When Teacher Training Shifts From Offline to Online: Roles,Efficacy Beliefs,and Cognitive Load of Collaboration

Abstract

As flexible transitions between online and offline teacher training modalities become ubiquitous, educators encounter multifaceted challenges. Their collaborative roles and learning experiences across these distinct CSCL environments remain under-explored. This study aims to map patterns of teacher adaptation and acquire responsive strategies for evolving professional development ecosystems by designing an offline-to-online (O2O) training program, which included two offline and two online sessions featuring collaborative tasks. A total of 30 teachers participated in the program. Teachers’ subjective experiences, including collaborative roles, efficacy beliefs, and cognitive load, were assessed after each training session. Thematic analysis and statistical methods were used to analyze the data. The results identified three collaborative roles—coordinator, integrator, and assistor—with coordinators demonstrating higher self-efficacy than assistors and greater collective efficacy than integrators. Additionally, a notable increase in mental load was observed as trainees transitioned from offline to online sessions, with no significant interaction effect between training modes and collaborative roles. These findings provide valuable insights for the design of collaborative activities and procedural interventions in teacher training, suggesting that role differentiation and careful management of cognitive load are essential for effective training programs.

Plain language summary

Study exploring the shift from offline to online teacher training, focusing on teachers’ collaboration, confidence, and mental effort before and after the shift

Nowadays, it’s common for teacher training programs to switch flexibly between online and offline sessions, but we don’t know much about how this shift affects teachers—especially when they work together on tasks. While many training programs now mix in-person (offline) and online sessions, we still need to understand how teachers’ roles, efficacy beliefs, and mental effort change when switching between these two modes. To explore this, the research team created a training program that combined offline and online sessions, where 30 teachers worked together on group tasks. After each session, the teachers shared details about their roles in the team, how confident they felt in their skills, and how mentally tiring the tasks were. The study found three main roles teachers took on: coordinators (who organize the team), integrators (who combine ideas), and assistors (who support the team). Coordinators felt more confident in their abilities than assistors, and they were more positive about the team’s overall performance than integrators. When moving from offline to online sessions, teachers reported feeling more mentally tired. However, the role each teacher took on didn’t significantly change how much mental effort they felt. This study shows that the design of teacher training programs matters—especially when switching between offline and online formats. Assigning clear roles and keeping mental effort manageable can help make collaborative learning more effective for teachers, whether they’re training in person or online.

Keywords

collaborative learning offline to online (O2O)cognitive load self-efficacy teacher education

Introduction

Online and blended professional development programs have gained growing popularity as flexible, cost-effective alternatives. Within this landscape, a novel teacher professional development model—termed the Offline to Online or Offline to Online (O2O) approach—has emerged as a blended learning innovation to leverage the strengths of online and offline learning (Wei et al., 2019). In the O2O model, the transition from offline to online contexts introduces new challenges, such as adapting to remote and distance learning environments (Alwafi, 2021). As a blended learning experience profoundly intertwined with O2O, computer-supported collaborative learning (CSCL) has gained increasing adoption for fostering collaboration among teachers in training programs (Lockhorst et al., 2010). CSCL also offers a solution to O2O implementation challenges by facilitating joint work, exchange of resources and knowledge co-construction (Jeong & Hmelo-Silver, 2016). During the past years, the role concept has become an important discussion topic in the CSCL field, a promising construct for analyzing and facilitating CSCL, and significant for group interaction, group regulation, and efforts coordination (Cesareni et al., 2016; Saqr & López-Pernas, 2022; Volet et al., 2017). Typically, collaboration roles can be scripted and assigned to participants or can emerge spontaneously during activities (Marcos-García et al., 2015; Strijbos & Weinberger, 2010). So, roles have been classified into two perspectives: emergent and scripted roles. Emergent roles are generated from spontaneous and conscious behavior in the collaborative process without pre-set roles (Strijbos & Weinberger, 2010). Moreover, efficacy beliefs and cognitive load constitute foundational constructs extensively examined within CSCL studies (see, e.g., Wang et al., 2020; Wang & Hwang, 2012). Efficacy beliefs have consistently been found to predict a variety of indicators of improved performance (Beauchamp et al., 2012). Normally, self-efficacy and collective efficacy are two important forms of human efficacy beliefs (Bandura, 2000). For teachers, efficacy beliefs exert greater influence than content knowledge in determining how they organize and approach tasks, serving as robust behavioral predictors (Albion, 1999). Furthermore, cognitive load significantly impacts teachers’ performance. Cognitive load represents the working memory resources required to process learning materials (Sweller & Chandler, 1994). Optimized cognitive load management accelerates task execution efficacy (Tasir & Pin, 2012), amplifies positive affect, and heightens pedagogical decision-making awareness (Blackley et al., 2021). However, research gaps persist regarding teachers’ roles in CSCL, particularly within the O2O framework, and the differences in their efficacy beliefs and cognitive load across offline and online settings. In China, the application of CSCL in O2O teacher training programs is increasingly widespread, where teachers play a critical role in guiding and facilitating students’ CSCL (Li et al., 2022). Their own experiences with CSCL significantly influence their ability to implement it effectively. In recent years, growing attention has been paid to Chinese teachers’ efficacy beliefs, cognitive load, and collaborative roles in CSCL settings, particularly within blended O2O models, as these factors are increasingly recognized as influencing teaching practice (Xie et al., 2023).

Consequently, this study addresses these gaps by examining the transition and differences in teacher training models from offline CSCL to online CSCL, focusing on teachers’ roles, efficacy beliefs, and cognitive load. The research questions to be investigated are as follows:

RQ1: What role types emerge during collaborative activities in teacher training? How do these roles change when teacher training shifts from offline to online?

RQ2: How do trainees’ efficacy beliefs and cognitive loads of collaboration vary by these roles and training models (offline vs. online)?

Literature Review

Teacher Training From Offline to Online

The comparison between offline and online learning environments has attracted significant research attention and generated extensive scholarly investigation. Empirical comparisons of online and offline learning model, especially in collaborative learning, yield nuanced insights: a quasi-experimental study demonstrated that face-to-face collaboration yielded marginally superior knowledge construction, whereas online CSCL groups developed significantly denser social networks (Yu & Yuizono, 2021). Complementing these findings, Zhang et al. (2023) identified four psychosocial advantages of offline settings over online CSCL through mixed-methods analysis: enhanced group familiarity, higher teamwork satisfaction, sustained behavioral engagement, and elevated perceived knowledge gains. Notably, similar studies further validate the offline advantage, demonstrating significantly lower extraneous cognitive load compared to the computer-mediated collaborative learning condition (Jiang et al., 2021). Existing research has clearly established differences between face-to-face and online collaborative learning, with CSCL studies recognizing that offline and online variations primarily stem from technologically mediated communication (Yang et al., 2022). Based on the above analysis, we can infer that transitioning from offline to online CSCL will bring differences for the trainees, particularly in their subjective experiences. This longitudinal shift differs from the horizontal comparison between offline and online training. Despite this, there is a lack of research focused on teachers’ experiences of this transition.

