The Effects of Three Instructor Participatory Roles on a Small Group’s Collaborative Concept Mapping

Abstract

Collaborative concept mapping, as one of the widely used computer-supported collaborative learning (CSCL) modes, has been used to foster students’ meaning making, problem solving, and knowledge construction. Previous empirical research has used varied instructional scaffoldings and has reported different effects of those scaffoldings on collaboration. To further examine the effects of instructional scaffoldings, this research implements three different instructor participatory roles (i.e., cognitive contributor, group regulator, and social supporter) to support online collaborative concept mapping (CCM). We use multiple learning analytics methods to examine the group’s CCM processes from the social, cognitive, and metacognitive dimensions, supplemented with assessments of the concept maps. The research reveals different effects of three instructor participatory roles on the group’s collaborative behaviors, discourses, and performances. When the instructor engaged as a cognitive contributor, the student group achieved a lowly-interactive, low-level metacognitive engagement and behavior-oriented knowledge construction; when the instructor engaged as a group regulator, the student group achieved a socially-balanced, high-level metacognitive engagement and behavior-communication-interrelated knowledge construction; and when the instructor engaged as a social supporter, the student group achieved a highly-interactive, medium-level metacognitive engagement and communication-oriented knowledge construction. Based on the results, this research proposed pedagogical, analytical, and theoretical implications for future empirical research of CSCL.

Keywords

computer-supported collaborative learning collaborative concept mapping instructor scaffolding learning analytics higher education

Introduction

Grounded upon the social, cultural, situated perspectives of learning (Vygotsky, 1978), computer-supported collaborative learning (CSCL) focuses on the collaborating groups’ meaning-making practices through peer interactions with the pedagogical and technological supports (Hmelo-Silver, 2004; Hmelo-Silver & DeSimone, 2013; Stahl, 2009). Collaborative concept mapping (CCM), as one of the CSCL modes, has been used in K-12 and higher education to engage students work together to build proximal, semantic, and conceptual relationships between ideas in order to improve their epistemic understandings and knowledge constructions (Chiou et al., 2020; Farrokhnia et al., 2019; Wang et al., 2017). Since the CCM has no predetermined or fixed answers, it has potential to transform students to work as active knowledge workers actively interacting with peers and information, rather than passive recipients of knowledge that is transmitted by the instructor (Damşa, 2014; Palincsar, 1998; Prawat, 1992). During the collaborative process, instructors usually need to provide some form of scaffoldings or interventions to facilitate a high quality of collaboration, such as cognitive scaffoldings (Clarke & Bartholomew, 2014), metacognitive regulations (Malmberg et al., 2017) or socio-emotional supports (Park et al., 2015). However, previous empirical results varied about the effects of different instructions on the collaborative quality (Hmelo-Silver & DeSimone, 2013; Kirschner et al., 2006; Stahl, 2009). To further examine the effects of instructional scaffolding, we designed and implemented three different instructor participatory roles (i.e., cognitive contributor, group regulator, and social supporter) to support a group of graduate students’ online CCM activities. Furthermore, we used multiple learning analytics methods to examine the group’s CCM processes from the social, cognitive, and metacognitive dimensions, supplemented with the micro-level, temporal transitional analyses of fragments emerged during the group’s collaboration. Based on the empirical research results, we provide pedagogical, analytical, and theoretical implications to promote CSCL in higher education.

Literature Review

Collaborative Concept Mapping

Grounded upon the sociocultural perspective of learning, computer-supported collaborative learning (CSCL) emphasizes that students work together to complete learning tasks, construct knowledge, or solve problems through sustained interactions and participations (Damon & Phelps, 1989; Dillenbourg, 1999). Compared to traditional instructor-directed learning, CSCL transfers learning into a socio-cognitive process through which learners collaborate together to achieve shared goals with instructional and technological supports (Brown et al., 1989; O’Donnell & Hmelo-Silver, 2013). Collaborative concept mapping (CCM), as a learning mode of CSCL, is an idea-centered knowledge construction activity that enables student groups to discover the proximal and semantic relationships between concepts, and to create, build on, and rise above ideas in a less-scripted and more self-organized fashion (Farrokhnia et al., 2019; Reiter-Palmon et al., 2017; Wang et al., 2017). CCM is grounded upon the philosophy that knowledge is not considered as a pre-defined information transferred from an instructor to students but is negotiated and constructed by students who are attuned to each other’s contributions in socially situated contexts (Bereiter, 2002; Sawyer, 2014). Since the CCM task has no predetermined or fixed answers, students need to actively structure and connect information, represent and construct knowledge in a structured manner, and create the collective, group-level concept map artifacts (Elorriaga et al., 2013; Wang et al., 2017). Therefore, the CCM activity enables students with different knowledge backgrounds and capacity levels to maintain communications and actions, work together to complete the concept map, and share and construct knowledge together (Chiou, 2009; Chiou et al., 2020).

More importantly, CCM usually requires students to coordinate their social, cognitive, metacognitive practices through peer communications and behaviors, which take place synergistically, intertwine inseparably, and influence each other. First, on the social dimension, a group of students must actively participate in the CCM activities and interact with each other frequently, since a CCM task cannot be completed with a low level of peer interactions and participations (Hakkarainen et al., 2013; Stahl, 2009). Second, the CCM activities require students to actively engage in cognitive processes in CCM needs to continuously produce new ideas and artifacts, through information sharing, cognitive elaboration, and knowledge synthesis and argumentation (Elorriaga et al., 2013; Hakkarainen et al., 2013; van Aalst, 2009). Third, to succeed in completing the CCM task, metacognitive strategies are supposed to be used by the student group to regulate the interactive, collaborative process. Students should negotiate what to achieve as a group, plan, and implement problem-solving strategies, and monitor and reflect on the working progress (Jarvela & Hadwin, 2013; Winne et al., 2013). More importantly, during a collaborative learning activity supported with technological tools or platforms, students usually need to interact and discuss with peers through oral or text communications and meanwhile take actions to externalize their knowledge in some forms of group knowledge artifacts (e.g., concept map) (Stahl, 2014). Therefore, CCM is a complex, multidimensional process that usually needs the instructor’s deliberate scaffoldings or interventions to assure a high quality of group collaboration.

Effects of Instructor Scaffoldings on CSCL

CSCL is an active process of constructing knowledge in socially situated contexts (Bereiter, 2002; Lave & Wenger, 1991; Vygotsky, 1978). Although the minimally guided instruction is intuitively appealing for the instructional design of CSCL, it is usually difficult for student groups to achieve a high quality of collaboration autonomously without the instructor scaffolding (Hermkes et al., 2018; Järvelä et al., 2016; Kaendler et al., 2015). In contrast to a traditional class where the instructor takes a substantive leader role to design and guide the instruction and learning, the instructor needs to transfer their roles to maintain a balance between instructor authority and student meaning making and facilitate the high quality of collaboration (Nel, 2017; Ouyang & Scharber, 2017; Robinson et al., 2019). Previous research has indicated that varied scaffoldings have been provided by the instructors to facilitate CSCL, such as cognitive scaffoldings (Tabak & Baumgartner, 2004; van de Pol et al., 2019), metacognitive scaffoldings (Ben-David & Zohar, 2009; Ouyang et al., 2021), and socio-emotional scaffoldings (Ouyang et al., 2020; Park et al., 2015).

Specifically, different instructor scaffoldings usually have different effects on the CSCL processes and performances. First, some studies have found that appropriate cognitive scaffoldings can promote students’ cognitive inquiry and meaning making. For example, Tabak and Baumgartner (2004) found the positive effects of cognitive scaffoldings provided by instructor in a high school biology course, which could balance instructor’s knowledge authority and student group’ knowledge construction. van de Pol et al. (2019) found that the cognitive supports provided by the instructor could improve the accurate answers in the student group work. Second, some studies have found that metacognitive scaffoldings can improve students’ thinking skills and guide their inquiry discourses. For example, Ben-David and Zohar (2009) found that metacognitive hints stimulated the student groups’ discussions through sharing ideas to group members and asking follow-up questions. Ouyang et al. (2021) compared the effects of three metacognitive scaffoldings (i.e., minimal-guided scaffolding, the task-oriented scaffolding, and the idea-oriented scaffolding) on students’ collaborative performances, processes, and perceptions. Results found that when the instructor provided metacognitive scaffoldings related to the ideas and perspectives, students’ cognitive contribution, metacognitive regulation, and knowledge artifact behaviors strengthened. Third, some studies have found that socio-emotional scaffoldings can create good communication atmosphere to facilitate learning. For example, Park et al. (2015) found that the socio-emotional encouragement provided by instructor in online forum effectively promoted students’ conversational moves. Ouyang et al. (2020) found that social support from the instructor, such as recognition, encouragement, or sharing of daily stories, successfully created a positive learning environment and facilitated active peer discussions. Therefore, previous empirical research has used varied instructional scaffoldings (e.g., cognitive scaffoldings, metacognitive regulations, or socio-emotional supports) and reported different effects of those scaffoldings on the student groups’ collaborations.

