1 Introduction

Since the early 1970s, collaborative learning has been a crucial teaching strategy whether in traditional classrooms, blended learning or distance education (Al-Samarraie & Saeed, 2018; Laal & Ghodsi, 2012; Mayes, 2018; Zhu, 2012). Collaborative learning encourages students to learn by exchanging ideas or experiences in small groups (Dillenbourg, 1999). Existing research has demonstrated that collaborative learning is positively associated with learning effectiveness (Sumtsova et al., 2018; Van Leeuwen & Janssen, 2019).

After the outbreak of COVID-19, the limitations of social distance force all classes to become online models (Surani & Hamidah, 2020). Compared with FtF setting, students in online environments are experiencing more challenges during computer-supported collaborative learning (CSCL) (Zheng et al., 2022). More specially, social interaction in CSCL are limited by technology, interaction patterns, and social distance, and students have a low sense of social presence (Zhang et al., 2023; Zhou, 2023). Furthermore, constrained by the limitations of computer mediation, building group norms, trust, and a sense of community belonging in an online environment is more challenging (Baturay & Toker, 2019; Fernandes, 2018). However, all of these factors are positively associated with the efficiency of collaborative learning, placing the CSCL in the online environment at a disadvantage (Zhang et al., 2022, 2023).

Researchers have long focused on ways to improve students’ collaborative learning experiences. Group composition has been emphasized as a crucial consideration when designing collaborative learning (Farland et al., 2019; Post et al., 2020). In traditional face-to-face (FtF) and blended learning settings, allowing students to collaborate with familiar people has consistently been reported to have a positive impact on students’ collaborative learning experiences, especially for affective learning outcomes (Chapman et al., 2006; Mahenthiran & Rouse, 2000). For example, promoting a higher level of psychological safety and group cohesion (Schepers et al., 2008; Van den Bossche et al., 2006). Driven by the benefits of group member familiarity, we suppose that grouping familiar people in the same group would help improve the efficiency of CSCL in an online environment.

The difference between FtF communication and CMC (computer-mediated communication) has been widely demonstrated from various perspectives (Ean, 2010; Rhoads, 2010; Romiszowski & Mason, 2013; Yang et al., 2022). Although there are inherent disadvantages to CMC, such as the lack of non-verbal clues, hyperpersonal model (Walther, 1996) showed that users in CMC environments may feel more unity, liking, and intimacy than they would in FtF interaction. CSCL in online settings and FtF collaborative learning as two different forms of mediated interaction, yet no research has explored the differences between them, particularly in terms of group member familiarity.

Therefore, the current study accomplishes two objectives. We first examine how group member familiarity affects students’ engagement and perceived knowledge construction during CSCL. In this way, we argue that group member familiarity has an indirect effect on student performance through teamwork satisfaction rather than a direct effect. Thus, teamwork satisfaction mediates the relationship between group member familiarity and students’ experiences with collaborative learning. Second, the study contrasts how group member familiarity functions in CSCL and FtF collaborative learning contexts.

2 Literature review

2.1 Group member familiarity

Group member familiarity is defined as the interpersonal familiarity with other group members (Rockett & Okhuysen, 2002; Yang et al., 2022). Various perspectives, such as group norms, communication, and collaborative activities, have shown the benefits of group member familiarity (Janssen et al., 2009). In established groups, developing norms and rules quickly requires minimal effort (Postmes et al., 2001). Group cohesion is likely to increase with increased familiarity, and group members can advance through the different phases of team development more quickly (Adams et al., 2005; Mennecke et al., 1995). Furthermore, according to the “information pooling strategy” (Gruenfeld et al., 1996), groups composed of familiar people are more effective in aggregating the information they individually learn and develop trust in the group (Wilson et al., 2006). In groups with more familiar members, students might feel more comfortable expressing their disagreements, leading to more effective communication (Gruenfeld et al., 1996). All of these imply that close group member knowledge will increase students’ collaboration’s efficiency and effectiveness.

2.2 Group member familiarity and teamwork satisfaction on CSCL in online environment

CSCL is the computer-mediated equivalent of conventional FtF collaborative learning. It refers to participants sharing and building knowledge with the help of technology as the primary means of communication or as a common resource (Nam & Zellner, 2011). CSCL can be implemented in both online and classroom learning environments (Jeong et al., 2019). However, due to the limitations of the new crown epidemic, the CSCL studied in this paper was only implemented in an online environment. Encouragement of active and constructive learning, in-depth information processing, critical thinking, and goal-based learning are considered equally effective in CSCL as they are in a conventional and blended collaborative environment (Chou & Chen, 2008; Graham & Misanchuk, 2004).

