected for its structured, iterative, and evidence-based development stages that produce pedagogically sound, contextually grounded instructional interventions. The Dick and Carey framework proceeds from initial needs analysis through systematic design, development, formative evaluation, revision, and summative evaluation, ensuring that each component of the model is grounded in theory, empirically validated, and aligned with learning objectives before large-scale implementation. This approach is recognized as particularly appropriate for the development of instructional frameworks in educational technology research ( Dick et al., 2015). The overall development procedure followed ten systematic stages, from identifying instructional goals and conducting learner analysis through to summative evaluation.

Participants consisted of 30 third-semester students enrolled in the Pendidikan Guru Madrasah Ibtidaiyah (PGMI) program in the Media Pembelajaran course at Universitas Pesantren Tinggi Darul Ulum (UNIPDU) Jombang, Indonesia, during the 2024/2025 academic year. Purposive sampling was employed on the basis that this population was simultaneously developing competencies in instructional design, media development, and technology integration and therefore represented the most contextually valid group for initial model evaluation. The sample comprised 22 female and 8 male students, with a mean age of 20.4 years (SD = 0.9). Informed consent was obtained from all participants prior to data collection. This study acknowledges that the use of a single cohort without a randomized control group constitutes a limitation in terms of attributional clarity, which is addressed in Section 5.5.

The CoCreAL (Collaborative, Creativity, Augmented Learning) model is a five-phase instructional framework grounded in Vygotskian social constructivism ( Gauvain, 2008), design thinking principles ( Liu et al., 2024), and the TPACK framework ( Mishra & Koehler, 2006). Each phase produces tangible outputs that build toward a final collaborative AR instructional product, ensuring both cognitive and applied skill development. The five phases are described below.

Phase 1 preparation (Digital readiness and orientation).

Students are oriented to 21st-century competency frameworks and introduced to the AR development platform (Assemblr EDU) through structured tutorials, live demonstrations, and a guided hands-on exercise. Baseline digital literacy is assessed, and learning objectives are made explicit through a competency transparency activity. The primary purpose of this phase is to reduce digital anxiety, identified as a barrier in 80.0% of the target population, before collaborative and creative demands are introduced.

Phase 2 collaboration (Social construction and idea development).

Students form heterogeneous groups of six, constructed to balance digital literacy levels based on baseline data. Groups identify authentic instructional challenges relevant to MI classroom contexts using structured scenario cards. Cooperative learning strategies as think-pair-share, jigsaw discussion, and structured role assignment, scaffold negotiation and co-construction of knowledge. The output is a collaborative idea map and project blueprint, which functions as the planning document for subsequent phases.

Phase 3 creativity (Divergent thinking and instructional design).

Groups engage in structured creativity scaffolding, applying SCAMPER ideation and design thinking problem-framing techniques to develop innovative instructional solutions. Two rounds of iterative peer critique (structured using a critique protocol adapted from design studio pedagogy) require students to give and receive evidence-based feedback before revising their storyboards. This phase explicitly operationalizes creativity as a process (ideation → prototyping → critique → revision) rather than a spontaneous trait.

Phase 4 implementation (AR development and iterative refinement).

Groups develop functional AR instructional modules using Assemblr EDU (version 3.2, compatible with Android and iOS devices). Modules incorporate 3D objects, multimedia overlays, text triggers, and interactive elements aligned with the instructional objectives formulated in Phase 2. A structured cross-group peer review session midway through this phase requires groups to evaluate one another’s prototype against the project rubric, generating written feedback used in a subsequent revision cycle. The output is a functional AR instructional module ready for presentation.

Phase 5 reflection (Presentation and metacognitive analysis).

Groups present their final AR modules to peers and the course instructor using a structured 10-minute presentation protocol. Post-presentation, students complete guided reflection journals using open-ended prompts targeting metacognitive awareness, perceived growth in collaboration and creativity, and self-identified areas for continued development. Post-tests are administered in this session.

The CoCreAL model was implemented over five sessions of 150 minutes each (3 × 50 minutes), spanning five consecutive weeks. Table 2 provides a detailed session-by-session implementation log documenting activities, AR tools used, student outputs, and session duration. Implementation fidelity was monitored by an independent observer using a structured checklist, confirming that 94.7% of planned activities were executed as designed; minor deviations involved time redistribution within sessions rather than omission of activities.