Multimodal Learning Analytics of CSCL

From the analytical perspective, the mixed methods have been promoted to collect and analyze process and performance data from multidimensional perspectives to understand the collaborative learning processes (Janssen et al., 2013; Medina & Stahl, 2021; Stahl, 2009). Existing empirical research has usually used traditional self-report data (e.g., student questionnaires and interviews) or summative data (e.g., the final knowledge products) to understand collaborative perceptions or performances. For example, Wang et al. (2017) collected questionnaire surveys and student interview responses to investigate the effects of web-based collaborative concept mapping on group interactions in collaborative task. Chiou et al. (2020) used questionnaire, pretest, posttest, and interviews to investigate the effect of structured computer-assisted collaborative concept mapping on student learning performance and motivation. However, analytics of those perception and product data may overshadow the student groups’ collaborative processes that can better reflect the groups’ collaborative learning characteristics and effects (Stahl, 2009).

With the development of learning analytics and educational data mining, relevant work has collected multimodal data during the collaborative process and has used multiple analytical methods to reveal the complex, multilevel characteristics of collaborative process. On the one hand, quantitative analysis methods (e.g., clickstream analysis, sequential analysis, clustering analysis, and social network analysis) are used to explain and understand the characteristics and patterns of students’ collaboration. The advantage of the quantitative method is that it can be applied to large-scale data to investigate the structure, pattern, or sequence of the learning process through standardized framework system (Cohen et al., 2013). On the other hand, given that quantitative methods might overshadow some subtle details, the qualitative, ethnographic approaches (e.g., conversation or discourse analysis) are used to examine the micro-level turn-taking relevancies between interactional, behavioral, and cognitive activities during a short time period of collaborative learning (Governor et al., 2021; Stahl, 2009; Zemel et al., 2009). Complementing with each other, qualitative and quantitative methods together can provide a more holistic, multilevel, and multidimensional analysis of group collaboration and cognition (Borge & Rose, 2021; Janssen et al., 2013; Suthers et al., 2013). Echoing this research trend, this research uses multiple learning analytics methods to analyze the group’s collaborative concept mapping process and final performances based on quantifiable measurements, complemented with the qualitative, fine-grained microanalysis of fragments.

Methodology

Research Purposes and Questions

This research used a multi-method approach to investigate the effects of three instructor’s participatory roles on the online CCM activities in a graduate-level course at a top research-intensive university in China. Specifically, with different participation modes, the instructor (the first author) took three different roles (cognitive contributor, group regulator, and social supporter) in three online CCM activities. This research was not a quasi-experiment research but occurred in an authentic, natural instructional context. Our research question is: Whether, to what extent, and how did three instructor roles influence a small group’s CCM processes and performances?

The Instructional Contexts

The instructional context was a series of online, synchronous CCM activities designed in a small-size graduate-level seminar course, titled Distance and Online Education, offered by the Educational Technology (ET) program at a top research-intensive university in China (8 weeks in total). Four graduate students enrolled in this course (see Table 1). The instructor (the first author) designed five open-ended, ill-structured problems related to the course content for the student groups to work on through the CCM activities. Due to technical issues, this research consisted of three dataset that fully recorded students’ collaborative process (1.5 hours/activity). These three CCM activities took place sequentially in order, which focused on three topics: Online Tool Design and Development (Activity 1), Relationships between Students and Data (Activity 2), and Robots in Education (Activity 3). Taking the topic of Online Tool Design and Development as an example, we asked the student group to work as a teaching team to summarize the current design and development of online tools, platforms, and systems, which they would further present to their students. The difficulty levels of three activities were similar, and all three topics were new to four students. In this way, we controlled the difficulty levels as well as topic familiarity to reduce the effects of influential factors in an authentic, natural learning context.

Table 1.

Student information.

Participant	Gender	Age	Status	Major
S1	Female	26	Full-time Ph.D. student	Educational technology
S2	Female	23	Full-time master student	Educational technology
S3	Female	23	Full-time master student	Educational technology
S4	Female	23	Full-time master student	STEM education

The online platform Huiyizhuo (https://www.huiyizhuo.com/), a visualized tool, was used to support the collaborative work. Huiyizhuo provides functions such as text chatting, audio and video communication, concept map, note and comment, resource sharing, etc. In the CCM process, group members first communicated through the audio and text chatting to determine how to proceed the problem; then, groups shared resources, continued communications, and constructed concept map to demonstrate their problem-solving processes; and finally wrote the groups’ solution proposals (see Figure 1). The concept map served as the main medium for the participants to interpret the problems, discuss and negotiate understandings, present knowledge from multiple perspectives, identify misunderstandings, and reach the group consensus of the solution (Novak & Cañas, 2008; Novak et al., 1984).

Figure 1.

A screenshot of a CPS activity on the Huiyizhuo platform. Note. Students’ contributions were shown in different colors in the concept maps.

The Design of the Instructional Participatory Roles

The instructor deliberately took three different participatory roles, namely, cognitive contributor, group regulator, and social supporter, in three CCM activities, respectively (see Table 2). The instructor scaffolded the CCM activity every 2–3 minutes by the specific role through oral participations without online operations of the concept maps. When the instructor played a cognitive contributor role, she provided conceptual or cognitive contributions to the discussion topics with a high frequency of CC (freq. = 48) (Activity 1). When the instructor played a group regulator role, she monitored, managed, and organized students’ collaborative process with a high level of GR (freq. = 38) (Activity 2). When the instructor played a social supporter role, she encouraged students’ engagement with a high level of SS (freq. = 42) (Activity 3). In three CCM activities, the instructor occasionally used other two types of scaffoldings (frequency < 10 times).

Table 2.

The coding scheme of instructor scaffolding (adapted from Dobber et al., 2017).

Instructor scaffolding	Code	Description	Example
Cognitive contribution	CC	The instructor provided information or content on the topic or shared her own conceptual or cognitive understandings of the topic	We need to define the concept first, for example, what is group cognition... I think social constructivism theory is appropriate to ground this concept…
Group regulation	GR	The instructor monitored and stimulated student’s task understanding, goal setting, as well as thinking and reflection process	Okay, anyone has any ideas about it?... What is the final product we want to present to students?
Social support	SS	The instructor provided social supports, encouragement and appreciations to students	I think this is a good beginning to start with exploring the idea of…

Data Collection and Analysis

The process data includes computer screen videos (including audios) in Huiyizhuo (about 1.5 hours/activity), as well as the group’s collaborative product (i.e., concept maps). An overall analytical framework of process data was proposed, which used multiple analytical approaches to analyze peer communication and online behavior data (see Figure 2). Furthermore, secondary analyses were conducted to examine the social, cognitive, and metacognitive dimensions, as well as the detailed temporal transition within each activity.

Figure 2.

The analytical framework.

First, the computer screen recording data (with audio) was transcribed by two raters to record participants’ discourse communications and online behaviors in a time-based fashion. A coding scheme was proposed to examine the group’s process data from the social, cognitive, and metacognitive dimensions (see Table 3). It should be mentioned that the social, cognitive, and metacognitive dimensions were all analyzed based on data from students’ peer communications and online behaviors. The social dimension recorded students’ social interactions through peer communications and online behaviors of the concept map (the interaction with the instructor was excluded). On the cognitive dimension, referring to a previously validated framework (Ouyang & Chang, 2019), we analyzed students’ knowledge contributions at the superficial, medium, and deep levels (see Table 3). The cognitive dimension represented students’ sharing, exploration, and elaboration of knowledge content related to the topics through either peer verbal discourses or concept mapping behaviors. For example, a student shared some information related to the topic through verbal communication; she could also revise the information as an element on the concept map. The unit of analysis was the unit of sentence (i.e., a full sentence spoken by a student) and the unit of concept map branch (i.e., the text on a branch of concept map typed by a student). Referring to a previous research (Malmberg et al., 2017), we analyzed the metacognitive dimension that represented students’ regulation of their collaborative processes. Three metacognitive codes were task understanding, goal setting and planning, and monitoring and reflection (see Table 3). Again, the metacognitive dimension was also demonstrated through either the peer communications or the concept mapping behaviors. For example, a student may express her understandings of the task with peers through discourses; she could also monitor the process by observing the knowledge built through the concept map. The unit of analysis was the unit of sentence (i.e., a full sentence spoken by a student) and the unit of concept map (i.e., a behavior on the concept map operated by a student).

