Relative valuation of oracy-related criteria among engineering learners following VR-based and traditional instruction: an AHP study

1 Introduction

Over the last decade, the use of virtual reality (VR) and extended reality (XR) has become more prevalent in higher education. The increase in this is due to the improvements in head-mounted display hardware, real-time rendering and the reduction in the cost of immersive platforms. There is a large amount of empirical evidence available, providing evidence of learning outcomes in several subjects such as science education (Yu, 2021), language acquisition (Parmaxi, 2023), and other subjects (Yang et al., 2021). In general, the findings from the studies on the use of VR-based instruction to improve the measurable outcomes of learning have been confirmed by the findings from the meta-analytic and systematic review studies (Yu, 2021; Petersen et al., 2022). A key finding that has remained consistent in the most stringent recent reviews, however, is that the field has been more concerned with showing the effectiveness of VR than it has been with understanding how it works (Buchner and Mulders, 2026; Oje et al., 2025; Reigeluth and Honebein, 2023). Buchner and Mulders (2026) refer to the dominant research paradigm at the time, which was focused on media comparison and focused on whether VR is superior to a control condition instead of on the theoretical mechanisms of VR’s effects. Oje et al. (2025) concur with this criticism, noting that evidence-based pedagogical principles rarely form the basis of designing educational VR environments, and that there is an ambiguous relationship between the immersive affordances and the learning outcomes. The overall context of educational technology research is structurally oriented towards the construction of theoretical foundations, often in the form of conflicting ones, as indicated by the inconsistent and sometimes contradictory results characteristic of VR learning research studies, as explained by Schneider et al. (2025). The current study directly contributes to the Frontiers in Virtual Reality research topic that focuses on Immersive Learning with Virtual Reality, which aims to research the mechanisms, conditions and human factors that impact upon the learning experience in the immersive environment. In particular, it argues that the impact of VR-based learning on first-generation engineering learners might not be on skills as most frameworks and measures imply, but on how the learners perceive the technology and its connection to their future careers.

1.1 The problem of oracy in engineering education

Oracy (the ability to effectively use spoken and multimodal communication in professional, technical and social contexts) is a skill always identified as an important graduate attribute in Engineering education. However, it is not structurally developed in curricula that focus on imparting technical knowledge rather than communicative competence. Engineers are often asked to communicate highly technical information to non-technical audiences, collaborate with others from different fields of engineering on projects, and advocate for their organisations in client and regulatory settings. When this happens, communicative problems often have work-related consequences that go beyond the effects of minor technical problems. The situation is worsened in the regional scenario of engineering education in India, where first-generation learners (those whose parents have not completed their tertiary education) face multiple disadvantages. The students are usually those who have attended secondary schools in regional language medium schools, and they enter undergraduate programmes with little previous experience of studying English-medium professional communication.

Furthermore, they do not have the same informal networks of professional socialisation from which students from more privileged backgrounds learn communicative norms and develop consciousness beyond the formal classroom setting. The traditional oracy pedagogy, with decontextualised grammar tasks and scripted presentation tasks as well as assessment systems based on standardised written English, has proven to be insufficient to promote the students’ communicative development, especially when it requires authentic, contextualised and low-stakes environments for practice. A pedagogically feasible solution to this deficit is the VR-based immersive learning that can create a communicative context of professional activities with a fidelity that cannot be achieved in the classroom.

1.2 VR as a pedagogical intervention for oracy development

Many theoretical models have been suggested to explain the pedagogical applications of ILs, especially in terms of communication skills development. Embodied cognition is the idea that cognition transcends abstract symbolic manipulation and involves the body, the sensorimotor system, and learning contexts. By immersing learners in virtual environments that replicate professional contexts, virtual reality (VR) can help encode communicative knowledge in a manner that cannot be achieved through classroom instruction (Spangenberger et al., 2024). Likewise, situated learning theory has asserted that competence is best developed through engaging in authentic communities of practice, and that the best transfer to the real world occurs when the learning context is similar to the real context (Parong, 2021). Petersen et al. (2022) provide empirical evidence for both immersion and interactivity, two important characteristics of VR, to be effective in improving learning outcomes, with interactivity showing a greater effect on knowledge transfer. For a communication-related domain, Yang et al. (2021) provide direct evidence that a VR intervention for improving students’ descriptive writing is effective and engaging over traditional instruction, suggesting that VR environments can facilitate students’ expressive language production and receptive language comprehension. Stenberdt et al. (2025) also demonstrate that generative learning activities in VR–ones that require learners to re-experience and self-explain content–can have different impacts on retention and transfer based on the type of generative demands placed. These frameworks and findings often assume that skill acquisition is the main and/or most significant result of the learning process and is a theoretically important outcome. Reactions to the immersive environment in terms of motivation and attitudes are frequently viewed as conditions that either moderate or enable the reactions, rather than reactions themselves. Weech et al. (2019) point out that the subjective experience of being in the virtual environment is a key construct in VR research, but there is a lack of consistency in theorising the relationship between this construct and learning outcomes. The present study suggests that attitudinal orientation towards the technology may be another important, but under-researched construct. This lack of existing theory is a gap that is identified in the present work and is informed by the analytical approach of this study.