Collaborative Roles in Teacher Education

It is noticeable that the analysis of emerging roles helps to understand the CSCL process, including the dynamics of CSCL, individual contribution, and patterns of interactions with each other (Sarmiento & Shumar, 2010). In the context of teacher training, Lockhorst et al. (2010) investigate eight learning tasks within teacher training programs to identify tasks that promote teachers’ collaboration in CSCL. The results showed that to trigger task-related communication, more structured input should be focused on, such as roles. However, scant research has examined teachers’ collaborative roles, particularly emergent roles. In addition, a role is a dynamic phenomenon; therefore, it can be changed by individuals in different circumstances (Jahnke, 2010). That’s to say, teachers’ collaboration roles in offline training might be different from online. There are some research questions that have been explored seldomly and deserve our effort to study, for instance, “what roles emerge as teachers involved in CSCL?,”“what are the differences of emergent roles when teacher training shifts from offline to online?,” and “what impact do different emergent roles bring to the teachers’ learning experience?.” This study tries to fill the gaps in prior research by answering the above three questions, contributing to the field of collaboration roles in teacher training.

Efficacy Beliefs in CSCL

Teacher efficacy beliefs has been shown to correlate with teacher motivation and perseverance (Clark & Newberry, 2019), group performance (Wang et al., 2014), group cohesion, and persistence in the CSCL environment (Wang & Lin, 2007). Multiple studies have provided compelling evidence that teachers’ efficacy beliefs are strong predictors of teaching behavior and commitment to the teaching profession (Chesnut & Burley, 2015; Dou et al., 2016). Previous studies are promising with regard to the effect of training on teachers’ efficacy beliefs no matter in the online or face-to-face training model. For example, Schina et al. (2021) proved that pre-service teachers’ self-efficacy toward educational rdobotics improved after they completed the teacher training course. In research by Corry and Stella (2018), it was suggested that online teacher education programs are beneficial to developing teachers’ self-efficacy, and self-efficacy is always used as one of the evaluative measures of training programs. That’s part of the reason why this study measured trainees’ efficacy belief. Another reason is that efficacy beliefs are context-specific and must be considered carefully as situations change (Hodges, 2008). Such change might happen with the mode of training shift, for example, from face-to-face to online.

Cognitive Load in CSCL

Cognitive load is divided into assessment and causal factors, and assessment factors are divided into two dimensions: mental load and mental effort (Paas & Van Merrienboer, 1994). The mental load is related to expected cognitive capacity. If individuals have too great of a mental load, then their function of learning, transferring, and constructing knowledge in the long-term memory will be negatively affected (Sweller et al., 2019). Mental effort refers to the cognitive capacity required to complete learning tasks, and a high mental load will often yield high mental effort (Paas & Van Merrienboer, 1994; Van Merrienboer & Sweller, 2005).

Apparently, under collaborative learning, group members can exchange information and knowledge, and then information is distributed among multiple working memories of the different group members, which helps to decrease the cognitive load in individuals who have limited working memory (Kirschner et al., 2018). That’s why research on cognitive load has expanded from individual learning to collaborative learning, especially CSCL. With respect to cognitive load in CSCL, Costley (2021) conducted research on the roles students take within the group, and collaboration roles were divided into high, moderate, or low contributors. The result showed that learners who make fewer contributions to group work have higher levels of the germane cognitive load than those who make greater contributions. Except for learning individually or collaboratively, does cognitive load also relate to learning formats, online or offline? In a study with EEG measuring cognitive load (Hsu, 2021), students were arranged to learn utilizing traditional face-to-face firstly and an online course afterward. The study found no significant difference in cognitive load between online and face-to-face, but the correlation between cognitive load and learning outcomes was significant. A similar result was also found in a mixed-methods study (Mills, 2016), which revealed that there were not any statistically significant differences in cognitive load between students taking online and face-to-face college mathematics courses. Although online or offline learning does not bring differences in cognitive load, does online or offline CSCL also bring no significant difference in cognitive load? As such, this research intends to understand the difference in cognitive loads from O2O CSCL, especially the difference derived from collaborative roles.

Methods

Participants

The participants in this study were 30 teachers (12 male and 18 female) from a city in Western China participated in both the online and offline training, helping us identify the differences brought by this transition, rather than simply comparing the online and offline modes. Among them, 8, 9, and 13 teachers came from vocational school, high school, and middle school, respectively. Moreover, mean age of male teachers was 39.92 years (SD = 6.00, n = 12), and for females was 34.83 years (SD = 6.40, n = 18), the overall mean age of the sample was 36.87 (SD = 6.72, n = 30). The teachers participated in a teacher training program with the theme of educational technologies. The program contained four learning themes: (1) development of instructional media, (2) online and mobile learning, (3) CSCL, and (4) artificial intelligence and learning analysis. The first two themes of training were in the offline model, wherein the trainees gathered together in a classroom. The last two themes were conducted in the form of an online model, and the trainees accessed the live training and took part in a group discussion via the instant messaging tool WeChat. Every two themes of training constitute offline or online training models, respectively. For personal reasons resulting in absences, 28, 28, 30, and 29 trainees participated in the four themes of training activities. In sum, 115 (person-time) trainees took part in all four training activities. To assess the adequacy of sample size, a sensitivity power analysis was conducted using G*Power 3.1 with an independent two-sample t-test. Given the attained sample size of 115 participants (56 online vs. 59 offline). With α = .05 (two-tailed) and power = 0.80, the analysis revealed adequate sensitivity to identify medium effects (d = 0.527 ≥ 0.5) based on Cohen’s (1988a) criteria. Consequently, the study is adequately powered to identify educationally meaningful impacts.

Procedure

All four training activities were organized by the same teaching assistant, and the training content was delivered by the same educational expert. The total training time for the activity was 120 mins. As shown in Figure 1, the training procedures of the offline and online models were the same. The first was the teaching session (40 min), in which the training expert gave a lecture on the corresponding learning theme. For instance, in the training theme “development of instructional media,” the training expert introduced the development history and corresponding teaching organization forms of oral communication, manual transcription, printing technology, radio and television, the Internet, and other instructional media. Second, the teaching assistant introduced the collaboration rules, and the trainees took a break before the collaborative activities. The trainees were randomly assigned an identification number ranging from 1 to 30. Every five trainees formed a group to conduct a 30-min collaborative learning activity with access to the Internet. For example, in the training theme “development of instructional media,” the training expert required groups to choose an instructional medium and discuss its role in teaching reform. After the collaboration, each group was asked to give a 5-min oral presentation to show what they had discussed on the learning theme. The last session was a 10-min summary by the training expert.

Figure 1.

The training procedure for each learning theme.

In addition, in the offline training model, the trainees could search for information on the Internet with their laptops or mobile phones. Regarding the devices the trainees adopted in the online models, 42.37% used mobile phones, followed by laptops (38.98%). The remaining 13.56% and 5.08% used tablet PCs and desktops, respectively.

Before the training and questionnaire survey, teachers were informed about the study and provided oral consent, followed by signed written consent confirming their voluntary participation. The study design minimized potential risks to participants (e.g., time burden, psychological stress) through voluntary participation, brief procedures, and anonymized data collection, thus minimizing harm risks. In addition, the training aimed to enhance teachers’ professional development, with only minimal survey related subjective experience, so benefits far outweighed risks.

Instruments

An electronic questionnaire was utilized to investigate the trainees’ roles, efficacy beliefs, and cognitive load after each collaborative learning activity. The trainees were provided a QR code linked to the questionnaire, and they used mobile phones to scan the code and complete the questionnaire. In addition to the trainees’ identification numbers, we collected demographic information, including gender, age, and the educational stage taught by the trainee. The role-related questions and self-report scales are described as follows.

