VET leaders are experienced professionals in leadership positions in governance, industry, international organisations, and higher education institutions. They oversee VET programs and have strategic or operational management responsibility for these programs’ national-or international-level goals (OECD, 2021). VET leaders may be school or higher education leaders with a teaching background (Ruiz-Valenzuela et al., 2017) or work in VET policymaking. Those working in educational institutions often focus on continuous educational advancements to maintain or enhance academic quality (Bouwmans et al., 2019), while those in governance and policymaking play a strategic role in introducing and implementing innovation (Arribas and Papadakis, 2019).

New megatrends that have emerged in recent years—like digitalisation, the transition to green economies, and the disruption due to the COVID-19 pandemic—have added to the complexity of VET leaders’ roles (Mathur et al., 2022). The impact of these trends on VET leaders is significant because they generate new demands in organisational, educational, and technological domains. Arribas (2016) emphasizes the crucial roles of individuals such as employers, union representatives, and members of civil society to actively promote inclusion and strengthen collaboration and coordination among regional and local VET stakeholders. Their efforts have the potential to build stronger social partnerships that actively shape and drive investments in skill development initiatives. For leaders in VET, responding to these demands necessitates the cultivation of resilient and interdisciplinary skill sets (Coates et al., 2013).

VET structures inherently require collaboration. Previous work in the field offers diverse insights into the predominant role of collaboration on a national level. Rusten and Hermelin (2017) discuss collaborative arrangements in Sweden and Norway, focusing on partnerships between industry and schools and their impact on the quality of VET programs. Bürgi and Gonon (2021) present the way in which professional associations (meso level) shape governance structures in Swiss VET, as well as examples of collaboration among state actors, unions, VET schools and single firms.

On the international level, a report for the European Training Foundation (Arribas and Papadakis, 2019) emphasise collaboration and synergy as integral characteristics of VET on several levels. First, within the school context, the VET dual model (which many countries consider optimal) is premised on collaboration between the workplace and school-based training programs. In Germany, for example, on the federal, state, and regional levels, the dual apprenticeship system shows a high degree of institutionalised collaboration, wherein the vocational schools are coupled with partners and companies (Gonon, 2014). Additionally, VET professionals working in schools must establish networks with social partners and develop the capacity to engage in VET policy development in cooperation with governments (Emmenegger and Seitzl, 2020). On a governmental level, VET requires vertical and horizontal coordination and cooperation (ETF, 2013). The horizontal dimension of multilevel governance involves coordination and collaboration between actors at the same organisational or institutional level, whereas the vertical dimension extends from higher governmental levels to the foundation of VET systems and vice versa. In the context of governmental coordination, the education and training policy sector has enacted laws to promote collaboration amongst EU Member States and other institutions (Arribas and Papadakis, 2019). VET leaders collaborate when, for example, they participate in advisory national or sectoral councils, coordinate between social partners and vocational schools, and take over new roles within territorial governmental departments and local offices on behalf of the central level.

In our increasingly complex professional lives, challenges necessitate combining individual expertise into a collective context (Graesser et al., 2018). Traditionally, CPS skills have been categorised into two fundamental domains: cognitive and social (Care et al., 2014). The framework for CPS skills explored in this study is grounded in this broader skill differentiation and builds upon this classification (Hesse et al., 2015; Liu et al., 2016). Social skills constitute the “collaborative” element of CPS, while cognitive skills represent the “problem-solving” dimension (Bagnou et al., 2022).

CPS research centres on two critical questions. The first considers how we can teach and learn to collaborate effectively in teams. Educational curricula across all levels increasingly recognise the significance of integrating CPS skills into training (Fiore et al., 2018). However, identifying effective pedagogical approaches to enhance CPS skills remains challenging (Graesser et al., 2018; Rosen, 2017). Merely assigning group activities does not guarantee improved collaboration skills (Karakostas and Demetriadis, 2010). Recent efforts have investigated instructional methods conducive to CPS skill development. Initiatives such as the Animalia Project (Rosen et al., 2020) have introduced performance indicators that shed light on collaborative learning processes leading to improved CPS in 14-year-old students. Additionally, Gu et al. (2015) have devised conversation exercises and problem-solving skills trainings to assist primary school students in planning, organising, and evaluating their collaborative tasks. Educational paradigms like this are critical for the early promotion of CPS skills. However, the practical application of these findings in corporate settings is questionable, given the complexities of group dynamics and the ill-defined tasks in professional contexts.

