AI-assisted animation versus text-based description: divergent pathways to intercultural understanding of Norwegian, Mising, and Javanese greetings on a 3R-virtual intercultural stage

Human greetings at initial encounters are often viewed as simple formulaic expressions, yet in many languages they embody culturally unique worldviews that are often lost in English translation. This study investigates three such greetings— Norwegian Det er hyggelig å møte deg, Mising kumríksudung, and Javanese sugeng —within a structured AI-supported virtual intercultural exchange on a 3R-Virtual Intercultural Stage based on “3Rs of Social Psychology” (i.e., Role, Rank, and Room) and an Equivalent Indigenous Concept Model. Thirty-one university students from Norway, India (Mising), and Indonesia (Javanese) were organized into six cohorts, each randomly assigned to one of two groups during a two-week exchange: an AI-animation Group (n = 16), which used AI-assisted scripted Vyond animations to visually depict their indigenous greetings, and a Writing Group (n = 15), which created text descriptions of the greetings in Google Docs. Analysis of post-task reflections suggests that participants in the AI-assisted Group reported stronger evidence of intercultural understanding through encoded, non-verbal cues (e.g., gestures and character posture) that emerged when critiquing inaccuracies in their AI-generated animations—a disruption that slowed and deepened processing, which the authors term a “Pedagogical Speed Bump.” Writing Group participants, by contrast, developed understanding primarily through sustained textual explanation, with comparatively less embodied elaboration. Although both groups appeared to reach a similar general level of intercultural understanding, they arrived there through markedly different processes, suggesting that the collaborative medium in virtual exchange (AI-assisted vs. text-based) functions not merely as a delivery method but as a cognitive mediator in its own right, shaping how intercultural understanding is constructed rather than how it is conveyed. This distinction carries important implications for intercultural education and the preservation of indigenous languages.

The exchange of greetings at first encounters is one of the most universal acts of human sociality, yet its meanings, expectations, and underlying worldviews are anything but uniform (Duranti, 1997). Far from being formulaic openers, greetings encode the historical, spiritual, and relational structures of the cultures that produce them. For example, in Mising, a Tibeto-Burman (Tani) language of Northeast India, the act of meeting is inseparable from prayer (Doley and Mittal, 2026). In Mising, the greeting kumríksudung, composed of kum (“pray”) + ríksu (“meet”) + dung (progressive aspect), conveys the message “I pray to welcome you into my/our lives” (Doley, 2025). In Southeast Asian Javanese (Indonesian) society—a majority-Islamic but historically syncretic society—the greeting sugeng encodes a wish for the other’s health and aliveness, prioritizes communal harmony (rukun), and foregrounds the social rank of the interlocutor through its placement in the high honorific register (krama inggil). These are not merely different words for the same social act. They are windows into fundamentally different ways of understanding the essence of interpersonal encounters. In contrast, some Western European languages can be interpreted as relatively straightforward. In Norway, for example, “Det er hyggelig å møte deg” could be relatively more accurately translated as “Nice to meet you.” (Yet we analyze the cultural category of this greeting later.)

One challenge for intercultural communication education is that the function (pragmatics) of greetings in some areas of the Global South can be easily lost. When these greetings are rendered into English, their deeper sentiments are frequently diminished. As alluded to above, whereas in some European languages, such as Norwegian, “Nice to meet you” translates relatively naturally, preserving the essential social function of the greeting (Det er hyggelig å møte deg), Javanese and Mising cases are far more resistant to this kind of transfer. Sugeng enjing/sugeng siang/sugeng dalu cannot be translated as “good morning/afternoon/night” without losing the prayer embedded in sugeng itself—its wish that the other may be healthy and alive. Kumríksudung fares no better: “Nice to meet you” evacuates the pragmatic depth of a greeting that frames the moment of encounter as a divine blessing.

Dalsky et al. (2025) suggest one possible solution to this problem: “liberating” these types of emic concepts, using a lingua franca through intercultural communication to understand the nuances underlying them. Derived from a decade of virtual exchanges based on practitioner research by Yang and Dalsky (2025) Equivalent Indigenous Concept Model framed and tested this process, and the research presented in this paper builds on it and extends it.

Yang and Dalsky (2025) framework posits that through the process of intercultural communication via a lingua franca—such as English (but not limited to)—the surface-level etic meanings (i.e., semantics) and implicit emic functions (i.e., pragmatics) of indigenous greetings (in this case) can be negotiated to achieve “intercultural understanding” (cf. Deardorff, 2006) as we define and explain later in terms of a particular pedagogical approach called Exploratory Practice (EP; Allwright, 2003).

In the present intercultural exchange, rather than flying from the Equator to the Arctic Circle (and making a stop in India), we use the methods of Collaborative Online International Learning (COIL), which has emerged as a scalable alternative to face-to-face international mobility since the disruptions involving the COVID-19 pandemic (Bali et al., 2021; Netswera et al., 2022; Rubin and Guth, 2022). We describe the project as a structured, virtual intercultural exchange designed using COIL principles, incorporating cross-institutional collaboration, synchronous intercultural contact, and shared task completion to align with specific pedagogical goals, institutional constraints, and time limitations.

Despite growing adoption, many COIL initiatives remain pedagogically thin, focusing on surface-level comparisons of cultural phenomena such as food, religion, customs, rituals, and geography, without a guiding theoretical model to facilitate deeper cognitive transformation (Fukkink et al., 2024; O’Dowd, 2021). Without such a model, these exchanges risk reinforcing the very stereotypes they set out to challenge. Moreover, in the present study, culturally grounded greetings in text-based exchanges may be misread as hesitation or awkwardness rather than recognized as a culturally appropriate communicative practice (Kecskes, 2013). Therefore, what is missing is not more cultural content, but a structured pedagogical process through which learners can move from surface-level semantic descriptions of cultural concepts to a deeper pragmatic emic understanding (of cultural greetings, in our case).

The intercultural exchange described in this paper brings together university students virtually, via Zoom and Microsoft Teams, from three geographically distant sites from the Arctic Circle to the Equator: UiT The Arctic University of Norway in Tromsø (representing the Norwegian linguistic community); Jawaharlal Nehru University in New Delhi, India (representing the Mising linguistic community); and Universitas Kristen Satya Wacana in Salatiga, Indonesia (representing the Javanese linguistic community). The three greetings examined— Det er hyggelig å møte degi (Norwegian), Kumríksudung (Mising), Sugeng (Javanese)—were selected to represent a spectrum of cultural orientations. The Norwegian greeting reflects egalitarian social schemas in which pleasantness (hyggelig) is emphasized. The Mising greeting encodes a non-dualistic integration of social recognition and spiritual prayer. The Javanese greeting prioritizes communal harmony and social hierarchy, foregrounding social symmetry. Together, we arrange them into a typology and a culturally diverse test case for a “Role, Rank, and Room (3R-) Virtual Intercultural Stage” (inspired by Dalsky, 2026).

Theoretical foundation

The theoretical architecture of this study is organized as a hierarchy rather than a set of co-equal frameworks. The Dynamic Constructivist Approach to Culture and Cognition (DCA; Hong et al., 2000) supplies the social-cognitive theory of how cultural schemas are triggered and shift. The 3R-Virtual Intercultural Stage (Role, Rank, Room) (Dalsky, 2026), operationalizes this theory as an instructional design, specifying the structured conditions under which schema activation is expected to occur. Exploratory Practice (Allwright, 2003) supplies the pedagogical stance through which these conditions are enacted and interpreted, treating moments of intercultural difficulty as puzzles for collaborative inquiry rather than problems to be solved. Finally, the Equivalent Indigenous Concept Model (Yang and Dalsky, 2025) supplies the analytic lens applied to participants’ reflections in the Results section, tracking movement from semantic–etic toward pragmatic–emic understanding. The subsections below introduce each level in turn, from theory through design to analysis.

Conceptualizing a dynamic culture

To begin, the theoretical foundation of the current virtual intercultural exchange requires an explanation of how culture is defined. Rather than defining culture as a fixed, monolithic entity with nationality treated as a proxy, we adopt an alternative approach: the Dynamic Constructivist Approach to Culture and Cognition (DCA; Hong et al., 2000). DCA conceptualizes culture as a network of domain-specific knowledge structures—including implicit theories, social schemas, and procedural scripts—that are not permanently active but are selectively triggered by contextual cues. Cultural behavior, on this account, is not an expression of stable cultural identity, but a context-dependent activation of one among multiple available cultural scripts. This process, known as cultural frame switching, is particularly evident in multilingual or multicultural individuals who move fluidly between different cultural orientations in response to situational and linguistic cues.

This approach to culture carries direct implications for intercultural communication education, which apply to the virtual intercultural exchange presented in this paper. If cultural schemas are triggered rather than fixed, then designing appropriate contextual cues—settings, relational dynamics, and communicative tasks—should actively facilitate the activation and re-evaluation of schemas in intercultural learning.

Within the broader context of cultural linguistics (Sharifian, 2017), the knowledge structures described by Hong et al. (2000) are further categorized into cultural categories and cultural scripts. Cultural categories are mental frameworks through which speakers group aspects of their social experience into meaningful units (Sharifian, 2017; Talla Makoudjou and Loumngam Kamga, 2017). Cultural scripts capture widely shared understandings of how people think, behave, and communicate in specific social situations (Hong et al., 2000; Prodanović Stankić, 2017). We suggest that the three indigenous (cultural) greetings central to this study each instantiate a distinct cultural category.