Role-Related Questions

After the demographic information questions, the questionnaire for the offline training model had one role-related question: What kind of role did you play in the collaborative activity? This question was used to collect the roles the trainees played in the collaborative activity and has been empirically validated as an effective means of identifying the emergent roles of participants within collaborative learning (Wang & Li, 2021).

Self-Report Scales

The self-report scales included in the questionnaire are an efficacy beliefs scale and a cognitive load scale. The scale items are all rated on a 7-point Likert-type scale. The efficacy beliefs scale was adopted from Liu & Wang (2022) originally established by Paas (1992), containing two items to investigate trainees’ self-and collective efficacy as follows: (1) “My performance in this task during group collaboration was …” and (2) “In this task, the performance of my group in the whole training class was ….” The options ranged from (1) “very poor” to (7) “very good.” This scale is the most commonly used subjective appraisal scale. The cognitive load scale consists of mental load and mental effort dimensions with four items, which were based on the measures proposed by Paas and Van Merriënboer (1994) and adapted from the study of Hwang et al. (2013) and Wang et al. (2018). The mental load dimension aims to investigate trainees’ perception of the difficulty of the corresponding training activity; an example item is “the difficulty of this information exchange process for me.” The mental effort dimension aims to obtain trainees’ subjective effort in the collaborative activity; an example item is “the degree of energy I devoted to the learning activity.” The Cronbach’s alpha coefficients for the dimensions of mental load and mental effort were .888 and .840, respectively, indicating that the scale used in the study had high reliability.

Data Analysis

Thematic Analysis

To explore the kinds of roles that emerge during collaborative learning activities, a qualitative thematic approach (Braun & Clarke, 2006) was applied to analyze the role information reported by the trainees in the role-related questions. The thematic analysis had two coding stages. The first was the open-coding stage, wherein the roles the trainees reported in the electronic questionnaire were labeled following similarities in the trainees’ self-reported contents. The coding unit was one self-reported text by the trainees in the four training activities, which means that trainees were repeatedly measured four times (N = 115). For instance, the text “I acted as a catalyst during collaboration” was labeled as a coordinator role, and “I summarised the group discussion” was labeled as an integrator role. The open-coding process was conducted by two researchers separately; then, an expert in teacher education was invited to examine the divergence between the two researchers’ labels and collaborate with the researchers to unify the differences. After the unified process, three role types (coordinator, assistor, and integrator) were generated. In the second coding stage, the two authors used the three role labels to code the text data collected from the online questionnaire.

Additionally, the expert was invited to examine and unify the coding results with the two authors. The inter-rater reliability kappa of the coding analysis was .853 (p < .001, N = 115), which is in the range from moderate to good (Landis & Koch, 1977). In sum, through the two coding stages of the thematic analysis, the validity and replicability of the coding were ensured.

Statistical Analysis

Descriptive statistics were used to depict the general characteristics of the trainees. A one-way analysis of variance (ANOVA) was used to determine whether there were any statistically significant differences between the roles. The effect size f was applied as a measure to quantify the magnitude of an effect in ANOVA. According to Cohen’s estimates, the effect sizes f for small, medium, and large are 0.10, 0.25, and 0.40, respectively (Cohen, 1988b). An independent-sample t-test was used to compare the difference between the offline and online training models. Cohen’s d was calculated to assess the effect size of the independent-sample t-test, with values of 0.2, 0.5, and 0.8 representing small, medium, and large effect size (Cohen, 1988a). Besides, a two-way ANOVA was conducted to test the interaction effect of roles and training models on trainees’ efficacy beliefs and cognitive loads. Partial eta squared (η²) was used as effect size index for two-way ANOVA, with thresholds of 0.01, 0.06, and 0.14 indicating small, medium, and large effects, respectively (Richardson, 2011). The statistical analysis was performed using SPSS 20.0 with p < .05 deﬁned as statistically signiﬁcant.

Results

Role Emergence and Distribution

Role Emergence

Through the qualitative thematic analysis, three role types were found: coordinator (R1), assistor (R2), and integrator (R3). Descriptions of the role functions and corresponding examples reported by the trainees are presented in Table 1. The table shows that the coordinator (R1) was similar to the leader of a group, even though no one was appointed as the leader of these training groups. The assistor (R2) was more like a follower, aiming to ensure the operation of the group function. Moreover, as the collaborative activities asked the groups to generate collective works based on searching and discussions, the integrator (R3) emerged in such conditions.

Table 1.

Role Types and Descriptions.

Type	Description	Examples
Coordinator	Lead group members in conducting tasks; assign tasks to group members; supervise discussion processes.	• “Organiser, give advice and suggestions.” • “I acted as a catalyst during collaboration.” • “The pillar of the group.”
Assistor	Make recommendations and undertake assistant tasks; cooperate with other team members to participate in discussions.	• “I was the atmosphere conditioner in our group.” • “I played a supporting role.” • “Auxiliary personnel.”
Integrator	Incorporate ideas and suggestions generated from group discussions; search for information, and collect possible materials.	• “I summarised the group discussion.” • “I’m in charge of searching for information.” • “Search for information on the Internet and organize selected information.”

Role Distribution

The distributions and percentages of the three roles in the offline and online training models are shown in Table 2. A chi-square test was adopted, and the results showed that there were significant differences in the distribution of the roles between the offline and online training models (χ² = 11.891, p = .003 < .01). The radar map in Figure 2 shows that the offline model had more coordinators (R1) than did the online model, while more integrators (R3) were generated when the teacher training shifted from the offline to the online model. In contrast, the offline and online models have similar percentages of assistors (R2).

Table 2.

Roles Distribution From Offline to Online.

Role	Offline	Online	Total
R1	32 (57.14%)	17 (28.81%)	49
R2	18 (32.14%)	23 (38.98%)	41
R3	6 (10.71%)	19 (32.20%)	25
Total	56 (100%)	59 (100%)	115

Figure 2.

Radar map of roles distribution.

Efficacy Beliefs and Cognitive Load

Effects of Roles

The mean and standard deviation values of different roles’ efficacy beliefs and cognitive loads are shown in Table 3. In terms of efficacy beliefs, a one-way ANOVA showed significant differences for both self- (F[2, 112] = 7.056, p = .001< .01, f = 0.394 > 0.25) and collective efficacy (F[2, 112] = 3.501, p = .033 < .05, f = 0.305 > 0.25) among the three roles, reaching a medium effect size level. For the cognitive loads, a one-way ANOVA indicated a significant difference in mental effort (F[2, 112] = 5.969, p = .003 < .01, f = 0.344 > 0.25), achieving a medium effect size. While no significance was found for the roles’ mental loads (F[2, 112] = 0.516, p = .598 > .05, f = 0.121 > 0.1). Furthermore, the results of the post hoc comparisons are presented in the following paragraphs.

Table 3.

Descriptive Analysis of Different Roles’ Efficacy Beliefs and Cognitive Load.