Multiple instruments were developed and validated to collect data across evaluation stages. All instruments underwent pilot testing with a comparable group of 15 PGMI students from a different institution, and internal consistency was verified using Cronbach’s alpha prior to the main study.

Expert validation sheets were developed with 38 items across six validation dimensions (see Table 3), derived from theoretical frameworks including constructivism, TPACK, and 21st-century skills frameworks. Items were formulated at the sub-criterion level to ensure granular diagnostic utility. Three validators for the research specialists in instructional design, educational technology, and MI teacher education, respectively that each of expert evaluated each item on a five-point scale. The Content Validity Index (CVI) was calculated at both item and scale levels using the formula proposed by Lynn (1986), with an item-level CVI threshold of ≥0.78 required for retention.

Student practicality questionnaires comprised 24 items across six dimensions (Cronbach’s alpha = 0.89), administered post-intervention using a five-point Likert scale. Collaboration Skills Rubrics (20 items, Cronbach’s alpha = 0.87) and Creative Thinking Assessments (20 items, Cronbach’s alpha = 0.84) were used for pre- and post-testing. Construct validity for all instruments was reinforced through expert judgment and confirmed CVI scores. The observation checklist (30 items, Cronbach’s alpha = 0.86) was used by the independent observer to document behavioral engagement and implementation fidelity. Reflection journals were completed by all 30 participants across five structured entries per student (150 journal entries in total), providing the qualitative data corpus.

Quantitative data from validation and practicality instruments were analyzed descriptively using mean scores, standard deviations, and percentage categories. Effectiveness was evaluated using paired-sample t-tests comparing pre- and post-test scores, with effect size calculated as Cohen’s d. The assumption of normality was verified using the Shapiro-Wilk test (W = 0.973 for collaboration; W = 0.968 for creativity; both p > 0.05). The alpha level was set at 0.05.

Qualitative data from the 150 reflection journal entries were analyzed using reflexive thematic analysis following Braun and Clarke’s (2006) six-phase framework. Two researchers independently coded a 30% random sample (45 entries) using an initial inductive codebook; inter-rater reliability was acceptable (Cohen’s Kappa = 0.81). Emergent codes were grouped into candidate themes through iterative discussion, and themes were finalized after member-checking with six student participants confirmed interpretive accuracy. Disconfirming evidence was specifically sought during the review phase to ensure analytical balance.

Prior to reporting model validity scores, Table 3 presents the Content Validity Index (CVI) for the expert validation instrument itself, establishing the credibility of the validation framework.

The overall CVI of 0.94 confirms that the validation instrument itself possesses strong content validity. All six validation dimensions achieved CVI values classified as Excellent, providing a sound methodological basis for interpreting the model validity scores reported in Table 4.

The overall validity score of M = 4.62 confirms that the CoCreAL model is theoretically robust and structurally sound for implementation in MI teacher education contexts. The highest sub-criterion score (4.78, clarity of learning outcomes) reflects the explicit alignment of each model phase with measurable performance objectives. The lowest score (4.45, student usability of AR tools) prompted a targeted revision: the Preparation Phase was extended by 20 minutes and an additional step-by-step AR tutorial handout was developed before the main implementation. All dimensions exceeded the minimum validity threshold of 4.00 established a priori by the research team. Inter-rater reliability of ICC = 0.91 indicates excellent agreement among validators, lending confidence to the consistency of expert judgment.

The independent observer’s checklist confirmed that 94.7% of planned CoCreAL activities were implemented as designed across all five sessions. The minor deviations observed (5.3%) were confined to Session 3, where the second peer critique round was shortened by approximately 15 minutes due to slower-than-anticipated storyboard completion by two groups. These groups were provided with additional asynchronous feedback via the course’s Google Classroom platform prior to Session 4 to compensate for the reduced in-class time. No activities were entirely omitted, and the core sequence of all five phases was preserved.

Post-intervention questionnaires (n = 30) assessed the model’s practicality across six dimensions. Table 5 presents results with item counts and standard deviations.

