1Virtual Exchanges (VE) for Second Language Learning (SLL) connect learners from diverse cultural and linguistic backgrounds for collaborative language learning and intercultural exchange via digital tools. In these exchanges, learners must balance individual learning with collaboration, adapt to evolving technological tools and to cultural diversity, and manage the tensions between personal goals and group dynamics. VE environments are dynamic and technologically mediated and require participants to navigate a complex interplay of affordances (i.e., possibilities for action provided by the learning environment, see Blin, 2016) and constraints. Key studies in VE (e.g., O’Dowd, 2018) emphasise the importance of learners’ active participation, often driven by their sense of agency, which facilitates autonomous learning, enhances motivation (O’Dowd & Waire, 2009), and fosters intercultural understanding (Helm, 2016). But what is learner agency? The concept of agency in SLL has been interpreted through various theoretical lenses, be they socio-cognitive (Bandura 2001), sociocultural (Lantolf and Pavlenko, 2001), ecological (van Lier, 2008), or post-structuralist (Norton, 2013), to mention but a few. While most would define learner agency as the capacity to actively shape one’s learning, each perspective offers distinct insights into how learners shape the latter, i.e., through motivation, autonomy, self-regulation, identity negotiation, sociocultural awareness, or by reshaping the power structures within the learning environment. However, these perspectives also come with specific limitations. For example, the broader social and cultural contexts may be overlooked by some and the complex interplay between individual agency and social influences may be oversimplified. Furthermore, the learning environment’s internal tensions and contradictions that influence learner decisions and actions, or the way learners navigate and reconcile internal contradictions arising from multiple, and sometimes conflicting identity positions, may not be fully accounted for.
2Overcoming these limitations requires a ‘meta-theory’ of sorts, which encompasses the various dimensions briefly outlined above, without reducing our understanding of agency to one or the other. For example, it can be argued that the concept of agency goes beyond the notions of autonomy and self-regulation in so far that agentive learners can not only make choices and show independence from external control (i.e., act autonomously) or systematically resort to strategic actions to manage their learning (i.e., self-regulate), but they can also navigate and reshape the constraints that are imposed on them by the learning environment and the sociocultural context they inhabit. Conversely, the learners’ capacity to navigate and reshape constraints through negotiation, resistance and adaptation is dependent on their capacity to self-regulate and to be autonomous as active members of a given learning community.
3This article proposes that a dialectical view of agency that emphasises the dynamic relationship between language learners and their learning and sociocultural environment, and is not restricted to prevalent views on autonomy, mastery, and control, offers conceptual tools that may assist researchers and practitioners firstly in their understanding of learner agency for SLL and, secondly, in the design and deployment of learning environments and pedagogies that will support learner agency. Drawing on Rainio and Hilppö’s (2017) dialectical framework of agency, it views agency as a dynamic phenomenon emerging from the interplay of opposing forces, such as autonomy versus constraint or individual versus collective action (Rainio and Hilppö, 2017). This perspective provides a nuanced understanding of how learners negotiate tensions in VE and manage technological affordances alongside cultural and task-based constraints. Unlike static sociocultural or cognitive approaches, it captures the fluid and complex nature of VE environments, offering unique insights into how learners navigate challenges and opportunities in cross-cultural, technology-mediated interactions. This paper argues that integrating internal factors (e.g., motivation, self-regulation) with external elements (e.g., partner contributions, task design, digital tools) offers a robust interpretive framework for exploring how agency unfolds in VE, and addresses the following research questions:
-
How do dialectical relationships of agency manifest themselves in VE?
-
How do learners navigate these relationships and tensions?
4The next section will further explore the central role of agency in VE and briefly present the dialectical framework of agency proposed by Rainio and Hilppö (2017). A case study methodology, focusing on three Irish students who took part in two Franco-Irish VE projects in 2022-2023, is then used to gain a more comprehensive understanding of how learners steer their language learning experience in complex pedagogical VE environments. Implications for designing VE activities that foster learner agency are proposed.
5Learners in VE environments need to negotiate a complex array of demands and opportunities. They interact with peers outside traditional classroom contexts, adhering to designed tasks and institutional rules or cultural norms, thereby balancing autonomy and constraint. While agency in VE plays a pivotal role in shaping how learners engage with their peers and the broader learning environment, its development and exercise also face challenges, such as the need to balance individual autonomy with collaborative interdependence (Helm, 2015); power imbalances between novice and expert speakers hindering learners’ confidence and willingness to participate (Chun, 2008); technology related issues limiting learners’ ability to take control of the learning process (Hampel & Stickler, 2012); the juxtaposition of one’s own culture against the other’s culture (Helm, 2016); the need to support learner engagement without inhibiting learner autonomy or undermining learners’ sense of ownership over their learning (Vinagre, 2017).
6A dialectical view of learner agency positions it as a product of the interplay between individual capacities (such as motivation and self-efficacy) and external factors (such as institutional rules, social norms, and peer interactions). Rooted in Vygotsky’s sociocultural theory, it acknowledges the importance of mediation in learning, while considering how learners’ agency evolves in response, or in opposition to external constraints (Lantolf & Thorne, 2006).