Variables	Role	Mean	SD	N	F	P	Effect size f
Self-efficacy	R1	5.796	1.020	49	7.056**	.001	0.394
	R2	5.024	1.172	41
	R3	4.960	1.241	25
Collective efficacy	R1	5.939	1.144	49	3.501*	.033	0.305
	R2	5.439	1.097	41
	R3	5.200	1.581	25
Mental load	R1	4.362	1.269	49	0.516	.598	0.121
	R2	4.402	1.272	41
	R3	4.090	1.311	25
Mental effort	R1	5.056	1.057	49	5.969**	.003	0.344
	R2	4.604	1.112	41
	R3	4.170	1.012	25

p < .05. **p < .01.

The self-efficacy, collective efficacy, and mental effort scores were homoscedastic, so Tukey’s honestly significant difference (HSD) method was used in a post hoc comparison to analyze the difference between the roles. Coordinators (R1) had a significantly higher level of self-efficacy than did assistors (R2) (mean difference = 0.772, p = .004 < .01) or integrators (R3) (mean difference = 0.836, p = .009 < .01), while the difference between assistors (R2) and integrators (R3) was not significant (mean difference = 0.064, p = .972 > .05). With respect to collective efficacy, a post hoc comparison revealed a significant difference between coordinators (R1) and integrators (R3) (mean difference = 0.739, p = .043 < .05), while neither the difference between coordinators (R1) and assistors (R2) nor the difference between assistors (R2) and integrators (R3) was significant (mean difference = 0.500, p = .140 > .05 and mean difference = 0.239, p = .727 > .05, respectively).

With regard to the cognitive load, a post hoc comparison shows that coordinators (R1) had a significantly higher level of mental effort than did assistors (R2) (mean difference = 0.886, p = .003 < .01). In contrast, no significant difference was found between coordinators (R1) and integrators (R3) (mean difference = 0.452, p = .117 > .05), nor was there a difference between assistors (R2) and integrators (R3) (mean difference = 0.434, p = .250 > .05).

Overall, these results indicate that the coordinator had the highest level of both self and collective efficacy among the three roles. Meanwhile, as the leading role, the coordinator engaged in the most mental effort during the collaborative learning activities in the teacher training of all three roles.

Effects of Training Models

After comparing the role differences in efficacy beliefs and cognitive load, we moved to the differences between the training models. The mean and standard deviation values of the trainees’ efficacy beliefs and cognitive load for the two training models are shown in Table 4.

Table 4.

Descriptive Analysis of Training Models’ Efficacy Beliefs and Cognitive Load.

Variables	Model	Mean	SD	N	t	p	Cohen’s d
Self-efficacy	Offline	5.518	1.062	56	1.588	.115	0.297
	Online	5.169	1.275	59
Collective efficacy	Offline	5.732	1.198	56	1.095	.276	0.204
	Online	5.475	1.318	59
Mental load	Offline	4.286	1.363	56	−2.507*	.014	0.471
	Online	4.814	0.812	59
Mental effort	Offline	4.888	1.164	56	1.758	.081	0.328
	Online	4.525	1.043	59

p < .05.

An independent-sample t-test was used to compare the difference between the offline and online training models. The results show that the trainees’ mental load in the offline model was significantly lower than they were in the online model (mean difference = −0.528, t[88.715] = −2.507, p = .014 < .05, d = 0.471 > 0.2), while there was no significant difference between the two models concerning mental effort (mean difference = 0.363, t[113] = 1.763, p = .081 > .05, d = 0.328 > 0.2). Cohen’s d values were all above the threshold for a small effect size. Moreover, concerning efficacy beliefs in the two models, no significant difference was found for self-efficacy (mean difference = 0.348, t[113] = 1.588, p = .115 > .05, d = 0.297) or collective efficacy (mean difference = 0.258, t[113] = 1.095, p = .276 > .05, d = 0.204). In summary, these results show that when teacher training shifts from offline to online, the mental load of the trainees observably increases. According to cognitive load theory (Paas & Van Merriënboer, 1994), a high mental load refers to a more complicated information interaction between tasks and trainees.

Interaction Effect of Training Models and Roles

Furthermore, to explore the interaction effect of training models and roles on efficacy beliefs and cognitive load, a two-way ANOVA test was conducted. Box plots of trainees’ efficacy beliefs and cognitive load in different models and roles sub-groups were depicted in Figure 3.

Figure 3.

Box plots of trainees’ efficacy beliefs and cognitive load.

As shown in Table 5, the test results indicate that, for the four dependent variables, there was no significant interaction effect between training models and roles. In other words, the above roles’ differences in self-efficacy, collective efficacy, and mental effort were not affected by training models, and the models’ differences in mental load also didn’t vary by roles. Further discussion is presented in the next section. The partial η² values were all above the threshold for a small effect size.

Table 5.

Two-Way ANOVA of Training Models and Roles.

Dependent Variables	SS	df	MS	F	p	Partial η²
Model × Role
Self-efficacy	5.523	2	2.761	2.214	.114	0.039
Collective efficacy	3.710	2	1.855	1.210	.302	0.022
Mental load	2.201	2	1.100	0.871	.421	0.016
Mental effort	2.590	2	1.295	1.134	.326	0.020

Discussion

Emerging Roles and Role Distributions

To address RQ1 regarding emergent role types, this study empirically identified three distinct collaborative roles: coordinator, assistor, and integrator. The coordinator and assistor roles are similar to the guidance and promoter proposed by Marcos-García et al. (2015) and similar to mediators proposed by Saqr and López-Pernas (2021). However, the role of the coordinator extends beyond both guidance mediator to include coordinating task among team members. For example, many trainees reported engaging in activities related to “planning” (e.g., Nos. 21 and 22) within their groups, while others perceived themselves as playing a “coordinating” role in the groups (e.g., Nos. 2 and 11). The role of an assistor, compared to that of a mediator, has a slight distinction. It emphasizes more on facilitating task completion, involving greater cognitive effort on the individual’s part, rather than focusing solely on social interaction efforts. The coordinator role differs significantly from traditional leadership roles, which are typically characterized by concentrated effort exertion, sustained influence, and extensive social networks, as established in previous research (Saqr & López-Pernas, 2022). Unlike leadership roles, the coordinator role operates within a distributed framework, where effort allocation, task engagement, and dominance dynamics are shared across three roles. The three emergent roles identified in this research also differ from those outlined by Volet et al. (2017), where roles were categorized as content-focused, performance-focused, evaluation-focused, and social roles. In contrast, the roles of coordinator, assistor, and integrator in this study encompassed aspects of content, evaluation, and social interaction, influencing the completion of learning tasks, social performance, and evaluation. For instance, some coordinators (e.g., No. 22) mentioned “making contributions by providing suggestions,” some assistors (e.g., No. 24) described “using imagination and coming up with creative ideas,” and some integrators (e.g., No. 18) viewed their role as that of a “commentator.” Overall, the three emerging roles identified in this study exhibit more balanced and comprehensive characteristics, reflecting a more equal distribution of responsibilities and interactions among participants. Synthetically speaking, both coordinators and assistors are responsible not only for leading and coordinating but also for engaging in content-related discussions. This dual responsibility helps prevent the risk of focusing solely on organizational tasks at the expense of learning tasks. Therefore, all teachers should be encouraged to adopt roles that are both content-oriented and process-oriented in CSCL, enabling them to invest significant effort in learning and fostering meaningful interactions.

