As culturally and linguistically diverse classrooms become more prevalent, pre-service teachers’ empathy toward multicultural students and multicultural attitudes are increasingly critical for interpreting classroom behaviors, building supportive re...
As culturally and linguistically diverse classrooms become more prevalent, pre-service teachers’ empathy toward multicultural students and multicultural attitudes are increasingly critical for interpreting classroom behaviors, building supportive relationships, and responding to participation barriers shaped by language demands and peer dynamics. Yet teacher education often offers limited opportunities for trainees to experience such barriers from within authentic, socially contingent interactions. While conventional instruction can foster conceptual understanding, it seldom provides opportunities for learners to inhabit students’ experiences in a socially situated way. Immersive virtual reality (VR) has been proposed as a promising alternative because it can offer embodied first-person immersion, thereby transforming abstract knowledge about students’ struggles into self-relevant and emotionally resonant learning.
However, VR in teacher education remains at an early stage, and evidence for its effectiveness is inconclusive. More importantly, much prior research has emphasized outcomes while leaving the learning process under-specified, providing limited insight into when change emerges, how it unfolds during the experience, and why some effects persist. Because empathic and attitudinal change may be expressed through in-situ process signatures that retrospective self-reports do not fully capture, multimodal process evidence is needed to open this “black box” and to identify conditions under which socio-affective change is most likely to occur and endure.
The purpose of this study was to examine the effectiveness of an AI-powered VR empathy simulation that places pre-service teachers in the first-person perspective of a multicultural student facing language-related barriers, relative to a text-based learning group using the same narrative script. The research questions included: (RQ1) compared with the text group, how does the VR simulation affect pre-service teachers’ empathy and multicultural attitudes over time (pretest, immediate posttest, delayed posttest), including (RQ1a) cognitive and emotional empathy, (RQ1b) multicultural attitudes, and (RQ1c) perceived changes; and (RQ2) within the VR group, how do engagement and embodiment influence empathy and multicultural attitudes, and what mechanisms do learners attribute to these effects.
To examine the effectiveness of the AI-based VR simulation, this study employed a repeated-measures experimental design with assessments at pretest, immediate posttest, and a two-week delayed posttest. Ninety-six pre-service teachers from South Korea participated (VR group: n = 64; text group: n = 32). Participants in the control group completed an empathy-focused activity using the developed text materials, whereas those in the experimental group completed the same activity through a Head Mounted Display-based VR simulation. Surveys measuring empathy, multicultural attitudes, and related constructs were administered at each time point. Engagement was measured in both groups, whereas embodiment and eye-tracking were collected only in the VR group; HRV was recorded in both groups to capture process-sensitive regulatory signatures. In addition, participants completed a prediction writing task prior to the intervention and a reflective writing task afterward to capture expectations and post-experience meaning-making. Finally, post-intervention interviews were conducted and analyzed using thematic analysis to elucidate perceived mechanisms underlying change and its persistence.
Results indicated significant between-group differences favoring VR for both empathy and multicultural attitudes. Both groups showed short-term increases in overall empathy, but the VR group demonstrated higher posttest levels that remained and became more pronounced at delayed posttest. When empathy was decomposed, cognitive empathy showed comparable short-term gains across groups, while a clearer advantage for VR emerged at the delayed assessment. Emotional empathy exhibited a more consistent and immediate VR advantage, with higher levels at both posttest and delayed posttest, whereas the text group tended to decline or plateau across time. Multicultural attitudes also improved more positively in the VR group at both post-intervention time points, demonstrating significant between-group differences beyond baseline levels.
Process analyses clarified how these outcomes were supported. Within the VR group, engagement robustly predicted immediate post-intervention empathy and multicultural attitudes, indicating that stronger cognitive and emotional involvement during the experience was closely associated with initiating socio-affective change. Embodiment provided additional explanatory value for empathy, consistent with the notion that feeling physically and self-relevantly “in” the situation can amplify affective resonance. Multimodal process data further complemented these findings. Physiological analyses showed higher HRV in the VR group than in the text group during learning, indicating distinct autonomic regulation. Eye-tracking did not show stable independent predictive power in the primary regression models; however, exploratory analyses suggested that higher AFD may differentiate learners who exhibited larger gains in emotional empathy, pointing to overall engagement as a potentially informative process marker. Meanwhile, delayed multicultural attitudes were not reliably explained by in-situ process indicators once immediate post-intervention levels were considered, suggesting that longer-term attitudinal maintenance may additionally involve post-experience meaning-making and reflective integration beyond experiential intensity during the simulation. Qualitative analyses converged on mechanisms including a shift from “understanding from the outside” to “experiencing from the inside,” reinterpretation of student behaviors as responses to structural constraints, and reflection-driven reappraisal linked to responsibility awareness and emerging action intentions.
In sum, this study provides time-sensitive evidence that a AI–powered VR simulation can produce stronger and more durable improvements in pre-service teachers’ empathy and multicultural attitudes than text-based learning. By integrating engagement and embodiment with HRV and eye-tracking indicators as process-sensitive evidence, alongside writing and interviews data, the study advances a process-oriented account of socio-affective change and strengthens methodological approaches for evaluating VR-based education in teacher preparation.