The second pivotal question about CPS pertains to its measurement, specifically focusing on how CPS skills can be assessed and quantified in tangible work and learning environments. A substantial body of literature has generated theoretical frameworks to address this issue. The most prominent are the large-scale international frameworks used in CPS assessments: PISA (PISA/ OECD, 2017) and Assessment and Teaching of 21st-Century Skills (Griffin and Care, 2015). In 2015, the Organization for Economic Cooperation and Development (OECD) measured CPS skills in students from 52 countries. The study revealed deficient cooperation skills, with fewer than 10% of pupils in OECD member countries achieving high scores. These results are a significant concern, as CPS skills are in high demand in the workforce (Fiore et al., 2018). Small-scale assessments have also explored various indicators for measuring performance in CPS, again in educational settings. The performance indicators include the quality of the solutions produced by collaboration (Avouris et al., 2003) and the level of team communication (Foltz and Martin, 2013). A combination of measurement tools has been used, including surveys, observations, and think-aloud protocols (Oliveri et al., 2017) in both online and face-to-face environments (Andrews-Todd et al., 2018; Siddiq and Scherer, 2017).

Moving beyond the educational realm, we need to investigate CPS as a driving force behind efficiency, effectiveness, and innovation (Fiore et al., 2018) in corporate and professional settings (Pappa et al., 2023). While working in groups, leaders discuss and analyse problems and their causes and identify solutions through conversations with those affected or involved (Saiti, 2014). Because CPS manifests in such practical applications, the present study focuses on conversations involving actual tasks. Research has already delved into how human-robot interaction transforms collaboration from the leadership perspective in the era of artificial intelligence (Sejera and Bocarnea, 2022; Tsai et al., 2022). This study addresses the gap in research focusing on human interactions and leaders’ everyday conversations (Meyer et al., 2020).

The present case study seeks to examine the CPS skills of VET leaders, with a specific emphasis on their team dynamics while handling tasks and challenges. To bridge the gap between laboratory measurements and real-world environments, we utilise an educational setting of an executive master’s program for experienced professionals.

Existing research on leaders’ collaboration makes use of hypothetical scenarios (Marcy and Mumford, 2010; Patuawa et al., 2022). Although these scenarios are based on realistic situations, they do not replicate the way in which real-life situations involve pressure and interpersonal dynamics (Graesser et al., 2018; Robinson et al., 2020).

Recognising the importance of CPS skills in enhancing leaders’ professional performance and employability prospects, we investigated the following research questions (RQs):

This single case study (Yin, 2018) recruited postgraduate students from a professional master’s degree program (n = 15) with a minimum of 5 years of work experience in leadership positions in VET. Some have positions managing or leading ministries, chambers, companies, and associations (n = 11), while others work in vocational education institutions such as vocational schools and colleges (n = 4). Participants who did not complete the self-reflection survey on their assessment of CPS skills (n = 3) were excluded. As a result, we report on data from nine female and three male participants with a mean age of 41.66 years (SD = 4.69). Nine have already obtained a master’s degree at the time of enrolment in the study. The average work experience is 17.1 years (SD = 6.3), and the average experience in leadership is approximately 7 years.

VET professionals in leadership positions participated in a master’s program aimed at upskilling experts in vocational education and training. We utilised videos of leaders working in groups of three or four on an assignment during the program’s first semester. The assignment required the groups to create a framework for peer-group mentoring (Prummer et al., 2023) to support them throughout their studies. Participants had 2 hours for this task, and they were asked to work on different elements and to prioritise the subtasks on which they could work as a team. All groups had access to the same supplementary materials and needed to choose what would be helpful for them to use. The videos were analysed to evaluate the participants interaction with a focus on their CPS skills.