Table 3.

The coding scheme of process data.

Dimension	Code		Descriptions
Social	Social interaction through communications	Soc-C	Student interactions through verbal communication (i.e., one student responded to others through audio) and texting (i.e., one student replied to others through text)
Social	Social interaction through behaviors	Soc-B	Student interactions through construction of the concept map artifact (i.e., one student built on or modified others’ work on the concept map)
Cognitive	Superficial-level Knowledge (KS)	KS-C	A student simply shared existing information about the topic without further explanations or elaborations
	Superficial-level Knowledge (KS)	KS-B	A student created the first-level concepts or ideas on the concept maps
	Medium-level Knowledge (KM)	KM-C	A student explained the details of the topic content without further elaborations
	Medium-level Knowledge (KM)	KM-B	A student added concept, arguments and explanations as the second level on the concept map
	Deep-level Knowledge (KD)	KD-C	A student elaborated their ideas, arguments, or opinions
	Deep-level Knowledge (KD)	KD-B	A student added examples to the concept map to further support the third level (or above)
Metacognitive	Task understanding (TU)	TU-C	A student thought about the purpose of the task, and read and explain the problems or questions
	Task understanding (TU)	TU-B	A student moved the mouse over the task area to read and understand the problems to be solved
	Goal setting and planning (GSP)	GSP-C	A student discussed the purpose of the task, divided the task into specific steps, and planed what to do next
	Goal setting and planning (GSP)	GSP-B	A student added a structure and framework to the concept map on the platform to help plan the next step
	Monitoring and reflection (MR)	MR-C	A student monitored the progress of tasks, evaluated the timeline for completing the task, and summarized what had been done and what needed to be done
	Monitoring and reflection (MR)	MR-B	A student moved the mouse to the concept map to observe its content

Note. C represented the analysis of students’ peer communication data; B represented the analysis of students’ online behavior data.

Three raters completed the analysis process. Rater 1 first coded 30% of the dataset based on the proposed coding scheme. Next, rater 2 (the first author) coded the data again and had multiple meetings with rater 1 to solve discrepancies. Krippendorff’s (2004) alpha reliability was 0.855 among two raters at this phase. Finally, rater 1 finished the rest of the dataset, and rater 3 doublechecked the analysis, and consulted with rater 1 to decide the final codes if there were conflicts.

Next, the goal of the secondary analysis was to further reveal details about the group’s collaborative characteristics. Analytical methods included social network analysis (SNA), epistemic network analysis (ENA), and temporal analysis (TA). SNA was used to analyze the multimodal interaction network that represented student interactions through verbal communication (i.e., one student responded to others through audio), texting (i.e., one student replied to others through text), and knowledge artifact (i.e., one student built on or modified others’ work on the concept map). The original data was transformed into the directed, weighted student–student network dataset, where the direction represented who responded to whom and built on whose work (i.e., bi-directional); tie weight represented the frequency of responses, replies and build-on work a student made to others (i.e., interaction frequency). We used the network-level SNA metrics to calculate the student group’s social characteristics (see Table 4). R\ packages sna (Butts, 2016), tnet (Opsahl & Opsahl, 2015), and igraph (Csardi, 2013) were used.

Table 4.

The descriptions of SNA metrics (see Ouyang, 2021; Ouyang & Scharber, 2017).

SNA metrics	Description
Average degree (outdegree, indegree)	Degree denotes the number of links pointing to or away from a node, including outdegree and indegree; average degree indicates the mean of all group members’ outdegree and indegree
Average closeness	The mean of all student members’ closeness; closeness is the length of paths from a student to all others in the network, defined as the inverse total length
Average betweenness	The mean of all student members’ betweenness; betweenness is the number of shortest paths that passes through a student
Density	The ratio of actual links between any nodes to all potential possible links, and it represents the whole network structure’s cohesion
APL	The average number of the shortest paths for all possible pairs of nodes
Reciprocity	Ratio of symmetric dyads to non-null dyads in a network
Transitivity	Ratio of transitive triads to all triads in a network
GCC	GCC reflects the degree of connectedness between this node and its neighborhood, ranged between 0 and 1
ICV	The inverse of the standard deviation of student–student interaction frequency divided by its mean
Centralization	The extent to which degree centrality is concentrated within a small number of participants in a network, ranged between 0 and 1
Outdegree/indegree centralization	The extent to which outdegree/indegree is concentrated within a small number of nodes

Epistemic Network Analysis (ENA) (Shaffer et al., 2016) was performed to demonstrate the accumulative connections of the social, cognitive, and metacognitive dimensions. An ENA Webkit (epistemicnetwork.org) was used here to perform ENA analysis and visualization (Marquart et al., 2018). However, the structure was not merely focused on the cognitive dimension but on all social, cognitive, and metacognitive dimensions. Finally, we conducted the qualitative, micro-level, temporal analysis of a fragment from each activity to explain the collaborative characteristics.

Finally, to evaluate the product data of concept map, we adapted a previously validated assessment approach (Novak & Cañas, 2008) in terms of three dimensions, that is, propositions, hierarchy, and examples (see Table 5). Two raters assessed the groups’ concept maps individually, and consulted with each other to reach the final scores.

Table 5.

Assessment standards of concept map (adapted from Novak & Cañas, 2008).

Dimension	Description	Example	Assessment rule
Propositions	Did the concept map reflect basic themes and concepts? Was the relationship between topics correct and appropriate?	A proposition that reflects students’ understanding of the topic is characterized as a concept, idea, or argumentation (e.g., “data-oriented objects,” “data types”)	A meaningful concept, idea or argumentation was scored 1 point
Hierarchies	Did the concept map show a certain hierarchy? Was each subordinate concept more specific than the previous concept?	The hierarchy (i.e., the concept map level) represents the depth of students’ thoughts (e.g., depth of five levels)	An effective hierarchical structure was scored 5 points
Examples	Did examples in concept maps reflect their corresponding themes, concepts, or labels? Were examples used effectively and properly?	An example that reflects the completeness of students’ knowledge construction of ideas is characterized as the detailed examples (e.g., “clickstream data”/“eye movement data” of data types)	A valid example was scored 1 point

Results

The Overall Results

According to the assessment of the group performance reflected by the concept map, Activity 1 ranked first among three activities (S_cm =162; propositions = 46, hierarchies = 30, examples = 86), followed by Activity 2 (S_cm = 137; propositions = 38, examples = 74, hierarchies = 25) and Activity 3 (S_cm = 104; propositions = 32, examples = 57, hierarchies = 15). In Activity 1, students contributed to the social dimension with a frequency of 210 (Mean = 52.50, SD = 28.62), cognitive dimension with a frequency of 246 (Mean = 61.50, SD = 39.00), and metacognitive dimension with a frequency of 88 (Mean = 22.00, SD = 15.20). In Activity 2, students contributed to the social dimension with a frequency of 296 (Mean = 74.00, SD = 28.95), cognitive dimension with a frequency of 307 (Mean = 76.75, SD = 28.56), and metacognitive dimension with a frequency of 128 (Mean = 32.00, SD = 16.63). In Activity 3, students contributed to the social dimension with a frequency of 306 (Mean = 76.50, SD = 22.94), cognitive dimension with a frequency of 329 (Mean = 82.25, SD = 43.18), and metacognitive dimension with a frequency of 97 (Mean = 24.25, SD = 5.56) (see Figure 3). Based on the following results, Activity 1 was characterized as lowly-interactive, low-level metacognitive engagement, and behavior-oriented knowledge construction (the instructor was a cognitive contributor); Activity 2 was characterized as socially-balanced, high-level metacognitive engagement, and behavior-communication-interrelated knowledge construction (the instructor was a group regulator); and Activity 3 was characterized as highly-interactive, medium-level metacognitive engagement, and communication-oriented knowledge construction (the instructor was a social supporter).