7Drawing on Rainio’s (2010) original dialectical conceptualisation of agency in educational contexts, Rainio and Hilppö (2017) propose a framework of the dialectics of agency that sees reality as “fundamentally consisting of opposing forces and their dynamics” and focuses on “contradictions as a unit of analysis” (p. 79). Contradictions are here understood as the dynamic back and forth movement between oppositional (or contradictory) pairs (Rainio and Hilppö, 2017, p. 80). Agency is said to emerge through the resolution of these contradictions or tensions (Rainio, 2010), which manifest themselves in practical activities and can trigger change and development (Sannino & Engeström, 2018). Rainio and Hilppö’s (2017) ethnographic case study of children’s agency in the context of a pedagogical playworld yielded five contradictory, yet connected, dimensions or “oppositional elements” of children agency (i.e., enacted – imagined, situative emergence – progressive development, dependence – separation, mastery – submission, and control – freedom) and seek “to depict how the movement between opposing poles of these dimensions constitutes [agency]” (Rainio & Hilppö, 2017, p. 80). The five dialectics are explicated below, with a particular focus on their potential use in the study of agency in Virtual Exchanges environments.
81. Agency as enacted and imagined. Enacted agency refers to the visible actions individuals take in shared activities. It involves the tangible, practical execution of choices and decisions. On the other hand, imagined agency is not easily captured in practical educational activities. It involves the creation of alternative realities where students can distance themselves from the current class activity and exercise agency on their own terms (Rainio & Hilppö, 2017, p. 82). In a VE context, some students may enact or imagine different scenarios. For example, a learner may actively participate in a synchronous online discussion (enacted agency), while also imagining themselves applying the intercultural communication skills gained from the experience, in a future study abroad program (imagined agency). At times, this may involve a temporary withdrawal from the on-going discussion or trigger a change of direction in the ongoing conversation.
92. Agency as situatively emergent and progressively developmental. For Rainio and Hilppö (2017, p. 83), the dialectics between situatively emergent agency (sociological view) and progressively developmental agency (psychological view) relates to the interplay between, firstly the manifestations of agency in situational micro-interactions emerging in “local and changing social, material, and discursive conditions” (situative emergence), and secondly, development through personal growth, experience and learning over a period of time (progressive development). VE environments offer ample opportunities for repeated interactions and thus for capturing and interpreting the development of situational manifestations of agency, such as learners’ creative responses to communication breakdowns, as their skills and confidence in actively participating in VE activities grow through the completion and transformation of the tasks assigned to them.
103. Agency as dependence and separation. Rainio and Hilppö (2017, p. 84-85) identified a dialectical contradiction, or struggle felt by children, between the need for belonging to a community (dependence) and the need for separating from it. They argue that agency requires both dependence and separation. The latter involves carving out a space where one can act without undue influence or hindering constraints. The notion of dependence acknowledges that agency is often exercised within a web of social, economic, and relational dependencies: Individuals rely on others, institutions, and systems to achieve individual and shared goals. As students strive to express their unique ideas (separation) in a VE setting, they need their peers’ input and feedback to refine their work and achieve the group’s shared goals (dependence).
114. Agency as mastery and submission. According to Rainio and Hilppö (2017, p. 85), agentive members of a community learn and internalise the practices, tools and discourses of that community, and can use them creatively to influence one's environment and circumstances (mastery). Mastery involves competence, skill, and the capacity to effect desired outcomes. However, the necessity or choice to yield or adapt to external conditions or rules is at times unavoidable. The agentive member may accept limits and constraints on one's actions as a strategic or pragmatic form of submission. For example, while a learner may master the language, discourses, and other tools and resources that enable them to lead fruitful discussions and to propose alternative ways of achieving the group’s shared goals, they may choose to adjust the discussion’s pace and content to accommodate less proficient members or to avoid potential conflicts that would slow down the group’s progress.
125. Agency as control and freedom. While the above dialectics of agency have been described from the point of view of learners, the dialectics between control and freedom is viewed from the point of view of teachers and designers. It refers to a well-known paradox of education: Although the end goal of education may be the development of children’s agency, pedagogical relationships presuppose “some form of coercion from the part of the adult” (Rainio & Hilppö, 2017, p. 87). Curriculum design for agency requires balancing necessary or context-dependent constraints, such as institutional policies and practices, with enabling and supporting learners’ freedom for unencumbered choices and actions, allowing them to pursue their own paths and desires. It also requires that pedagogical designers accept that the learning environment they construct is likely to create “new limits and boundaries for agency that cannot be controlled” (Rainio & Hilppö, 2017, p. 88). This is particularly true of pedagogical designs in the context of VE, that often grant learners freedom to choose the topics or cultural aspects they want to explore and how they present their findings, while limiting their choice to specific themes. Other unplanned limitations and contradictions will emerge when the design is enacted.
13Having reinforced the importance of examining agency in VE through the dialectical view of agency, the next section will firstly describe the context and design choices that have underpinned the deployment of a VE project between France and Ireland, and secondly the methods used to capture and interpret the dialectical contradictions that emerged during the enactment of the design.