Because the absence of nonverbal cues limits CMC, students in online collaboration inevitably face more challenges (Anderson et al., 2010; Robinson et al., 2017). For example, the lack of non-verbal interaction makes it difficult for students to make connections with each other (Wang & Zou, 2021). In addition, due to the restrictions of the COVID-19, students have to stay at home and attend classes alone. Compared with FtF environment, students experience greater psychological stress (Xiong et al., 2020). There is almost zero emotional support from the school, teachers and classmates (Van Leeuwen & Janssen, 2019). These are not conducive to CSCL in an online environment.

According to Walther (1992), as people learn about their partners, they may develop deindividuating impressions of their peers, which may help them overcome the inherent limitations of the medium, such as a lack of verbal cues, voice inflection, and gestures. Furthermore, according to Walther (1996) hyperpersonal model, users in CMC environments may feel more unity, liking, and intimacy than they would in FtF interaction. Empirical research has suggested that the CMC environment promotes more effective and equitable interaction (Flanagin et al., 2002; Walther, 1996). However, information on group member familiarity in online CSCL settings is limited. Furthermore, it remains unclear, whether group member familiarity promotes greater teamwork satisfaction in online CSCL compared with FtF settings.

To address this lack of pertinent research, the current study poses the following research questions.

RQ1: How do group member familiarity and the associated outcome variables differ between CSCL and FtF collaborative learning?

RQ1(a): How does the impact of group member familiarity on students’ teamwork satisfaction differ between CSCL and FtF collaborative learning?

RQ1(b): What is the difference in the impact of teamwork satisfaction on student engagement in CSCL and FtF cooperative learning?

RQ1(c): What is the difference in the impact of teamwork satisfaction on students’ perceived knowledge construction in CSCL and FtF cooperative learning?

2.3 Mediating role of teamwork satisfaction

Some studies have examined the implications of group member familiarity on collaborative learning in FtF and blended environment (Chapman et al., 2006; Crompton et al., 2022; Janssen et al., 2009; Ku et al., 2013). However, the findings suggest that familiarity of group members has a positive outcome on affective learning, but the relationship between familiarity and cognitive outcomes has not been found. More specially, according to Janssen et al. (2009) and Ku et al. (2013), familiarity plays a significant role in teamwork satisfaction, but there is no evidence of a significant relationship between familiarity and improved group performance.

Previous research has shown the role of group member familiarity from an affective learning perspective (Janssen et al., 2009; Ku et al., 2013), but both affective and cognitive learning are significant indicators for evaluating student learning outcomes. When examining the factors influencing student learning outcomes, existing research has argued that affective learning as a mediating variable influences cognitive learning (Bolkan, 2015; Rodríguez et al., 1996). To evaluate the role of group member familiarity, we use teamwork satisfaction as a mediating variable and assessed learning outcomes with two cognitive learning dependent variables: student engagement and perceived knowledge construction, respectively. The following hypothesis and research question are proposed (shown in Figure 1).

Fig. 1
figure 1

Hypothesized model

Full size image

H1

Group member familiarity influences teamwork satisfaction, which in turn promotes higher student engagement and better perceived knowledge construction.

RQ2: Does the mediation effect of teamwork satisfaction differ between CSCL and FtF collaborative learning?

3 Method

3.1 Participants

The participants in this study were students from different universities in Seoul, South Korea. Suitable respondents were selected by asking if they had experience with collaborative learning. Subjects who were eligible for the study were instructed to recall a recent CSCL or FtF collaborative learning experience, including the type of collaborative learning (CSCL or FtF collaborative learning), duration, task, group members and to complete the questionnaires based on that experience.

The questionnaires on CSCL and FtF collaborative learning were completed by those who had experience with respective method. For the students who had participated in both CSCL and FtF collaborative learning, the system randomly assigned the content of the questionnaires they answered. All participants were provided with information about the study and agreed to participate before starting the survey.

To recruit participants, we used a qualified research panel provided by Macromill Embrain, a reputable online questionnaire company in South Korea. The link to the questionnaire was send to university students by email. A total of 436 valid questionnaires were collected, with 221 participants in the CSCL survey and 215 in the FtF collaborative learning survey. Demographic information for all participants is shown in Table 1.