The overall practicality score of M = 4.55 (SD = 0.37) confirms that students found CoCreAL comprehensible and executable in authentic learning conditions. The highest dimension score — engagement and motivation (M = 4.61, SD = 0.33) — aligns with the model’s deliberate use of AR as a novelty-driven engagement mechanism. Crucially, the ease of understanding AR tools (M = 4.48, SD = 0.41) demonstrates that the scaffolded digital onboarding in Phase 1 succeeded in reducing the technological barriers identified in the needs analysis, where 80.0% of students initially fell below the digital literacy threshold. The standard deviations across all dimensions are narrow (0.33–0.41), indicating consistent perceptions across the cohort rather than polarized responses.

Pre- and post-tests were administered to all 30 participants. Normality of score distributions was confirmed by the Shapiro-Wilk test prior to parametric analysis (all p > 0.05). Table 6 presents descriptive statistics and effect sizes.

The composite mean gain of 19.20 points (29.7% improvement) represents a practically substantial advancement from below the institutional competency threshold (pre-test M = 64.60) to well above it (post-test M = 83.80). Both skill dimensions showed improvements of similar magnitude (collaboration: 28.6%; creative thinking: 30.9%), which is theoretically consistent with the interdependent scaffolding design of CoCreAL’s phases. Notably, the post-test standard deviation decreased relative to the pre-test (SD = 5.68 vs. 6.63), indicating that CoCreAL produced not only mean-level gains but also a reduction in inter-individual competency dispersion, a finding that supports the model’s equity-relevant design principle of scaffolding for diverse ability levels.

The Cohen’s d values of 1.88 (collaboration) and 1.94 (creativity) both exceed Cohen’s (1988) threshold for a large effect (d ≥ 0.80) by a substantial margin. Such effect magnitudes, while rarely achieved in controlled trials, are not atypical in single-group pre-post designs where participants begin well below competency threshold and receive intensive, purpose-designed instruction. The absence of a control group limits causal inference and is addressed in Section 5.5.

As shown in Table 7, the paired-sample t-test results confirm that pre-to-post score improvements are statistically significant for all dimensions (all p < 0.001). The t-values of 12.82 and 13.25 for collaboration and creativity, respectively, are notably consistent, corroborating the argument that both competencies developed as co-dependent outcomes rather than independently. These findings provide robust quantitative evidence that the CoCreAL intervention produced measurable, statistically significant improvements in the target competencies within the study population.

Thematic analysis of 150 student reflection journal entries (five entries per participant, n = 30) yielded four primary themes and one disconfirming theme. Each theme is reported with frequency data, illustrative evidence, and analytical interpretation grounded in both the qualitative corpus and the quantitative findings. Table 8 presents the full thematic analysis summary.

The qualitative findings provide explanatory depth for the quantitative outcomes. Collaborative engagement was the most prevalent theme (90.0% of participants), consistent with the statistically significant collaboration gains (t = 12.82, p < 0.001) and suggesting that the structured group activities in Phases 2 and 4 functioned as intended. The second most prevalent theme creativity development (83.3%), corroborates the creative thinking post-test improvements and reveals that students attributed their growth explicitly to the iterative, process-oriented design of Phase 3 rather than to individual ability.

The digital confidence theme (73.3%), while not directly assessed through a quantitative post-test, is analytically significant: its higher prevalence among students who scored lowest on the digital literacy baseline suggests a targeted effect on the most technologically vulnerable learners in the cohort. This finding has implications for the model’s relevance in resource-constrained MI educational settings. The reflective awareness theme (66.7%) was the least prevalent, which warrants interpretive caution: the structured reflection prompts may have produced metacognitive articulation in some students but not in others, suggesting that Phase 5 reflection scaffolding could be strengthened in future implementations.

The disconfirming theme (26.7%) that encompassing device compatibility issues and intra-group conflict, is reported as analytically essential rather than incidental. The five students who documented AR application failures on their personal devices represent a real implementation boundary condition: Assemblr EDU’s compatibility requirements may constitute a practical barrier in contexts where students do not have access to devices meeting minimum specifications. These cases do not invalidate the model’s effectiveness, but they do identify a critical prerequisite for implementation that must be addressed in scaling decisions.

Final AR instructional modules were assessed by two independent raters using a standardized rubric. Table 9 presents results across evaluation criteria and sub-criteria.