14This study examines a VE pedagogical project between an Irish and a French university, implemented over two semesters, and focusing on the development of learner agency in language learning.
15In the first semester (VE1), the VE involved an e-tandem model, where Irish students engaged in bilingual exchanges with French students. Each session was divided into two parts: The first half in French and the second half in English. The discussions centred on neutral intercultural topics, such as university life, work, idiomatic expressions, artificial intelligence, and topics of the students’ choice. Alongside these synchronous interactions, students participated in peer-feedback, session-building activities, and reflection-based tasks and assessments.
16In the second semester (VE2), the same Irish students collaborated with a different group of French undergraduate students specialising in teaching French as a Foreign Language (FLE). Unlike the first semester, this phase involved monolingual exchanges in French, addressing both neutral and socially charged topics including Erasmus experiences, glottophobia, racism, and topics chosen by the students. The second VE iteration introduced more complex learning activities, including peer- and tutor-feedback, podcast creation (with Irish students acting as experts), and reflection-based tasks like post-project peer- and self-reviews that were integral to the assessment strategy. The French students in VE2 focused on pedagogical conversations and asynchronous feedback.
17A progressive complexification of tasks and assignments was designed not only within the projects but also from one semester to the next. In VE1, the Irish students, pursuing a Bachelor of Education, interacted with second-year French students studying Intercultural Communication. These exchanges were held during timetabled, teacher-organised sessions. In VE2, the Irish cohort continued their interactions with third-year French students from a Didactics of FLE (Français langue étrangère – French as a foreign language) program, in self-organised sessions conducted independently from the classroom environment. The progressive transition from a tandem, bilingual format (VE1) to a monolingual, French-only format (VE2) aimed to challenge the Irish students’ linguistic abilities while fostering deeper reflection on their intercultural experiences. Table 1 below shows the progression of task complexity between VE1 and VE2.
Table 1 – Progressive task complexity from VE1 → VE2 for Irish students.
|
Aspect
|
VE1 (5 weekly sessions)
|
VE2 (6 weekly sessions)
|
|
Participants
|
Irish second-year Bachelor of Education students interacting with French second-year Intercultural Communication students.
|
Irish second-year Bachelor of Education students interacting with French third-year Didactics of FLE students.
|
|
Organization
|
Teacher-organized, timetabled synchronous sessions.
|
Self-organized, independent synchronous sessions.
|
|
Language use
|
Bilingual tandem: Half the session in French, half in English.
|
Monolingual: Entire session conducted in French.
|
|
Discussion Topics
|
Neutral intercultural topics: University life, work, idiomatic expressions, artificial intelligence, etc.
|
Socially charged topics: Glottophobia, racism, Erasmus experiences, and student-chosen social justice-related topics.
|
|
Feedback
|
Post-session peer-feedback for each partner.
|
Post-session peer-feedback & post-project peer-reviews and self-reviews conducted in French.
|
|
Reflective Tasks
|
Individual reflection on interactions.
|
Collaborative podcast creation with French partners.
|
|
Feedback Type
|
Direct oral & written feedback during synchronous sessions.
|
Detailed written feedback sent asynchronously by French students to Irish partners.
|
18From the point of view of the designer or teacher, the question of fostering freedom for students within the institutional constraints of curriculum and assessment is at the crux of learner agency (van Lier, 2008). VE by its very nature requires students to be highly agentive: setting up meetings (often outside class hours), proactively communicating with foreign partners through unsupervised channels of communication, moderating their own sessions and post-session feedback, etc. However, varied amounts of control and freedom were designed within the proposed pedagogical and technological tools.
19While thematic session plans were suggested for each session thus exerting a certain degree of control, students were encouraged to deviate from the suggested prompts to align themselves with their interests or the spontaneous course of the conversation. Moreover, for the final sessions in each project, students had the freedom to come up with their own discussion themes (in VE1) and compile a podcast on a social justice theme of their own choice (in VE2). Students chose to focus on cultural aspects such as sports, festivals, religion, films, literature and cuisine as discussion themes in VE1, and on social justice topics such as sexism in the workplace, homophobia, feminism, etc. in VE2.
20Despite being conducted within the structured environment of a pedagogical project with a set calendar, number of sessions, themes, tasks, and assignments, students could liaise with their partners to reschedule their meetings, improvise conversations within the proposed themes, and tailor their productions for assessment to suit their own needs and desires. Freedom was intertwined in the controlled design for teaching through a flexible curriculum that learners could adapt based on their needs and interests through authentic interactions and assessments that aligned with the learners’ awareness of their own progress.
- 1 Loop is the Irish institution’s expanded version of the learning management system Moodle.
21The VE project was supported by a number of technological tools. These included institutional platforms such as Loop Forum (for asynchronous discussions between students), Loop Reflect (for portfolio use), and Loop Workshop (for anonymous peer-reviews)1, as well as tools like Padlet (for pre-project multimodal presentations), Google Docs (for post-session asynchronous feedback), Zoom (for synchronous videoconferencing interactions), and L2L. The L2L tool (Figure 1), developed at the Irish institution, provided students with visualised conversation metrics, including audiographs, conversation shares, and conversation flow, to facilitate self-reflection and peer-feedback (Dey-Plissonneau, Gomez-Soler, Lee, Liu, Scriney, & Smeaton, 2023).