To address RQ1 concerning offline-online CSCL role shifts, this study revealed that the distribution of the three roles was different in the online and offline contexts. The number of coordinators in the offline model was significantly greater than that of the online model, while the number of integrators in the offline model was significantly less than that of the online model. Such differences may have been affected by the features of online and offline training modes. In offline CSCL, classroom interactions are widely recognized as a fundamental component that requires substantial coordination efforts to facilitate effective learning processes (Buhl-Wiggers et al., 2023). When participants transition to online CSCL environments, group familiarity tends to increase while communication becomes more task-focused rather than coordination-oriented, largely due to the enhanced accessibility and efficiency of social networking tools. For example, some trainees gave feedback in open-ended questions, and they argued that the online CSCL had “more freedom and participation,”“more convenient and free discussion,” etc. This kind of freedom and convenience led to every teacher being able to speak actively without having to deal with the work of mobilizing members to discuss and assign tasks. As a result, coordinators shifted into the integrator role. For example, trainees No. 1 and 5 were self-rated as coordinators in the first two offline cycles but as integrators after the two online cycles of CSCL. This role change is consistent with the research of Strijbos and De Laat (2010). The roles’ behaviors change with time, and the roles’ conversion is affected by the interaction with the other roles in the group. This finding aligns with the results of Song and Elftman (2024), who similarly identified significant differences in participants’ collaborative experiences between face-to-face and virtual environments.

Efficacy Beliefs and Cognitive Load Vary by Roles and Training Models

To address RQ2 on how trainees’ efficacy beliefs and cognitive loads of collaboration vary by emergent roles, significant role-based variations emerged in self-efficacy. Specifically, coordinators exhibited significantly higher self-efficacy than assistors and integrators. This result is consistent with existing research demonstrating that CSCL participants who were particularly active on the social level of regulation tend to maintain higher self-efficacy beliefs (De Backer et al., 2022). This difference can be attributed, in part, to the varying nature of the roles: teachers in different collaborative learning roles are assigned different tasks, which are based on distinct abilities, and these abilities are determined by varying skill sets (Bawane & Spector, 2009). Additionally, there was a notable difference in collective efficacy, mainly between coordinators and integrators, with coordinators showing higher collective efficacy. Two primary factors contribute to the coordinator’s higher sense of collective efficacy. First, there is a positive correlation between collective efficacy and interaction within a group (Moolenaar et al., 2012). Coordinators engage in activities such as knowledge exchange, experience sharing, problem-solving, organization, and coordination, which likely enhances their confidence in the group’s collective capabilities. Coordinators’ central position in task management and group coordination also fosters stronger perceptions of group capability (Angelle & Teague, 2014). And individuals in coordination roles tend to develop stronger collective efficacy beliefs due to their increased responsibility for group processes and outcomes (Järvelä et al., 2016). Second, individual self-efficacy significantly influences collective efficacy (Wang & Lin, 2007). Therefore, the higher self-efficacy of coordinators contributes to their elevated sense of collective efficacy.

The three roles did not show significant differences in mental load. However, significant differences were observed in mental effort. Coordinators reported significantly higher mental effort compared to assistors. This difference is primarily due to the correlation between mental effort and self-efficacy (Author, 2018). Therefore, the high levels of self-efficacy and collective efficacy among coordinators may lead them to believe that investing more mental effort will result in greater success. Thus, rotating roles at different stages of CSCL is recommended to allow more teachers to assume leadership roles, enhancing their subjective experiences.

To address RQ2 about differences between offline and online training models, comparative analysis indicated that there were no significant differences in self-efficacy and collective efficacy between online and offline settings. This finding aligns with previous research showing no significant differences in collaborative learning effectiveness between online and offline environments (Solimeno et al., 2008). And this finding is consistent with a study by Francescato et al. (2006). They compared the efficacy of collaborative learning in face-to-face and online groups and showed no significant difference in the increase in self-efficacy between the offline and online groups. This is different from a previous study by Artino (2010), who found that online and offline training models bring the difference in self-efficacy. The findings of this study support the view that online learning can sustain trainees’ self-efficacy, which is largely attributed to the inherent advantages of online learning. The absence of significant differences may be attributed to several factors. First, the well-designed online collaboration tools have significantly reduced the psychological distance between virtual and physical learning spaces (Zheng et al., 2015). Second, modern platforms provide features that effectively support both individual and collective efficacy development, mirroring the social dynamics of offline settings (Martin & Bolliger, 2018). Regarding the trainees’ cognitive load, online and offline training exhibited a significant difference in mental load, which was significantly higher online, whereas the mental efforts exhibited no significant difference online and offline, which is consistent with the findings of Lan et al. (2019) and Andersen and Makransky (2021). The mental load of online training is higher than that of face-to-face training, partly because the trainees have to learn how to use the online collaboration platform and they have to experience extraneous cognitive load related to media, network, and devices (Andersen & Makransky, 2021). According to studies on the influencing factors of cognitive load, cognitive load is affected by learners’ cognitive ability, the complexity of learning tasks, and the learning environment (Lin & Kao, 2018). In this study, the teachers’ mental load was mainly due to being unfamiliar with the technology and the massive amount of online information. For instance, some trainees claimed that the online learning environment “is more prone to be absent-minded”; the “operation is not skilled”; or they had “technical problems, the chat is not convenient.” This is different from previous studies (Hsu, 2021; Mills, 2016), in which undergraduate students were participants and didn’t perform differences in cognitive load between online and offline learning. Such inconsistency probably results from differences between students’ and teachers’ competency with the online platform. To alleviate teachers’ higher cognitive load in online learning, external scripts should be provided to reduce the cognitive load associated with online training models. These could include tips on collaboration skills and knowledge, scaffolding for problem-solving, and technical support.

Notably, there was no significant interaction effect found between training models and emergent roles. This suggests that the type of online or offline training model did not significantly influence the influence of roles on efficacy beliefs and mental effort within the group. Additionally, the different roles did not affect the impact of the training model on mental load. The findings indicate that various emerging roles in CSCL, whether in online or offline settings, experience similar information processing systems in which group members collectively process capacity, communicate information, and coordinate actions, ultimately leading to a unified collective work (Kirschner et al., 2014). Self-efficacy primarily arises from an individual’s assessment of their abilities (Bandura, 2000), while roles are similarly based on such self-evaluations. Although online and offline training models differ in their communication mediums, they follow the same processes, which may explain the minimal interaction between efficacy beliefs and the training model. Similarly, cognitive load theory (Sweller, 1988) suggests that extraneous mental load is primarily determined by task complexity, resulting in minimal interaction with roles. To enhance interaction within O2O training model, synchronous features should be strategically leveraged. For instance, employing synchronous video-conferencing tools for collaboration enables real-time knowledge co-construction among participants, thereby significantly enhancing learning effectiveness in O2O training contexts (Peterson & Roseth, 2016). In this study, many teachers also suggested that “video calls can reduce psychological and social distance.” Establishing interdependent relationships among roles is important in both online and offline settings and is crucial for promoting both individual and group success, as several trainees recommended creating more opportunities for “everyone to exchange ideas” and for “all members to express their opinions more effectively.”