Det er hyggelig å møte deg (Norwegian): This greeting reflects a cultural category grounded in egalitarian social schemas. Applying the principles of cultural linguistics (Sharifian, 2017), Norwegian linguistic categories often prioritize social symmetry and “pleasantness” (hyggelig). The script focuses on establishing a sense of equality, operating within a schema that views social interaction as a shared, non-hierarchical experience.

Kumríksudung (Mising)ː This greeting embodies a cultural framework emphasizing respect rather than strict hierarchy. For example, in contrast to Japanese, which typically encodes respect through a formalized honorific system (keigo), Kumríksudung expresses respect through a ritualized, prayerful welcome that integrates social recognition with divine blessing—reflecting a Mising understanding of “meeting” as inherently sacred. Reserved for formal or respectful contexts rather than for use with friends or close relatives, it illustrates what Sharifian (2017) describes as language that encodes specific expectations for senior–junior or teacher-student interactions.

Sugeng (Javanese)ː Classified within the krama inggil (high honorific register), sugeng is used to address interlocutors of higher rank or elders, usually accompanied by a two-handed handshake and a slight bow—a gesture-linguistic combination reflecting both hierarchy and community harmony (rukun) (Brown and Levinson, 1987). As a base, it forms phrases like sugeng rawuh (welcoming guests) and time-of-day greetings (sugeng enjing/siang/dalu). As a cultural category based on social status (Yu, 2017), sugeng emphasizes recognition of social rank over personal feelings, demonstrating a schema where social maintenance and harmony are central to every encounter.

These three descriptions represent commonly documented usages rather than fixed, homogeneous systems. Each greeting varies regionally, generationally, and situationally within its community—Mising greeting practice, for instance, varies by seniority (including between the Pagro variety and other Mising varieties) and by the generation and religious background of the speaker—and the descriptions above should be read as characteristic reference points rather than as claims of cultural uniformity.

The 3R-virtual intercultural stage

Dalsky et al. (2025) decade of work on virtual intercultural exchanges with indigenous concepts (i.e., culturally unique concepts with no adequate English translations) suggests that these conceptual greetings can play vital roles in culturally scripted behavior. We interpret this literally, applying Dalsky (2026) work inspired by the Shakespearean metaphor, “All the world’s a stage” (Goffman, 2023, p. 450). The virtual exchange described in this paper sets the stage for intercultural communication through Dalsky’s “3Rs of social psychology”: Role, Rank, and Room, drawing on a dramaturgical, dynamic constructivist approach to culture that performs in a Virtual 3rd Space. More specifically, this “3R-Virtual Intercultural Stage” places Mising, Javanese, and Norwegian linguistic actors at the center of a virtual intercultural negotiation hub (i.e., the AI-mediated Virtual Room) depicted in Figure 1.

Descriptions and explanations of the components of the 3R-Virtual Intercultural Stage are as follows: Role: A social position independent of rank. In this study, students assume both “teacher” and “learner” roles (i.e., Team Teaching/Learning; Tajino et al., 2015) as they navigate indigenous greetings, enacting these roles within the AI-mediated Virtual Room and fostering mutual intercultural understanding based on the Equivalent Indigenous Concept Model (Yang and Dalsky, 2025).

Rank: Refers to the appropriate scripts used in hierarchical relationships, including understanding vertical relationships and adjusting language accordingly. The dynamic constructivist approach (Hong et al., 2000) emphasizes that such scripts are not fixed but activated by situational and cultural cues. On the 3R-Virtual Stage (Figure 1), Rank is managed through each indigenous greeting as an equal linguistic marker, using English (in our virtual exchange) as a lingua franca, as students negotiate meanings (semantics) and functions (pragmatics) following the Equivalent Indigenous Concept Model (Yang and Dalsky, 2025).

Room: The Room functions as a Virtual 3rd Space in our intercultural exchange—a digitally mediated environment (using MS Teams or Zoom) that helps “neutralize” initial social biases (see Bretag, 2006; Ma, 1996). It can be viewed as activating specific cultural schemas and scripts (Lu, 2017). This virtual setting provides the essential contextual cues for participants (i.e., students) to shift from traditional classroom behaviors to following an (implicit) virtual intercultural script.

The three cultural orientations central to this study—Norwegian, Mising, and Javanese—are conceptualized through Dalsky (2026) 3Rs of Social Psychology, which can be interpreted to understand cultural variation in terms of distinct configurations of Role (social identity and interactional positioning), Rank (hierarchical distance and power distribution), and Room (social space and relational boundaries). These dimensions serve as foundational cultural schemas that shape greeting practices differently across cultures.

Mising culture is rooted in community-based hierarchy (Rank), with kinship and age relations (Role), structured around communal harmony (Room) (Doley, 2022). In contrast, Norwegian culture is characterized by egalitarian power distance (related to Rank) (Hofstede, 2001), individual agency (flexible Role boundaries), and an orientation toward personal autonomy (in a Room) (Wang and Fränti, 2022). Javanese culture emphasizes the importance of Rank and Role through strict hierarchical structures, evidenced by formal language registers (ngoko / krama), and Room governed by rukun (communal harmony) (Isodarus, 2020). The 3R-Virtual Intercultural Stage, operationalized in the present study, defines these three dimensions within an AI-mediated Virtual 3rd Space, creating a framework for intercultural negotiation across linguistic and cultural differences.

The AI-mediated virtual room

“Technology-powered” intercultural pedagogical practices (e.g., Godwin-Jones, 2022; Yang et al., 2024) have been shown to allow learners to express and investigate the complex feelings embedded in indigenous (cultural) concepts at the implicit (emic) pragmatic level (i.e., function), which exist beneath the explicit (etic) semantic level (i.e., meaning) (Yang and Dalsky, 2025). In our virtual exchange curriculum, technology is not just a delivery tool but a safe, iterative space where learners can externalize their cognitive processes. By integrating an AI-assisted visualization tool, we facilitate deeper exploration of both the meanings (semantics) and the situational, context-specific uses (pragmatics) of indigenous (cultural) greetings. This digital platform renders abstract cultural nuances visible, open for questioning, debate, and flexible interpretation.

Specifically, our virtual exchange uses Vyond,1 a commercial cloud-based animation software developed by GoAnimate, Inc. We chose Vyond over standard video-editing or presentation platforms because of its asset customization, which allows users to select the design, non-verbal behaviors, and physical postures of characters, as well as environmental settings, using AI and guided by a user-written script. This feature serves as a “Pedagogical Speed Bump” in the synchronous virtual classroom. For instance, animations created by Vyond require students to translate an indigenous (cultural) concept, such as (Mising) kumríksudung, as “praying to welcome,” into precise visual parameters specifying Role, Rank, and Room. By having users designate the Role of specific digital avatars, detailing physical gestures and proximity to denote relative hierarchy (Rank), and describing culturally authentic settings (Room), instructional requirement—not the Vyond software itself—creates a necessary cognitive tension. This tension encourages a shift from the passive reception of a cultural fact to the active multimodal construction of a culturally informed narrative, effectively circumventing reliance on superficial stereotypes or generalized caricatures and creating the conditions for intercultural understanding on the 3R-Virtual Stage through the creation of actual intercultural scripts rather than mere performance.

It is worth separating the components of this design explicitly: the technological affordance is Vyond’s capacity for asset customization; the pedagogical scaffolding is the requirement to specify Role, Rank, and Room parameters before producing output; the collaborative learning occurs in cohort discussion of those parameter choices; and the intercultural interaction and teacher facilitation occur across both sessions described below. The cognitive tension we describe is attributable to the second of these—the scaffolding requirement—rather than to the software as such; a text-based task with an equivalent parameter-specification requirement could in principle produce a comparable effect, a possibility we return to in the Discussion.

The proposed mechanism by which this scaffolding requirement promotes deeper intercultural understanding is schema activation: whereas a written description allows a participant to state a cultural fact without necessarily engaging the tacit knowledge underlying it, specifying an avatar’s posture, spatial arrangement, and setting requires the participant to externalize implicit knowledge —embodied, spatial, and relational—that may not be readily accessible through verbal description alone (Hong et al., 2000). This is consistent with claims that multimodal production tasks increase perspective-taking and reflective noticing by requiring learners to make normally implicit cultural knowledge explicit and decidable (e.g., a specific gesture, a specific spatial distance) rather than leaving it at the level of general description. We treat this as the working hypothesis for why the AI-animation task should produce richer reflections, rather than as an established causal mechanism, given the qualitative and exploratory nature of the present study.

Pedagogical foundation

The pedagogical principles of this virtual exchange are based on Exploratory Practice (EP), a practitioner-research methodology developed by Allwright (2003) that reframes “understanding” as a quality of classroom life to be cultivated throughout the learning process. EP is organized around seven principles, three of which are particularly important here. Principle 1 emphasizes the quality of life during the learning encounter through collaborative creation. Principle 2 encourages participants to work with “puzzles,” which are the genuine, lived wonders that emerge in the course of intercultural engagement, rather than focusing on problem-solving. Principle 3 involves all participants, including students, as “inclusive practitioner-researchers” of their own learning, viewing the classroom not as a site solely for data collection but as a community of inquiry. Applied to intercultural communication, EP provides a corrective to the tendency to judge success based on satisfaction scores or overall intercultural awareness without paying attention to the quality of the cognitive and relational processes through which understanding is—or is not—achieved (Fernández-Cézar et al., 2024). This study highlights participants’ lived, reflective experiences as the primary evidence.