1.3 The attitudinal gap in VR education research

Based on the results of the systematic reviews on VR-based learning conducted by Yu (2021), Oje et al. (2025), and Buchner and Mulders (2026), the key measured variables in empirical studies of VR-based learning are knowledge retention, task performance accuracy, and reduction of cognitive load. Valtonen et al. (2022) identified a tendency in the field of educational technology research to copy dependent variables used in earlier research, without using a theoretical approach to develop a new one. This inclination may contribute to the continued emphasis on skill-based measures of outcomes, even in settings where motivational or attitudinal outcomes may be more relevant. Although the authors have not found any specific study that has placed attitudinal changes toward the learning technology itself (in terms of beliefs about the usefulness of the learning tool, willingness to recommend its use to others, or perceived relevance of learning to future career plans) as primary outcome variables in educational research on VR, they have found that perceptions of usefulness and perceived relevance of the technology itself are seldom treated as primary outcome variables in educational research on VR. If attitudinal measures are also incorporated, these are usually used as secondary indices to confirm the skill-based results or as covariates to account for the inter-individual differences in the skill-based results. In a systematic literature review on VR use in L2 learning, Parmaxi (2023) points out a common pattern whereby motivational and engagement outcomes are observed as an ancillary result, not as a primary outcome. Reigeluth and Honebein (2023) contend that educational technology research has not advanced to a level of complexity that warrants the creation of an instructional theory framework that is large enough to capture the complexity of media effects, and that results are not easily generalisable or applicable to the design of technology. This study highlights a theoretically important gap in which, if students were asked to explicitly compare and contrast the relative importance of different aspects of their immersive learning experience, they would consistently rate the attitudinal/motivational aspects higher than the direct skill-effectiveness aspects. Then, pedagogical design frameworks that focus more heavily on optimising for measurable competency outcomes may not align with the mechanism by which VR works. Traditional Likert scale instruments require the respondents to rate all of the criteria favourably (Saaty, 2008; Forman & Selly, 2001) while the Analytic Hierarchy Process (AHP) is a methodologically appropriate instrument to uncover these latent valuation structures by asking respondents to explicitly rate the relative importance of a number of criteria. For ease of understanding, attitudinal inversion in this manuscript is defined as the implicit null hypothesis, which would be that the priority weight of the direct skill-effectiveness criterion would be expected to be the greatest within each hypothesis, and that the inversion occurs when the actual empirical priority weight is the opposite (consistent with the traditional focus of the field on measurable competency outcomes; Yu, 2021; Valtonen et al., 2022).

1.4 The present study has the following goals:

To analyse the priority weights assigned by the 144 first-generation engineering students from three sites of Tamil Nadu in four hypotheses for five oracy-related criteria with the help of AHP-generated pairwise comparison matrices and normalised priority vectors.

Whether attitudinal and motivational criteria (e.g., attitude change toward immersive e-learning and willingness to recommend immersive e-learning to peers) are consistently found to be more important than direct skill-effectiveness criteria in all four hypothesis frameworks.

To validate the priority pattern across Tiruvallur, Kancheepuram and Gummidipoondi sites to see if it is a structural attitudinal feature of the learner cognition or because of the local institutional conditions.

Findings are discussed in the context of a theoretical framework that is based on self-determination theory and attitudinal orientation towards technology. These goals make the study a contribution towards theorising immersive learning mechanisms and towards the evidence base for pedagogically appropriate integration of VR in regional engineering education contexts, as the first generation learners are the main population of students.

2 Theoretical framework

The empirical pattern found in this study, where attitudinal and motivational criteria are always the highest criteria in learners’ evaluative hierarchy in comparison to their skill-effectiveness criteria, requires a theoretical explanation. This explanation should include the reasons that VR learning mainly leads to attitudinal change, and that this attitudinal change is specifically expressed as peer advocacy behaviour. Two complementary theoretical models are used to offer this explanation: the self-determination theory (SDT) and the technology acceptance model (TAM) are used together with diffusion of innovations theory. These frameworks provide a comprehensive understanding of the attitudinal inversion that was found in the AHP priority data. The level of self-determination need satisfaction in the learning experience is one factor that influences a person’s attitudinal orientation towards technology.