Conclusions

In summary, this study examined the transition and differences in teacher training models from offline CSCL to online CSCL, focusing on teachers’ roles, efficacy beliefs, and cognitive load. By investigating these elements, the study aims to uncover patterns of change and provide insights to better adapt to evolving circumstances in teacher professional development. Specifically, to address RQ 1, the study found that teachers in both online and offline CSCL environments emerged in three distinct roles: coordinator, assistor, and integrator. Among these, the coordinator role was more prevalent in offline CSCL settings. To address RQ 2, the study revealed that, among the three roles, coordinators exhibited higher levels of self-efficacy and collective efficacy, as well as greater mental effort. Additionally, online CSCL was associated with higher cognitive load for participants. However, transitioning from offline to online settings did not result in significant differences in efficacy beliefs or cognitive effort. Furthermore, no interaction effect was found between emerging roles and the mode of training (online or offline). In general, this study offers several important perspectives that contribute to the understanding of roles, efficacy beliefs, and cognitive load in CSCL environments. Theoretically, the identification of three emergent roles extends the application of role theory of CSCL contexts. The higher self-efficacy and collective efficacy observed among coordinators further support social cognitive theory, emphasizing the importance of efficacy beliefs in role adoption and task engagement. Practically, the findings underscore the need to support teachers in managing their cognitive load, particularly in online CSCL environments where mental load is higher. Furthermore, this study offers three key innovations. Firstly, and most importantly, it explores the impact of O2O training on teachers’ CSCL learning. Secondly, by examining emerging roles, the study demonstrates how teachers can have greater initiative and autonomy in collaborative learning, analyzing the different roles they assume. Thirdly, the study examines the psychological learning experiences associated with different roles, finding significant differences in self-efficacy, collective efficacy, and mental effort.

Limitations

While highlighting the importance of the current study, it is also essential to acknowledge its limitations, which should be addressed in future research, as outlined below. First, while the existing study by Ouyang et al. (2023) employed discourse analysis to trace the temporal evolution of group discourse moves, our study did not analyze the discourse within teacher collaborations to examine group interaction patterns. Second, while collaborative multimodal interaction data (e.g., eye-tracking, user logs, and clickstream data) can be employed to investigate teachers’ mental load (Hakami et al., 2024), we did not utilize this approach in the present study to preserve the ecological validity of the natural learning context. Third, the sample size of teacher participants may be a limitation, despite conducting four themes of training activities. For future work, we plan to analyze teachers’ conversational, social interaction data, and electrodermal activity during the O2O training model and expand the sample size to verify the generalizability of the research findings. While this study could not determine which specific CSCL factors (online or face-to-face) influence roles and learning experiences, future controlled experimental studies are needed.

Footnotes

Acknowledgements

The authors thank all the participants.

ORCID iDs

Ruonan Hu

Cixiao Wang

Ethical Considerations

This study was reviewed and approved by the University Committee on Human Research Protection at East China Normal University (approval: HR223-2024) on April 3, 2024.

Consent to Participate

Participants gave oral consent and signed to confirm their voluntary participation before starting training and survey.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research is funded by the National Natural Science Foundation of China (NSFC) [Grant No. 62207003] and “the Fundamental Research Funds for the Central Universities.”

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

References

Albion

P. R.

(1999). Self-efficacy beliefs as an indicator of teachers’ preparedness for teaching with technology. In Society for Information Technology & Teacher Education International Conference (pp. 1602–1608). Association for the Advancement of Computing in Education (AACE). https://www.learntechlib.org/primary/p/8156/

Alwafi

(2021). Tracing changes in teachers’ professional learning network on twitter: Comparison of teachers’ social network structure and content of interaction before and during the covid-19 pandemic. Journal of Computer Assisted Learning, 37(6), 1653–1665. https://doi.org/10.1111/jcal.12607

Andersen

M. S.

Makransky

(2021). The validation and further development of the Multidimensional Cognitive Load Scale for Physical and Online Lectures (MCLS-POL). Frontiers in Psychology, 12, Article 642084. https://doi.org/10.3389/fpsyg.2021.642084

Angelle

Teague

G. M.

(2014). Teacher leadership and collective efficacy: Teacher perceptions in three US school districts. Journal of Educational Administration, 52(6), 738–753. https://doi.org/10.1108/JEA-02-2013-0020

Artino

A. R.

Jr. (2010). Online or face-to-face learning? Exploring the personal factors that predict students’ choice of instructional format. The Internet and Higher Education, 13(4), 272–276. https://doi.org/10.1016/j.iheduc.2010.07.005

Bandura

(2000). Exercise of human agency through collective efficacy. Current Directions in Psychological Science, 9(3), 75–78. https://doi.org/10.1111/1467-8721.00064

Bawane

Spector

J. M.

(2009). Prioritization of online instructor roles: Implications for competency-based teacher education programs. Distance Education, 30(3), 383–397. https://doi.org/10.1080/01587910903236536

Beauchamp

M. R.

Jackson

Morton

K. L.

(2012). Efficacy beliefs and human performance: From independent action to interpersonal functioning. In Murphy

S. M.

(Ed.), The Oxford handbook of sport and performance psychology (pp. 273–293). Oxford University Press. https://doi.org/10.1093/oxfordhb/9780199731763.013.0014

Blackley

Redmond

Peel

(2021). Teacher decision-making in the classroom: The influence of cognitive load and teacher affect. Journal of Education for Teaching, 47(4), 548–561. https://doi.org/10.1080/02607476.2021.1902748

10.

Braun

Clarke

(2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77–101. https://doi.org/10.1191/1478088706qp063oa

11.

Buhl-Wiggers

Kjærgaard

Munk

(2023). A scoping review of experimental evidence on face-to-face components of blended learning in higher education. Studies in Higher Education, 48(1), 151–173. https://doi.org/10.1080/03075079.2022.2123911

12.

Cesareni

Cacciamani

Fujita

(2016). Role taking and knowledge building in a blended university course. International Journal of Computer-Supported Collaborative Learning, 11, 9–39. https://doi.org/10.1007/s11412-015-9224-0

13.

Chesnut

S. R.

Burley

(2015). Self-efficacy as a predictor of commitment to the teaching profession: A meta-analysis. Educational Research Review, 15, 1–16. https://doi.org/10.1016/j.edurev.2015.02.001

14.

Clark

Newberry

(2019). Are we building preservice teacher self-efficacy? A large-scale study examining teacher education experiences. Asia-Pacific Journal of Teacher Education, 47(1), 32–47. https://doi.org/10.1080/1359866X.2018.1497772

15.

Cohen

(1988a). Chapter 2—The t test for means. In Cohen

(Ed.), Statistical power analysis for the behavioral sciences (2nd ed., pp. 19–74). Academic Press. https://doi.org/10.1016/B978-0-12-179060-8.50007-4

16.

Cohen

(1988b). Chapter 8—F tests on means in the analysis of variance and covariance. In Cohen

(Ed.), Statistical power analysis for the behavioral sciences (2nd ed., pp. 273–406). Academic Press. https://doi.org/10.1016/B978-0-12-179060-8.50013-X

17.

Corry

Stella

(2018). Teacher self-efficacy in online education: A review of the literature. Research in Learning Technology, 26, Article 2047. https://doi.org/10.25304/rlt.v26.2047

18.

Costley

(2021). How role-taking in a group-work setting affects the relationship between the amount of collaboration and germane cognitive load. International Journal of Educational Technology in Higher Education, 18(1), 1–13. https://doi.org/10.1186/s41239-021-00259-w

19.