Consistent with Exploratory Practice (EP), participants in our intercultural exchange play dual roles: both “cultural experts,” teaching their own indigenous greetings (e.g., Mising, Javanese, or Norwegian), and “collaborative learners,” exploring the nuances of their partners’ worldviews. They concentrate on examining the quality of their shared experience rather than simply completing assigned tasks: a learner who is both an active participant and an agent in the educational process, capable of gaining insights from within (Hanks, 2017). The Role component of our Virtual Stage enacts this idea. By positioning each student as the local expert of their own cultural greeting, the design ensures that all participants enter the virtual intercultural exchange with something genuinely at stake—a worldview to represent, articulate, and ultimately see reflected through someone else’s perspective by working through “intercultural puzzles” in creating AI-powered animations.

The theoretical-pedagogical Nest

In summary, the theoretical architecture of this virtual intercultural exchange comprises a social-cognitive psychological approach influenced by sociocultural conditions and situated within a pedagogical epistemology. The dynamic constructivist approach to culture and cognition (Hong et al., 2000) explains how and why cultural schemas shift in response to new contextual cues. The 3R-Virtual Intercultural Stage describes the specific sociocultural conditions—Role, Rank, and Room—that set the stage for that shifting (Dalsky, 2026). Dynamic constructivism provides the social-cognitive theory; the 3Rs provide the instructional design. Furthermore, selected principles of Exploratory Practice (Allwright and Hanks, 2009; Hanks, 2017) view moments of encountering unfamiliar greetings not as problems to solve but as opportunities for intercultural understanding through AI-assisted animations structured on the 3R-Virtual Stage. Finally, in the analysis to evaluate intercultural understanding, we apply Yang and Dalsky (2025) Equivalent Indigenous Concept Model as a framework to test how participants move from a superficial etic (semantic) understanding of indigenous greetings to a deeper, implicit emic (pragmatic) comprehension through intercultural communication.

Research question

The theoretical-pedagogical nest outlined above leads to the following research question, which operationalizes two dimensions through which the 3R-Virtual Stage is evaluated in this study: How will the structured conditions of the 3R-Virtual Stage—instantiated through AI-mediated animation (Vyond) versus a text-based collaboration (Google Docs)—shape participants’ movement from explicit (semantic-etic) toward implicit (pragmatic-emic) intercultural understanding (of indigenous greetings), as evidenced in their reflections about the virtual exchange?

This question tests the 3R-Virtual Stage’s central pedagogical claim: that deeper, pragmatic-level emic understanding requires not only appropriate intercultural engagement but a structured virtual environment—constituted through Role, Rank, and Room—that affords lived cultural (AI-assisted) creation rather than mere description.

We note that Role, Rank, and Room are treated here as the structuring conditions of the intervention itself, the design features that differentiate the two collaborative media, rather than as separate outcome variables to be independently measured. Exploratory Practice and the Equivalent Indigenous Concept Model, correspondingly, function only as interpretive lenses applied to participants’ reflections. The single research question is accordingly scoped to what a qualitative study of this size (N = 31, two sessions) can support, rather than to a claim that each theoretical component has been separately tested.

Participants and setting

Participants were recruited from three universities representing the three main linguistic and cultural communities involved in the study. Fifteen students from UiT The Arctic University of Norway in Tromsø represented Norwegian language and culture. Six students from Jawaharlal Nehru University (JNU) in New Delhi, India, represented the Mising linguistic community. Ten students from Universitas Kristen Satya Wacana (UKSW) in Salatiga, Indonesia, represented the Javanese linguistic community. To promote close collaborative interaction, participants were organized into six groups (cohorts): five cohorts of five members each and one cohort of six members, aiming for a balanced, international mix within each cohort. Each small team of participants was arranged to include at least one member of each cultural group in the online sessions. The participants’ (students’) distribution across institutions is detailed in Table 1.

The pairing of Norwegian, Mising, and Javanese was motivated primarily by an existing institutional partnership among the three universities rather than by a typological research design; the three languages are not typologically matched, and we do not treat them as a controlled comparison set. Typologically and structurally, the three differ considerably: Norwegian is a Germanic language with a comparatively flat register system reflecting Norway’s cultural emphasis on social equality; Javanese has an elaborate, grammaticalized speech-register system (e.g., krama vs. ngoko) that encodes social hierarchy directly in morphology and lexical choice; and Mising, a Tibeto-Burman language of the Eastern Tani branch, encodes social and spiritual respect relations more through pragmatic and ritual convention than through a comparably grammaticalized register system. What the three do share, and what motivated their inclusion together, is that each greeting under study resists straightforward translation into English and indexes a socially or spiritually significant relationship that a purely semantic gloss does not capture—the shared property the study investigates is this translational and pragmatic depth, not typological similarity among the languages themselves. We treat this diversity as a deliberate feature of the exploratory design, allowing us to observe whether the AI-mediated task supported comparable processes of intercultural noticing across structurally disparate cases, rather than as a controlled variable whose effects can be isolated from language-specific differences.

Study design

The study uses a research design based on three principles of Exploratory Practice described above (Allwright, 2003; Allwright and Hanks, 2009). Moving away from strictly controlled laboratory settings, this research occurs in a naturalistic (virtual) educational environment to “puzzle” through the complex, dynamic nature of intercultural understanding. For participants in the AI-assisted Group, the Room comprises not only the synchronous virtual meeting environment but also the specific features of the Vyond platform, which operationalizes all three R dimensions simultaneously. This process transforms the virtual exchange from a space of cultural description into a space of cultural construction.

To examine the effect of the technological medium on transcultural negotiation, the cohorts were split into two groups. During a two-week intervention, all six cohorts met online simultaneously for two sessions, each lasting 60–90 min. Writing Group (n = 15; 3 cohorts): Participants in this condition used Google Docs for a traditional collaborative task, focusing on text-based descriptions of indigenous cultural greetings such as the Norwegian hyggelig, the Mising greeting kumríksudung, and the Javanese sugeng. This group did not have the visual–spatial mediation of the 3R-Virtual Stage.

AI-Animation Group (n = 16; 3 cohorts): Participants in this condition engaged with indigenous cultural greetings by creating scripts for Vyond animations, which required them to turn their cultural scripts into visual formats. This process involved selecting specific Roles (e.g., social position), Ranks (e.g., senior–junior), and Rooms (settings) to depict their indigenous greetings on the 3R-Virtual Stage.

In line with the principles of Exploratory Practice, the research team—comprising five practitioner-researchers (authors), three of whom were affiliated with the participating institutions—did not position itself as neutral outsiders observing the exchange. Instead, we (the authors) engaged as co-investigators, attending synchronous sessions, moving between groups to listen and take notes, and collaboratively reflecting on what we observed across both conditions. Each of us brought a different linguistic and cultural background to this role: as speakers of Mising, Norwegian, Javanese, Mandarin Chinese, and Japanese (aside from English), we were navigating the same challenge we had set for our students—the challenge of finding authentic points of cultural contact across worldviews that do not align neatly. Our own puzzles—moments of uncertainty about whether a particular Vyond avatar effectively conveyed the pragmatic depth of kumríksudung, or if a student’s written summary of sugeng remained superficial—were regarded not as methodological inconveniences but as data in their own right, consistent with EP Principle 2’s call to work with puzzles rather than problems to solve (Allwright and Hanks, 2009).

This positionality has direct implications for how the findings should be interpreted. The analysis relies not only on student reflections but also on the collaborative field notes created by the research team during both sessions. These notes were generated in real time, shared asynchronously among the team between sessions, and used to guide the facilitation of Session 2—transforming the research process into an iterative, reflective practice rather than a detached post-hoc analysis.

This positionality raises a legitimate concern about confirmation bias: as practitioner-researchers invested in the 3R-Virtual Stage model, we were positioned to notice and record observations consistent with our theoretical expectations more readily than disconfirming ones. Two aspects of the design partially mitigate this risk without eliminating it. First, field notes were shared asynchronously among the team between sessions, so observations were visible to co-authors with different disciplinary and cultural vantage points rather than remaining solely with the researcher who recorded them. Second, as detailed in the Limitations section, the categorization of reflections into the Equivalent Indigenous Concept Model’s dimensions was performed independently by each facilitator for their own session, without a shared coding meeting or cross-checking step; we note this openly because it means our shared field-note review, while valuable for triangulating observations during data collection, did not extend to a formal check on interpretive bias at the analytic coding stage.

All interactions occurred within a Synchronous Virtual Environment (SVE), using Zoom (AI-animation Group) and Microsoft Teams (Writing Group) for real-time communication, Google Docs for Writing Group collaboration, and Vyond for the AI-animation Group. English served as the working lingua franca across all sessions—a choice that facilitated communication across institutions but also influenced participation and meaning-making.

