Pair Interactions and Emergent Roles in Problem-Solving with Modular Robotics in Primary Education

The ability to solve problems collaboratively has become essential for success across various personal and professional sectors, as it is a core competence required for navigating the complex, ill-structured challenges of modern workplaces and societies (Chai et al., 2023). Exhibiting proactivity as a collaborative team member, alongside strong verbal communication and problem-solving competencies, is essential in technology-mediated contexts (McGunagle & Zizka, 2020; Zhou & Ye, 2024), particularly in the utilization of robotic systems (Kurtz & Kohen-Vacs, 2024) and artificial intelligence for human-AI problem resolution (Schelble et al., 2024). Collaborative problem solving (CPS) is essential in disciplines such as engineering and technology, where innovation increasingly depends on diverse groups collaborating to develop solutions. Consequently, comprehending the dynamics of individual collaboration and problem-solving in real-world contexts is essential for educational research. CPS in education has been examined across various computer-supported collaborative learning (CSCL) tasks and scripts regarding students’ self-regulation (Miller & Hadwin, 2024), social interaction (Asfani & Chen, 2025), critical thinking (Xu et al., 2023), and learning performance (Tian & Zheng, 2024).

The present work focuses on students’ CPS skills in the context of educational robotics. OECD (2017) defines CPS “as the capacity of an individual to effectively engage in a process whereby two or more agents attempt to solve a problem by sharing the understanding and effort required to come to a solution and pooling their knowledge, skills, and efforts to reach that solution” (p. 5). Hesse et al. (2015) underline that CPS is a very promising activity that builds upon various social and cognitive skills and can be measured and taught in educational contexts. OECD (2017) proposes the assessment of CPS by evaluating the following competencies: (i) establishing and maintaining shared understanding; (ii) taking appropriate action to solve the problem; and (iii) establishing and maintaining team organisation. CPS is of crucial importance to face technology challenges in which competencies from different teammates are required to think critically (Xu et al., 2023) and to regulate the complexity of CPS in technology-rich environments (Polyakova et al., 2024). The study of CPS has been developed in the last few years in computer science education (Priemer et al., 2020; Umutlu, 2021) and with an increasing number of studies in educational robotics (Ates & Aktamis 2024; Avcı, 2024; Chevalier et al., 2022; Çakır et al., 2021). Over the past five years, classroom settings have increasingly incorporated educational robotics to improve problem-solving and collaboration skills. The meta-analysis of Zhang and Zhu (2024) observed a positive effect of educational robotics in creativity and problem-solving in kindergarten but also primary and secondary students. The research emphasises the value of educational robotics for promoting hands-on, interdisciplinary learning experiences that foster teamwork.

Despite growing interest in CPS, most studies focus on screen-based environments, leaving a gap in understanding how young learners interact around tangible, modular robotics tasks. This study addresses this gap by focusing on pair-based CPS, which offers a controlled context to examine fine-grained CPS processes.

Educational robotics encompasses a wide range of pedagogical approaches, from guided activities to ill-defined problem solving (Atman Uslu et al., 2022; Komis et al., 2017). Our study is situated within the latter, focusing on the use of robotics as a tool for children to engage in problem-solving through manipulation and social interaction, consistent with the principles of constructivism. According to Zhang and Zhu (2024), educational robotics “essentially a constructionist tool which students interact with and utilise their knowledge and experience to solve real problems” (p. 1540).

Ouyang and Xu (2024) found that incorporating robotics into primary science, technology, engineering, and mathematics (STEM) curricula significantly improved students’ cognitive flexibility and engagement in collaborative tasks. Similarly, Xu et al. (2024) observed that robotics-based activities promote co-construction of knowledge and negotiation of roles among students in problem-solving contexts. These findings suggest that educational robotics supports individual learning and creates authentic opportunities for students to develop key 21st-century skills such as cooperation and communication.