The study utilised content analysis on video data and a survey tool asking VET leaders to self-reflect on their CPS skills. At the beginning of the first semester, participating leaders attended face-to-face classes. No previous knowledge was required for the group assignment, which demanded various stages of work and problem-solving skills. The assignment required participants to give both constructed responses and creative solutions (Puccio et al., 2018). The group activity was video recorded with the participants’ consent and the university’s approval, and the videos were transcribed verbatim. One rater coded the videos (Table 1) with a coding scheme based on the Hesse et al. (2015) rubric. The assessment noted each time that the participants showed behaviour relevant to a criterion, which sometimes occurred several times during the task. Two months after the video-recorded sessions, participants watched the videos of their group activities and used a simplified version of the evaluation rubric (Table 2) to reflect on their roles as collaborative problem-solvers.

To investigate VET leaders as collaborative problem-solvers, we employed a combination of concept-driven and inductive approaches and followed the five steps for validity in content analysis devised by Rourke and Anderson (2004). We draw from previous work on skills assessment frameworks, particularly relying on the rubrics developed by Hesse et al. (2015). Hesse and his team highlight that success in collaborative problem-solving (CPS) requires two key types of skills: social skills, which enable individuals to coordinate their actions effectively with others, and cognitive skills, which closely align with those used in individual problem-solving tasks. The social and cognitive skills outlined in this framework are essential for leadership roles in VET. According to Hesse et al.’s framework, the social dimension of CPS encompasses three categories: participation, perspective taking (Trötschel et al., 2011), and social regulation (Peterson and Behfar, 2004), which in turn link to nine corresponding codes altogether. The cognitive dimension includes two categories: task regulation and learning and knowledge-building skills, which are also associated with nine codes. In Table 2, we include six indicative codes from the overall 18 codes of the five social and cognitive categories, along with their indicators, descriptions of the inferences derived from the data to understand participants’ behaviours, and coded segments. After completing the deductive coding, we excluded the code “action” because the focus of the research is the quality and not the quantity of the interactions.

The simplified version of the rubric that the participants used for the self-reflection included items for the same dimensions and codes as the one used by the coder. This ensured that the independent coding of video data and the participants’ self-assessments were comparable. For example, the coding scheme for the coder included the following inferences: The participant acknowledges communication directly or indirectly (1 point), responds to or initiates interaction (2 points), or promotes interaction (3 points). For the same code, participants answered the following in the self-report survey: “Reflecting on my interaction, helping and responding to the contributions of others: a. I accepted communication directly or indirectly, b. I responded to cues in communication, or c. I started a conversation or activity and supported it.”

We used the MAXQDA (2022) software to analyse the video-recorded discussions during the assignment, focusing on the content and relevance of the individual contributions in relation to the task. We developed a coding scheme including low-, medium-, and high-inference codes, using one to three points depending on the speaker’s compliance with the predefined quality criteria (Table 2). This method enabled us to consider and account for the distinct characteristics and the quality of the interactions in a detailed way. For the coding, we used event sampling. The coding was conducted by an independent researcher to mitigate bias. To assess the intrarater reliability, a single rater coded the data on two occasions (Harvey, 2021). The second coding occurred 2 months after the initial assessment and followed the same coding scheme. It involved the random selection of four out of the 12 participants. We compared the scoring obtained on the two occasions and quantified the degree of agreement with the intraclass correlation coefficient (ICC) to assess the reliability of measurements made by the coder. Using the two-way random effects, consistency, single rater formula of ICC = (MSR - MSE) / (MSR + (k-1) * MSE), we obtained the ICC value ≈ of 0.78. This value indicates a good level of consistency, aligning with the thresholds suggested by Koo and Li (2016).

We calculated separately the frequencies, means, and standard deviations for the five CPS dimensions of the coded video transcripts (Table 1) and from the point of view of the participants’ responses from the self-report survey tool (Table 3). For both sources of data, we calculated the mean percentage correct for each dimension by converting mean scores to percentages based on the number of points available for that dimension.

To understand the relationship between the social and cognitive dimensions of VET leaders’ CPS skills (RQ2), according to the scoring of the independent coder, we performed a correlation analysis and calculated the Pearson correlation coefficient (r). In turn, we calculated how the social aspects of the participants’ CPS skills predicted their performance in the cognitive dimensions by running linear regression to measure the strength and direction of the relationship between these two continuous variables.

Finally, for the third RQ, we ran descriptive statistics on participants’ sociodemographic data, professional experience, and experience in leadership positions. We then conducted a correlation analysis between a. the sociodemographic characteristics and the different dimensions from the video content analysis conducted by the coder and b. the sociodemographic characteristics and the participants’ self-reflection surveys.