Instruments

To capture the rich, subjective aspects of the intervention, a qualitative survey was administered post-intervention using a structured Reflection Guide via Google Forms. Adapted from several virtual exchanges by Dalsky et al. (2025), this instrument was created to gather meaningful participant reflections on three specific areas: Knowledge, Awareness, and the Lived Experience of the AI-mediated interaction. Importantly, in line with the core principles of EP, this study emphasizes “understanding” over traditional interventionist problem-solving. In this approach, understanding is viewed as a fundamental, non-reductive epistemic principle vital for maintaining the quality of classroom life (Allwright and Hanks, 2009). We also used open-ended reflective prompts to encourage students to actively voice and explore their own “puzzles” about the intercultural greeting exchange process and their self-view as practitioner-learners.

Procedure

The study was conducted over two synchronous sessions, exactly 1 week apart, facilitated through a Synchronous Virtual Environment. While the cultural content remained identical for both conditions, the pedagogical mediation differed: The AI-animation Group used AI-assisted scripting, which created Vyond animations, while the Writing Group produced text-based artifacts in Google Docs.

Session one: conceptualization and scripting

In the initial meeting, both the AI-animation Groups (n = 3) and Writing Groups (n = 3) engaged in a “meet and greet” to establish social rapport.

Writing Group: Sessions were designed to elicit the emic content of each participant’s indigenous greeting and to identify interpretive “puzzles” framed by the 3Rs, with discussions documented in Google Docs. In both sessions, the three cohorts met via a Microsoft Teams meeting facilitated by the same practitioner researcher (the fourth author).

In Session 1, the practitioner researcher (fourth author) first provided an overview of the meeting structure in the main room. Then, cohorts moved into breakout rooms for a quick icebreaker, where they introduced themselves, performed their indigenous greeting, and shared its literal English translation. After the icebreaker, the facilitator briefly explained, using Japanese examples (the facilitator was Japanese), that cultural greetings can vary depending on the situation, relationship, and role or identity, implicitly introducing the dimensions of 3Rs to prepare participants for deeper analysis. Each cohort in their breakout room was asked to discuss the deeper meanings of each other’s greetings, using prompts related to lexical meaning (e.g., “What are the dictionary equivalents in English?” “What is being communicated on the surface?”), contextual factors such as settings, situational conditions, proxemics, and gestures (e.g., “Where and in what settings is it typically used?” “What are the rules?”), social relationships (e.g., “Who uses it and to whom?” “Is it for elders, friends, or strangers?”), and underlying worldviews (e.g., “Beyond the dictionary, what prayer, sentiment, or cultural values are conveyed in the greeting?”). They were also asked to create different scenarios where the greeting would naturally be used and record the main points in a shared Google Doc. For example, Cohort 1 compared Norwegian handshake norms (firm, brief, direct eye contact), Indonesian salim and thumb-pointing rules, and the Mising kumríksudung as a prayerful welcome. Cohort 2 contrasted the directness of Norwegian greetings, the Javanese politeness registers and social ranks (ngoko, ngoko alus, krama inggil), and the respectful social acknowledgment of the Mising kumríksudung. Cohort 3 noted Norway’s generally uniform, non-hierarchical greetings, compared India’s general namaste with the community-specific kumríksudung used in formal or communal settings, and reiterated Javanese register distinctions. These discussions aimed to encourage deep reflection and to highlight the interpretive “puzzles” that EP asks participants to explore. However, only one cohort (Cohort 1) recorded an explicit, though general, interpretive question, asking whether other greetings are commonly used in Norway besides handshakes. Most of the Session 1 notes were descriptive, and no cohort recorded puzzles explicitly framed within the 3R dimensions.

The practitioner researcher led the session from the main room, using PowerPoint, provided instructions, managed the breakout rooms, and guided the discussion. She offered discussion prompts and occasionally joined breakout rooms to provide technical support, but did not give cultural answers, leaving interpretive work to the participants. To encourage participation, the practitioner researcher assigned rotating roles in each cohort (discussion leader, script editor, timekeeper/note-taker). Participants also completed a short peer impression.

AI-animation Group: Following the same icebreaker and 3R framing used in the Writing Group, AI-animation Group cohorts moved into AI-assisted scripting and animation production.

In Session 1, the three AI group cohorts followed the same opening structure as the Writing Group: a practitioner-researcher-led overview in the main room, followed by breakout-room introductions in which participants shared their names and nationalities and performed their indigenous greetings. The cohorts then engaged in a structured discussion of lexical meaning, contextual factors, social hierarchy, and the underlying cultural values associated with their greetings—using the same four-category prompt structure employed in the Writing Group.

Following this discussion, the cohorts moved on to AI-assisted scripting. Each cohort created a prompt for ChatGPT-5 describing a social scenario explicitly structured around the 3Rs—a Rank relationship (e.g., employer-employee), a Role for each participant (e.g., teacher-student), and a Room or setting (e.g., a conference hall, restaurant, or street encounter). Participants then collaboratively revised the AI-generated script, correcting linguistic and cultural inaccuracies before finalizing it. (The ChatGPT-5 prompts and the scripts from the cohorts can be found in Supplementary material.) Vyond animations based on these finalized scripts were then produced by the practitioner researchers before Session 2, incorporating culturally appropriate avatars, gestures, and settings selected in consultation with each cohort’s script. Variation within cohorts in this process demonstrates how the 3Rs were adopted differently across groups, even within a shared structural task. Cohort 1 staged a horizontal, peer-to-peer Rank relationship at an academic conference, catching and self-correcting a conflation of Indonesian and Javanese greeting forms in their AI-generated script—an unprompted moment of emic monitoring. Cohort 2 instead built an explicit hierarchical Rank relationship into an employer–employee–manager scenario, and framed the asymmetry as mutual teaching rather than hierarchy alone. Cohort 3 focused on embedding Rank and Room directly into stage directions (gestural annotations for each character) in a lecturer–student encounter.

Session two: refinement and summary

The second meeting concentrated on strengthening intercultural understanding: The deeper cross-cultural recognition that emerges when participants move beyond surface description toward genuine emic engagement with their partners’ indigenous greetings through the various collaborative artifacts.

Writing Group: In Session 2, the facilitating researcher began with a brief recap of the previous session. Afterward, each cohort reviewed its Session 1 logs, which summarized the main discussion points. During these reviews, participants were asked to reflect on their partners’ cultural greetings and discuss how these greetings reflect the situation, the relationship, and the role or identity taken in the interaction. They were then asked to identify any confusing or particularly interesting aspects of their partners’ greetings, compare them with those of their own cultures, and explore the cultural beliefs or values underlying these differences. To support their reflections, the participant researcher provided guiding prompts such as “What aspects of your partner’s greeting were most surprising or difficult to understand?”, “How does this greeting differ from greetings in your own culture?”, and “What cultural values or beliefs might explain these differences?” Each cohort collaboratively wrote a summary of their discussion and interpretations of the puzzles in shared Google Docs. For example, Cohort 1 explicitly recorded a proxemics puzzle: “What aspects of your partner’s greeting were most surprising or difficult to understand?—Surprised by the use of directional gestures in Indonesia, where it seems unusual because people use their thumb rather than the index finger to point the way.” Cohort 2 offered more analytical comparative reflections. As one participant noted, “The most surprising or difficult to understand from a Norwegian perspective is the different social layers and how to choose the appropriate greeting or words.” The group added that “greetings differ mainly in the layers of social context. While Norwegians are more direct and prefer to avoid overcomplicating social interactions,” and observed that “cultural norms in Norway are often egalitarian and power distance is relatively small.” Cohort 3 did not submit a Session 2 summary. Participants also completed a Google Form to share their impressions of all teammates and themselves before concluding the session, which ended with a wrap-up where participants shared one or two key insights from their discussions.

AI-animation Group: In Session 2, the facilitating practitioner researchers (the first, second, and third authors) began by showing each cohort their completed Vyond animation and replaying it as needed when participants had difficulty viewing it clearly. Participants were then asked to assess the accuracy of the animation’s portrayal of their own and their partners’ cultural greetings, considering clothing, body language, accent, physical appearance, and gestural fidelity, before comparing these observations to their own lived cultural knowledge.

Cohort 1 focused on visual accuracy: the Mising participant believed that formal attire should have included a traditional coat along with Western clothing, and pointed out that the animation’s darker skin tone was an inaccurate, potentially offensive generalization of Mising appearance—a concern that the Javanese participants did not share about their own depiction. This illustrates how evaluative standards were shaped by participants’ cultural perspectives rather than a uniform standard of accuracy.

Cohort 2 critiqued accents and physical representation overall: Norwegian participants felt that the accent and appearance (skin tone, hair) did not match, noting that Norwegians tend to default to English greetings in practice; the Mising participant found the accent inauthentic and preferred a softer, more traditional dress portrayal, while emphasizing that Mising shyness should not be mistaken for a lack of confidence; the Javanese participants pointed out a technical issue (script slashes read aloud literally) along with cultural concerns—a “too-Japanese-like bow” and mismatched hair/complexion.

Cohort 3 focused on pronunciation and body language: the Indonesian participant noted that the greeting was spoken with a native-English-speaker accent rather than the speaker’s claimed background and that a culturally appropriate gesture (hand to an elder’s forehead or cheek) was missing. This sparked a broader discussion about formality, comparing Norwegian first-name informality with Javanese honorifics (bapak, ibu) that indicate social distance.