Educational robotics is also an opportunity to evaluate CPS from the perspective of learning theories that are important for the study. In relation to self-efficacy theory, educational robotics permits observing students’ willingness to engage with CPS (Jaipal-Jamani, 2024). The social nature of CPS in robotics also aligns with social constructivism, where knowledge is co-constructed through peer interaction and tangible engagement with educational robotics (Pelizzari et al., 2024). Activity Theory further deepens this understanding by framing the robotics classroom as an evolving system in which tools, rules, and community mediate collaborative outcomes (Romero & Barma, 2024). In this study, from this theoretical perspective, we aim to evaluate the evolution of collaborative roles within these pairs of learners, identifying the shifts in role adoption as the tasks progressed.

Studies on collaborative problem-solving (CPS) in screen-based learning environments have advanced significantly over the last decade, supported by learning analytics and diversifying CPS tasks for evaluation in computer-based environments (Chevalier et al., 2022; Çakır et al., 2021). In this study, CPS is analysed through a task mediated by “visuo-spatial constructive play objects” (VCPOs; Ness & Farenga, 2016), a type of constructible playful objects that include Lego bricks but also modular interactive objects such as the modular robots in this study (Ates & Aktamis 2024; Avcı, 2024; Kalmpourtzis & Romero, 2024, Lin et al., 2024). Modular robotics are relevant for the study of CPS because they enable hands-on, embodied interactions that make collaborative problem-solving processes observable in physical space. Their open-ended and constructible nature allows researchers to capture rich multimodal behaviours such as manipulation, negotiation and coordination, which are central to understanding how young learners collaborate in real-world settings in interaction with tangible artefacts. However, the analysis of CPS through VCPOs is still limited and faces challenges in identifying appropriate observables to develop behavioural learning analytics methodologies for non-screen interactions with VCPOs (Köhler & Romero, 2023). This study engages students in educational robotics to examine the specificities of manipulative pair problem-solving, the emergence of roles within CPS tasks, and the types of difficulties primary education learners encounter in these tasks.

In order to support students’ development of CPS through educational robotics activities, we need to understand the interaction through cooperative activities as a process (Pásztor-Kovács et al., 2023). For this objective, we examine students’ interactions in a CPS activity with educational robotics. Students’ interaction includes the manipulation of the robots, their emerging roles, and their verbal exchanges. In this paper, we closely examine students’ interactions in a CPS activity. We analyse the CPS process through the learners’ manipulations, emerging collaborative roles, and problems throughout their interaction.

To meet these goals, this study aims to examine the learners’ pair dynamics in CPS using educational robotics. We focus on how physical actions and conversations combine to shape learning, and they aim to uncover the intricate ways learners interact with modular robotic systems. This approach allows us to improve our understanding of how teamwork skills develop through practical, hands-on activities and how these skills can be best supported and evaluated in primary education settings.

This study contributes original insights into the field of collaborative problem solving by examining how young learners engage with modular robotics in unstructured, hands-on tasks using visuo-spatial constructive play objects (VCPOs; Ness & Farenga, 2016). While CPS has been widely studied in digital environments, there is limited research addressing how these processes unfold in non-screen, tangible contexts. The rationale for this study lies in addressing this gap by providing an in-depth behavioural analysis of pair interaction patterns and emergent roles during physical collaboration. The use of modular robotics as a CPS medium is particularly significant, as it introduces a tangible, interactive component that requires negotiation, role distribution, and real-time adaptation due to the material limitation using modular robotics.

We conduct a mixed-methods interaction analysis approach to examine how learners in pairs engage in collaborative problem-solving, the roles they assume, and the challenges they encounter during the CreaCube activity. We chose to focus on pairs of learners, rather than larger groups, to closely observe and track the CPS process. This structure maximizes the opportunity for both participants to be actively engaged in the task. The analysis is developed through the video recording of the creative problem-solving task and the subsequent analysis of the coding schema applied to the video.

