Embedding Attributional Retraining in Voice-Responsive Low-Code Virtual Reality for Lower-Performing EFL Students: An Exploratory Implementation Pilot

This exploratory implementation pilot examined selected implementation and feasibility indicators for embedding structured Attributional Retraining (AtR) within a voice-responsive low-code virtual reality (VR) artifact-design environment and described preliminary outcome patterns to inform a future cluster-based trial. Over a 12-week artifact-design and exhibition-preparation program, 30 lower-performing EFL undergraduates from nursing, electronic engineering, and communication engineering formed six five-member teams using Meta Quest 3 headsets. Of 37 students screened, 30 met the inclusion criterion of scoring below 60 on the English subtest of the TVE Joint College Entrance Examination, and all eligible students enrolled. The shared VR environment combined learner-editable visual scripting, spoken-input triggers, and conditional response pathways, allowing interaction mismatches to be inspected, revised, and re-tested. Three teams received structured AtR feedback emphasizing specific and modifiable causes of difficulty, whereas three comparison teams received conventional technical and linguistic feedback without explicit attributional reframing. Implementation indicators included screening-to-enrollment yield, retention, team-level VR exposure, artifact completion, and outcome-data completeness. Secondary outcomes included the Personal Control subscale of the Revised Causal Dimension Scale (CDS-II), condition-blinded expert ratings of post-program exhibition performance, and participant interviews. Screening-to-enrollment yield was 81.1%, with 100% enrollment among eligible students and 100% retention. All six teams completed at least 10 documented VR design and/or rehearsal sessions, produced and presented functional VR artifacts, and provided complete CDS-II, expert-rating, and interview data. These results met the study-defined benchmarks for recruitment, retention, team-level VR exposure, artifact completion, and outcome-data completeness. However, overall feasibility, intervention fidelity, instructional-dosage equivalence, participant burden, and VR tolerability could not be fully established because relevant implementation, dosage, and safety records were not sufficiently standardized. Descriptively, the VR-AtR condition showed a larger pre/post difference in perceived Personal Control (1.53 points) than the comparison condition (0.11 points), although the two assessments were anchored to contextually related but non-equivalent performance events. Mean post-program communication-related exhibition scores were 58.20/75 for VR-AtR and 43.50/75 for the comparison condition. Participants were nested within only six project teams, standardized baseline communication performance was unavailable, exhibition questions and speaking opportunities were not fully standardized, and differences in instructor attention and feedback dosage could not be excluded. Accordingly, these outcome patterns should be interpreted as descriptive and hypothesis-generating rather than as evidence of causal intervention efficacy.