Across cohorts, participants criticized the animation’s fidelity more than they accepted it, relying on embodied and linguistic knowledge—accent, gesture, posture, attire, skin tone—not explicitly elicited in Session 1. Sessions concluded with a break and an introduction to the Reflection Guide.

Post-task reflections

Following the second session, all participants submitted a detailed Reflection Guide via Google Forms, collecting open-ended responses on their perceived learning, the interpretive “puzzles” encountered during the exchange, and their evaluation of the 3R-Virtual Stage as a space for intercultural discovery. These puzzles varied between the two groups: Writing Group participants more often described difficulties finding a suitable English equivalent for culturally specific terms—for example, one participant mentioned difficulty translating kumríksudung because of its deeper cultural meaning compared to “hello,” while another struggled to express the Javanese concept of gotong royong (social cooperation and mutual support) without an exact English match. In contrast, AI-animation Group participants more frequently identified puzzles related to the authenticity of the AI-generated representations—such as questioning whether an animated avatar’s accent, gestures, or appearance genuinely reflected their own or their partner’s culture, or noticing and correcting linguistic inaccuracies in AI-generated dialogue. These puzzles are revisited and analyzed in detail as follows.

Data analysis

The primary analysis applied Yang and Dalsky (2025) Equivalent Indigenous Concept Model to the full corpus of open-ended Reflection Guide responses submitted by participants in both conditions. Reflections were categorized according to the model’s four interrelated knowledge dimensions—semantic, etic, pragmatic, and emic—by the practitioner-researcher who had facilitated that participant’s session: the fourth author for Writing Group participants, and the first, second, and third authors for AI-animation Group participants, each coding the reflections of the participants they had facilitated. Coding was performed individually by each facilitator rather than through a shared coding meeting or a cross-checking step, and no intercoder agreement statistic was calculated. This procedure is consistent with the study’s Exploratory Practice epistemology (Allwright, 2003; Allwright and Hanks, 2009), which treats practitioner interpretation as legitimate data in its own right rather than a source of error to be eliminated through external validation; as discussed in the Limitations, however, it also means the categorizations reported here have not been checked for consistency across facilitators. Coding attended both to the content of participants’ reflections (what cultural knowledge was expressed) and to the process by which that knowledge appeared to be reached (e.g., through embodied critique of AI-generated animations, or through sustained textual elaboration). Coded reflections were read alongside the collaborative field notes generated by the full research team during both sessions (see Participants and Setting), which provided contextual grounding for interpretation, though this shared field-note review did not extend to a formal check on interpretive bias at the analytic coding stage. Given the study’s qualitative, naturalistic design and modest sample size, the analysis does not employ inferential statistical testing; understanding is treated, consistent with EP, as a quality of practice to be interpretively developed rather than a quantity to be statistically verified (Allwright and Hanks, 2009).

The following analysis applies Yang and Dalsky (2025) Equivalent Indigenous Concept Model (EIC) to open-ended reflections from participants in both conditions. As depicted in Figure 2, rather than treating the model’s four knowledge dimensions—semantic, etic, pragmatic, and emic—as fixed, sequential stages, this analysis treats them as fluid analytical lenses through which various aspects of intercultural understanding become visible.

Because these categories are applied interpretively rather than through a standardized instrument, we specify here the criteria used to distinguish them in coding: a reflection was treated as semantic when a participant reported the literal, dictionary-level meaning of a greeting; as etic when a participant described a greeting practice from an outsider’s comparative standpoint without claiming insider authority; as pragmatic when a participant identified a context-dependent behavioral rule (when, to whom, or how a greeting is used) without asserting insider correction of an outsider view; and as emic when a participant asserted insider cultural knowledge to correct, complicate, or go beyond an outsider representation. These criteria are applied consistently to the examples discussed below.

The reflection data demonstrate that participants did not always start at the semantic surface and move linearly toward emic depth. Some began with a surface-level recognition of equivalence before working toward cultural differences; others began with pragmatic use or emic insider experience and worked outward toward cross-cultural comparison. This non-linear movement across dimensions is itself seen here as evidence of intercultural understanding actively developing, aligning with the principles of Exploratory Practice (Allwright and Hanks, 2009), where understanding emerges through the act of organic puzzling and wonder rather than through a fixed sequence.

Depth of comparative reflection

A descriptive pattern emerged from the Reflection Guide item asking participants whether they had noticed cultural concepts or greetings that seemed equivalent across cultures but carried different underlying meanings. Responses to this item varied in length and detail across both conditions: several AI-animation Group participants provided extended comparative reflections involving two or three cultures at once, while responses in the Writing group tended to be shorter, with some participants noting they had not noticed such equivalences during the exchange. This variation in response length and detail is noted here not as a measure of one condition’s superiority over the other but as a descriptive observation about how differently the two collaborative media may have prompted participants to identify and express their intercultural puzzles—in line with the Exploratory Practice principle emphasizing the quality of the pedagogical process over measurable outcomes (Allwright and Hanks, 2009).

AI-animation group reflections

A key feature of the AI-animation Group’s reflective data is that participants often engaged with the EIC model’s cultural knowledge dimensions through the non-verbal and embodied aspects of cultural practice—details about accent, gesture, posture, physical appearance, and material culture that the Vyond platform required them to explicitly operationalize during the animation task. This pattern aligns with understanding AI-mediated animation as a form of multimodal scaffolding: by asking participants to make precise choices about how their cultural identity should be visually and aurally represented, the platform encouraged a level of embodied cultural specificity that might not have been as easily elicited through text-based description alone. Instead of starting with the semantic surface (What does this greeting mean in English?), AI-animation Group participants more often began by asking a different question: Does this representation of my culture look, sound, and move the way my culture actually does?

This emic-first, embodied approach is most clearly seen in participants’ critical evaluations of the AI-generated animations—evaluations that serve as a “Pedagogical Speed Bump”: a moment of productive friction in which participants express what the AI has failed to represent, thereby externalizing cultural knowledge that might otherwise have remained implicit. The following reflections demonstrate this pattern across the Norwegian, Mising, and Javanese participants in the AI-animation Group. One Mising participant offered the following critique of their avatar:

“Regarding the AI-assisted video, there was a clear skin tone bias. The skin tone for the person belonging to the Mising Tribe is depicted with a dark complexion, which is not aligned with the real-life case. It seems like there is a clear bias and generalisation towards the physical appearance specifically for tribal or minority people.” (AI-animation Group, R2).

Observation: The participant states that the animation gave the Mising character a darker skin tone than is accurate and names this as a generalization applied to tribal or minority appearance specifically.

Interpretation: This reflection operates entirely within the emic dimension. The participant is not translating a word or comparing a cultural practice; they are asserting insider knowledge of how Mising people actually look and correcting the AI’s outsider approximation, which has reduced a diverse community to a generic “tribal” visual stereotype. The term “generalization” is analytically precise: the participant has pinpointed exactly how an imposed etic fails—it applies a broad external category (dark-skinned tribal person) where insider knowledge requires specificity. This is a striking example of emic knowledge being activated not through dialogue with a partner but through confronting a technological representation. An alternative, non-exclusive reading is also available: this reflection could equally be read as evidence of etic awareness of how the AI system represents minority groups in general, rather than emic knowledge of Mising appearance specifically—the participant’s wording (“tribal or minority people”) frames the concern partly in general, comparative terms rather than exclusively in terms of Mising particularity. We favor the emic reading because the participant grounds the critique in a direct claim about “the real-life case” of Mising appearance, but we note the etic dimension is also present. A Javanese participant’s critique of their own animation shifted from an emic to a pragmatic understanding with a single observation.

“The Javanese accent didn’t sound quite right—it didn’t really feel Javanese, and the gestures seemed more Japanese than Javanese. Also, Javanese people don’t usually use many gestures, so we don’t move too much while speaking or having a conversation.” (AI-animation Group, R3).

Observation: The participant states that the AI-generated accent did not sound authentically Javanese, that the gestures resembled Japanese conventions rather than Javanese ones, and that Javanese speakers typically use fewer gestures while speaking.

Interpretation: The first part of this reflection is emic—the participant knows from the inside that the accent was inauthentic. The second part shifts to pragmatic knowledge: the claim that Javanese people “don’t move too much while speaking” is a contextual-behavioral rule, a culturally specific norm that governs body language in interactions. The participant is not only correcting a visual inaccuracy but also articulating an implicit social script—exactly the kind of knowledge that the EIC model identifies as pragmatic, because it explains how a concept is enacted in context rather than what it means in the dictionary. One Norwegian participant’s Vyond critique approached the same issue from an etic vantage point:

“I didn’t think it matched us; they all had kind of an Indian accent.” (AI-animation Group, R1).

Observation: The participant states that the animated characters’ accents sounded more Indian than Norwegian to them.

Interpretation: This comment serves a specific purpose: the participant claims that the AI’s audio representation combined three distinct linguistic communities into a single generic accent, removing the phonological details unique to each. This is an etic observation—produced by a participant who came to the exchange with no particular insider knowledge of Mising or Javanese phonology—prompted by the animation’s failure. The participant has learned that such specificity exists and matters, without yet being able to articulate what it is. This reflects a shift from etic awareness to the brink of emic understanding, driven directly by the multimodal scaffolding of the Vyond task.