Table 1 Participants’ demographic information (N = 436)
Full size table

3.2 Measures

The questionnaire contains three parts. The first section collected demographic data about the participants (e.g. age, gender, level of education and preferences for collaborative learning). In the second part, participants were instructed to recall a recent collaborative learning experience and based on their experience, to answer questions relating to group members familiarity. The third part assessed student learning outcomes, containing both cognitive and affective outcomes. All the questions involved in the questionnaires were derived and modified from existing research. For all measures in the third part, responses were gathered using a 7-point Likert-type scale (1 = strongly disagree, 7 = strongly agree). To ensure that the two datasets, CSCL and FtF collaborative learning, were comparable, we used a standardized data collection method and the same questionnaire.

In this study, group member familiarity was measured as an independent variable. Based on Janssen et al. (2009) work, students were asked to rate the familiarity of group members on a 7-point scale ranging from 1 (not familiar at all) to 7 (very well known) based on their recent CSCL or FtF collaborative learning experience. This question was preceded by three specific yes/no questions (for example, “I have connections with some of these group members in my life”) designed to help students in assessing familiarity with group members. The validity of the familiarity measure was tested by correlating the sum of the three “yes/no” questions with the familiarity measure. A significant correlation was found in CSCL (r = .68, p < .002) and FtF collaborative learning (r = .71, p < .002). This demonstrated the validity of the overall familiarity scores (Mean = 4.018, SD = 0.95 in CSCL, Mean = 4.312, SD = 0.90 in FtF collaborative learning).

Teamwork satisfaction (α = 0.92 in CSCL, α = 0.90 in FtF collaborative learning) was measured using three items, (“I enjoy collaborating with my teammates to solve problems,” “I enjoy the peer interaction in collaborative learning,” and “I enjoy the collaborative learning environment”) adapted and modified from Ku et al. (2013). Engagement (α = 0.80 in CSCL, α = 0.85 in FtF collaborative learning) was calculated using three items, (“I am very active in interacting with group members in collaborative learning,” “My teammates and I are prompt in responding to messages in collaborative learning,” and “I am actively involved in all aspects of collaborative learning”) adapted and modified from Post et al. (2020). Three items were used to calculate perceived knowledge construction (α = 0.85 in CSCL, α = 0.90 in FtF collaborative learning), namely “Collaborative learning helps me construct knowledge more effectively,” “My teammates, and I are prompt in responding to messages in collaborative learning,” and “Collaborative learning improves my ability to communicate with others,” which were adapted and modified from Sanders and Wiseman (1990).

A multiple-item online survey measuring collaborative learning experiences was distributed to university students in Seoul, Korea. To ensure that the translated Korean version of the questions remained true to the originals, two qualified translators were invited to translate the questionnaire into Korean, which was then reviewed by two bilingual researchers. Finally, to ensure that there were no ambiguities, two researchers independently checked the questions.

4 Results

A one-way multivariate analysis of variance was conducted (F4, 431 = 5.636, p < .001) to investigate the differences between CSCL and FtF collaborative learning in group member familiarity, teamwork satisfaction, student engagement, and perceived knowledge construction. The results shown in Table 2 indicate that CSCL and FtF collaborative learning are significantly different in group member familiarity (F4, 431 = 5.745, p < .05), teamwork satisfaction (F4, 431 = 8.144, p < .001), student engagement (F4, 431 = 22.046, p < .001), and perceived knowledge construction (F4, 431 = 8.007, p < .05).

Table 2 Means, standard deviations, and MANOVA analysis variables by collaborative learning type
Full size table

4.1 Multiple group analysis

4.1.1 Reliability and validity examinations

Composite reliability (CR), average variance extracted (AVE), and Cronbach’s alpha coefficient were calculated to evaluate the reliability of the proposed model and questionnaire design.

As shown in Table 3, the CRs of all structures were greater than the recommended value of 0.7; all AVEs were greater than the recommended value of 0.5; and all Cronbach’s values were greater than the recommended value of 0.7 (Tavakol & Dennick, 2011). Therefore, we believe that the reliability, convergent validity, and internal consistency of our model were well acceptable.

Table 3 Reliability and discriminant validity results by collaborative learning type
Full size table

Discriminant validity was up to standard when the minimum value of the square root of AVE is greater than the maximum value of the correlation coefficient between the factors (Fornell & Larcker, 1981). In Table 3, the minimum value of the square root of AVE in CSCL is 0.89 greater than the maximum value of the correlation coefficient of 0.80. Similarly, in the FtF setting, 0.90, the minimum value of the square root of AVE, is greater than 0.62, the maximum value of the correlation coefficient. Therefore, discriminant validity was found.

4.1.2 Model fit

As shown in Table 4, by calculating the structural and measurement model fit indices without requiring the model’s parameters to be equal in both CSCL and FtF collaborative learning, we further verified the model fit. From earlier studies (Byrne, 2001; Nan et al., 2022), the model fits were acceptable.

