JMIR Publications

ABSTRACT

Background:

Serious game-based training modules are pivotal for simulation-based healthcare training. With advancements in artificial intelligence (AI) and natural language processing (NLP), voice command interfaces offer an intuitive alternative to traditional VR controllers in VR applications.

Objective:

The aim of this study is to compare AI-supported voice command interfaces and traditional VR controllers in terms of user performance, exam scores, presence and confidence in Advanced Cardiac Life Support )ACLS trainings.

Methods:

Sixty-two volunteer students from Acibadem Mehmet Ali Aydinlar University Vocational School for Anesthesiology, aged 20-22, participated in the study. All the participants completed a pretest consisting of 10 multiple-choice questions about Advanced Cardiac Life Support (ACLS). Following the pretest, participants were randomly divided into two groups: the Voice command Group (VG) (n=31) and the Controller Group (CG) (n=31). The VG Group members completed the VR-based ACLS serious game in training mode twice, using an AI-supported voice command as the game interface. The CG Group members also completed the VR-based ACLS serious game in training mode twice, but they used VR controllers as the game interface. The participants completed a survey for estimating the level of presence and confidence during game play. Following the survey, participants completed the exam module of the VR-based serious gaming module. At the final stage of the study, participants completed a post-test, which had the same content of the pretest. VR-based exam scores of VG group and CG group were compared using an independent samples t-test and linear regression analysis was conducted to examine the effect of presence and confidence rating.

Results:

Both groups showed an improvement in performance from pretest to posttest, with no significant difference in the magnitude of improvement between the two groups. When comparing presence ratings, there was no significant difference between the VG and CG groups (VG: M = 5.18, SD = 0.83; CG: M = 5.42, SD = 0.75). However, when comparing VR-based exam scores, the CG group significantly outperformed the VG group (CG: M = 80.47, SD = 13.12; VG: M = 66.70, SD = 21.65), despite both groups having similar time allocations for the exam (VG: M = 18.59 min, SD = 5.28; CG: M = 17.33 min, SD = 4.83). Confidence levels were similar between the groups (VG: M = 0.23, SD = 0.18; CG: M = -0.09, SD = 0.18), but VG group displayed a significant overconfidence bias (VG: M = 0.09, SD = 0.24; CG: M = -0.09, SD = 0.18). Regression analysis revealed a marginally significant positive effect of presence on confidence scores (B = 0.23, t = 1.858, p = .07), suggesting that increased immersion in VR environments may enhance learners' confidence.

Conclusions:

AI voice recognition advancements could enhance presence and learning outcomes in VR-based training. Further research should explore ways to optimize AI's role in education through VR. Clinical Trial: ClinicalTrials.gov NCT06458452