A different AI-animation Group participant offered the most sustained example of non-linear, flexible movement through all four EIC dimensions. Beginning at the semantic level with a surface recognition of cross-cultural equivalence, the reflection moves through pragmatic rules, emic distinctions, and a derived etic conclusion: “Yes, I noticed that some greetings were treated as equivalent across cultures, but actually they are used differently and carry different meanings depending on the cultural context. For example, in Javanese culture, saying “sugeng enjang” is not only a way to say “good morning,” but it also reflects politeness and respect, especially when combined with appropriate body gestures like slightly bowing or lowering the head. Meanwhile, in Norwegian culture, people usually greet each other by saying their name, shaking hands, and saying “nice to meet you.” At first, these greetings may seem equivalent because they both function as polite ways to start an interaction. However, the level of formality and cultural meaning behind them is quite different. Javanese greetings emphasize respect and social hierarchy, while Norwegian greetings tend to reflect equality and informality. Another example comes from the Mising tribe in India. In this culture, people greet each other by saying “kumríksudung” while joining their palms together and slightly bowing their heads. At first glance, this greeting might seem similar to Javanese greetings because both use body gestures like bowing. However, the meaning behind them can be different. In the Mising culture, the gesture represents respect and cultural identity, while in Javanese culture, the bow is often more related to politeness and awareness of social hierarchy.” (AI-animation Group, R5)

Observation: The participant compares Javanese, Norwegian, and Mising greetings, noting that greetings which look superficially equivalent (e.g., bowing gestures in Javanese and Mising practice) carry different cultural meanings.

Interpretation: This reflection opens with a semantic observation— sugeng enjang translates as “good morning”—before immediately noting that this translation is insufficient. It transitions from pragmatic knowledge, which identifies the contextual rules governing the greeting (such as body gestures, bowing, formality level, and relationship type), to emic cultural knowledge, which distinguishes the insider meanings of the Mising and Javanese bowing gestures. These gestures seem identical on the surface but encode different cultural values. It ends with a derived etic generalization: that what appear to be similar greetings across cultures are actually surface-level convergences concealing bigger differences. This is the most comprehensive example in the AI-assisted Group dataset of a participant traversing all four dimensions of the EIC model in a single reflection—and it is notable that this participant also produced the most detailed non-verbal analysis of any respondent, comparing hugging norms, palm-joining gestures, and bowing forms across all three cultures in a single, extended response. The multimodal scaffolding of the Vyond task appears to have primed this participant to attend to the embodied, gestural aspects of greetings, shaping their subsequent written reflection. We present this as the most parsimonious reading of the sequence as written; a different, equally defensible segmentation could group the bowing-gesture comparison under a single sustained emic contrast rather than as a separate etic conclusion. We flag this to avoid implying that the four-dimension structure is the only correct parse of this reflection, rather than one useful lens applied to it.

Writing group reflections

Writing Group participants worked through a text-based Google Docs exchange rather than an AI animation showing a different characteristic entry point into the EIC model’s dimensions. Without the multimodal scaffolding of the Vyond platform (which requires specifying accent, gesture, posture, and physical appearance), Writing Group participants were not prompted to externalize embodied cultural knowledge in the same way. In other words, rather than engaging the non-verbal levels of emic concepts, their discussion stays at the verbal level, more often entering through the pragmatic or semantic levels and building toward emic depth through sustained descriptive exchange and collaborative writing with their partners. Where AI-animation Group participants were pushed toward emic knowledge by confronting an inadequate external representation of their own culture, Writing Group participants were more often pulled toward it by the act of explaining their culture to someone who genuinely did not share it. A Javanese participant’s reflection on the register system of their language illustrates this pragmatic-first entry clearly:

“When I’m talking to my parents I will use ‘Krama Inggil’ meanwhile when I’m talking to my grandparents I will use ‘Krama Alus’… the difficulty is that the difference between the hierarchy is not totally visible… people in Norway treat and talk to everyone equally regardless of age and title, which is different from what we learn in Java.” (Writing Group, R10).

Observation: The participant explains which register (Krama Inggil vs. Krama Alus) they use with parents vs. grandparents, and notes that the hierarchy distinction is not fully visible or explainable, contrasting it with Norwegian egalitarian address.

Interpretation: The reflection begins pragmatically—articulating the specific relationship-register mapping that governs Javanese language use. It then moves into emic territory with the admission that “the difference between the hierarchy is not totally visible”: a recognition that the distinction between registers is partly tacit, embedded in relational knowledge that cannot be fully externalized in English. The etic dimension appears last, as the participant holds this implicit Javanese norm against the Norwegian egalitarian alternative to produce a cross-cultural contrast. The sequencing here—pragmatic first, emic second, etic third—is structurally distinct from the AI-assisted Group examples above but produces a comparable range of knowledge dimensions. A Mising participant in the Writing Group articulated the emic dimension of kumríksudung in the English Skills section of the Reflection Guide:

“I struggled to translate ‘kumríksudung’ because it has a deeper cultural meaning than just ‘hello.’ I explained it using examples. This project made me feel more confident as a Global English Speaker and taught me to express ideas clearly across cultures.” (Writing Group, R5).

Observation: The participant states that kumríksudung resists direct translation, explains it through examples instead, and describes the exercise as building confidence in expressing ideas across cultures in English.

Interpretationː The participant begins at the semantic level—identifying the translation gap—and resolves it through a pragmatic move: rather than finding an English equivalent, they resort to contextual examples to convey the meaning. The emic dimension is embedded in the recognition that the concept resists direct translation, that its meaning exceeds what “hello” can contain. While briefer than the AI-animation Group’s sustained emic reflections on kumríksudung, this response traces the same semantic-to-pragmatic path—and the participant’s closing observation that the project made them more confident as a Global English Speaker connects directly to the lingua franca dimension of the EIC model: the experience of struggling to translate an emic concept through English appears to have deepened rather than diminished this participant’s sense of English as a tool for intercultural meaning-making. A Norwegian participant’s reflection showed a different trajectory—beginning with an etic observation and moving toward an implicit emic recognition:

“Not me, but the others had some struggles with translating exactly what their greetings meant, as they seemed to have many layers, different levels of respect and a warmth that perhaps is difficult to describe.” (Writing Group, R8).

Observationː The participant reports observing their partners struggle to explain their greetings, describing the difficulty itself as reflecting layers of respect and a warmth that is hard to put into words.

Interpretationː The participant is reporting emic depth from a third-person vantage point—they have observed their partners struggling to externalize implicit cultural knowledge, and have registered that the difficulty itself is meaningful, as shown in the kumríksudung example above. The phrase “a warmth that perhaps is difficult to describe” is an intuition about the emic dimension of their partners’ greetings, arrived at by witnessing the failure of semantic translation rather than by possessing insider knowledge. This is a second-hand emic awareness, produced by proximity to the translation puzzle rather than membership in the culture being described. It is worth noting that this kind of awareness—recognizing that depth exists without yet being able to articulate what that depth contains—does not appear in the AI-animation Group data in the same form. In the Writing Group, emic awareness tends to be first-person and specific; in this Writing Group example, it is third-person and intuitive. Both are meaningful, but they represent different positions relative to the EIC model’s emic dimension. An Indonesian participant in the Writing Group offered a reflection that moved from etic observation to pragmatic and emic grounding:

“Eastern countries, especially Asian countries, tend to have a collective hierarchy kind of culture that is reflected even in the language, which has many layers for different people and occasions. While in Western countries like Europe, the culture is more individualistic and egalitarian, so the language is simple and doesn’t have any layers, because everyone has the same position in society. Indonesia and India have similar cultures where both have layers in the language that can be used for different occasions. It is a reflection of the hierarchy and collective culture.” (Writing Group, R6).

Observationː The participant contrasts collective, hierarchical language layering in East Asian/South Asian cultures (including Indonesia and India) with the egalitarian language of Western Europe.

Interpretationː This reflection opens at the etic level—a broad cross-cultural observation about East–West communication styles—before moving inward to pragmatic knowledge: the register layering system is tied to specific social occasions and relationships, not just to general politeness. The emic dimension emerges in the explanation of why those layers exist, grounding the pragmatic rule in a collective cultural worldview. The movement here is etic first, then pragmatic, then emic—a third distinct entry-point trajectory in the dataset, different from both the AI-assisted Group’s emic-first path and R10’s pragmatic-first path described above. That three distinct trajectories appear across the dataset is itself consistent with the EIC model’s claim that intercultural understanding does not follow a single fixed path through the model’s dimensions.

Cross-condition patterns: intercultural understanding and shared challenges

As shown in Figure 2, the four dimensions of the Equivalent Indigenous Concept Model—semantic, etic, pragmatic, and emic—are arranged in a circle rather than a linear sequence to reflect what the data consistently indicate: participants entered these dimensions from different starting points, moved in various directions between them, and reached some level of intercultural understanding through different pathways depending on their cultural background, the medium they were working with, and the specific puzzles the intercultural exchange revealed for them. The bidirectional arrows between adjacent dimensions in Figure 2 illustrate this non-linearity directly: there is no single path through the model that defines the data, and no one dimension functions as either a universal entry point or a guaranteed endpoint.

Intercultural understanding across both conditions

When asked in the Reflection Guide whether they and their group members had achieved intercultural understanding during the exchange, the majority of participants in both conditions responded affirmatively. An AI-animation Group participant described the process in relational terms:

“Sometimes I didn’t quite catch the conversation, but my group members kindly explained it again… misunderstandings can be minimized by active listening.” (AI-animation Group, R8).