2.1 Self-determination theory and attitudinal orientation

Self-determination theory (Deci and Ryan, 1985) and subsequent empirical work suggest three universal human motivational needs: autonomy (experience of volition and self-endorsement in their actions), competence (experience of effectiveness and mastery in their environment), and relatedness (experience of meaningful connection with others). As long as these needs are met, intrinsic motivation is turned on and maintained; if these needs are not met, motivation becomes extrinsic (controlled) or amotivated. In a systematic review and meta-analysis of SDT interventions in education, Wang et al. (2024) found that need satisfaction in general is a consistent predictor of positive motivational and attitudinal outcomes in a variety of educational environments, with autonomy support being the strongest predictor of continued engagement. To extend this body of evidence, Hagger and McAnally Star (2026) showed that SDT constructs predicted behaviour in a range of technology adoption contexts, further suggesting that SDT is relevant in contexts other than education.

SDT has been found to be especially fruitful as an explanation model in the context of technology-mediated learning. Luo and Day (2026) showed that need satisfaction variables obtained from SDT significantly predicted lecturer readiness to adopt generative AI in higher education, and Chiu et al. (2026) studied teacher education for AI literacy using the SDT lens and revealed that the autonomous motivation variable was the mediator between competence support and technology adoption intention. In a gamified intelligent tutoring context, Hsia et al. (2025) demonstrated convergent evidence for measuring engagement and perceived competence improvements resulting from SDT-based instructional design over traditional instructional design. What is directly implied for the present study is that VR-based immersive learning environments can stimulate learners’ intrinsic motivation and positive attitude towards the technology more than other types of learning environments, as they allow the learner to experience the growth of competence in a realistic professional context without facing real consequences, while enabling the learner to make choices and decisions on their own. In this regard, attitudinal change can be thought of not as a consequence of VR skill acquisition but rather as a preconditioning mechanism, as this process was also found in the control group (as reported in Section 4.4), making VR skills more likely to be responsible for the degree of engagement rather than its presence. Here, this attitudinal activation (explicit comparative valuations) is interpreted as empirical evidence for the attitudinal activation predicted by SDT, as the consistently high AHP priority weight (0.45–0.51) across all sites and hypotheses suggests that attitude toward the utility of immersive learning is not only a background moderating condition, but also an attitudinal activation.

2.2 Technology acceptance and peer diffusion

The technology acceptance model (TAM) was first proposed by Davis (1989) and later extended in several ways, which states that perceived usefulness and perceived ease of use are the most important factors driving technology adoption intention. In this list, perceived usefulness has a more significant effect on behavioural intention, especially in an educational environment (Venkatesh and Davis, 2000; Venkatesh et al., 2003). The high analytic hierarchy process (AHP) weighting students assigned to the likelihood of recommending immersive learning to peers (0.26–0.27 across all four hypotheses and three sites) indicates that students’ positive acceptance of technology has led to their intentions to recommend it to peers. High perceived usefulness is a downstream consequence of this endorsement intention, as identified by Abdullah and Ward (2016) and Zou et al. (2025) in their TAM-based research. A complementary view, proposed by Ajzen (1991), suggests that subjective norms, which are the perceived social pressures due to important others’ behaviour and endorsement, directly influence the intention to engage in the behaviour. Peer recommendation acts as a mechanism for transmitting subjective norms in students’ communities.

The diffusion of innovations theory is best suited to explain the peer diffusion part of this finding. In the context of academics’ use of generative AI, Singh and Strzelecki (2026) show that innovators and early adopters serve as informal advocates whose support of behaviour helps to propel the adoption process beyond what institutional mandates can do. Gutiérrez-Leefmans et al. (2026) expand this analysis to the broader use of AI in higher education and conclude that the diffusion dynamics are heavily dependent on the perceived relative advantage and compatibility of the technology with the current practices. These attributes are directly met by immersive learning’s professional scenario simulation with engineering students. In their study of digital service adoption, Sandoval-Almazan et al. (2026) also find peer influence to be one of the main mechanisms of digital service transmission when institutional infrastructure for formal adoption is weak. Peer-driven diffusion may therefore offer a more realistic and theoretically sound pathway for the adoption of virtual reality in resource-limited regional engineering institutions, where the adoption of VR may be constrained by curricular mandates that come from the top down and infrastructural and budgetary limitations. This is viewed as evidence of a first-generation engineering student internalising an informal diffusion role in the context of the high and geographically stable priority weight assigned to peer recommendation in Tiruvallur, Kancheepuram and Gummidipoondi. They do not only represent students, but also advocates, and their endorsement actions are a “scaling up hand” of immersive learning in the regional engineering education system. It is important to note that the constructs of SDT, TAM and diffusion of innovations were used in this study as interpretive frameworks for the AHP priority pattern but were not directly measured. Autonomy, perceived usefulness and innovator status were not directly measured in this study.