Figure 1 – L2L dashboard showing conversation metrics for 3 students (pseudonymised) interacting via Zoom.
22While the linguistic and technological tools afforded a structured environment that aligned with institutional goals and supported task completion (control), they collectively allowed learners the freedom to exercise interaction, collaboration, creativity and reflection. Capturing and understanding the continuous dialectical movement between ‘control and freedom’, and how it may impact on the realisation of agency, is key to better understanding the interplay between the sociological and psychological aspects of learner agency as they manifest themselves. The next sections will give insight into the participants and how their emerging actions mediated the dialectics of agency within the designed curriculum.
23Fifty students participated in the VE pedagogical project. The analysis focused on three Irish students and their respective French partners who all consented to fully participate in the research study. This small sample size nevertheless aligns well with the case study approach, which prioritizes in-depth exploration over breadth. By narrowing the focus, it was possible to conduct a granular and multimodal analysis of these students’ interactions, capturing nuanced manifestations of agency that might otherwise be missed in larger-scale studies.
24The three cases also reflect a balance of homogeneity and diversity within the broader cohort, representing a range of French proficiency levels: Erin at the lower-intermediate (A2), Ellie at the intermediate (B1+), and Annie at the advanced (C1+) levels, as shown in Table 2. The French students in semester 1 had a higher intermediate level of English. This variation enabled us to examine how agency unfolded across different linguistic competencies, while maintaining consistency in other key variables, such as educational background, triadic setting (imposed due to the number of students on either side), and participation in the same VE project.
Table 2 – Irish and French student participants.
|
Irish University (2nd Year BEd)
|
French University (Licence Communication, 2nd Year – VE1)
|
French University (Licence FLE & Didactique, 3rd Year – VE2)
|
|
Group 1: Annie (C1+)
|
Cathy, Jeanne
|
Solange, Adele
|
|
Group 2: Ellie (B1+)
|
Alexia, Tina
|
Zahra, Fatima
|
|
Group 3: Erin (A2)
|
Christine, Paul
|
Dalia, Aurélie
|
25Data was gathered through several methods, including:
-
Zoom recordings of synchronous interactions between Irish and French students, capturing the development of agency over two semesters, totaling 38 hours of recorded sessions for the three cases;
-
Post-project written reflections and podcast productions created by both Irish and French students;
-
Semi-structured interviews conducted after the projects with three Irish students to explore their experiences in more depth;
-
Conversation metrics generated by the L2L tool, reflecting students' participation levels during each session across the two projects.
26This study required a detailed, multimodal examination of how agency manifests in specific interactional contexts. Focusing on the interactions of three students allowed us to apply qualitative coding techniques with precision to annotate verbal, non-verbal, and textual interactions at a level of detail that would not have been feasible with a larger dataset. Elan (Brugman & Russel, 2004) was used to code and annotate video recordings of the synchronous sessions. Elan allowed for synchronised multi-tiered annotation of verbal speech, speech turns, non-verbal gestures, and any textual interaction (e.g. chat) between participants in each triad enabling a holistic analysis of multimodal interactions. Elan was particularly useful for capturing the complexity of agency as it unfolded dynamically across multiple modalities in real time, offering a granular level of detail that would not have been possible with traditional transcription methods alone. Annotation tiers based on Rainio and Hilppö's (2017) five dialectics of agency were created for each participant in each triad, namely control and freedom, dependence and separation, mastery and submission, situatively emergent and progressively developmental, and enacted and imagined. Segments pertaining to thematic shifts, conversational turns, and types of interactive tasks, such as asking for clarification, providing feedback, or engaging in social justice discussions, etc. were also identified and annotated, as well as turn-taking patterns, interruptions, overlaps, and moments of silence, which provided critical insight into the conversational dynamics and the ways in which agency was enacted or constrained.
27Other data sets were used for consolidation and triangulation purposes. Irish students’ interviews were analysed to identify their perception of the interaction dynamics, manifestations of agency, and tensions and struggles. Visualisations generated by L2L showing students’ levels of participation in each session for the two projects were used to note the tendencies in their speech densities. The online data from recordings and L2L metrics were compared with students’ testimonials of their perceived levels of success in their actual interactions.
28Table 3 below clarifies the distinction we see between enacted agency and situatively emergent agency in the context of a VE setting and this guided our coding. An example of ‘enacted agency’ was when a student initiated a discussion on a predetermined topic (e.g., cultural differences) as part of the task, demonstrating deliberate control over the conversation. In contrast, ‘situatively emergent agency’ was coded when a student shifted the conversation in response to their partner's unexpected question or comment, displaying a reactive and contextually bound form of agency.