De Backer

Van Keer

De Smedt

Merchie

Valcke

. (2022). Identifying regulation profiles during computer-supported collaborative learning and examining their relation with students’ performance, motivation, and self-efficacy for learning. Computers & Education, 179, Article 104421. https://doi.org/10.1016/j.compedu.2021.104421

20.

Dou

Devos

Valcke

(2016). The effects of autonomy gap in personnel policy, principal leadership and teachers’ self-efficacy on their organizational commitment. Asia Pacific Education Review, 17(2), 339–353. https://doi.org/10.1007/s12564-016-9428-7

21.

Francescato

Porcelli

Mebane

Cuddetta

Klobas

Renzi

(2006). Evaluation of the efficacy of collaborative learning in face-to-face and computer-supported university contexts. Computers in Human Behavior, 22(2), 163–176. https://doi.org/10.1016/j.chb.2005.03.001

22.

Hakami

Hernández-Leo

Amarasinghe

Sayis

(2024). Investigating teacher orchestration load in scripted CSCL: A multimodal data analysis perspective. British Journal of Educational Technology, 55(5), 1926–1949. https://doi.org/10.1111/bjet.13500

23.

Hodges

C. B.

(2008). Self-efficacy in the context of online learning environments: A review of the literature and directions for research. Performance Improvement Quarterly, 20(3–4), 7–25. https://doi.org/10.1002/piq.20001

24.

Hsu

(2021). A tale of two classes: Tourism students’ cognitive loads and learning outcomes in face-to-face and online classes. Journal of Hospitality, Leisure, Sport & Tourism Education, 29, Article 100342. https://doi.org/10.1016/j.jhlste.2021.100342

25.

Hwang

G.-J.

Yang

L.-H.

Wang

S.-Y.

(2013). A concept map-embedded educational computer game for improving students’ learning performance in natural science courses. Computers & Education, 69, 121–130. https://doi.org/10.1016/j.compedu.2013.07.008

26.

Jahnke

(2010). Dynamics of social roles in a knowledge management community. Computers in Human Behavior, 26(4), 533–546. https://doi.org/10.1016/j.chb.2009.08.010

27.

Järvelä

Kirschner

P. A.

Hadwin

Järvenoja

Malmberg

Miller

Laru

(2016). Socially shared regulation of learning in CSCL: Understanding and prompting individual- and group-level shared regulatory activities. International Journal of Computer-Supported Collaborative Learning, 11(3), 263–280. https://doi.org/10.1007/s11412-016-9238-2

28.

Jeong

Hmelo-Silver

C. E.

(2016). Seven affordances of computer-supported collaborative learning: How to support collaborative learning? How can technologies help?. Educational Psychologist, 51(2), 247–265. https://doi.org/10.1080/00461520.2016.1158654

29.

Jiang

Kalyuga

Sweller

(2021). Comparing face-to-face and computer-mediated collaboration when teaching EFL writing skills. Educational Psychology, 41(1), 5–24. https://doi.org/10.1080/01443410.2020.1785399

30.

Kirschner

P. A.

Kirschner

Janssen

(2014). The collaboration principle in multimedia learning. In Mayer

R. E.

(Ed.), The Cambridge handbook of multimedia learning (2nd ed., pp. 547–575). Cambridge University Press. https://doi.org/10.1017/CBO9781139547369.027

31.

Kirschner

P. A.

Sweller

Kirschner

Zambrano

R, J

. (2018). From cognitive load theory to collaborative cognitive load theory. International Journal of Computer-Supported Collaborative Learning, 13(2), 213–233. https://doi.org/10.1007/s11412-018-9277-y

32.

Lan

C. H.

Sheng

M. H.

Hsu

Y. C.

Shiue

Y. M.

(2019). Influence of online and face-to-face collaboration and learning style on cognitive load and engagement in a health introductory course. In Proceedings of the Third International Conference on Medical and Health Informatics 2019 (pp. 142–148). Association for Computing Machinery, New York, NY, United States. https://doi.org/10.1145/3340037.3340069

33.

Landis

J. R.

Koch

G. G.

(1977). The measurement of observer agreement for categorical data. Biometrics, 33, 159–174. https://doi.org/10.2307/2529310

34.

Zhang

Bao

Xing

(2022). Examining teachers’ behavior patterns in and perceptions of using teacher dashboards for facilitating guidance in CSCL. Educational Technology Research and Development, 70, 1035–1058. https://doi.org/10.1007/s11423-022-10102-2

35.

Lin

F. R.

Kao

C. M.

(2018). Mental effort detection using EEG data in E-learning contexts. Computers & Education, 122, 63–79. https://doi.org/10.1016/j.compedu.2018.03.020

36.

Liu

Wang

(2022). How roles in collaboration respond to the exchange of device-student ratio under the impact of external scripts? Educational Technology & Society, 25(2), 15–30. https://www-jstor-org-443.web.bisu.edu.cn/stable/48660121

37.

Lockhorst

Admiraal

Pilot

(2010). CSCL in teacher training: What learning tasks lead to collaboration? Technology, Pedagogy and Education, 19(1), 63–78. https://doi.org/10.1080/14759390903579190

38.

Marcos-García

J. A.

Martínez-Monés

Dimitriadis

(2015). DESPRO: A method based on roles to provide collaboration analysis support adapted to the participants in CSCL situations. Computers & Education, 82, 335–353. https://doi.org/10.1016/j.compedu.2014.10.027

39.

Martin

Bolliger

D. U.

(2018). Engagement matters: Student perceptions on the importance of engagement strategies in the online learning environment. Online Learning, 22(1), 205–222. https://doi.org/10.24059/olj.v22i1.1092

40.

Mills

J. J.

(2016) A mixed methods approach to investigating cognitive load and cognitive presence in an online and face-to-face college algebra course [PhD dissertation, University of Kentucky]. https://doi.org/10.13023/ETD.2016.069

41.

Moolenaar

N. M.

Sleegers

P. J.

Daly

A. J.

(2012). Teaming up: Linking collaboration networks, collective efficacy, and student achievement. Teaching and Teacher Education, 28(2), 251–262. https://doi.org/10.1016/j.tate.2011.10.001

42.

Ouyang

Tang

Cheng

Chen

(2023). Using an integrated discourse analysis approach to analyze a group’s collaborative argumentation. Thinking Skills and Creativity, 47, Article 101227. https://doi.org/10.1016/j.tsc.2022.101227

43.

Paas

F. G. W. C

. (1992). Training strategies for attaining transfer of problem-solving skill in statistics: A cognitive-load approach. Journal of Educational Psychology, 84(4), 429–434. https://doi.org/10.1037/0022-0663.84.4.429

44.

Paas

F. G. W. C.

Van Merrienboer

(1994). Instructional control of cognitive load in the training of complex cognitive tasks. Educational Psychology Review, 6(4), 351–371. https://doi.org/10.1007/BF02213420

45.

Peterson

A. T.

Roseth

C. J.

(2016). Effects of four CSCL strategies for enhancing online discussion forums: Social interdependence, summarizing, scripts, and synchronicity. International Journal of Educational Research, 76, 147–161. https://doi.org/10.1016/j.ijer.2015.04.009

46.

Richardson

J. T. E.

(2011). Eta squared and partial eta squared as measures of effect size in educational research. Educational Research Review, 6(2), 135–147. https://doi.org/10.1016/j.edurev.2010.12.001

47.