3 Methods

3.1 Research design

The study was of two-group comparative design (Creswell and Creswell, 2018) and carried out in three different geographical sites of Tamilnadu, namely, Tiruvallur, Kancheepuram and Gummidipoondi sites. Students were randomly assigned to either an experimental group (VR-based oracy skills development) or to a control group (traditional classroom-based communication skills development) for each site, with the same number of students in each group and the same amount of time spent on the intervention. Comparative logic of the design was designed to allow for an analysis of the comparison of the priority weights given to the oracy-related criteria across instructional modality, and the replication across three sites was designed to allow for an analysis of the geographical stability of the pattern observed. AHP priority data was gathered from all participants at one time after the intervention. The detailed characteristics of the participants are provided in Section 3.2.

3.2 Participants and research context

The study was carried out at three geographically distinct instructional sites in Tamil Nadu, India, in the localities of Tiruvallur, Kancheepuram, and Gummidipoondi. Kancheepuram constitutes an independent administrative district; Tiruvallur and Gummidipoondi lie within the same district, the former being the district headquarters and the latter a taluk within it. Specific institutional names are not disclosed, consistent with the confidentiality assurances given to participating institutions at the point of data collection. A total of 144 first-year engineering students participated in the programme, out of which 50 were from the Tiruvallur site (24 male and 26 female), 48 from the Kancheepuram site (26 male and 22 female), and 46 from the Gummidipoondi site (21 male and 25 female) (Table 1). The operational definition of first-generation learners was used and adopted by all participants, which is a student whose parent(s) did not go beyond tertiary education. The students were in the age group of 18–20 years when data were collected and were studying in regional engineering institutions for Bachelor of Engineering programmes.

3.3 Instrumentation: AHP questionnaire design

Four oracy skill development evaluative domains were identified, and structured questionnaires were developed for each of these four hypotheses. The participants rated each item on a nine-point scale ranging from 1 to 9 as per Saaty’s (2008) standard AHP intensity scale, with one representing “equal importance” between two criteria, and nine representing “extreme importance” between two criteria. The criterion sets differed for each hypothesis, depending on the dimension of oracy being studied (Table 2). Hypotheses one and two were operationalised by five criteria for general oracy effectiveness and five criteria for the comparison between the experimental and control group, respectively, sharing a common set of criteria. The criteria for Hypothesis three were the application of oracy skills in actual engineering situations. For Hypothesis 4, the criteria included the impact of positive attitudes towards immersive learning on career aspirations. The full sets of criteria for each of the four hypotheses are given in Table 2. The five criteria were the same for both hypotheses. Hypothesis one was intended to address general oracy effectiveness and Hypothesis two was intended to address the comparison between the experimental and control groups (Table 2); however, the archived instructions do not preserve a version of the two prompts that differs in wording, and it is not possible to confirm retrospectively whether the framing given to participants for the two hypotheses was in fact distinguishable in practice. This ambiguity is acknowledged as a limitation of the study (Section 5.6). Based on a literature search, five criteria were developed for each hypothesis and arranged in four sets of five criteria spanning VR-based oracy learning, engineering communication, and the Analytic Hierarchy Process.

TABLE 1. Participant demographic characteristics by site and group.

TABLE 2. AHP criterion sets by hypothesis.

3.4 Procedures

VR-based immersive learning sessions for oracy-skills development were delivered to the experimental group throughout the intervention period, while the control group received parallel communication instruction in a traditional classroom setting over the same period as represented in Figure 1. Each participant in the experimental group completed eight 45-min sessions, delivered on the ENGAGE VR platform using the Meta Quest two head-mounted display, within custom-built immersive oracy scenarios. These scenarios modelled engineering workplace communication situations, comprising presentations, interviews, team-work discussions, and client meetings, and were experienced from a first-person, unembodied perspective, with no participant avatar customisation. Each scenario involved a single virtual interlocutor, controlled live by the instructor rather than scripted. Interaction was structured through branching dialogue and voice-based interaction, and the instructor delivered feedback verbally, in real time, during the session; this feedback protocol was standardised across every instructor, site, and session. The control group received the same number and duration of sessions, directed toward an equivalent communicative objective, comprising classroom discussion, presentation, and role play conducted in pairs; participants in this arm likewise received individual, real-time feedback from the instructor. The same instructors delivered both arms across all three sites. The control-arm syllabus was developed independently of the VR scenarios rather than matched topic-for-topic to their specific learning objectives, so content equivalence between arms should be interpreted at the level of task type, duration, and feedback structure rather than session-by-session correspondence; this is acknowledged as a limitation (Section 5.6). Detailed scenario scripts and session-specific learning objectives beyond the four task-type categories described above were not retained in a form that could be reported here, which is likewise acknowledged as a limitation. AHP priority data were collected from all participants once, at the conclusion of the intervention period, in both arms.