Table 4 Fit indexes
Full size table

To examine RQ1a-c, by exploring whether the effects between different variables differ in CSCL and FtF collaborative learning, we conducted a model comparison with multiple group analyses (Table 5).

Table 5 Model comparison
Full size table

There are significant differences between unconstrained and measurement weights (χ2/df = 5.14, p < .001) and unconstrained and structural weights (χ2/df = 4.51, p < .001), demonstrating that the differences between CSCL and FtF collaborative learning are clear from the factor loadings and path analysis perspective. We also performed pairwise parameter comparisons to determine where the path differences are significant (Table 6).

Table 6 Critical ratios for differences between parameters (Measurement weights)
Full size table

As shown in Table 6, the effect of group member familiarity on student engagement in FtF collaborative learning was significantly stronger than the effect in CSCL (βCSCL = − 0.05, βFtF = 0.25, Z = -2.394, p < .05). The effect of group member familiarity on perceived knowledge construction also yielded significant results (βCSCL = 0.12, βFtF = 0.48, Z = -3.249, p < .05). However, the effects of group member familiarity on teamwork satisfaction (βCSCL = 0.68, βFtF = 0.77, Z = -0.872, p > .05) (RQ1a) and teamwork satisfaction on student engagement (βCSCL = 0.58, βFtF = 0.45, Z = 1.337, p > .05) (RQ1b) did not significantly differ between the two groups. The effect of teamwork satisfaction on perceived knowledge construction in CSCL settings was significantly stronger than that in FtF collaborative learning environments (βCSCL = 0.55, βFtF = 0.28, Z = 3.091, p < .05) (RQ1c).

4.2 Mediational analysis

To test H1 and RQ2, we computed the direct, indirect, and total effects using bootstrap in multiple group analyses. Table 7, Figures 2 and 3 show the importance of the mediating variable of teamwork satisfaction in both CSCL and FtF collaborative learning. Group member familiarity influenced teamwork satisfaction, which in turn increased student engagement and enhanced perceived knowledge construction. Both collaborative learning settings supported H1. Regarding RQ2, the two direct effects were not statistically significant in CSCL; therefore, teamwork satisfaction fully mediated student engagement and perceived knowledge construction. In FtF collaborative learning, teamwork satisfaction partially moderated student engagement and perceived knowledge construction as the direct effects of group member familiarity and engagement were statistically significant.

Table 7 Mediating role of teamwork satisfaction
Full size table
Fig. 2
figure 2

Mediating role of teamwork satisfaction in CSCL

Full size image
Fig. 3
figure 3

Mediating role of teamwork satisfaction in FtF setting

Full size image

Note

Bolded words indicate statistical significance at the 0.05 level. The mediating effect of teamwork satisfaction is significant in CSCL.

Note

Bolded words indicate statistical significance at the 0.05 level. The mediating effect of teamwork satisfaction is significant and plays a partially mediating role in FtF collaborative learning.

5 Discussion

The current study examined and compared the implications of group member familiarity and teamwork satisfaction in student engagement and perceived knowledge construction in online CSCL and FtF collaborative learning. The study also investigated the potential mechanisms that could explain how and why group member familiarity affects students’ collaborative learning experiences. The current study is the first to conduct a systematic comparison of the value of teamwork satisfaction and group member familiarity in online CSCL and FtF collaborative learning settings.

Our research suggests that students’ group member familiarity, teamwork satisfaction, student engagement, and perceived knowledge construction are significantly lower in CSCL settings compared with FtF collaborative learning. Our findings could be attributed to two factors: first, CMC has many inherent drawbacks compared with FtF communication, including lack of non-verbal cues, interaction, and psychological intimacy (Andersen, 1979; Ean, 2010). This may also be the cause for more students preferring FtF collaborative learning in our study (CSCL: 32.11%, FtF: 58.26%, no idea: 9.63%). Alternatively, restrictions such as social distancing because of COVID-19, where most students have never even met classmates, made group members’ familiarity even lower. Meanwhile, low social interaction in CSCL environment also negatively impacts student engagement and academic performance (El-Sayad et al., 2021; Natarajan & Joseph, 2022).

