Collaboration is often envisioned as a way of working where individuals rely on each other, leading to greater engagement and better outcomes than if they worked alone. This principle is fundamental to education, where collaboration fosters shared knowledge and collective values. Collaborative Problem Solving (CPS) has been widely recognized as more effective than individual or competitive approaches, benefiting both academic and social development (Eleftheriadou, 2022). Moreover, CPS skills are increasingly valued in modern workplaces (Hesse et al., 2015), positioning them as essential competencies for the 21st century. Despite these advantages, integrating CPS into education remains challenging (Le et al., 2018). CPS is not merely about solving problems—it encompasses both cognitive and social dimensions, requiring students to communicate, negotiate, and engage in critical thinking (Hesse et al., 2015). Effective collaboration fosters a dynamic learning environment where students actively build on each other's ideas, deepening their understanding and enhancing their problem-solving abilities. However, even though CPS is often associated with adults and adolescents (Veldman & Kostons, 2019), high school students frequently struggle with group tasks (National Research Council, 2012). Recognized as a key 21st-century skill (Sun et al., 2021), CPS has been emphasized in international education frameworks such as those developed by the OECD (OECD, 2017). While CPS is a widely endorsed educational goal, its implementation and assessment pose significant challenges, particularly in identifying the interactional dynamics that lead to productive collaboration. To address this gap, a measurement instrument for assessing productive interaction in CPS was developed based on findings from Baucal et al. (2023). The instrument is grounded in four key components that define productive collaboration: 1) Socio-Cognitive Engagement – The extent to which students contribute ideas, challenge reasoning, and build on each other’s thoughts, facilitating knowledge co-construction; 2) Socio-Emotional Regulation – The ability to manage emotions within the group, resolve conflicts constructively, and foster a supportive social climate and; 3) Task Regulation – The ability of group members to establish shared goals, plan strategies, and adapt approaches based on task demands;  4) Group Regulation – The extent to which a group collectively manages interactions, balances participation, and maintains a shared commitment to collaboration. These dimensions capture the interplay between cognitive, emotional, and regulatory processes that contribute to effective CPS. By quantifying these aspects, the instrument enables researchers and educators to examine how different collaborative behaviors impact problem-solving success.

The main aim of this study is to determine whether distinct patterns or groups of collaborative interaction can be identified based on this instrument, alongside the quality of the problem solution. To achieve this, we employ cluster analysis in order to identify interaction profiles. In this way we would deeper understand how students engage in CPS and the conditions under which collaboration leads to successful problem-solving. 

Participants: The research was conducted with a sample of 120 second-grade students (aged 16) from six grammar schools and six vocational secondary schools in a metropolitan area. These students were organized into 42 subgroups, consisting of triads and dyads. The composition of student groups varied in terms of gender (female 72 %) and school type (vocational 50%). Procedure: Tenth-grade students were invited to participate in the study through school counselors. Those who volunteered formed triads based on personal preference, with each triad consisting of same-gender classmates. All groups engaged in a CPS session, solving a complex scientific task. The tasks were adapted from OECD/PISA science literacy assessments (Pavlović Babić et al., 2009), selecting higher proficiency items and modifying them to require constructed responses and argumentation. Seven tasks were randomly assigned to triads, and measures were taken to prevent cross-group discussion within schools. Students used tablets with internet access and were encouraged to consult various resources. Each triad produced a written solution, later evaluated by independent coders. Their interactions were video recorded. Sessions took place during school hours, with an experimenter present to introduce tasks, monitor progress, and address technical issues. Measure: For cluster analyses we used measures of quality of collaboration (CPS Observational Grid (CPS-OG) and quality of solution (CPS Product Assessment Protocol, CPS-PAP 2). CPS-OG was assessed based on video recordings of CPS sessions, with two independent reviewers assessing each session using a 20-item observational grid reflecting the four key CPS dimensions (Baucal et al., 2023). In addition to the items covering these four dimensions, CPS–OG also includes items related to group engagement, the quality of discussion, and the extent to which three types of dialogue were present during the group interaction. Each item was rated on a five-point Likert scale (0 = not at all to 4 = to a very large extent). Item-level intraclass correlation (ICC) indexes reached excellent values (Cicchetti, 1994), ranging from .82 to .91. Composite reliability indicated good internal consistency across all CPS-OG dimensions (SC = 0.94, SE = 0.90, TM = 0.87, GM=0.86). CPS-PAP 2 measured product quality by evaluating each group's written solution at the end of the session. Two independent raters scored each solution based on (1) accuracy (0–4 points) and (2) quality of argumentation (0–2 points), with total scores ranging from 0 to 6. The ICC for solution accuracy was excellent (.84), while the ICC for argumentation quality was good (.64).