3.5 AHP computation procedure

The analytical method used was the standard three-step Analytic Hierarchy Process (AHP) computation procedure developed by Saaty (2008) and elaborated in the following methodological reviews (Ishizaka and Labib, 2011; Vaidya & Kumar, 2006). A 5 × 5 pairwise comparison matrix was first developed for each hypothesis and site; a_ij represents the relative importance of criterion i over criterion j, as determined by the member of the participant group. The reciprocal values were placed in the corresponding transposed cells to maintain the internal consistency of the matrix. The second stage involved adding the numbers in each column of the pairwise matrix and dividing each cell by the sum of its column to produce the normalised comparison matrix with all column numbers adding to 1, so that the numbers in each column can be compared across criteria. The third stage involved calculating the priority vector for each criterion by taking the row mean of the normalised matrix, that is, by averaging the five column values for each criterion row in the normalised matrix. The priority vector indicates the relative importance of each criterion in the evaluative hierarchy, with all five priorities adding up to 1.00 (Subramanian and Ramanathan, 2012). The three steps were performed independently for each of the four hypotheses, in each of the three sites, for a total of 12 sets of priority vectors to be analysed. The number of people involved in each group of this study (23–26 people per group per site) is within the recommended range for group decision-making (n = 10–20), so the geometric mean of individual judgements (Aggregation of Individual Judgements, AIJ) was used to maintain the reciprocal property of the pairwise comparison matrix at the group level. It was considered to be the optimal way of combining the group AHP judgements for the group decision making of this scale. The Consistency Ratio (CR) was calculated for each of the 12 pairwise comparison matrices and all matrices were below 0.10, the recommended lower limit for a 5 × 5 matrix (Saaty, 2008) (Table 3). This threshold was not exceeded in any matrices. These 12 matrices correspond to the experimental group only. Control-group judgements were not retained as site-level pairwise comparison matrices and could not be recomputed for this revision; the control-group values reported in Section 4.4 and Table 4 were instead obtained by averaging individual priority vectors. The absence of site-level control-group matrices is acknowledged as a limitation (Section 5.6).

TABLE 3. Consistency Ratio (CR) values for the 12 AHP pairwise comparison matrices (experimental group).

All CR, values are below Saaty’s (2008) recommended threshold of CR ≤0.10 for a 5 × 5 matrix, indicating an acceptable level of consistency in the pairwise comparison judgements.

TABLE 4. Comparison of mean priority weights for the highest- and lowest-weighted criterion types, experimental versus control group.

3.6 Analytical approach for cross-site comparison

A descriptive cross-site comparison of priority vectors was conducted, which analysed the differences in priority weights for each criterion for each hypothesis between Tiruvallur, Kancheepuram, and Gummidipoondi. This comparison was not intended for inferential hypothesis testing, but to locate stable priority patterns. This is because this distinction emphasises that AHP is a decision-modelling tool and not an inferential tool (Saaty, 2008; Forman & Selly, 2001). A pattern was considered to be geographically stable if the difference in the priority weights for a particular criterion between sites was less than 0.025 for the three sites in the same hypothesis. This threshold corresponds to the level of precision used in the computation of the AHP priority vector at the sample scale (Ishizaka and Labib, 2011). Detailed priority vectors for all four hypotheses across all three sites are presented in Table 5 (Section 4.1), which allows for a direct visual comparison of consistency across sites. The 0.025 threshold is a descriptive convention rather than a statistically derived bound, and is not accompanied by an uncertainty estimate (e.g., a bootstrap confidence interval); this is acknowledged as a limitation (Section 5.6).

TABLE 5. Consolidated AHP priority vectors across the four hypotheses and three sites (experimental group).

The priority weights sum to 1.00 within each hypothesis-site combination. Higher values indicate greater relative importance assigned by participants.

4 Results

4.1 Priority vectors across four hypotheses

The number of priority vectors computed for each hypothesis corresponds to the number of criteria specified in the corresponding criterion set (Table 2).

Priority vectors for all four hypotheses were computed in the three sites of Tiruvallur, Kancheepuram, and Gummidipoondi for the Analytic Hierarchy Process (AHP). This was accomplished by creating pairwise comparison matrices and then normalising the columns. The consolidated priority vectors are presented in Table 5.

4.2 The attitudinal inversion pattern

The results of this analysis, presented in Table 5, show three consistent patterns of structure. First, the most general and future-oriented criterion in each set, across the three sites, was given the highest priority weight (between 0.4514 and 0.5090). In particular, for Hypotheses 1 and 2, highlighting the importance of developing oracy skills in engineering education was the predominant emphasis, with mean weights of 0.490 and 0.489, respectively, across the sites. In Hypothesis 3, the benefits that would contribute to future careers were very important, with a mean weight of 0.497. The criterion related to attitude change towards the usefulness of immersive learning had the highest weight for Hypothesis 4, with a mean value of 0.457. Second, the immediately related skill-effectiveness criterion always had the smallest weightings (0.0348–0.0542) in all hypotheses and sites. It is important to note that this low relative weighting is a subjective valuation of the criterion in the AHP comparative framework, and not a measure of the lack of learning gain, as the AHP priority weights are measures of relative valuation among criteria and not direct measures of learning gain. This is a key empirical finding of the study, as it is a systematic inversion in which the attitudinal and prospective aspects of a skill are given greater weight than the direct skill-effectiveness dimensions. Third, the priority vectors of Kancheepuram and Gummidipoondi were the same for all five criteria in Hypothesis four and almost the same for Hypotheses two and 3, thus indicating maximum stability between sites.