By examining possible mechanisms for the relationship between group member familiarity and learning experiences, our study advances the research on hypotheses about group member familiarity in educational settings. Prior studies primarily investigated the direct relationships between group member familiarity and learning outcomes (Crompton et al., 2022; Janssen et al., 2009; Janssen & Kirschner, 2020; Ku et al., 2013). However, in this study, we obtained different outcomes when we included the mediating variable of teamwork satisfaction. In online CSCL, group member familiarity was indirectly related to learning experience. Instead, teamwork satisfaction mediated this relationship. Thus, group member satisfaction could only positively impact the online CSCL learning outcomes if it enhances teamwork satisfaction. By contrast, group member familiarity was directly and positively correlated with learning experiences in FtF collaborative learning, and teamwork satisfaction partially mediated this relationship. This finding may explain why some earlier studies found a relationship between group member familiarity and learning outcomes (Janssen & Kirschner, 2020; Ku et al., 2013), whereas others did not (Crompton et al., 2022; Janssen et al., 2009). Additionally, this outcome highlights the significance of including satisfaction variables when evaluating group member familiarity and offers guidelines for future research in collaborative learning satisfaction studies.

During the investigation on mediating effects, teamwork satisfaction was found to be positively influenced by group member familiarity and positively influenced the learning experience in both online CSCL and FtF collaborative learning. However, in online CSCL, the mediating effect of teamwork satisfaction played a more significant role because it fully mediated student engagement and perceived knowledge construction. Furthermore, while there were no significant differences between online CSCL and FtF collaborative learning in terms of the impact of group member familiarity on teamwork satisfaction and that of teamwork satisfaction on student engagement, the impact of teamwork satisfaction on perceived knowledge construction in online CSCL showed stronger significance. Our findings further confirm that in CMC, although many non-verbal cues limit people’s communication, intimacy is effective in removing individualized impressions and mediating people’s experiences, which improves reinforcement effects (Jiang et al., 2013).

5.1 Practical implications

This survey’s findings provide some insights for teachers. First, to encourage more effective collaborative learning experiences, teachers should consider various strategies to enhance familiarity between students and their teamwork satisfaction. According to the results of H1, teamwork-satisfaction mediated group member familiarity and students’ learning experiences. Therefore, teachers should understand the significance of teamwork satisfaction in collaborative learning.

Second, according to current research, educators should adopt different teaching strategies to various collaborative learning contexts. According to the results of RQ1, teachers in online CSCL settings should focus on the impact of teamwork satisfaction because the direct impact of group member familiarity on students’ learning experiences is limited compared with FtF collaborative learning.

6 Limitations and future research directions

Although the current study has produced some intriguing results, there are still some restrictions. First, as Seifert & Bar-Tal, (2022) recommended that the teaching for an online course should be planned gradually from personal assignments and familiarization of both students and lecturers with the learning environment, continuing on to work in pairs and getting used to online collaboration and then forming collaborative work in groups. In the current study, however, we did not consider the effect of this dynamic teaching process on the familiarity of group members in CSCL. Future research should consider this dynamic change and investigate the role of group member familiarity accordingly.

Furthermore, because participants’ perceptions are entirely self-reported, our dependent variable may not reflect reality. Although Richmond et al. (1987) argued that university students are seasoned adults who can fairly and accurately assess their own learning, future research should use data such as final course grades to investigate the impact of group member familiarity beyond students’ perceptions.

Lastly, the questionnaire questions are very general and do not allow us to identify the differences between the conditions of activity in FtF and CSCL. It would have been advisable to ask participants to describe the task they were thinking about when completing the questionnaire. This would allow us to know the differences between the activity conditions and how they affect the variables studied.

7 Conclusion

This study examined how group member familiarity and related variables differ in CSCL and FtF collaborative learning and emphasized the mediating role of teamwork satisfaction.

The findings suggested that students had higher group member familiarity in FtF collaborative learning compared to CSCL, leading to more significantly impacted student engagement and perceived knowledge construction. Interestingly, although the effect of group member familiarity on teamwork satisfaction was significant in both CSCL and FtF collaborative learning settings, the difference between the two was not found. Regarding the mediating variables, the mediating role of teamwork satisfaction was more significant in CSCL, in contrast, in FtF collaborative learning, which plaeds only a partially mediating role.

Although the lack of communication of non-verbal cues in CMC may pose a challenge to collaborative learning, once group members get to know each other, as described in the hyperpersonal model (Walther, 1996), CMC helps to build close interpersonal relationships and increases students’ teamwork satisfaction. Teamwork satisfaction in turn has a positive impact on student engagement and the construction of perceived knowledge. In conclusion, the role of familiarity of group members cannot be overlooked in either CSCL or FtF collaborative learning.

Given the significance of collaborative learning and the popularity of online courses, teachers should be aware of the distinctive differences in various learning environments to tailor strategies to improve the efficacy of collaborative learning. In addition to focus on the significance of teacher-student relationship strategies, researchers should further investigate student-student relationships to facilitate effective collaborative learning experiences in CSCL and FtF educational contexts.