Table 3 – Differences between enacted and situatively emergent agencies.
|
Enacted agency
|
Situatively emergent agency
|
|
Enacted agency is the result of concrete, intentional actions that the learner takes to achieve specific goals in VE.
|
Situatively emergent agency arises from context-specific factors. It is contingent on the immediate social and linguistic context, is spontaneous and can change depending on the dynamics of the VE.
|
|
Enacted agency is observable and can be directly measured. It involves tangible actions—speaking, writing, or creating content.
|
Situatively emergent agency is not always immediately observable, as it might manifest in subtle shifts in the learner’s engagement or responsiveness to external factors like peer behaviour or environmental constraints.
|
29The ‘enacted agencies’ and context-specific ‘situatively emergent agencies’ were identified in the online interactions. However, due to the ‘subtle’ nature of the situatively emergent agencies, we also relied on the post interviews to identify some salient situatively emergent features based on the students’ perceptions of their experiences. These accounted episodes were then traced back to the online interactions to see how the related actions unfolded. The enacted agencies (Table 4) and situatively emerging agencies (Table 5) were identified as follows.
Table 4 – Observed actions coded as enacted agencies.
|
Enacted agencies
|
Observed discursive actions in online interactions
|
|
Small Talk
|
Greetings and fixing sound
Improvised social interaction
|
|
Establishing rules for the interaction
|
Negotiating the session objectives and instructions
|
|
Asking questions & giving responses
|
Prompted by each session
|
|
Asking for & giving feedback
|
-
Programmed at the end of each session
|
|
Co-constructing lexical and thematic explanation
|
-
Session 3 in VE1 on idiomatic expressions
|
|
Co-constructing intercultural interaction
|
-
All sessions in VE1 and notably session 5 designed by the students
|
|
Co-constructing interaction on social justice issues
|
-
All sessions in VE2 and notably session 6 designed by the students
|
Table 5 – Observed actions coded as iteratively emergent agencies.
|
Situatively emergent agencies
|
Observed discursive actions in online interactions
|
|
Co-constructing simple and complex linguistic production
|
-
Compensating for lack of linguistic knowledge (incomprehension or incapacity to convey meaning) by asking for clarification or help
-
Compensating for lack of general knowledge
|
|
Co-constructing intercultural interaction
|
-
Explaining to the other a new concept that emerges
-
Spontaneous building up on peer’s response to relate personal experiences
|
|
Co-constructing interaction on social justice issues
|
-
Raising critical impromptu questions
-
Spontaneous building up on peer’s response to relate personal experiences
|
30We first captured the educational and technological affordances that were designed by the teachers, and that translated as ‘control and freedom’ agency as per Rainio and Hilppö’s (2017) framework. Then, through the lens of ‘situatively emergent’ agency, we examined the ways in which students navigated this environment and the contradictions that emerged in the process.
31Control and freedom in the pedagogical design manifested into other dialectical forms of agency when enacted by the students as discussed in the following sections. It was noted that ‘control and freedom’ as designed in the curriculum, when enacted by the learners in the course of the online interactions, manifested itself in terms of the different components of the dialectical agencies of Rainio and Hilppö’s (2017) framework. In the following case studies, the most salient emergent agentive actions of three Irish students with varying levels of proficiency in the target language are observed.
32The most striking agentive action in Ellie’s case (intermediate proficiency), especially in VE1, was her eagerness to get critical feedback (situatively emergent agency) from her French peers and her desire to improve her fluency and accuracy (imagined agency). In VE1, she insists in her online interactions that her peers not be worried about hurting her feelings as she values critical feedback. In the final session of VE1, unsatisfied by her peers’ sparse feedback, she even starts listing her own impressions of her linguistic deficiencies and errors to get her partners’ reactions on them (situatively emergent). In her interview, she evaluates her second VE as more successful because of the denser feedback received from her teacher trainee French partner and also because L2L showed a higher participation rate despite being “forced to speak in French for the full hour”: “I find the hour-long sessions [...] the more challenging of the two and I didn't want to do it, but I do think it helped my French the most” (04:13:51 Ellie).
33Ellie attributes her difficulty in oral interaction to her language training in secondary school where the focus had mainly been on writing (historical context). Her high self-awareness of her own shortcomings in oral French and her willingness to be perceived as a more fluent and accurate speaker was corroborated in her interview. She embraced the challenge despite feeling forced to do VE2 in French only (submission). However, she also noted that L2L showed an increase in her participation rate from 39% to 50% and she reasoned that that reflected an improvement not only in spoken but also in her aural skills (progressively development).
34For Ellie, the discomfort experienced in interacting with an expert speaker of the language was exacerbated when watching herself interacting in the video recordings as a less proficient speaker (dependence) of the language in her triad. Although students were encouraged to watch their videos back for auto-confrontation and reflection, Ellie chose not to engage with them (separation) until she was forced to do so for her reflection and podcast assignments (submission). This demonstrates a choreography between dependence, separation, and submission to achieve one’s goals.