The overall project quality score of M = 4.56 demonstrates that the CoCreAL model cultivated not only measurable competencies but also tangible applied capabilities in creating technology-enhanced instructional media. The highest sub-criterion score (4.65, novelty of instructional concept) confirms that the explicit creativity scaffolding in Phase 3 produced original instructional designs rather than reproductions of existing media. The slightly lower score on congruence with MI learning objectives (4.45) suggests that the integration of Islamic educational content requirements with AR design demands is an area warranting additional scaffolding in future iterations — a practical refinement point consistent with the limitation noted by the reflective awareness theme.

The collaboration post-test improvement of 28.6% (t = 12.82, p < 0.001, d = 1.88) reflects a convergence of two design features that distinguish CoCreAL from less structured collaborative models. First, heterogeneous group formation based on baseline digital literacy data ensured that collaboration involved genuine interdependence rather than default division of labor by self-selected ability. Students who entered below the digital literacy threshold (80.0% of participants) were distributed across all five groups, creating conditions in which knowledge-sharing was structurally necessary. Second, the explicit collaborative role assignments in Phase 2 (navigator, designer, critic, integrator, and presenter) operationalized Vygotsky’s zone of proximal development ( Gauvain, 2008) at the group level, ensuring that peer scaffolding was directionally embedded into the task structure rather than left to emerge spontaneously. The reflection journals support this interpretation: 90.0% of participants documented collaborative engagement experiences, and the sub-theme of role clarity appeared in 17 of 30 journals, a frequency consistent with the deliberate role-assignment design.

These findings align with ( Ceballos et al., 2026) systematic review finding that collaborative problem-solving produces higher-order thinking gains when task design specifies both interdependence structures and individual accountability. They also extend ( Bach & Thiel, 2024) work on digital collaborative learning quality by demonstrating that structured role assignment in AR development tasks generates the kind of ‘quality interaction’ associated with significant learning gains, even in populations with initially low collaborative competence.

The creative thinking post-test improvement of 30.9% (t = 13.25, p < 0.001, d = 1.94) exceeded the collaboration gain by a modest margin, a finding that is theoretically interpretable rather than incidental. The Creativity Phase of CoCreAL subjects students to two full ideation-prototype-critique-revision cycles before they reach the AR development phase. This iterative structure directly addresses what ( Beghetto et al., 2015) identified as the central limitation of creativity in educational settings: the tendency to treat creative output as a first-draft phenomenon rather than a developmentally scaffolded process. The SCAMPER framework and design thinking protocols in Phase 3 provided students with structured divergent thinking tools that reduced the blank-page paralysis commonly observed when pre-service teachers encounter open-ended design tasks.

The pattern in the reflection journals supports this interpretation: creativity development was the second most prevalent theme (83.3%), and students consistently attributed their gains to the iterative process rather than to innate ability. The quote from S22 that creativity is ‘built step by step’ encapsulates the process orientation that Phase 3 was designed to produce. This finding has a practical implication: creativity can be systematically cultivated in teacher education through structured iteration, provided the instructional design makes the iterative process explicit and time-protected rather than compressed or treated as optional.

The practicality data (AR usability M = 4.48) and the digital confidence theme (73.3% prevalence) together indicate that CoCreAL’s AR integration functioned as intended: not as a demonstration of technological capability but as a design medium that expanded the creative and collaborative possibilities available to students. This distinction is critical. Prior research has documented that AR in teacher education frequently fails to produce meaningful learning gains when introduced as a supplementary demonstration rather than as a required component of instructional design ( Apriyani et al., 2023). CoCreAL inverts this arrangement: students are not shown what AR can do but are required to use AR as their primary design medium, which means that their engagement with the technology is necessarily purposeful and iterative.

Three specific mechanisms appear to explain AR’s contribution to the observed outcomes. First, the 3D object manipulation and interactive overlay capabilities of Assemblr EDU provided a design vocabulary that students found generative, reflection journal entries in the creativity development theme frequently referenced the AR platform’s capacity to represent abstract MI curriculum concepts spatially. Second, the requirement to develop a functional AR module as the primary course output created a high-stakes, authentic performance task that sustained motivation over five sessions, consistent with ( Prasetya et al., 2024) meta-analytic finding that AR-based motivational design effects are stronger when AR production (not merely AR consumption) is the required student activity. Third, the cross-group peer review of AR prototypes in Phase 4 created a distributed evaluative community in which students judged one another’s work against pedagogical and creative criteria.