We ran means and standard deviations for each CPS dimension to answer the first research question on how VET leaders act as collaborative problem-solvers (Table 1), according to the coder. In total, with a wide range of earned points (39–275) and a small standard deviation (SD = 4.64), the data appears to be less dispersed around the mean (M = 134.33), implying that most participants scored relatively close to the average. Most participants collected higher scores on the social dimensions of participation and social regulation (M % correct = 46.11 and 46.16, respectively). The high standard deviation (SD = 35.86) in participation reveals that participation levels among the sample vary considerably. Perspective taking (M % correct = 29.69) had the lowest total score amongst the dimensions, with eight participants scoring below average and a minimum of zero points (maximum = 39). Similarly, almost two-thirds of the participants collected scores below average in the cognitive dimension of learning and knowledge building (M % correct = 36.45).

Table 3 describes how participants evaluated themselves as collaborative problem-solvers through the self-report survey. Perspective taking stands out as particularly strong (M % correct = 90), indicating that participants believe that they are highly self-efficacious. Overall, the total score of M % correct = 82 reflects participants’ assessment of their performance as strong across all dimensions (Table 4).

We performed a Pearson correlation analysis to answer the second research question exploring the relationship between the social and cognitive dimensions of VET leaders’ CPS skills according to the scoring of the coder. In summary, we found a significant, strong positive relationship between the two variables [r (12) = 0.705, p = 0.05].

To further investigate the nature and strength of the relationship between the two variables, we ran a linear regression analysis (Table 5). A significant regression was found: F (1.10) = 9.88, p < 0.05. The R² was 0.497, indicating that social dimensions explained approximately 49.7% of the variance in the cognitive dimensions. The standardised coefficient of beta = 0.705 suggests that these social dimensions have a strong positive standardised effect on cognitive dimensions, and the t-statistic had a value of 3.143 > 2.

The correlation analysis between the participants’ sociodemographic data and the scoring of the coder’s video content analysis, revealed a significant moderate positive correlation [r (12) = 0.589, p = 0.044] level (2-tailed) between participants’ age and their performance in the cognitive dimensions. The years of work experience also had a significant, moderate positive correlation to the cognitive dimension of task regulation [r (12) = 0.577, p < 0.05]. In contrast, the correlation between participants’ attributes and the results from the self-reflection survey was found to be weak and nonsignificant in most cases, with one exception: Participants’ age and the self-reflection survey general score were moderately negatively correlated and statistically significant [r (12) = −0.633, p < 0.05] level (2-tailed).

The limitations of the current study mainly stem from the fact that the fields of psychology and education are still in the early stages of addressing evaluation methods in the context of CPS. The Hesse et al. (2015) framework used for the content analysis is detailed and robust but also complex to administer and time-consuming for small settings. The framework was initially used to evaluate participants’ performance in several well-articulated, computer-based tasks. For this reason, a shortened version of the instrument should be used for small-scale future studies.

As Rosnow and Rosenthal (1991) highlight, the test–retest reliability method also comes with notable limitations. The purpose of our study and the behavioural and situational nature of the data we analysed led us to choose intrarater reliability (Harvey, 2021). We addressed reliability concerns by carefully considering the time gap between the initial and subsequent tests, ensuring that it was not too short in order to avoid coder recall bias. Additionally, it is essential to note the constraints of a small sample size like ours. To counterbalance this limitation, we augmented our approach by collecting rich video data and implementing triangulation in the analysis.

Despite these limitations, the current study’s findings have implications for research and practice. By examining leaders’ interactions in complex, practical situations, this study sheds light on how VET leaders navigate challenges and contributes to a clearer understanding of effective leadership in this context. Moreover, by obtaining detailed empirical data on CPS skills and identifying areas where leaders excel and areas for improvement, practitioners can develop targeted interventions and training programs to enhance these skills among VET leaders. Future research should actively address this gap, employing diverse assessment methods. We also suggest exploring CPS from various perspectives, such as the viewpoint of peers assessing their peers’ CPS skills. Finally, our findings emphasise the need to further study the reasons for the significant differences in leaders’ performance in CPS.