Figure 3.

The bar plots of three CCM activities. Note. The colors of the codes are set according to the dimensions (codes’ colors are consistent below).

Activity 1: Lowly-Interactive, Low-Level Metacognitive Engagement, and Behavior-Oriented Knowledge Construction

Activity 1 was characterized as lowly-interactive, low-level metacognitive engagement, and behavior-oriented knowledge construction. First, on the social dimension, the social interaction frequency ranked last among three activities (freq. = 210). The group-level SNA results indicated Activity 1’s social characteristics were lowly-interactive (reflected by the lowest score of average degree, closeness, and density), poorly-cohesive (reflected by the lowest score of transitivity, GCC as well as the highest score of APL), unbalanced (reflected by the lowest score of ICV), and centralized (reflected by the highest score of centralization) (see Table 6).

Table 6.

The group-level SNA measures.

SNA metrics	Activity 1	Activity 2	Activity 3
Average degree (outdegree, indegree)	7.36 (7.41, 7.30)	13.21 (13.30, 13.12)	13.35 (13.42, 13.27)
Average closeness	0.256	0.321	0.290
Average betweenness	0.750	0	0.250
Density	0.833	1	1
APL	2.02	1.28	1.61
Reciprocity	1	1	1
Transitivity	0.75	1	1
GCC	0.73	1	1
ICV	0.97	1.69	1.26
Centralization	0.333	0	0
Outdegree/Indegree centralization	0.220, 0.220	0, 0	0, 0
SNA network diagram

Second, Activity 1’s overall cognitive contribution also ranked last among three activities (freq. = 246). Compared to other two activities, Activity 1 had the most frequent cognitive code of KD-B (freq. = 93), which indicated that students contributed to the deep-level knowledge construction through building the concept map. Third, on the metacognitive dimension, Activity 1 also ranked last among three activities (freq. = 88), among which the most frequent metacognitive code was MR-B (freq. = 48).

ENA results showed the overall connection structure of all codes (see Figure 4). In Activity 1, behavior-related codes shared relatively strong connections (e.g., connection of soc-B & KD-B = 0.94, connection of soc-B & MR-B = 0.84, connection of soc-B & KM-B = 0.44), while communication-related codes shared weak connections (e.g., connection of soc-C & KM-C = 0.34, connection of soc-C & KS-C = 0.28, connection of soc-C & KD-C = 0.19). Consistent with the results abovementioned, the ENA results emphasized Activity 1’s behavior-oriented knowledge construction attribute.

Figure 4.

The epistemic networks of three CMM activities. Note. The colors of the codes are not automatically generated through ENA Webkit, but set manually according to the code color.

Finally, we conducted the micro-level, temporal transition analysis of a fragment from Activity 1 (see Table 7 & Figure 5). The instructor engaged in the Activity 1 as a cognitive contributor, that is, she mainly guided, supported, and corrected the student group’s cognitive engagement (line 1, 6, 8, 9, and 15). However, the instructor and students did not perform equally in the cognitive contributions: the instructor’s professional knowledge was richer than the students', therefore she performed in an authoritative way; as a result, students tended to answer the instructor’s questions most of the time (e.g., line 2, 7, and 12). Although students did not achieve the deep level of knowledge construction through communications, they spent much time on building and modifying the concept map through online behaviors (e.g., line 3, 4, 5, and 11). Taken together, we concluded Activity 1 as lowly-interactive, low-level metacognitive engagement, and behavior-oriented knowledge construction.

Table 7.

Fragment A from Activity 1

Line	Participant	Content	Code
1	Ins	According to the situation, so what is the learning situation?	CC
2	S1	For example, … it is like mobile learning, problem-oriented learning, collaborative learning, project-based learning ...	KD-C
3	S3	Modifying the concept map	KM-B, Soc-B
4	S1	Observation	MR-B
5	S2	Moving the concept map	GSP-B, Soc-B
6	Ins	The following one is based on object-oriented, as we often mention…… So what do you mean by the system tools?	CC
7	S3	It’s the developer kits	KM-C
8	Ins	So it’s under the developer kits, right? Just like java programming, Python, and R… so what common technologies are available?	CC
9	Ins	And what’s the difference between the collection, choice of database tools, and tool development per se?	CC
10	S2	Observation	MR-B
11	S4	Modifying the concept map	KD-B, Soc-B
12	S3	This should be the same	KS-C
13	S4	…Choosing a database and then building the online platform	KM-C, Soc-C
14	S3	The platform building should consider the technologies used	KM-C, Soc-C
15	Ins	The database includes the data storage. For example, learning analytics may automatically analyze data, right? In this case, a database needs to store those online data	CC

Figure 5.

The temporal transition of fragment A, Activity 1. Note. The temporal transition of codes among participants through the dotted lines (i.e., Ins represents the instructor, S1, S2, S3, S4 represent four students).

Activity 2: Socially-Balanced, High-Level Metacognitive Engagement, and Behavior-Communication-Interrelated Knowledge Construction

Activity 2 was characterized as socially-balanced, high-level metacognitive engagement, and behavior-communication-interrelated knowledge construction. First, on the social dimension, the social interaction frequency ranked second among three activities (freq. = 296). The group-level SNA results indicated that the Activity 2’s social characteristics were interactive (reflected by the medium score of average degree), highly-cohesive (reflected by the highest score of closeness density, transitivity, and GCC as well as the lowest score of APL), highly-balanced (reflected by the highest score of ICV), and evenly-distributed (reflected by the lowest score of centralization) (see Table 6). Although the overall interactive characteristics ranked medium, Activity 2 was highly-cohesive, balanced, and evenly-distributed.

Second, on the cognitive dimension, Activity 2’s overall cognitive contribution ranked second among three activities (freq. = 307); all cognitive codes ranked at the medium or low levels among three activities. Third, on the metacognitive dimension, Activity 2 ranked first among three activities (freq. = 128). Compared to other two activities, Activity 2 had the most frequent metacognitive codes of MR, including MR-B (freq. = 64) and MR-C (freq. = 14). This indicated that students were metacognitive-engaged in the CMM activity to monitor and reflect on the collaborative process.

Activity 2’s ENA connection structure indicated a strong connection between behavior-related and communication-related codes (e.g., connection of Soc-C & KD-B = 0.67, connection of Soc-B & KM-C = 0.55) (see Figure 4). Therefore, Activity 2 had the characteristics of behavior-communication-interrelated knowledge construction.

Finally, according to the micro-level, temporal transition analysis of a fragment from Activity 2 (see Table 8 & Figure 6), the instructor engaged as a group regulator, regulating the students’ collaboration through group regulation strategies (line 1, 3, 5, and 14). Under the instructor’s regulation, student S3 put forward an article she read before (line 7) and triggered a discussion among students on how students produced data initiatively (see line 8–13). This triggered students’ further cognitive contributions. Meanwhile, students also used the metacognitive strategies actively, such as task understanding (line 2 and 11), goal setting and planning (line 4 and 13), and monitoring (line 12). Taken together, we concluded Activity 2 as socially-balanced, high-level metacognitive engagement, and behavior-communication-interrelated knowledge construction.

Table 8.

Fragment B from Activity 2

Line	Participant	Content	Code
1	Ins	We should read the question and see what it asks and what we are discussing now… Have you answered this question?	GR
2	S1	Oh… it is about relationship...	TU-C
3	Ins	What you guys discussed were based on the perspective of the traditional teaching contexts, right? … Now, you can go to the left branch of the concept map and work on it	GR
4	S4	Modifying the concept map	GSP-B, Soc-B
5	Ins	Then you can work on the right branch, and create another dimension. Or you can create it in a creative way	GR
6	S1	Did they become the producer of the data ...	KS-C
7	S3	I have read an article about when students face difficulties, they take the initiative to ask questions and think of specific solutions. This seems like the students producing data actively	KM-C, Soc-C
8	S1	What does this mean ...?	KS-C, Soc-C
9	S3	Ask questions initiatively	KS-C, Soc-C
10	S4	How to define this initiative? Will students consider they take the initiative to collect the data? Or ...?	KM-C, Soc-C
11	S1	The relationship between students and data ...	TU-C, Soc-C
12	S2	Observation	MR-B
13	S3	Can this be categorized as an exploration process?	GSP-C, Soc-C
14	Ins	You guys can read the article and discuss it again	GR

Figure 6.