35In the absence of a recording, it is generally difficult to objectively perceive progressive development for students in such student-led ephemeral synchronous videoconferencing environments. However, in our study, the L2L graphs allowed students to visualise progression thus enabling self-awareness and improved self-assessment. The graphs below show Ellie’s final sessions in VE1 and VE2 where she designed the session plan for her group. Based on Ellie’s testimonial in her interview, these were the sessions that were the most free-flowing for her as she mastered the topics (instead of working on teacher-proposed topics). VE1 (Graph 1) shows more density in Ellie’s participation (blue audiograph) in the second half (in English) as compared to the first half (in French). A qualitative evaluation of the interaction recording does reveal that the first half in French was laborious due to Ellie’s hesitant interlanguage which affected the whole group’s conversation dynamic. However, in the interaction (whole hour in French) from VE2 (Graph 3), we notice that both the Irish and French students’ participation is uniformly dense, and their conversation share metrics reveal an almost equal percentage, thus showing a more balanced interaction with greater mastery and control by Ellie. A qualitative evaluation of this interaction indeed reveals a certain flow in Ellie’s participation where she sounds passionate and well-informed on her chosen topic of feminism and explains its different aspects to her French partner. This increased efficacy and confidence is corroborated by Ellie in her interview where she said that the sheer volume of spoken French in the second semester (controlled by the design) forced her to prepare and progress more. The ability to observe progression in the quality and quantity of their interactions with the help of the video recordings, audiographs, and conversation metrics allowed students to have a more informed understanding of progress in their desired outcome.
Graph 1 – Ellie (B1+) VE1. Final session: Intercultural comparisons (Irish student’s thematic choice) (part 1).
Graph 2 – Ellie (B1+) VE1. Final session: Intercultural comparisons (Irish student’s thematic choice) (part 2).
Graph 3 – Ellie (B1+) VE2. Final session: Podcast discussion (Irish student’s thematic choice) (part 1).
Graph 4 – Ellie (B1+) VE2. Final session: Podcast discussion (Irish student’s thematic choice) (part 2).
36The other striking phenomenon of Ellie’s learning experience was her discomfort in discussing social justice topics (submission). This contradicts her envisioned role as a future teacher where she is required to be aware of such issues. She did not prefer relating personal stories to her peers, however she preferred VE2 as her peer opened up about her own personal vulnerabilities (situative agency). Such agentive actions depended on the peer or group dynamic (dependency).
37Gaining oral fluency and accuracy emerged as Ellie’s most predominant needs. She enacted self-criticality and self-awareness as situative agencies that made her progress toward her goals. However, these agencies were not necessarily induced by freedom in this instance, but more so by control in the curriculum design. Ellie was forced to confront them by the actual interactions and then post reflections, L2L metrics, recordings, and the podcast assignment. She felt discomfort regarding her slow and hesitant interlanguage, discussing social justice themes, and podcast building. These highlighted the tensions and contradictions between her enacted agencies and imagined role as a teacher. Despite her negative emotions, she boosted her participation rate and her confidence. These contradictions highlight the complex relationship between achieving tangible progress in language skills and the emotional challenges of communication in a second language, especially with expert speakers of the language in VE.
38Annie (advanced proficiency), in contrast to Ellie, followed a different dialectical trajectory, displaying a strong situatively emergent agency that was most apparent in her proactive approach to VE discussions. She consistently steered conversations toward politically charged social justice issues, revealing her deep personal investment in topics related to inclusion and equality. Annie's interactions were driven by her imagined agency as a future teacher who not only teaches language but also addresses broader societal issues. This proactive stance highlights how Annie navigated the dialectic between freedom and control, operating within the structured VE environment while simultaneously pushing the boundaries to discuss issues of personal relevance.
39Annie noted in her interview that her discussions with her partners in VE2 were more meaningful as she was paired with two French peers who were highly politically aware and had very informed stances on social justice issues, such as xenophobia and homophobia. On the other hand, in VE1, while Annie made attempts to separate herself from the proposed neutral session plans by introducing politically charged viewpoints, and enacting critical and creative agencies, the lack of general knowledge in such topics of her VE1 French partners did not allow the discussion to grow. This is an example of critical and creative ‘separation from the system’ as being dependent on peers to draw out the best experience for the agentive subject. The subject navigates between the mutually constitutive dynamic dialectical agencies of separation and dependence.
40The following example suggests that Annie was working in a free-flow mode (freedom) within the controlled environment of imposed assignments (control). This motivation for the enactment of agency was drawn from one’s imagined role as a teacher of French (enacted vs imagined agency). Annie speaking of the peer-review assessment notes that “I don't know exactly how much time, but I spent a lot of time because I was interested in that, and because I thought that it was beneficial, and I don't care how long I spend on it. I want to do a good job, you know”.
41Another instance of enacted and imagined agency manifested in VE1 session 2 when Annie’s peer Jeanne voiced a real dilemma following the mention of Mahsa Asimi in Iran. Their conversation revolved around social justice issues and the dilemma between being an empowered women and talking about it but also feeling disempowered and not being able to act on it to make a difference for women in need. This highlights the dilemma between enacted and imagined agencies as is the case in all VE conversations where the actions remain at the level of words and do not necessarily materialise into concrete actions.
42Another striking agentive quality of a highly motivated learner like Annie was her ability to view herself as a language learner. It not only heightened her critical assessment of her own capabilities but also propelled her to take the required action as a responsible learner. The ensuing agentive actions were reported as trying to take corrective actions and master the conversation dynamics. This is reflected in Annie’s interview with regard to the technological affordance of recording:
The recording was actually really helpful, because when you look back on yourself, you can identify errors that you make where you didn't in the moment, which is really good, because you're kind of treating yourself as your own student so as someone who's like training to be a French teacher, I was kind of my own student.