4.3 Cross-site stability

The inter-site ranges for each criterion were computed as the differences between the top and bottom of the priority weight ranges for the three sites and are shown in Table 6. The inter-site ranges for all 20 criteria at all four hypotheses were below 0.025, indicating that the priority ordering was stable across Tiruvallur, Kancheepuram, and Gummidipoondi.

TABLE 6. Inter-site range of priority weights across all four hypotheses.

Range = maximum minus the minimum priority weight across three sites. All values were below 0.025, indicating a high geographical stability.

The high uniformity of inter-site ranges, spanning 0.0032 to 0.0219 and all remaining below the 0.025 threshold used here, is consistent with the pattern not being a reflection of the local institutional conditions in any given site but is a structurally stable aspect of attitudes toward immersive-learning experiences held by first-generation engineering students in this regional context. In all the hypothesis-site combinations, the dominant criterion is of an attitudinal or prospective type. The second or subordinate criterion in each combination is, however, the most directly skill-oriented criterion in the criterion set of the hypothesis.

4.4 Comparison with the control group

This subsection reports the AHP priority vectors obtained from the control group separately, so that the attitudinal inversion pattern observed in the experimental group can be evaluated against the pattern present in the absence of VR instruction. The control-group values reported in Table 4 were obtained by averaging individual participants’ priority vectors; site-level control-group matrices, comparable to those reported for the experimental group in Table 5, were not retained and are not available for this revision, which is acknowledged as a limitation (Section 5.6).

In all four hypotheses, the aggregated control-group means, for both the highest- and lowest-weighted criterion type, are numerically close to the corresponding experimental-group means (differences of approximately 0.005 or less). No formal statistical test was conducted on these aggregated values, and none is reported here, because a valid test requires participant-level priority vectors rather than group-aggregated weights; participant-level data were not available for this revision, which is acknowledged as a limitation (Section 5.6). The numerical similarity nonetheless suggests that the pattern reported in Section 4.2 was present in both the experimental and control groups, and therefore does not appear to be a specific effect of VR instruction; a more general effect associated with structured oracy instruction, present in both arms, is at least as plausible an explanation. This observation directly limits any claim that VR-specific affordances (e.g., embodied cognition, immersion-driven attitudinal activation) are responsible for the pattern, and the Discussion below is framed accordingly.

5 Discussion

The results of this study clearly support a consistent and theoretically relevant pattern that emerged for all four hypotheses and three geographically separate sites. Engineering first-year students always rate the attitudinal and prospective criteria–attitude change towards the utilities of immersive learning, its role in engineering education, and its advantages for future professional activities–higher than the direct skill-effectiveness criteria. This section introduces the implications of this pattern with respect to the existing literature, its utilization for the design of a virtual reality pedagogical system in engineering education and the limitations of this study. The current study is important because it tries to draw a pattern of prioritisation among the first generation engineering learners by replacing the Likert-type instruments with the structure of forced pairwise comparison and it also demonstrates that the pattern of prioritisation was consistent across three geographically different sites.

5.1 Relative valuation of attitudinal and skill-effectiveness criteria

The most interesting theoretical result of the study is the opposite of the expected priority ordering found in both the hypotheses and the sites. Research in educational technology has been dominated by the concept of skills being the sole or at least the main outcome of educational interventions, with attitudes and intentions being treated as intervening variables or secondary outcome measures. The data presented here seriously question this ordering. In every case of a combination of criterion and hypothesis, the criterion with the lowest priority weight was the one that was the furthest removed from the criterion of skill effectiveness. In contrast, the farthest removed criterion with the highest attitudinal or prospective criterion was the one that received the highest weight. This result is consistent with the possibility that responses to instruction in this population, whether delivered through immersive or conventional formats, are principally motivational and dispositional rather than reflecting competency transfer; as reported in Section 4.4, the pattern was not confined to the VR condition, and no formal test excludes chance or shared-context explanations, so the phenomenon is more cautiously read as a broader response to structured oracy instruction than as a mechanism unique to immersive technology. In their study of immersive VR serious games designed to facilitate autonomous learning in computer assembly tasks, Checa et al. (2023) discovered that both attitudinal dimensions of learning (namely, learner engagement and self-directed exploration) were more strongly related to positive learning outcomes than were task performance measures. Liu et al. (2024) offer direct empirical evidence to support this interpretation, concluding in their fuzzy-set qualitative comparative analysis that autonomy and learner characteristics in a VR learning environment together affect cognitive load and learning outcomes, but not via the instructional content itself, but via the learner’s dispositional orientation towards the learning environment. The attitudinal shift that was identified in the present study is neither limited to the oracy domain of learning nor to the Indian context of regional engineering education, but is rather a generic structural phenomenon of immersive learning affecting learners who have a formal educational relationship that is defined by aspiration instead of a professional identity.