A Writing Group participant offered a parallel reflection:

“Yes, I think we achieved mutual understanding. We explored how greetings reflect our values and tried to understand each other’s perspectives even when they were very different from our own.” (Writing Group, R4).

A second Writing Group participant described the experience of intercultural understanding in terms that connect directly to the EIC model’s central claim about intercultural communication through a lingua franca:

“I got to know the difference between the greetings of different cultures and also got to know the values of those cultures. I think we did achieve intercultural understanding through this exchange.” (Writing Group, R3).

We use “consistent” and “similar general level” descriptively rather than as a measured equivalence claim: both conditions produced self-reports of achieved intercultural or mutual understanding (as illustrated above), and categorized reflections from both groups included examples reaching the emic dimension of the EIC Model. No quantitative parity metric was calculated, and given the small clusters of participants discussed in the Limitations, this should be read as a qualitative impression drawn from the categorized reflections rather than a statistically validated claim of equivalent outcomes.

While the general outcome of intercultural understanding was consistent across conditions, the process by which participants arrived at that understanding differed substantially between groups, as detailed in sections 3.2 and 3.3 and examined further in section 4.4. The consistency of this pattern across groups (AI-assisted vs. Writing) suggests that the structured intercultural exchange—defined by the Role, Rank, and Room components of the 3R-Virtual Stage—may have supported meaningful intercultural understanding regardless of the collaborative medium. This aligns with EP’s focus on the quality of the shared learning experience as the main goal (Allwright and Hanks, 2009), and it reinforces the paper’s main argument that the medium influences the process of understanding rather than whether understanding is ultimately achieved.

Translation difficulty across both conditions

Participants in both conditions also reported encountering moments where English was insufficient to carry the full meaning of an indigenous greeting—what Wierzbicka (2013) describes as the experience of being “imprisoned in English.” An AI-animation Group participant captured this in general termsː.

“Yes, English felt limiting at times because some greetings do not translate perfectly. I explained them by focusing on their meaning and cultural context instead of translating them directly.” (AI-animation Group, R8).

A Writing Group participant reported a parallel difficulty with the Javanese register system:

“There was a hard time for me to explain about the hierarchy of language used on Java because sometimes the difference between the hierarchy is not totally visible, but at the end I used the subject of the hierarchy to explain to my partner about the difference.” (Writing Group, R10).

Both participants bridge the translation gap using the same approach: instead of finding a direct English equivalent, they explain through meaning, cultural context, and examples—a pragmatic move that aligns with the EIC model’s explanation of how intercultural understanding via a lingua franca functions in practice. The fact that this approach appears in both conditions suggests that the semantic-to-emic gap is inherent to the concepts themselves rather than an artifact of the pedagogical design. The uneven distribution of translation difficulty across the three linguistic communities is itself a finding worth carrying into the discussion: the “imprisonment in English” identified by the EIC model as a driver of intercultural understanding seems most strongly felt where cultural and linguistic distances from English are greatest—among speakers of Mising and Javanese rather than Norwegian—and this asymmetry has implications for how COIL exchanges are designed when one of the participating languages is typologically and culturally closer to the working lingua franca than the others.

The 3R-virtual stage as a third space for intercultural exchange

This paper introduces a novel 3R-Virtual Stage model for AI-assisted intercultural understanding through the virtual exchange of indigenous (cultural) greetings. The findings illustrate how this 3R-Virtual Stage, drawing on the socio-psychological dimensions of Role, Rank, and Room, can be used to interpret the virtual greeting exchange as a dynamic intercultural interaction where participants act as both teachers and learners, working to understand the meanings and functions of unfamiliar indigenous greetings. Rank is balanced by treating each greeting as equal linguistic currency, with English serving as a lingua franca that equalizes hierarchical relationships between participants from different linguistic backgrounds. Room is conceptualized as a Virtual 3rd Space—an AI-mediated online environment (MS Teams/Zoom) that functions as a collaborative discursive zone where new meanings emerge at cultural interfaces (Bhabha, 1994). This setting may help neutralize initial social biases and activate cultural schemas, enabling learners to transition from classroom behavior to intercultural communication scripts. Key to this approach is the instructional requirement to specify Role, Rank, and Room parameters (not the Vyond software as such), functioning as a Pedagogical Speed Bump: a structured creative task inside the Virtual Room that intentionally slows learners during the critical moment of first cultural contact. This creates a space where the semantic, pragmatic, etic, and emic aspects of indigenous greetings can develop through exploration during the virtual intercultural exchange.

Indigenous greetings as cultural resources for intercultural learning

Indigenous greetings serve as the starting point of this exchange because they are expressed through indigenous languages such as Mising and Javanese, which carry cultural values embedded in the mental and social lives of indigenous communities. In English translation, these cultural meanings and nuances are often omitted, altered, or even lost. This project, therefore, sought to recover the original force of indigenous greetings that is often diminished in translation, while also promoting greater respect for indigenous languages and the cultural values of indigenous communities. Introducing indigenous languages and cultural values into intercultural communication is an exploratory process, through which we have found that meaning is not simply transferred but co-constructed through interaction, reflection, and negotiation. This dynamic was not limited to the participants. As practitioner-researchers who are ourselves speakers of Mising, Javanese, Norwegian, Mandarin, and Japanese, we found ourselves grappling with our own thoughts—moments of uncertainty about whether an English gloss adequately carried the weight of a greeting like kumríksudung or whether a Vyond avatar’s gesture genuinely represented a cultural practice one of us had grown up performing. These moments reminded us that the loss we set out to examine in our participants’ reflections was one we had already experienced ourselves while designing the study.

Fluid meaning-making in intercultural exchange

The Dynamic Constructivist Approach (Hong et al., 2000) suggests that cultural schemas are triggered by contextual cues; our data suggest that the intercultural greeting exchange served as such a cue, activating and reshaping practitioners’ cultural schemas in real time rather than merely transmitting fixed knowledge. A concrete instance of this is the Norwegian participant (R1) who, on encountering the Mising and Javanese hierarchical greeting norms, moved from describing Norwegian egalitarianism as a simple, stable fact (“we don’t really think about hierarchy when greeting”) to explicitly contrasting it against the newly encountered alternatives—a shift consistent with cultural frame switching, in which the contrast itself, not new information in isolation, made the participant’s own default schema visible to them for the first time. Similarly, the Mising participant’s insistence (Cohort 1, Session 2) that a specific animation detail misrepresented Mising practice reflects the activation of an existing cultural schema being used to evaluate and reject an externally generated representation, rather than the passive reception of new information. In this project, meaning was not just a one-way, linear transmission from cultural insider to outsider; rather, it emerged and was shared in multiple directions through iterative exchange. Here, the aspects of language and culture are no longer treated as static knowledge but are dynamically constructed and transformed through the practitioners’ engagement in discussions. We also note that these aspects were expressed through both verbal and non-verbal resources, emerging from the interaction of spoken explanation, embodied gestures, visual representations, and culturally grounded interpretations. This leads us to reflect on how and why intercultural understanding can be achieved through the exploration of indigenous greetings, following Yang and Dalsky (2025) EIC model. We propose that discussions about indigenous language and cultural knowledge can begin with indigenous concepts (such as greetings), with meanings fluidly transitioning as they move between semantic and pragmatic (language dimensions) and between etic and emic (cultural dimensions). Understanding then develops as practitioners compare and contrast each other’s language and cultural frameworks evoked by the greetings—covering both universal and culturally specific viewpoints—until they reach a point where they feel genuine understanding has been achieved, grounded in critical exploration and reflection.

The medium as cognitive architecture: AI and google docs as contrasting affordances

The most notable finding of the study is the difference in participant experience between the two conditions. Participants in the AI-assisted Group consistently offered richer, more embodied reflections on their partners; indigenous greetings that more often moved beyond surface descriptions to explore the relational, spiritual, and social aspects of cultural practice. This is not simply a story about AI tools outperforming text tools. It is a story about what happens to learners when they are asked to produce a multimodal, embodied representation of the meanings and functions of indigenous cultural greetings (in this case), versus when they are asked to describe them in a medium that may not adequately support their expression.

The Dynamic Constructivist Approach (Hong et al., 2000) posits that cultural schemas are triggered by contextual cues. The Vyond environment, by requiring students to make precise decisions about visual, spatial, and gestural parameters, constituted a contextual environment rich in schema-activating cues: decisions about avatar posture activated embodied knowledge of social hierarchy (Rank); decisions about setting (Room) activated cultural scripts about appropriate contexts for greeting; decisions about the spatial relationship (Role) between characters activated schemas of relational warmth. The Google Docs environment, by requiring students to produce written descriptions of the same concepts, provided fewer such cues—and may, paradoxically, have intensified the experience of linguistic imprisonment described by Wierzbicka (2013), particularly for participants whose first languages are not well served by English semantic categories.

This interpretation is supported by the qualitative data. Writing Group participants more frequently described their experience in terms of the difficulty of finding suitable English words, while AI-assisted Group participants more frequently described breakthroughs—moments of genuine recognition—associated with the visual construction of culturally informed representations within the 3R Virtual Stage. The implication for COIL design is significant: the choice of collaborative medium is not merely a logistical decision but a cognitive architecture decision that may carry consequences for the depth of intercultural understanding the exchange can produce.