43In general, Annie positioned herself as the one who made decisions about the tasks. She asked more questions and filled in the gaps and pauses more often and was more critical in her stances. Creative, communicative, critical, and collaborative agencies emerged in the process. Despite being a proficient speaker, Annie talks about the cognitive load that such fast-paced interactions generate on her brain, but that also creates new circuits that enable her to enact emergent agencies iteratively and to improve from one session to the next, thus enacting progressive development. As evidenced in graphs 3 and 4, Annie’s triad’s participation in VE1 was quite flat in the first half (in French). This was attributed to end-of-day fatigue during the feedback session. Nevertheless, Annie picks up speed in the second half (in English), as evidenced by the red audiographs (60%). It is noteworthy that the preceding sessions’ L2L graphs show that Annie’s participation was the highest in her triad in VE1. This was because Annie was the most talkative in the triad and had a high general knowledge and would tend to dominate the conversations (mastery).
Graph 5 – Annie (C1+) VE1. Final session: Literature & films (Irish students’ thematic choice) (part 1).
Graph 6 – Annie (C1+) VE1. Final session: Literature & films (Irish students’ thematic choice) (part 2).
Graph 7 – Annie (C1+) VE2. Final session: Podcast discussion (Irish students’ thematic choice) (part 1).
Graph 8 – Annie (C1+) VE2. Final session: Podcast discussion (Irish students’ thematic choice) (part 2).
44In her interview, Annie noted that she was aware of her high conversation share, thanks to the L2L graphs, and was constantly endeavouring to reduce her participation to make it more equal among all three participants (submission) even though she really wanted to speak more French. The opportunity to speak in French for the whole hour in VE2 was therefore a boon for her, and despite her relatively dense and high conversation share in the final session of the second semester (40%) (blue in Graph 8), she managed to moderate her conversation share to make it more equal. This shows her willingness to submit to the implicit rule of equal participation between participants that was conveyed by the L2L graphs (submission). A high participation rate was not seen as a sign of mastery by students, but rather the capacity to modulate one’s participation to suit the triads’ needs and capacities primed. Thus, students constantly navigated the pull between manifesting their agencies as mastery and submission to support the group dynamics.
45By virtue of the controlled/designed technological affordance of the viewability of conversation shares and recordings that manifested as a situatively emergent agency (inherent to this learning environment only), it was possible to capture the emerging progressive development over the two projects in an otherwise uncontrolled ephemeral learning environment.
46The most salient finding in Erin’s (lower-intermediate proficiency) case was that VE heightened her awareness of her shortcomings in French, especially in view of her future Erasmus mobility. This challenged her to actively seek linguistic help and advice during her online interactions with her French peers.
47In session 2 of VE1 while discussing jobs, Erin talked about her future dreams and desires for a job as a translator, her imagined future job. This seems to be Erin's motivation to learn French as a foreign language despite the huge gap that she still must overcome to reach her goal. She asked her French peers what advice they had for her to improve her French and which place would be the best Erasmus destination to improve her French. Erin’s French peer Paul advised her that she needs to talk as much as possible with French people. Paul further explained that textbook French is very different from day-to-day French. Erin strongly agreed with Paul (dependence on peer).
48Even though Erin’s interlanguage was sometimes unintelligible, Paul reassured her that he understood her and that she could take her time to express herself when she struggled to speak in French. This positive empathetic feedback from her peer strongly encouraged Erin whose interactions in French became more and more fluent despite her frequent code-switching between French and English (progressively developmental).
49Furthermore, relating her experience as a learner of French, Erin noted: “When I talk to myself in my head in French, it's fine. When I go to talk to someone, I go blank” (Zoom interaction, session 2, 21 mins 52 secs). Thus, outlining the divide between the psychological perception (imagined agency) and the social enactment (enacted agency) of her interactive competence in the target language. Erin’s conflicting competency levels (between imagined and enacted) faithfully captures the type of predicament that lower-intermediate learners are forced to confront in VE learning environments.
50A careful observation of the verbal data revealed that Erin’s interactions were more accurate and meaningful in VE2. She was also able to bounce off her French partner’s interjections to offer corrective feedback to her Irish peers. For example, she explains the French word “flemmard” (lazy) to her Irish partner when one of the French peers uses it in VE2. It was Paul who had taught her this word in VE1, thus evidencing an uptake. This perception of an improved interaction quality is reflected in Erin’s own post-project interview, thus, highlighting her perception of progressive development at the end of her second project despite her struggles to interact with her partners in VE1.
51In all three cases, it was found that dependence vs separation was not only manifested in terms of material resources but also human resources. The latter is a unique affordance of VE pedagogy as interactions are not only mediated by technological and pedagogical resources, and language, but also by peers who serve as learning resources themselves. Moreover, their levels of motivation and investment influence the others in a group. All three participants note in their interviews that one’s engagement is dependent on the partners’ interest on the proposed topic of discussion.