Video analysis serves us in this study as a multimodal data source capturing learners and robot interaction. Following the methodological principles articulated by Heath (2025), we treat video not simply as a record of events, but as a resource for the fine-grained analysis of social action, enabling us to observe interaction as it unfolds in real time and within the material environment of the activity. Consistent with the learning sciences tradition outlined by Goldman et al. (2014), video research allows us to trace learning as a process situated in activity, embedded in context, and mediated by tools and social relationships. The interaction analysis is structured around a coding schema that integrates emerging collaborative roles specific to the CreaCube context, drawing on and adapting categories from Cassone et al. (2021) and Chiu (2001). The coding process is applied to the video-recorded problem-solving sessions, allowing us to systematically examine patterns of collaboration and interactional dynamics as learners engage with each other and the modular robot cubes across the activity. We choose to use modular robotics for facilitating hands-on, embodied interactions that make collaborative problem-solving processes directly observable in a physical space. This tangible, non-screen-based context permits study collaboration dynamics that are not fully captured in purely digital environments, but require the observation of human-robotic interaction.The open-ended and constructible nature of these systems allows researchers to observe the behaviors such as manipulation, negotiation, and coordination, which are central to understanding CPS in real-world settings.

The problem-solving task was intentionally designed to be ill-structured. Unlike well-defined problems with a single correct solution, ill-structured problems lack a clear path to resolution, requiring participants to define the problem, generate multiple solutions, and evaluate their effectiveness (Romero et al., 2022). This type of task is particularly valuable for studying collaborative dynamics because it necessitates extensive communication, negotiation, and the co-construction of knowledge. The use of modular robotics, in this context, allows for open-ended exploration and multiple possible solutions, providing a rich dataset for observing how pairs coordinate their actions, manage disagreements, and create a shared understanding to reach a final outcome. This approach aligns with a social constructivist perspective, where problem-solving is viewed as a process of co-construction.

In the first task, participants assemble four pre-programmed cubes—a white motor cube, a red inverser cube, a black distance sensor cube, and a dark blue battery cube—to construct a vehicle that navigates from a red to a black point, marked either on a mat or by tokens on the table. The second task replicates the conditions of the first, with participants instructed to reassemble the cubes to either replicate or modify the initial vehicle design. We conduct both tasks under uniform instructions and record them for analysis.

To achieve this, we adopt a micro-analytic approach that considers both verbal and non-verbal interactional features, including speech, gestures, gaze coordination, and manipulative actions with the cubes. By analyzing these multimodal interactions, we aim to understand how participants construct collaborative engagement in real-time.

The following research questions drive the study:

This section presents the results according to each research question (RQ).

This study examines how six pairs engaged in CPS interact with modular robotics in a dynamic, unstructured activity. This analysis revealed that there is no single cooperative strategy for solving the CreaCube activity, but rather a variety of strategies, each composed of different cooperative manipulation forms, that can lead to the completion of the activity. In the same vein, the emerging collaborative roles assumed by the participants in every pair were quite different from one pair to the next. Strijbos and Weinberger (2010) explain the diversity in the roles assumed throughout ill-structured activities, supporting the idea that the roles emerging in CSCL closely relate to participants’ personal skills. Therefore, as participants have different personal traits and skills, it is reasonable that they will adopt different collaborative roles in the activity. Despite the diversity in participants’ roles, there were some common points in the roles that emerged over the course of the activity; in pairs’ first contact with the cubes, the most common role emerged was that of the ‘explorer’, while towards the end of the first and second tasks, most of the participants behaved as ‘proposers’. The assumed roles align with the essence of the proposed task, which requires first exploring the cubes’ characteristics to build a solution. Our analysis revealed that the most successful pairs navigated the problem-solving process by fluently transitioning between emergent roles and cooperative forms that supported each stage of the cycle. During problem definition, roles such as Proposer (R1) and Critic (R2) were essential as they used Acquiescent co-elaboration (M2) to establish a shared understanding of the CreaCube task. This collaborative foundation allowed them to transition effectively into the solution generation phase, where they engaged in Mutual co-construction (M3) to develop a joint plan. In contrast, pairs who struggled often demonstrated Individual manipulation (M0) during the implementation stage, a behavior that bypassed both solution generation and evaluation, ultimately leading to less favorable outcomes and highlighting a breakdown in the cooperative process.