AI bias and representation in generative visual media

Beyond its pedagogical implications, the AI-animation Group’s Session 2 critiques constitute a distinct empirical finding in their own right: a documented pattern of representational bias in generative visual AI toward smaller and indigenous cultural communities. The Mising participant in Cohort 1 identified the animation’s darker skin tone as an inaccurate and potentially offensive generalization of Mising appearance, a concern the Javanese participants in the same cohort did not raise about their own depiction—suggesting the bias was not applied uniformly but fell more heavily on the less commonly represented community. In Cohort 2, Norwegian participants judged both accent and appearance (skin tone, hair) inauthentic, the Mising participant found the generated accent inauthentic and the dress insufficiently traditional, and Javanese participants identified a gesture—a bow—that read to them as “too-Japanese-like” rather than reflective of their own greeting practice. In Cohort 3, the animation rendered an Indonesian greeting in a native-English-speaker accent and omitted a culturally significant gesture (hand to an elder’s forehead or cheek).

Taken together, these are not isolated technical glitches but a consistent pattern: the generative model defaulted toward more dominant or more widely represented cultural referents (Western dress, native-English accents, East Asian gestural conventions) when rendering smaller or less digitally represented communities, and participants from those communities were the ones positioned to notice and name the distortion. This asymmetry itself is pedagogically productive—it was, in each cohort, students’ identification of inaccuracy that prompted the deepest engagement with their own cultural knowledge—but it also raises a separate, important concern about culturally responsible AI design. Tools built for intercultural education carry a particular obligation to avoid reproducing the very homogenization they are meant to help learners move beyond. Future development and deployment of generative animation tools in intercultural pedagogy should include representational auditing with community members from underrepresented linguistic groups, and instructors should treat AI-generated inaccuracies as a designed-for pedagogical resource rather than a regrettable limitation to be minimized.

This raises a question we cannot fully resolve with the present data: whether the Pedagogical Speed Bump reflects a general cognitive friction that any sufficiently demanding multimodal specification task would produce, or whether it is partly an artifact of Vyond’s own asset library, which likely offers richer, more varied, and more accurate assets for Western European contexts than for Mising or Javanese ones. Under the second reading, AI-animation Group participants from underrepresented communities may have engaged more deeply not because the task structure itself deepened processing, but because the tool’s inaccuracies gave them more to correct—a difference in the amount of corrective work required, rather than in the task’s pedagogical design. We cannot cleanly separate these explanations with the current design, since the same participants who experienced richer critique opportunities were also the ones whose cultures Vyond represented least accurately. Disentangling this would require a follow-up design that varies representational accuracy independently of task structure—for example, comparing reflections on AI outputs of matched accuracy across communities, or manipulating asset quality directly. We flag this as an open question rather than a resolved finding, and note that either explanation is compatible with our core observation that critiquing AI-generated inaccuracies prompted deeper engagement; what remains unresolved is whether that engagement is attributable to the AI medium in general or to this specific tool’s uneven representational quality across the three communities studied.

Limitations and future directions

Several limitations of the present study should be recognized. The small sample size (N = 31) and quasi-experimental design, although suitable for a naturalistic setting, restrict generalizability. Not all participants attended both sessions: in the AI-assisted Group, 13 of the 16 participants attended both time points, and in the Writing Group, 13 of the 15 participants attended both meetings. Unequal representation of the three linguistic communities across the six groups, caused by differences in institutional enrollment, creates a potential confound that future studies should address through more controlled sampling. The two-week intervention measures short-term changes but does not assess long-term persistence—an important aspect of the broader goal of applying COIL principles for sustainable intercultural development.

Given these constraints, the findings reported here should be understood as context-specific to similarly structured, short-term COIL exchanges among university-age participants, rather than as broadly generalizable claims about AI-mediated intercultural pedagogy across settings, age-groups, or longer engagement periods.

A related concern is the number of participants in each particular condition: with three linguistic communities crossed with two conditions, each cluster of participants contained approximately five participants. At this scale, individual differences in prior intercultural experience, English proficiency, or personality are difficult to distinguish from effects attributable to the collaborative medium itself; a pattern we describe as characteristic of one condition could, in a cluster of participants this small, instead reflect the idiosyncrasies of one or two vocal participants. We treat the reported patterns as illustrative of what qualitatively distinct medium use can look like rather than as evidence that the medium reliably produces these effects at the level of a community or condition as a whole.

The two conditions were also not equivalent beyond the medium alone: the Vyond task took significantly longer than the Google Docs task, so AI-group participants spent more time on the artifact—a difference that could itself have influenced engagement depth, regardless of the medium’s specific features. Engagement also varied across cohorts, reflecting the practical realities of a three-institution, multi-time zone exchange. Future studies should use more controlled time-on-task designs to separate the effects of the medium from those of engagement duration. The two conditions also differed in facilitation: the Writing Group was facilitated throughout by a single researcher, while the AI-animation Group was facilitated by three different researchers (the first, second, and third authors). Facilitator differences, alongside the time-on-task and engagement differences noted above, mean this study cannot isolate the effect of the AI medium itself from the effects of facilitation, activity structure, and time spent on task; these variables were confounded rather than controlled.

Using English as the main lingua franca, while practically necessary and theoretically motivated, can both facilitate cross-cultural communication and limit emic expression—a tension the AI task partly addresses but does not fully resolve. Future research might explore other lingua francas (e.g., Hindi or Bahasa Indonesia) or multilingual prompts that allow participants to use their native language before translating into the shared medium.

Finally, the analysis would benefit from a more systematic approach, such as an Interpretive Phenomenological Analysis (IPA), to provide a more detailed account of individual experiences on the 3R Virtual Stage—including cases where emic understanding does not emerge, which are arguably as informative as those where it does. A further limitation concerns the coding procedure: reflections were categorized by the practitioner-researcher who had facilitated that participant’s session, applied individually rather than cross-checked by a second coder, and no intercoder agreement statistic was calculated. This is consistent with Exploratory Practice’s treatment of practitioner interpretation as legitimate data rather than a source of error to be eliminated through external validation, but it also means the categorizations reported here have not been checked for consistency across facilitators, and future iterations of this analysis would benefit from cross-coder review.

Implications for intercultural pedagogy and indigenous language preservation

Taken together, these findings suggest that AI-mediated visual collaboration, structured through a theoretically grounded framework such as the 3Rs and the Equivalent Indigenous Concept Model, can produce meaningful shifts in both the depth and the psychological experience of intercultural understanding—implications that extend beyond virtual university exchanges.

In intercultural pedagogy, the design of the collaborative medium is just as important as the cultural content. COIL programs that treat Role, Rank, and Room as pedagogical variables rather than neutral logistics are likely to generate richer, more lasting learning than contact alone. Incorporating AI visualization tools offers a promising direction—not because AI is inherently culturally intelligent, but because prompting, evaluating, and correcting its outputs fosters the deliberate, reflective engagement necessary for deeper intercultural understanding.

For indigenous language preservation, the implications are different but equally important. One of the three greetings at the center of this study— kumríksudung —represents a worldview under constant pressure from global English. UNESCO (2023) recognizes indigenous languages as gateways to cultural diversity, warning that losing them means losing cultural knowledge and relational meaning. AI-assisted multimodal pedagogy could offer a new way to preserve and share these worldviews—not just by documenting them as artifacts, but by creating environments where they are encountered, puzzled over, and genuinely understood by outsiders. The Virtual Stage, in this context, is a space where indigenous concepts are acknowledged, valued, and kept alive.

The research reported in this paper explores how a novel 3R-Virtual Stage model can be used to interpret AI-assisted intercultural understanding through the virtual exchange of indigenous greetings. Drawing on the social-psychological dimensions of Role, Rank, and Room, the 3R-Virtual Stage structures the virtual greeting exchange as a dynamic intercultural encounter in which participants simultaneously assume the roles of both teachers and learners as they attempt to understand the definitions and functions of unfamiliar indigenous greetings. Rank is managed by treating each greeting as equal linguistic currency, with English functioning as a lingua franca that flattens the hierarchical relationship between participants from asymmetrically positioned linguistic communities. Room is conceptualized as a Virtual 3rd Space—an AI-mediated online environment (MS Teams/Zoom) that neutralizes initial social biases, activating the cultural schemas that enable learners to shift from classroom-bound behavior to an intercultural communication script. Central to this design is the Role of AI-mediated animation technology (Vyond) as a Pedagogical Speed Bump: a structured creative task embedded within the Virtual Room that deliberately slows learners at the critical moments of initial intercultural encounters, creating the conditions for etic-semantic to emic-pragmatic understanding of indigenous greetings to emerge.

This study has attempted to demonstrate that the mode of intercultural collaboration is not pedagogically neutral. AI-mediated visual animation, when organized through a 3R-Virtual Intercultural Stage and experienced through principles of Exploratory Practice, produces meaningful shifts in how participants engage with unfamiliar cultural concepts—shifts that, we argue, represent a genuine deepening of intercultural understanding rather than merely a more pleasant social experience. The Mising greeting kumríksudung, the Javanese sugeng, and the Norwegian hyggelig are not simply different words for the same social act. They are invitations to inhabit, however briefly, a different way of understanding what it means to meet another person. The Virtual Stage, at its best, is a space where these invitations can be genuinely received.