The temporal transition of fragment B, Activity 2.

Activity 3: Highly-Interactive, Medium-Level Metacognitive Engagement, and Communication-Oriented Knowledge Construction

Activity 3 was characterized as highly-interactive, medium-level metacognitive engagement, and communication-oriented knowledge construction. First, on the social dimension, the social interaction frequency ranked first among three activities (freq. = 306). The group-level SNA results indicated the social interaction attribute of Activity 3 was highly-interactive (reflected by the highest scores of average degree and density). Although the overall interactive characteristic was active in Activity 3, its other SNA metrics were either equal to (e.g., density, reciprocity, transitivity, GCC, and centralization) or lower than those of Activity 2 (e.g., ICV) (see Table 6). Therefore, the student group in Activity 3 was not as balanced as Activity 2.

Second, on the cognitive dimension, Activity 3’s overall cognitive contribution ranked first among three activities (freq. = 329). Compared to other two activities, Activity 3 had the most frequent cognitive codes of KS-C (freq. = 88), KM-C (freq. = 103), and KD-C (freq. = 36), which indicated that students contributed to all levels of knowledge construction through peer communications. Moreover, students contributed most frequently to KS-B (freq. = 20) and KM-B (freq. = 46), and least to KD-B (freq. = 33). Third, on the metacognitive dimension, Activity 3 ranked second among three activities (freq. = 97). Compared to other two activities, students had the highest metacognitive code of GSP in Activity 3, including GSP-B (freq. = 30) and GSP-C (freq. = 15). This indicated that students tended to set goals and make planning during the collaborative process. But Activity 3 had a relatively low frequency of MR, indicating students did not monitor and reflect on the process frequently.

ENA results showed that in Activity 3 (see Figure 4), communication-related codes shared strong connections (e.g., connection of Soc-C & KM-C = 0.91, connection of Soc-C & KS-C = 0.77, connection of KS-C & KM-C = 0.63), while behavior-related codes shared weak connections (e.g., connection of Soc-B & KM-B = 0.39, connection of Soc-B & MR-B = 0.28, connection of Soc-B & GSP-B = 0.28). Consistent with the results abovementioned, the ENA results indicated that Activity 3’s characteristics of communication-oriented knowledge construction.

Finally, according to the micro-level, temporal transition analysis of a fragment from Activity 3 (see Table 9 & Figure 7), when the instructor engaged in Activity 3 as a social supporter, she usually encouraged students in the socio-emotional dimension (SS; line 2 and 8) and seldom involved in the cognitive engagement directly. As a result, it created a good and relaxed atmosphere, where students were willing to express and discuss. A highly-interactive discussion was formed among S1, S3, and S4 in fragment C (line 3–7 and line 10–13). However, they seldom contributed to the concept map while communicating with peers, which resulted in the lowest score of the final concept map. Taken together, we concluded Activity 3 as highly-interactive, medium-level metacognitive engagement, and communication-oriented knowledge construction.

Table 9.

Fragment C from Activity 3

Line	Participant	Content	Code
1	S4	I have a question, is the robot tangible, or is it like designed in the web page?	KM-C
2	Ins	Uh-huh. This is a very good question	SS
3	S4	AI, taking the speech recognition as an example… it gives you feedback when you need…	KM-C
4	S1	Because the AI has an identification function in it	KM-C, Soc-C
5	S4	But it focuses on robots…What is a robot?	KM-C, Soc-C
6	S3	I feel that most of them are physical	KM-C, Soc-C
7	S4	Yes, if it is a robot, I think it should have an entity, not just the feedback from the machine, such as the feedback from my mobile phone like Siri?	KD-C, Soc-C
8	Ins	I think this is a very good question. I didn’t think about it before. Your question is: is the robot tangible, or is it necessary for the robot to be tangible? Or embedded in the web page or the system?	SS
9	S1	I think it may be different from those intelligent agents based on artificial intelligence… if so, a mechanical arm can also be regarded as a robot. Yes, it needs to simulate some characteristics of living entities, so it is called a robot	KD-C
10	S1	The technology may involve what we just said, and it may be just an online web version of artificial intelligence, which can give intelligent recognition and feedback, but if it does not have a physical entity, it may not be a robot	KD-C, Soc-C
11	S3	It may be a virtual robot without any physical entities	KS-C, Soc-C
12	S1	What I mean is that the definition of robot should involve a physical machine	KM-C, Soc-C
13	S3	What we have questioned about just now is the definition of a robot… it may exist in a virtual environment or in a real environment, is that Okay?	KM-C, Soc-C

Figure 7.

The temporal transition of fragment C, Activity 3.

Discussions

Addressing the Research Questions

Since previous studies indicated different effects of the instructor participatory roles on collaborative learning (Hmelo-Silver & DeSimone, 2013; Kirschner et al., 2006; Stahl, 2009), this research examined a small group’s collaborative processes and performances when the instructors scaffolded with three different roles (cognitive contributor, group regulator, and social supporter) in online CCM activities. Specifically, we used multiple learning analytics methods to analyze the multidimensional characteristics of the group’s collaborative processes, complemented with the micro-level, temporal transitional analyses of fragments emerged during the collaboration. The research results revealed discrepancies of three CCM activities performed by the same group of students.

When the instructor engaged as a cognitive contributor in Activity 1, the student group achieved the highest score of concept map, and the collaborative process was identified as lowly-interactive, low-level metacognitive engagement, and behavior-oriented knowledge construction. In Activity 1, students’ social, cognitive, and metacognitive contributions all ranked last among three activities. But ENA results indicated that students contributed to the deep-level knowledge construction through the concept map behaviors, which resulted in a best score of the final concept map. The micro-level, temporal analysis further revealed that students tended to answer the instructor’s questions and write down on the concept map, rather than initiating peer knowledge constructions. One of the main reasons was that when the instructor scaffolded as a cognitive contributor, she was more authoritative than the students on the knowledge which resulted in unequal knowledge contributions (Tabak & Baumgartner, 2004).

When the instructor engaged as a group regulator in Activity 2, the student group achieved the medium score of concept map, and the collaborative process was identified as socially-balanced, high-level metacognitive engagement, and behavior-communication-interrelated knowledge construction. In Activity 2, although students’ social contributions did not rank first, they were highly-cohesive, balanced, and evenly-distributed. ENA results emphasized that the students constructed knowledge through an interrelated online behaviors and discussion communications. But this interrelation did not result in the best final performance of the concept map product. Moreover, students were most metacognitive-engaged in Activity 2, particularly using the metacognitive strategies frequently to monitor and reflect on the collaborative process. The micro-level, temporal analysis further revealed that, when the instructor engaged as a group regulator, students tended to imitate her use of regulation strategies, which might result in their frequent metacognitive engagement (Park et al., 2015).

When the instructor engaged as a social supporter in Activity 3, the student group gained the lowest score of concept map, and the collaborative process was identified as highly-interactive, medium-level metacognitive engagement, and communication-oriented knowledge construction. In Activity 3, students’ social and cognitive contributions ranked first. ENA results emphasized that students contributed to the knowledge construction through the oral communications; but compared to the other two activities, they took less actions on the concept map building, which also resulted in the lowest score of the concept map. Moreover, students’ metacognitive contributions ranked second among three activities; but they set goals and made plans frequently during the collaborative process. The micro-level, temporal analysis of a fragment further revealed that when the instructor engaged as a social supporter, students became highly-interactive, and tended to express and discuss their perspectives.

Consistent with previous research results (Hermkes et al., 2018; Ouyang et al., 2020; van de Pol et al., 2019), this research revealed discrepancies of collaborative processes and performances among three CCM activities performed by the same group, which indicated a complication of the instructor’s scaffoldings on the group’s collaborative effects. Based on the results, this research proposes pedagogical, analytical, and theoretical implications of design, instruction, and research of online collaborative learning.