Saqr

López-Pernas

(2021). Modelling diffusion in computer-supported collaborative learning: A large scale learning analytics study. International Journal of Computer-Supported Collaborative Learning, 16(4), 441–483. https://doi.org/10.1007/s11412-021-09356-4

48.

Saqr

López-Pernas

(2022). How CSCL roles emerge, persist, transition, and evolve over time: A four-year longitudinal study. Computers & Education, 189, Article 104581. https://doi.org/10.1016/j.compedu.2022.104581

49.

Sarmiento

J. W.

Shumar

(2010). Boundaries and roles: Positioning and social location in the Virtual Math Teams (VMT) online community. Computers in Human Behavior, 26(4), 524–532. https://doi.org/10.1016/j.chb.2009.08.009

50.

Schina

Valls-Bautista

Borrull-Riera

Usart

Esteve-González

(2021). An associational study: Preschool teachers’ acceptance and self-efficacy towards Educational Robotics in a pre-service teacher training program. International Journal of Educational Technology in Higher Education, 18(1), 1–20. https://doi.org/10.1186/s41239-021-00264-z

51.

Solimeno

Mebane

M. E.

Tomai

Francescato

(2008). The influence of students and teachers’ characteristics on the efficacy of face-to-face and computer-supported collaborative learning. Computers & Education, 51(1), 109–128. https://doi.org/10.1016/j.compedu.2007.04.003

52.

Song

Elftman

(2024). Beyond collaborative learning: A comparison of small groups in face-to-face and online settings. Medical Science Educator, 34(2), 379–385. https://doi.org/10.1007/s40670-024-01983-4

53.

Strijbos

J. W.

De Laat

M. F.

(2010). Developing the role concept for computer-supported collaborative learning: An explorative synthesis. Computers in Human Behavior, 26(4), 495–505. https://doi.org/10.1016/j.chb.2009.08.014

54.

Strijbos

J. W.

Weinberger

(2010). Emerging and scripted roles in computer-supported collaborative learning. Computers in Human Behavior, 26(4), 491–494. https://doi.org/10.1016/j.chb.2009.08.006

55.

Sweller

(1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257–285. https://doi.org/10.1207/s15516709cog1202_4

56.

Sweller

Chandler

(1994). Why some material is difficult to learn. Cognition and Instruction, 12(3), 185–233. https://doi.org/10.1207/s1532690xci1203_1

57.

Sweller

van Merriënboer

J. J.

Paas

(2019). Cognitive architecture and instructional design: 20 years later. Educational Psychology Review, 31(2), 261–292. https://doi.org/10.1007/s10648-019-09465-5

58.

Tasir

O. C.

(2012). Trainee teachers’ mental effort in learning spreadsheet through self-instructional module based on cognitive load theory. Computers & Education, 59(2), 449–465. https://doi.org/10.1016/j.compedu.2012.01.009

59.

Van Merrienboer

J. J.

Sweller

. (2005). Cognitive load theory and complex learning: Recent developments and future directions. Educational Psychology Review, 17(2), 147–177. https://doi.org/10.1007/s10648-005-3951-0

60.

Volet

Vauras

Salo

A. E.

Khosa

(2017). Individual contributions in student-led collaborative learning: Insights from two analytical approaches to explain the quality of group outcome. Learning and Individual Differences, 53, 79–92. https://doi.org/10.1016/j.lindif.2016.11.006

61.

Wang

Fang

Miao

(2018). Learning performance and cognitive load in mobile learning: Impact of interaction complexity. Journal of Computer Assisted Learning, 34(6), 917–927. https://doi.org/10.1111/jcal.12300

62.

Wang

Fang

(2020). Learning performance and behavioral patterns of online collaborative learning: Impact of cognitive load and affordances of different multimedia. Computers & Education, 143, 103683. https://doi.org/10.1016/j.compedu.2019.103683

63.

Wang

(2021). The trade-off between individuals and groups: Role interactions under different technology affordance conditions. International Journal of Computer-Supported Collaborative Learning, 16(4), 525–557. https://doi.org/10.1007/s11412-021-09355-5

64.

Wang

S. L.

Hsu

H. Y.

Lin

S. S. J.

Hwang

G. J.

(2014). The role of group interaction in collective efficacy and CSCL performance. Educational Technology & Society, 17(4), 242–254. https://www-jstor-org-443.web.bisu.edu.cn/stable/10.2307/jeductechsoci.17.4.242

65.

Wang

S. L.

Hwang

G. J.

(2012). The role of collective efficacy, cognitive quality, and task cohesion in computer-supported collaborative learning (CSCL). Computers & Education, 58(2), 679–687. https://doi.org/10.1016/j.compedu.2011.09.003

66.

Wang

S. L.

Lin

S. S.

(2007). The effects of group composition of self-efficacy and collective efficacy on computer-supported collaborative learning. Computers in Human Behavior, 23(5), 2256–2268. https://doi.org/10.1016/j.chb.2006.03.005

67.

Wei

Luo

Chen

(2019). The reform of computer experiment teaching based on O2O model. Computer Applications in Engineering Education, 27(1), 102–111. https://doi.org/10.1002/cae.22060

68.

Xie

Sui

Liu

(2023). Effects of teacher collaboration on teaching practices in China and England: A structural equation model with TALIS 2018 data. SAGE Open, 13(2), 1–13. https://doi.org/10.1177/21582440231177908

69.

Yang

Luo

Sun

(2022). Investigating the combined effects of group size and group composition in online discussion. Active Learning in Higher Education, 23(2), 115–128. https://doi.org/10.1177/1469787420938524

70.

Yuizono

(2021). Opening the “black box” of cooperative learning in face-to-face versus computer-supported learning in the time of COVID-19. Education Sciences, 11(3), 102. https://doi.org/10.3390/educsci11030102

71.

Zhang

Che

Nan

Kim

J. H.

(2023). I know my teammates: The role of Group Member Familiarity in Computer-Supported and face-to-face collaborative learning. Education and Information Technologies, 28(10), 12615–12631. https://doi.org/10.1007/s10639-023-11704-w

72.

Zheng

Niiya

Warschauer

(2015). Wikis and collaborative learning in higher education. Technology, Pedagogy and Education, 24(3), 357–374. https://doi.org/10.1080/1475939X.2014.948041

When Teacher Training Shifts From Offline to Online: Roles,Efficacy Beliefs,and Cognitive Load of Collaboration

Abstract

Plain language summary

Keywords

Introduction

Literature Review

Teacher Training From Offline to Online

Collaborative Roles in Teacher Education

Efficacy Beliefs in CSCL

Cognitive Load in CSCL

Methods

Participants

Procedure

Instruments

Role-Related Questions

Self-Report Scales

Other Open Questions

Data Analysis

Thematic Analysis

Statistical Analysis

Results

Role Emergence and Distribution

Role Emergence

Role Distribution

Efficacy Beliefs and Cognitive Load

Effects of Roles

Effects of Training Models

Interaction Effect of Training Models and Roles

Discussion

Emerging Roles and Role Distributions

Efficacy Beliefs and Cognitive Load Vary by Roles and Training Models

Conclusions

Limitations

Footnotes

Acknowledgements

ORCID iDs

Ethical Considerations

Consent to Participate

Funding

Declaration of Conflicting Interests

Data Availability Statement

References