The disconfirming evidence from device compatibility challenges (five students, 16.7% of the cohort) introduces an important boundary condition to this otherwise positive picture: AR as a pedagogical driver is contingent on adequate device access. In the current study, eight participants used institutional tablets that met Assemblr EDU’s minimum specifications, but five participants experienced intermittent application failures on personal devices. This finding aligns with ( O’Connor & Mahony, 2023) caution that AR’s positive effects on academic self-efficacy are mediated by technology access equity. Future implementations of CoCreAL in resource-constrained MI institutions will need to address device specification requirements as a prerequisite condition.

The CoCreAL model’s composite effect size of d = 1.91 exceeds those reported by the most directly comparable interventions. ( Li et al., 2025) meta-analysis of AR in higher education reported a pooled effect size of d = 0.64 for AR-integrated courses, and ( Liu et al., 2024, p. 202) study of design thinking models in pre-service teacher creativity reported d = 0.88. While direct comparison is complicated by differences in outcome measures, population characteristics, and design, the CoCreAL findings suggest that the combination of explicitly scaffolded collaboration, structured creativity development, and AR production in a unified instructional framework may produce additive or synergistic effects beyond those achievable by any single component in isolation.

Where CoCreAL most clearly differentiates itself from PjBL and PBL approaches is in its explicit treatment of creativity as a sequenced pedagogical target rather than an emergent outcome. ( Bell, 2010)‘s analysis of PjBL noted that creativity tends to appear in project-based contexts only when students have sufficient prior creative experience, a condition not reliably present in pre-service teacher populations with limited instructional design backgrounds. CoCreAL addresses this by dedicating an entire instructional phase to creativity scaffolding before project execution begins, thereby removing the pre-experience prerequisite. The project output evaluation scores (creativity and originality M = 4.62, innovation category) provide product-level evidence that this scaffolding approach produces substantially original designs, not merely technically competent reproductions of existing media.

The CoCreAL model’s alignment with Indonesia’s Merdeka Belajar policy, SN-Dikti standards, and KKNI competency frameworks is not incidental. The model’s design explicitly incorporates the competency taxonomy prescribed by these frameworks (collaboration, creativity, critical thinking, and technology integration) and operationalizes them through an instructional sequence that mirrors the project-based learning principles that Merdeka Belajar advocates. This alignment means that CoCreAL can be integrated into existing PGMI curricula without requiring structural program reform, a practical advantage that lowers the adoption threshold for institutions with limited curriculum development capacity.

The specific application within the MI context reveals an additional finding that the quantitative data alone do not capture: the reflection journals contain no evidence of perceived tension between AR-based creative design and Islamic educational values. Students’ descriptions of designing AR modules for Quranic vocabulary, Islamic history, and fiqh instruction suggest that the model’s open-ended design space accommodates culturally and religiously grounded content naturally, without requiring value modification. This contextual compatibility is an important signal for policymakers considering the scalability of CoCreAL to other MI institutions in Indonesia.

This study has five identifiable limitations that constrain the generalizability and causal strength of its findings. First, the absence of a randomized control group means that the observed pre-post gains cannot be attributed exclusively to CoCreAL as an isolated causal mechanism: maturation effects, instructor enthusiasm (Hawthorne effect), and practice effects from repeated testing may have contributed to score improvements. Future research should employ a waitlist control or parallel-group design to strengthen causal inference.

Second, the sample size (n = 30) and single-institution context limit statistical power for subgroup analyses and restrict the generalizability of findings to other PGMI institutions, regional contexts, or subject areas. Third, the overlap between the model developer and the course instructor introduces experimenter bias: students may have responded more favorably to the model’s demands because of their relationship with the instructor rather than because of the model’s design properties. Fourth, the self-report nature of the practicality questionnaire and reflection journals introduces social desirability bias, particularly in a course where the instructor both designed and evaluated the model. Fifth, the five-session implementation window does not permit assessment of skill retention or transfer to authentic classroom practice, a limitation that only longitudinal follow-up studies can address.

These limitations do not invalidate the study’s contributions but define the precise scope within which its findings can be interpreted with confidence. The data support the conclusion that CoCreAL produces statistically significant, practically large improvements in collaboration and creative thinking within a specific, well-characterized population under the described implementation conditions. Claims beyond this scope require the replication evidence that future research should provide.