Pedagogical Implications

To address challenges for improving groups’ CSCL quality, instructors should utilize varied scaffoldings in the classroom practices to support student groups’ collaborative work, particularly paying attention to the dynamic changes of the scaffolding strengths based on the groups’ collaborative procedures. First, the instructors can provide varied scaffoldings to guide students’ CSCL process, like what the instructor did in this research to offering cognitive, metacognitive, and socio-emotional scaffoldings. For example, the research results showed that cognitive scaffoldings improved the quality of the group’s knowledge product in Activity 1; metacognitive scaffoldings promoted students' metacognitive engagement in Activity 2; and socio-emotional scaffoldings made students highly-interactive in Activity 3. Second, the strengths of the instructor scaffoldings are not fixed and constant throughout the CSCL process; in other words, the instructor scaffoldings are supposed to be changeable in terms of the student groups’ dynamic changes during CSCL (Park et al., 2015; van de Pol et al., 2010). To be specific, consistent with previous research (Kaendler et al., 2015; Ouyang & Scharber, 2017; VanLeeuwen et al., 2015), when instructors provide scaffoldings to support students’ collaboration, at the beginning, by setting up a role model, the instructor’s active involvement can stimulate students’ peer interactions and knowledge constructions; and as the collaboration progresses to the middle or later phases, the instructor can change to metacognitive facilitator to monitor and regulate the group’s work. Third, the instructor should not presume that more diverse scaffoldings are better for promoting collaborative quality; on the contrary, excessive scaffoldings from the instructor may not be beneficial sometimes (Hmelo-Silver & DeSimone, 2013; Kirschner et al., 2006). As our results show, the instructor’s high strength of cognitive support as an authoritative (in Activity 1), could result in a low quality of the student group’s collaborative process. The instructor’s high frequency of social supports (in Activity 3) to some extent stimulate the student group’s frequency communications, but result in the lowest concept map score. The instructor’s group regulations (in Activity 2) caused a balanced, behavior-communication-interrelated knowledge construction. When the group faces cognitive barriers on some topics, the instructor can strength the cognitive support scaffolding to stimulate students’ thinking, but at the same time, she should be aware that the over-scaffolding of cognitive contributions may result in a weakened cognitive engagement from students (Clarke & Bartholomew, 2014; van de Pol et al., 2019); when the group students form active peer interactions and sustained cognitive engagement in collaboration, the instructor’s use and strength of scaffoldings can gradually fade with time (Ouyang et al., 2020; Ouyang & Scharber, 2017; van de Pol et al., 2010). Taken together, from a pedagogical perspective, the instructor should be aware of the student groups’ collaborative status, intervene the student group’s work at the appropriate timing, and maintain a proper strength of scaffoldings in terms of students’ collaborative dynamics in order to avoid undesirable effects.

Analytical Implications

From the analytical perspective, future work can extend the multimodal learning analytics (MLA) to collect and analyze process data from multidimensional perspectives in order to deeply understand the collaborative learning processes (Janssen et al., 2013; Medina & Stahl, 2021; Stahl, 2009). The current research integrates the quantitative, statistical approaches (e.g., content analysis, social network analysis, epistemic network analysis) with the qualitative, ethnographic approaches (e.g., micro-level temporal analysis) to examine the effects of different instructor participatory roles on a student group’s collaborative concept mapping process. These process-oriented learning analytics reveal the student group’s social interactions, cognitive, and metacognitive contributions supplemented with temporal sequences, which may be overshadowed by summative analysis (Stahl, 2009). But from a MLA perspective, we are not able to capture the group’s offline behaviors and facial expressions synchronously with their communicative discourses and online behaviors, such that we cannot analyze students’ discourses, online and offline behaviors to construct knowledge and artifacts in an integrated way (Stahl, 2009). With the development of learning analytics and educational data mining techniques, recent work has collected multimodal data (e.g., discourse, body movement, eye tracking, and facial expression) and used quantifiable algorithm-enabled models to analyze collaborative patterns. For example, in terms of the concept of distributed cognition, Amon et al. (2019) used multidimensional recurrence quantification analysis to examine fine-grained collective patterns of regularity across team members’ speech rate, body movement, and team interaction during collaborative problem solving. Gorman et al. (2020) used mixed quantitative models and techniques (including discrete recurrence, non-linear prediction algorithm, and average mutual information) to detect real-time changes in team cognition in the form of communication reorganization patterns during collaborative training. Future work should move towards this direction: the collection and analytics of multimodal process-oriented data can provide a more holistic, multilevel, and multidimensional analysis of CSCL to reveal the characteristics of collaborative interactions at the group level (Borge & Rose, 2021; Janssen et al., 2013; Suthers et al., 2013). In addition, traditional qualitative methods (e.g., observation, interview, survey, and reflection) can also be used to understand students’ perceptions about individual and group knowledge before, during, and after the collaborative learning. Overall, this research argues that a multimodal learning analytics method, particularly a combination of the quantitative equation and qualitative microanalysis, serves as an appropriate method for making a better understanding of the CSCL activities.

Theoretical Implications

From the theoretical perspective, agency is demonstrated by the individuals’ intentionality for and their action of taking initiations to achieve self-development (Bandura, 2001; Eteläpelto et al., 2013), which is essential for achieving the high quality of CSCL. On the instructor’s level, in the CSCL scaffolding, the instructors should transform their authoritative, directive roles to facilitators, learners or collaborators (Ouyang et al., 2020; Park et al., 2015; Tabak & Baumgartner, 2004), like what this research revealed of three facilitator roles (i.e., cognitive contributor, group regulator, and social supporter) the instructor can take. As our results show, the instructor’s excessive social or cognitive guidance might decrease students’ collaborative contribution and cause undesirable effects. Therefore, the instructors can relinquish some parts of the control of the instructional design, direction, and evaluation, and encourage students to take agency for learning (Bandura, 2001; Ouyang et al., 2020). With the instructor’s appropriate scaffolding, students tend to imitate and apply instructor strategies and scaffoldings, like what we found in this research. Although previous studies indicated a positive correlation between students’ application of instructor scaffoldings and their learning effects (Jadallah et al., 2011; van de Pol et al., 2019), our research showed that the students’ imitation and application of scaffoldings would not necessarily improve the final collaborative quality. Therefore, students should not rely too much on instructors, but exert their own agency, actively engage in the collaborative activity to improve group cognition and problem-solving quality (Clarke et al., 2016; Scardamalia & Bereiter, 2014). Furthermore, when students bring personal culture, prerequisite knowledge, and skills and capacities into the group work, the student groups have the ability to self-organize and self-regulate collaborative activities, even without instructor’s support (Clarke et al., 2016). In summary, instructors and students in CSCL should take agency to change their roles, achieve instructor-student collaboration, and foster the student-centered collaborative learning.

Conclusions

In CSCL, varied instructional scaffoldings (e.g., cognitive scaffoldings, metacognitive regulations, or socio-emotional supports) have been used, which have different effects on students’ collaboration. This research implements three different instructor participatory roles (i.e., cognitive contributor, group regulator, and social supporter) to support a group of graduate students’ online collaborative concept mapping (CCM) process. Multiple learning analytics reveals the different effects of three instructor participatory roles on a student group’s collaborative behaviors, discourses, and performances. When the instructor engaged as a cognitive contributor, the student group was characterized as lowly-interactive, low-level metacognitive engagement, and behavior-oriented knowledge construction, with the highest score of concept map; when the instructor engaged as a group regulator, the student group was characterized as socially-balanced, high-level metacognitive engagement, and behavior-communication-interrelated knowledge construction, with the medium score of concept map; when the instructor engaged as a social supporter, the student group was characterized as highly-interactive, medium-level metacognitive engagement, and communication-oriented knowledge construction, with the lowest score of concept map. Based on the results, this research proposed pedagogical, analytical, and theoretical implications for future empirical research of CSCL. A major limitation of this research is the small sample size of four students, with a limited range of demographic backgrounds and experiences; future empirical research needs to expand the research sample size of more small groups to further test, validate, the research results. Another limitation is that there may exist confounding effects of the instructor role with students’ acclimation to the environment, to their peers and the instructor, as well as to the nature of the task. Specifically, both interactive and cognitive contributions were low in Activity 1 and became to a high level in Activity 3, suggesting a possible progression of the group of students, independent of the instructor role. Therefore, to establish effect with statistical rigor, further research should conduct multiple experiments in which the order of the instructor roles is randomized (i.e., the cognitive role is not always assigned to the first activity to validate the research results. In addition, a more thorough multimodal learning analytics research is needed to collect and analyze discourse, behavior, and movement data in order to analyze the synergistic coordination of multimodal interactions, referential behaviors, and knowledge contribution discourses (Ochoa, 2017). Overall, considering the complexity of the CSCL process, this research takes a process-oriented perspective to explore the effects of the instructor’s scaffoldings on the group’s CSCL processes and proposes pedagogical, analytical, and theoretical implications for future empirical research of CSCL.