52Rainio and Hilppö’s (2017) framework offers a powerful dialectical view on learner agency. However, our findings show that situatively emergent agencies do not necessarily manifest themselves in a structured well-defined dialectical relationship as proposed in Rainio and Hilppö’s framework, but rather through evolving combinations of the framework’s dialectical components. For example, a learner might suddenly shift the direction of a conversation in response to a peer's unexpected question, demonstrating quick adaptability to the unfolding dialogue, showing both dependence on peer and mastery at the same time, and thus they do not fall into neat categories as outlined by the dialectical framework. The subjects’ expected (mastery and submission) and unexpected turns (separation and dependence) are part of their enacted and situatively emergent agencies too. We note that these emerge in the moment as a result of the subjects’ interactions with the tools, environment and peers and the explicit and implicit rules of the learning system.
53The different components of the dialectical framework work more as mutual enablers rather than polarized categories. Progressive developmental agency depends on situatively emergent agency that evolves over time through the accumulation of experiences, skills, and knowledge. It emphasises the learner's gradual progression toward greater independence, competence, and self-direction. As learners build on prior experiences, their capacity to act agentively becomes more complex and refined. For example, Erin, who initially struggled with fluency in French, showed situatively emergent agency by constantly asking her peers for help. She gradually improved her ability to engage more meaningfully in VE discussions by the second semester. Her participation, while still challenging, became more confident and accurate, showing progressive development as she adapted to the VE environment.
54Control in the curriculum can give rise to enacted and situatively emergent agencies, however, not all situative emergent agencies lead to progressive development. They could also lead to unexpected outcomes. For example, although the scope for co-construction of language and thematic explanations as well as intercultural and social justice topics was deliberately designed in the session plans and so appear in the list of enacted agencies, these categories also gave rise to situatively emergent agencies, notably the spontaneous building up on peer’s response to relate personal experiences. While this aspect was greatly appreciated by Annie and Erin, Ellie didn’t feel comfortable discussing social justice topics. Situatively developmental agency thus gave rise to separation that did not necessarily lead to progressive development in Ellie’s case.
55Furthermore, indeed, motivation came from sessions that students had designed themselves. This had been deliberately designed by the teachers in the project to empower the students. However, a learner may submit to authority in one instance (e.g., by adhering to teacher instructions) while exercising creative control within those constraints (e.g., by innovating within the task). Learners deviated from the script to make it more suitable for themselves. In this way, they enacted separation. However, they also had to depend on the positive interest and wilful participation of their partners to follow certain scripts or modify them. In this way, situatively emergent and progressively developmental, dependence and separation, as well as mastery and submission agencies were at play simultaneously. This allows for a more nuanced understanding of the complex evolving nature of learner agency within structured environments like VE.
56In terms of practical implications of the study findings, we propose that VE environments must balance structure/control (e.g., predefined themes, timelines) and flexibility/freedom (e.g., self-selected discussion topics, flexible formats) to allow for the emergence of agency within a guided environment. Encouraging negotiation of constraints by allowing students to deviate from scripted tasks to pursue their own interests help them navigate the constraints of group dynamics to collaboratively reshape VE tasks and meet both individual and group needs. Furthermore, integrating reflective tasks (e.g., journals, peer feedback, post-project interviews, and the use of technology for self-assessment) help learners become more self-aware, align their enacted actions with their imagined goals, and navigate tensions effectively. For detailed pedagogical strategies, see Table X in the Appendix.
57This study addresses the gaps in the existing literature on agency in virtual exchange. While individual frameworks offer valuable insights, few studies integrate the diverse dimensions of agency, limiting their ability to fully capture its complexity in VE contexts. Moreover, there is a notable lack of longitudinal research tracking the development of agency over time, which is crucial for understanding its evolution. Furthermore, the role of technological mediation is explored, with focus on how L2L shapes and influences agency in VE. Moreover, this study demonstrates that learner agency in VE is complex, fluid, and emerges dynamically through interactions with their learning environment. Learners navigate various dialectical tensions, between control and freedom, dependence and separation, mastery and submission, and enacted and imagined agency, in response to the structured yet flexible VE environment. The dialectical framework provides a valuable lens for understanding how learners’ agency emerges dynamically as they confront the cognitive, social, and tool-related challenges of VE. Implications for VE pedagogy include the need to design flexible, authentic learning environments that balance structure and autonomy, allowing learners to experience both control and freedom. Task complexity should be progressively increased to foster developmental agency, and reflective practices should be integrated to help learners navigate the tensions they encounter. Tools like L2L empower learners by allowing them to concretely visualise their progress in such fast-paced learning environments.
58In terms of limitations of this study, the context-specific nature of agency means that results may vary across different linguistic, cultural, and institutional settings. Furthermore, despite rigorous triangulation methods and robust definitions, the interpretation of agency remains subject to researcher bias in coding and categorization. Lastly, the study primarily focuses on Irish learners’ agency, with limited exploration of French students' perspectives. Future research will explore how different cultural or linguistic contexts shape the dialectical emergence of learner agency in VE and examine how the French students’ perspectives on these interactions can further enhance and provide deeper insights into the development of agency as a co-constructed phenomenon in VE.