Our preliminary study indicates that groups engaged in solving this task can be classified based on collaboration quality. Cluster analysis (Ward’s method, Euclidean distance) was used to identify group patterns, revealing four distinct groups. These groups differ significantly in their scores across all four dimensions of collaboration. To better understand these differences, a nonparametric test (Independent Samples Kruskal-Wallis) was conducted. The second group exhibits the highest scores in all four dimensions, whereas the fourth group has the lowest scores and shows minimal team member engagement. The third group demonstrates low scores on dimensions related to the social aspects of collaboration, specifically socio-emotional and group regulation aspects. The first group scores high across all four dimensions, though not as high as the second group. Regarding the types of dialogue present in these groups, analyses indicate that disputational dialogue is more prevalent in the third group, while cumulative dialogue is most prominent in the fourth group. The solution proposed by the members of the second group is more accurate compared to the solutions of the other groups. Surprisingly, the groups do not differ in the quality of the proposed arguments. However, it is important to note that these groups were not homogeneous in terms of their members' initial competencies. Additionally, we will examine whether the same group patterns emerge in a less structured task that involves discussing a broader everyday issue. Future analyses will incorporate qualitative insights from video-recorded collaborative processes to better understand differences and similarities between these groups and their relationship to performance. The findings suggest that different groups may encounter distinct obstacles during the collaboration task, potentially reducing the effectiveness of this learning method. Therefore, teachers' feedback and scaffolding during collaborative problem-solving should be tailored to address the specific difficulties identified in each group.

Baucal, A., Jošić, S., Ilić, I. S., Videnović, M., Ivanović, J., & Krstić, K. (2023). What makes peer collaborative problem solving productive or unproductive: A qualitative systematic review. Educational Research Review, 100567. https://doi.org/10.1016/j.edurev.2023.100567 Cicchetti, D. V. (1994). Guidelines, criteria, and rules of thumb for evaluating normed and standardized assessment instruments in psychology. Psychological Assessment, 6(4), 284–290. https://doi.org/10.1037/1040-3590.6.4.284 Eleftheriadou, S. (2022). Assessment of students’ collaborative problem-solving competence: A systematic literature review. European Conference on Educational Research (ECER). European Educational Research Association. Yerevan, Armenia, 23–25 August 2022. Hesse, F., Care, E., Buder, J., Sassenberg, K., & Griffin, P. (2015). A framework for teachable collaborative problem solving skills. Assessment and Teaching of 21st-century skills: Methods and approach, 37–56. https://doi.org/10.1007/978-94-017-9395-7_2 Le, H., Janssen, J., & Wubbels, T. (2018). Collaborative learning practices: teacher and student perceived obstacles to effective student collaboration. Cambridge Journal of Education, 48(1), 103–122. https://doi.org/10.1080/0305764X.2016.1259389 National Research Council. (2011). Assessing 21st Century Skills: Summary of a Workshop. National Academies Press (US). http://www.ncbi.nlm.nih.gov/books/NBK84218/ Pavlović Babić, D., Baucal, A., Kuzmanović, D. (2009). Naučna pismenost: PISA 2003 i PISA 2006. Beograd: Ministarstvo prosvete Republike Srbije, Zavod za vrednovanje kvaliteta obrazovanja i vaspitanja Sun, D., Ouyang, F., Li, Y., & Chen, H. (2021). Three contrasting pairs’ collaborative programming processes in China's secondary education. Journal of Educational Computing Research, 59(4), 740–762. https://doi.org/10.1177/0735633120973430