Footnotes

Acknowledgments

The authors acknowledge assistances from Jiawen Zhou for preliminary data analysis.

Declaration of Conflicting Interests

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

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research is financially supported by the National Natural Science Foundation of China (62177041; 61907038) and 2021 University-level Educational Reformation Research Project for Undergraduate Education, Zhejiang University (zdjg21033).

ORCID iD

Fan Ouyang

Author Biographies

Fan Ouyang is an associate researcher at College of Education, Zhejiang University. Her research interests are computer-supported collaborative learning, learning analytics, and artificial intelligence in education.

Weiqi Xu is a master student under Fan Ouyang’s supervision at College of Education, Zhejiang University. His research interests are knowledge building, learning analytics and educational data mining, online and blended learning.

References

Amon

M. J.

Vrzakova

D’Mello

S. K.

(2019). Beyond dyadic coordination: Multimodal behavioral irregularity in triads predicts facets of collaborative problem solving. Cognitive Science, 43(10), e12787. https://doi.org/10.1111/cogs.12787

Bandura

(2001). Social cognitive theory: An agentic perspective. Annual Review of Psychology, 52(1), 1–26. https://doi.org/10.1146/annurev.psych.52.1.1

Ben-David

Zohar

(2009). Contribution of meta-strategic knowledge to scientific inquiry learning. International Journal of Science Education, 31(12), 1657e1682. https://doi.org/10.1080/09500690802162762

Bereiter

(2002). Education and mind in the knowledge age. Routledge.

Borge

Rose

C. P.

(2021). Quantitative approaches to language in CSCL. In Cress

Rosé

Wise

A. F.

(Eds.), The international handbook of computer-supported collaborative learning. Springer.

Brown

J. S.

Collins

Duguid

(1989). Situated cognition and the culture of learning. Educational Researcher, 18(1), 32–42. https://doi.org/10.3102/0013189X018001032

Butts

C. T.

(2016). sna: Tools for social network analysis. (version 2.4) [Computer software]. https://cran.r-project.org/web/packages/sna/index.html

Chiou

C. C.

(2009). Effects of concept mapping strategy on learning performance in business and economics statistics. Teaching in Higher Education, 14(1), 55–69, https://doi.org/10.1080/13562510802602582

Chiou

C. C.

Tien

L. C.

Tang

Y. C.

(2020). Applying structured computer-assisted collaborative concept mapping to flipped classroom for hospitality accounting. Journal of Hospitality, Leisure, Sport and Tourism Education, 26, 100243. https://doi.org/10.1016/j.jhlste.2020.100243

10.

Clarke

L. W.

Bartholomew

(2014). Digging beneath the surface: Analyzing the complexity of instructors’ participation in asynchronous discussion. Online Learning Formerly The Journal of Asynchronous Learning Networks, 18(3), 1–2. https://dx-doi-org.web.bisu.edu.cn/10.24059/olj.v18i3.414

11.

Clarke

S. N.

Howley

Resnick

Penstein Rosé

(2016). Student agency to participate in dialogic science discussions. Learning, Culture and Social Interaction, 10, 27–39. https://doi.org/10.1016/j.lcsi.2016.01.002

12.

Cohen

Manion

Morrison

(2013). Research methods in education. Routledge.

13.

Csardi

M. G.

(2013). R package “igraph” (version 1.2.6) [Computer software]. https://igraph.org

14.

Damon

Phelps

(1989). Critical distinctions among three approaches to peer education. International Journal of Educational Research, 13(1), 9–19. https://doi.org/10.1016/0883-0355(89)90013-X

15.

Damşa

C. I.

(2014). The multi-layered nature of small-group learning: Productive interactions in object-oriented collaboration. International Journal of Computer-Supported Collaborative Learning, 9(3), 247–281. https://doi.org/10.1007/s11412-014-9193-8

16.

Dillenbourg

(1999). What do you mean by collaborative learning? In Dillenbourg

(Ed.) Collaborative-learning: Cognitive and computational approaches (pp. 1–19). : Elsevier.

17.

Dobber

Zwart

Tanis

van Oers

(2017). Literature review: The role of the teacher in inquiry-based education. Educational Research Review, 22, 192–214. https://doi.org/10.1016/ j.edurev.2017.09.002

18.

Elorriaga

J.A.

Arruarte

Calvo

Larrañaga

Rueda

Herrán

(2013) Collaborative concept mapping activities in a classroom scenario, Behaviour & Information Technology, 32(12), 1292–1304, https://doi.org/10.1080/0144929X.2011.632649

19.

Eteläpelto

Vähäsantanen

Hökkä

Paloniemi

(2013). What is agency? Conceptualizing professional agency at work. Educational Research Review, 10, 45–65. https://doi.org/10.1016/j.edurev.2013.05.001

20.

Farrokhnia

Pijeira-Díaz

H. J.

Noroozi

Hatami

(2019). Computer-supported collaborative concept mapping: The effects of different instructional designs on conceptual understanding and knowledge co-construction. Computers and Education, 142, 103640. https://doi.org/10.1016/j.compedu.2019.103640

21.

Gorman

J. C.

Grimm

D. A.

Stevens

R. H.

Galloway

Willemsen-Dunlap

A. M.

Halpin

D. J.

(2020). Measuring real-time team cognition during team training. Human Factors, 62(5), 825–860. https://doi.org/10.1177/0018720819852791

22.

Governor

Lombardi

Duffield

(2021). Negotiations in scientific argumentation: An interpersonal analysis. Journal of Research in Science Teaching, 58(9), 1389–1424. https://doi.org/10.1002/tea.21713

23.

Hakkarainen

Paavola

Kangas

Seitamaa-Hakkarainen

(2013). Sociocultural perspectives on collaborative learning toward collaborative knowledge creation. In Hmelo-Silver

C. E.

Chinn

C. A.

Chan

C. K. K.

O’Donnell

A. M.

(Eds.), The international handbook of collaborative learning (pp. 69–85). Routledge.

24.

Hermkes

Mach

Minnameier

(2018). Interaction-based coding of scaffolding processes. Learning and Instruction, 54, 147–155. https://doi.org/10.1016/j.learninstruc.2017.09.003

25.

Hmelo-Silver

C. E.

(2004). Problem-based learning: What and how do students learn? Educational Psychology Review, 16(3), 235–266. https://doi.org/10.1023/B:EDPR.0000034022.16470.f3

26.

Hmelo-Silver

DeSimone

(2013). Problem-based learning: an instructional model of collaborative learning. In Hmelo-Silver

Chinn

C. A.

Chan

O’Donnell

(Eds.), The international handbook of collaborative learning (pp. 370–385). Routledge.

27.

Jadallah

Anderson

R. C.

Nguyen-Jahiel

Miller

B. W.

Kim

I. H.

Kuo

L. J.

Dong

(2011). Influence of a teacher’s scaffolding moves during child-led small-group discussions. American Educational Research Journal, 48(1), 194–230. https://doi.org/10.3102/0002831210371498

28.

Janssen

Cress

Erkens

Kirschner

P. A.

(2013). Chapter 6 Multilevel analysis for the analysis of collaborative learning. In Hmelo-Silver

C. E.

Chinn

C. A.

Chan

C. K. K.

O’Donnell

A. M.

(Eds.), The international handbook of collaborative learning (pp. 124–137). Routledge.

29.

Järvelä

Hadwin

A.F.

(2013). New frontiers: Regulating learning in CSCL. Educational Psychologist, 48(1), 25–39. https://doi.org/10.1080/00461520.2012.748006

30.

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, 263–280. https://doi.org/10.1007/s11412-016-9238-2

31.

Kaendler

Wiedmann

Rummel

Spada

(2015). Teacher competencies for the implementation of collaborative learning in the classroom: A framework and research. Educational Psychology Review, 27(3), 1–32. https://doi.org/10.1007/s10648-014-9288-9

32.

Kirschner

P. A.

Sweller

Clark

R. E.

(2006). Why minimal guidance during instruction does not work: An analysis of the failure of constructivist, discovery, problem-based experiential and inquiry-based teaching. Educational Psychologist, 41(2), 75–86. https://doi.org/10.1207/s15326985ep4102_1

33.