5.2 The role of virtual learning environments in shaping educational dispositions

The findings from the three sites suggest that the immersive learning intervention not only offered communicative practice but also had a lasting impact on the participants’ thinking about educational technology. In their study on the effects of VLEs on student engagement and educational outcomes, Torres Martín et al. (2021) found that the perceived quality of the virtual learning experience and not the actual content of the virtual course was the key factor in students’ prolonged engagement and positive attitudinal shift. Likewise, Queiroga et al. (2021) analysed the data of virtual learning environments for the creation of institutional educational policies. They showed that the engagement patterns used by learners in VR-supported settings were a better predictor of educational outcomes than performance measures taken at the time of the intervention. In an earlier fundamental study on distance language learning in the virtual learning environment, Comas-Quinn et al. (2012) pointed to the transformation of the learner’s role, from a passive recipient of the learning technology to an active participant in the learning process, as the pedagogical key for immersive environments. Taken together, these results indicate that the high priority weighting on attitudinal criteria in the current study is unlikely to be an artefact of the AHP instrument alone; however, because priority weights were collected once, after the intervention, with no pre-intervention baseline on this measure, they document a post-intervention state rather than a shift, and any attribution to the instruction period as a whole rather than to the instructional delivery in VR specifically remains an interpretation rather than a demonstrated finding.

5.3 Career aspiration and prospective professional identity

Hypothesis 3, which prioritisation of long-term career goals was 0.497 with mean priority weights across all three sites, and Hypothesis four that shifts in attitude towards the utility of immersive learning was 0.457 with mean priority weights across all three sites, suggest a theoretically important characteristic of the first-generation learner population studied. Fernández-Arias et al. (2023) revealed that while communication skills are consistently valued in Latin American engineering students and faculty, they are more likely to be valued in career development stories than as an independent learning goal. This is particularly seen in learners from socioeconomically disadvantaged backgrounds who tend to have a career orientation. The results of the present study confirm this: First-generation engineering students in Tamil Nadu are more likely to value immersive learning experiences depending on their future career aspirations and value the relevance of technology to their future careers over their relevance to the present learning process. The orientation towards the construction of a career is theoretical and practical, with an emphasis on the practical use of pedagogical methods for this segment of the population, which is reflected in the pedagogical strategies.

5.4 VR as a didactic resource in engineering education

The consistent geographical pattern of the priority across the three sites (Tiruvallur, Kancheepuram, and Gummidipoondi), with all inter-site ranges being below 0.025, may have implications for how VR is considered for use in engineering education in the region, though the equivalent pattern observed in the control group (Section 4.4) means these implications concern oracy instruction generally rather than VR specifically. Vergara et al. (2022) conducted a survey of engineering teachers on the use of VR as a didactic resource. They found that it was mostly associated with the simulative capability of VR and the motivation that could be achieved by offering a hands-on experience. The features highlighted by the participants in the AHP in the current study are the same as those emphasised by the affordances. Ergan et al. (2022) created a unified VR-based engineering education platform. They highlighted that the motivation and attitudes of immersive engineering learning environments are dependent on the design of the environment. In particular, the dimensions can be intentionally strengthened by the platform architecture that highlights the role of the learner and authentic scenario simulations. The pattern of results has been consistent across the three sites, suggesting that a pedagogical approach that focuses on the motivational and attitudinal aspects of learning, whether in VR or traditional instruction, can be adopted across the three sites without adaptation to the specific context of each site; the specific benefits of VR affordances have not been established for the current pattern of results, as the equivalent pattern was observed in the conventional instruction group (Section 4.4).

5.5 Peer recommendation as a scalability mechanism

The probability of recommending immersive learning to others was the second most consistent criterion across all four hypotheses, which had mean weights between 0.265 and 0.274. The results provide implications for the institutional transfer of the teaching of oracy by using VR. In small-scale regional engineering colleges where the implementation of top-down VR into the curriculum might be faced with infrastructural and financial issues, it is possible that a more viable path for adoption is peer-driven advocacy. Students who said they were very engaged with the immersive learning environment were much more likely to recommend the technology to their fellow students and to seek further engagement with VR-based instruction on their own (Checa et al., 2023). In the same way, Torres Martín et al. (2021) found that attitudinal change (related to the factor in this study of student satisfaction with virtual learning environment) was most predictive of peer recommendation behaviour. The present results suggest that the respondents’ answers to the questions regarding high priority weight of the peer recommendation not only reveal a willingness to recommend but also make the respondents aware that they could become informal promoters of immersive learning in their institutional communities. In resource-constrained regional institutions, this informal advocacy might be realised practically in a variety of ways, such as the formalisation of peer mentoring or student-ambassador programmes, where experimental-group students are matched with incoming cohorts, or through demonstration sessions by students at the time of orientation or induction, or by purposeful integration of peer testimony into institutional communication about VR-based teaching, without requiring the capital investment of a top-down curricular mandate.