Furthermore, the study of participants’ emerging collaborative roles revealed that some participants had a passive attitude, assuming the role of the follower in some parts of the task (Pair 2, participant A, and Pair 3, participant B) or throughout the whole course of the activity (Pair 4, participant A). Cassone et al. (2021) have previously observed and documented submissive and domineering behaviours in collaboration during the CreaCube activity.

This study backs up what was already known about dominant behaviours in CreaCube and shows that when two people play together, they show individualistic and competitive behaviours, both verbally and physically. Instead of showing cooperative behaviour to collectively reach a solution, some participants pursue their own solution to the task at an individual level. As reported by Popov et al. (2019), “individualists are more likely to exhibit competitive behaviour focused on personal achievement” (p. 103). For instance, some pupils insisted on being the ones to lead the pair to the solution of the activity, and as Cassone et al. (2021) reported, some pupils seemed to be disappointed in not being the ones who led the pair to success. Some pupils, when engaged in the CreaCube task, do not perceive the task as a cooperative learning situation but rather as a competitive and individualistic situation; according to Johnson et al. (1986), “in a competitive learning situation, when one student achieves their goal, all others with whom they are competitively linked fail to achieve their goals, and in an individualistic learning situation, students’ goal achievements are independent” (p. 383). Therefore, cooperative learning is not always present in CPS activities. Participants’ unfamiliarity with collaborative tasks and unstructured CPS activities could explain this. While this exploratory study’s scope and sample size limit its broad generalizability, this micro-analytic analysis (Palincsar, 2012) was essential for a fine-grained understanding of how collaborative roles and interaction patterns emerge and evolve during a CPS robotic activity, a key contribution that provides a foundational model for future, larger-scale research.

The gender composition of the pair is another aspect that requires further study. The study’s results demonstrated that in same-gender pairs, a lot of competition occurs when the participants are male (Pairs 1, 5, and 6), but no competition occurs when all participants are female (Pair 2). When it comes to the mixed gender pairs (Pairs 3 and 4), the two cases in the sample are contradictory; in Pair 3, there was a very dominant female participant, while in pair 4, the female was unwilling to actively participate in the task. Given the limited sample size, we need to conduct additional research on the gender composition of pairs in CPS to obtain reliable results on potential gender differences in manipulation processes and emerging collaborative roles. Previous studies in this area have found interesting but sometimes conflicting results. For example, Holmes-Lonergan’s study (2024) on traditional gender stereotyped behaviours found that preschool girls use mitigation more frequently than boys when engaged in problem-solving tasks, and that girls in same-gender pairs are more likely to agree with each other than boys. On the contrary, Tomai et al. (2014) found that computer-supported collaborative learning (CSCL) promotes counter-stereotypical gender communication styles in male and female university students, with males exhibiting more social-emotional communication styles and females showing task-oriented communication styles.

The observed prevalence of competitive behaviors suggests that educators should scaffold CPS activities by assigning flexible but predefined roles. Such scaffolding can help balance participation and foster positive interdependence. Based on our finding that pairs with complementary and balanced roles showed better coordination and outcomes, we also propose a structured pedagogical intervention. Teachers could implement a short pre-task role-definition exercise to mitigate competitive behavior and the lack of a leading figure. This would involve introducing and having students explicitly assign themselves roles like Proposer, Critic, Hand, and Follower before starting a task. The intervention would also include a mandatory role rotation, ensuring both partners experience different functions, thereby fostering more equitable and effective collaboration.

This study provides a multimodal framework for analyzing CPS in modular robotics tasks, highlighting the importance of co-construction and balanced role distribution. Future studies should investigate scalable scaffolding strategies and explore gender-related dynamics in larger samples. Additionally, investigating the impact of predefined role assignments on collaboration effectiveness could provide further insights.