Krippendorff

(2004). Reliability in content analysis: Some common misconceptions and recommendations. Human Communication Research, 30(3), 411–433. https://doi.org/10.1093/hcr/30.3.411

34.

Lave

Wenger

(1991). Situated learning: Legitimate peripheral participation. Cambridge University Press.

35.

Malmberg

Järvelä

Järvenoja

(2017). Capturing temporal and sequential patterns of self-, co-, and socially shared regulation in the context of collaborative learning. Contemporary Educational Psychology, 49, 160–174. https://doi.org/10.1016/j.cedpsych.2017.01.009

36.

Marquart

C. L.

Hinojosa

Swiecki

Eagan

Shaffer

D. W.

(2018). Epistemic network analysis [Software] Version 1.6.0. Website: epistemicnetwork.org

37.

Medina

Stahl

(2021). Analysis of group practices. In Cress

Rosé

Wise

Oshima

(Eds.), International handbook of computer-supported collaborative learning. Springer.

38.

Nel

(2017). Students as collaborators in creating meaningful learning experiences in technology-enhanced classrooms: An engaged scholarship approach. British Journal of Educational Technology, 48(5), 1131–1142. https://doi.org/10.1111/bjet.12549

39.

Novak

J. D.

Cañas

A. J.

(2008). The theory underlying concept maps and how to construct and use them, Technical Report IHMC CmapTools, Florida Institute for Human and Machine Cognition, 2008. http://cmap.ihmc.us/Publications/ResearchPapers/TheoryUnderlyingConceptMaps.pdf

40.

Novak

J. D.

Gowin

D. B.

Kahle

J. B.

(1984). Learning how to learn. Cambridge University Press.

41.

Ochoa

(2017). Chapter 11: Multimodal learning analytics. In Lang

Siemens

Wise

Gašević

(Eds.), Handbook of learning analytics (1st ed., pp.143–150). Creative Commons License 4.0.

42.

O’Donnell

A. M.

Hmelo-Silver

C. E.

(2013). Introduction: What is collaborative learning? An overview. In Hmelo-Silver

C. E.

Chinn

C. A.

Chan

C. K. K.

O’Donnell

A. M.

(Eds.), The international handbook of collaborative learning (pp. 93–111). Routledge.

43.

Opsahl

M. T.

(2015). R package “tnet”. (version 3.0.16) [Computer software]. http://toreopsahl.com/tnet/

44.

Ouyang

(2021). Using three social network analysis approaches to understand computer-supported collaborative learning. Journal of Educational Computing Research, 59(7), 1401–1424. https://doi.org/10.1177/0735633121996477

45.

Ouyang

Chang

Y. H.

(2019). The relationship between social participatory role and cognitive engagement level in online discussions. British Journal of Educational Technology, 50(3), 1396–1414. https://doi.org/10.1111/bjet.12647.

46.

Ouyang

Chang

Y. H.

Scharber

Jiao

Huang

(2020). Examining the instructor-student collaborative partnership in an online learning community course. Instructional Science, 48(2), 183–204. https://doi.org/10.1007/s11251-020-09507-4

47.

Ouyang

Chen

Cheng

Tang

C-Y.

(2021). Exploring the effect of three scaffoldings on the collaborative problem-solving processes in China’s higher education. International Journal of Educational Technology in Higher Education, 18(1), 1–22. https://doi.org/10.1186/s41239-021-00273-y

48.

Ouyang

Scharber

(2017). The influences of an experienced instructor’s discussion design and facilitation on an online learning community development: A social network analysis study. The Internet and Higher Education, 35, 33–47. https://doi.org/10.1016/j.iheduc.2017.07.002

49.

Palincsar

A. S.

(1998). Social constructivist perspectives on teaching and learning. Annual Review of Psychology, 49(1), 345–375. https://doi.org/10.1146/annurev.psych.49.1.345

50.

Park

J. B. H.

Schallert

D. L.

Sanders

A. J. Z.

Williams

K. M.

Seo

L. T.

Vogler

J. S.

Song

Williamson

Z. H.

Knox

M. C.

(2015). Does it matter if the teacher is there? A teacher’s contribution to emerging patterns of interactions in online classroom discussions. Computers & Education, 82, 315–328. https://doi.org/10.1016/j.compedu.2014.11.019

51.

Prawat

R. S.

(1992). Teachers’ beliefs about teaching and learning: A constructivist perspective. American Journal of Education, 100(3), 354–395. https://doi.org/10.1086/444021

52.

Reiter-Palmon

Sinha

Gevers

Odobez

J.-M.

Volpe

(2017). Theories and models of teams and groups. Small Group Research, 48(5), 1–24. https://doi.org/10.1177/1046496417722841

53.

Robinson

C. D.

Scott

Gottfried

M. A.

(2019). Taking it to the next level: A field experiment to improve instructor-student relationships in college. AERA Open, 5(1), 233285841983970. https://doi.org/10.1177/2332858419839707

54.

Sawyer

K. R.

(2014). The Cambridge handbook of the learning sciences. Cambridge University Press.

55.

Scardamalia

Bereiter

(2014). Knowledge building and knowledge creation: Theory, pedagogy, and technology. In Sawyer

R. K.

(Ed.), Cambridge handbook of the learning sciences (2nd ed, pp. 397–417). Cambridge University Press.

56.

Shaffer

D. W.

Collier

Ruis

A. R.

(2016). A tutorial on epistemic network analysis: Analyzing the structure of connections in cognitive, social, and interaction data. Journal of Learning Analytics, 3(3), 9–45. https://doi.org/10.18608/jla.2016.33.3

57.

Stahl

(2009). Studying virtual math teams. Springer. http://GerryStahl.net/vmt/book

58.

Stahl

(2014). Chapter 32. The constitution of group cognition. In Shapiro

(Ed.), Handbook of embodied cognition (pp. 335-346). Routledge. http://GerryStahl.net/pub/embodied.pdf

59.

Suthers

Lund

Rosé

C. P.

Teplovs

Law

(2013). Productive multivocality in the analysis of group interactions. Springer.

60.

Tabak

Baumgartner

(2004). The teacher as partner: Exploring participant structures, symmetry, and identity work in scaffolding. Cognition and Instruction, 22(4), 393–429. https://doi.org/10.1207/s1532690Xci2204_2

61.

van Aalst

(2009). Distinguishing knowledge-sharing, knowledge-construction, and knowledge-creation discourses. International Journal of Computer-Supported Collaborative Learning, 4(3), 259–287. https://doi.org/10.1007/s11412-009-9069-5

62.

van de Pol

Mercer

Volman

(2019). Scaffolding student understanding in small-group work: Students’ uptake of teacher support in subsequent small-group interaction. Journal of the Learning Sciences, 28(2), 206–239. https://doi.org/10.1080/10508406.2018.1522258

63.

van de Pol

Volman

Beishuizen

(2010). Scaffolding in teacher–student interaction: A decade of research. Educational Psychology Review, 22(3), 271–296. https://doi.org/10.1007/s10648-010-9127-6

64.

VanLeeuwen

Janssen

Erkens

Brekelmans

(2015). Teacher regulation of cognitive activities during student collaboration: Effects of learning analytics. Computers & Education, 90, 80–94. https://doi.org/10.1016/j.compedu.2015.09.006

65.

Vygotsky

L.S.

(1978). Mind in society—The development of higher psychological processes. Harvard University Press.

66.

Wang

Cheng

Chen

Mercer

Kirschner

P. A.

(2017). The use of web-based collaborative concept mapping to support group learning and interaction in an online environment. Internet and Higher Education, 34, 28–40. https://doi.org/10.1016/j.iheduc.2017.04.003

67.

Winne

P. H.

Hadwin

A. F.

Perry

N. E.

(2013). Chapter 26. Metacognition and computer-supported collaborative learning. In Hmelo-Silver

C. E.

Chinn

C. A.

Chan

C. K. K.

O’Donnell

A. M.

(Eds.), The international handbook of collaborative learning (pp. 462–479). Routledge.

68.

Zemel

Xhafa

Çakir

(2009). Chapter 23. Combining coding and conversation analysis of VMT chats. In Stahl

(Ed.), Studying virtual math teams, (pp. 532–568). Springer. http://GerryStahl.net/vmt/book