5.6 Limitations

Several limitations to this study warrant mention. The AHP questionnaire used a set of criteria developed by the researcher and did not allow for open-ended responses from participants, limiting their evaluative judgements to these predetermined criteria; qualitative interviews exploring aspects of the immersive learning experience not captured by the current instrument would strengthen future research using this approach. The study relied solely on subjective perceptual judgements, collected once, after the intervention, with no pre-intervention baseline on this measure; the reported priority pattern is accordingly a post-intervention snapshot of relative valuation rather than a measured change in attitude, and no direct link to subsequent behavioural outcomes, such as continued engagement with immersive learning or active peer advocacy, was assessed. The full AHP questionnaire, including exact item wording, participant instructions, and the randomisation procedure applied to criterion order, was not archived in a form that could be reported in this revision; in particular, the wording intended to distinguish the Hypothesis one and Hypothesis two prompts is not preserved, and it cannot be confirmed retrospectively whether participants received a framing that in fact differentiated the two hypotheses (Section 3.3). While the VR and control-group protocols are described in more detail in Section 3.4 than in the original submission, the specific scenario scripts and session-by-session learning objectives underlying the four broad task-type categories were not retained, and the control-arm syllabus was developed independently of the VR scenarios rather than matched topic-for-topic to their specific objectives; equivalence between arms should accordingly be read at the level of task type, session count and duration, and feedback structure, rather than as content-matched equivalence. The Methods report twelve pairwise-comparison matrices, covering four hypotheses across three sites for the experimental group; the corresponding site-level matrices for the control group were not retained, and the control-group values reported in Section 4.4 and Table 4 are pooled, averaged priority vectors rather than a full set of group-specific matrices with their own Consistency Ratios. For the same reason, the numerical similarity between experimental- and control-group means reported in Section 4.4 could not be tested statistically: a valid test requires participant-level priority vectors, which were not available for this revision, so the similarity is reported descriptively rather than as an inferential finding, and should not be read as evidence that VR-specific affordances play no role, only that this design cannot establish that they do. The 0.025 threshold used to characterise stability across sites (Section 3.6) is a descriptive convention rather than a statistically derived bound and is not accompanied by an uncertainty estimate such as a bootstrap confidence interval; a formal measurement invariance test (e.g., confirmatory factor analysis or two-group modelling) was likewise not conducted in the present AHP-based design, and future research using scale-based instruments alongside AHP is recommended to incorporate such testing. Finally, because the pattern reported in Section 4.2 was also observed in the control group (Section 4.4), the present design cannot isolate VR-specific mechanisms as its cause; only a design capable of isolating the effect of instructional modality, ideally with a pre-intervention baseline and participant-level data, would be able to address that question. Taken together, these limitations mean the findings are best read as a description of relative valuation among the given criteria at a single post-intervention time point, for this sample and these three sites, rather than as evidence of learning gain, attitude change, durability, intervention effectiveness, or a VR-specific causal mechanism.

6 Conclusion

This study’s AHP results indicate that, in this sample, attitudinal and prospective criteria—attitude towards the utility of immersive learning, its importance in engineering education, and long-term career goals—received the highest relative priority across the three sites in Tamil Nadu, while direct skill-effectiveness criteria received the lowest. This pattern held across all four hypotheses and all three sites, with a range between sites of under 0.025 for all measures, suggesting a structurally consistent feature of how this group of learners valued these criteria at a single post-intervention time point, rather than a result of local institutional conditions. Because a closely similar pattern was also observed in the control group (Section 4.4), the pattern does not appear specific to VR-based instruction, and because priority weights were collected once, after the intervention, the results describe relative valuation rather than a measured change in attitude, a learning gain, or a causal mechanism. Within these constraints, the findings suggest that pedagogical design for this population might usefully attend to motivational and career-orientation dimensions of instruction, alongside skill-related learning goals, though this possibility would need a pre-post or experimentally controlled design to establish more firmly. Future studies should investigate whether the pattern holds in other disciplinary settings, platform types, and learner groups, using designs that include a pre-intervention baseline, participant-level data, and a formal comparison between VR and non-VR conditions, so that the questions of durability and VR-specificity left open here can be addressed directly.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Veselina Nedeva, Trakia University, Bulgaria

Reviewed by: Abílio Afonso Lourenço, University of Minho, Portugal

Thiti Jantakun, Roi et Rajabhat University, Thailand

Vera Septi Andrini, Universitas PGRI Mpu Sindok, Indonesia

Harald Kitzmann, University of Tartu, Estonia