Exploring Effectiveness of Immersive Virtual Reality-Based STEM Instruction Using the Cognitive-Affective Model of Immersive Learning (CAMIL)

Immersive Virtual Reality (IVR) has emerged as a promising instructional technology for STEM education, yet relatively few classroom-based studies have examined student learning experiences through the lens of the Cognitive-Affective Model of Immersive Learning (CAMIL). This study compared undergraduate students’ perceptions of IVR and traditional PowerPoint (PPT)-based instruction across four CAMIL-related constructs: immersion, presence, agency, and intrinsic motivation. A counterbalanced crossover design was implemented with 40 undergraduate STEM students enrolled in a leadership course at a large Midwestern research university. Participants experienced both IVR and PPT instructional modalities across three learning modules. Student perceptions were measured using an adapted questionnaire consisting of validated subscales administered on a six-point Likert scale. Because questionnaire completion varied across instructional sessions, analyses were conducted using 64 complete IVR-PPT paired observations. Repeated-measures ANOVA was used to analyze motivation, presence, and immersion, while the Wilcoxon signed-rank test was used to analyze agency because of non-normality. Results indicated a significant difference in immersion between instructional modalities, whereas motivation and presence did not differ significantly. Agency also differed significantly between conditions based on the Wilcoxon signed-rank test. These findings demonstrate that instructional modality influences selected dimensions of the student learning experience while suggesting that the relationships proposed by the CAMIL framework may depend on implementation context and instructional design. Rather than validating the CAMIL model, this study illustrates its usefulness as an interpretive framework for examining learner perceptions in authentic undergraduate STEM instruction. Limitations include the modest sample size, incomplete questionnaire responses across sessions, the absence of formal testing for sequence effects, and the use of paired observations rather than participant-level analyses.