VISUALIZATION OF COMPUTER-

SUPPORTED COLLABORATIVE LEARNING

MODELS IN THE CONTEXT OF

MULTIMODAL DATA ANALYSIS

Jianqiang Mei*

School of Electronic Engineering, Tianjin University of Technology and Education,

Tianjin, 300222, China - Tianjin Engineering Research Center of Fieldbus Control

Technology, Tianjin, 300222, China

meijianqiang@tute.edu.cn

Wanyan Chen

School of Electronic Engineering, Tianjin University of Technology and Education,

Tianjin, 300222, China

Biyuan Li

School of Electronic Engineering, Tianjin University of Technology and Education,

Tianjin, 300222, China - Tianjin Engineering Research Center of Fieldbus Control

Technology, Tianjin, 300222, China

Shixin Li

School of Electronic Engineering, Tianjin University of Technology and Education,

Tianjin, 300222, China. Tianjin Engineering Research Center of Fieldbus Control

Technology, Tianjin, 300222, China

Jun Zhang

School of Electronic Engineering, Tianjin University of Technology and

Education, Tianjin, 300222, China - Tianjin Engineering Research Center of

Fieldbus Control Technology, Tianjin, 300222, China

Reception: 06/11/2022 Acceptance: 01/01/2023 Publication: 23/01/2023

Suggested citation:

M., Jianqiang, C., Wanyan, L., Biyuan, L., Shixin and Z. Jun (2023). Visualization

of computer-supported collaborative learning models in the context of

multimodal data analysis. 3C Empresa. Investigación y pensamiento crítico,

12(1), 87-109. https://doi.org/10.17993/3cemp.2023.120151.87-109

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

ABSTRACT

Deep learning evaluation is a new direction formed by the intersection of multiple

domains, and the core issue is how to visualize collaborative learning models to

motivate learners. Therefore, this paper realizes real-time knowledge sharing and

facilitates learners' interaction through computer-supported collaborative learning

(CSCL) technology. In this paper, we collect, label, and analyze data based on five

modalities: brain, behavior, cognition, environment, and technology. In this paper, a

computer-supported collaborative learning process analysis model is developed under

the threshold of multimodal data analysis. The model is based on roles and CSCL for

intelligent network collaboration. This paper designs and develops an interactive

visualization tool to support online collaborative learning process analysis. In addition,

this paper conducts a practical study in an online classroom. The results show that the

model and the tool can be effectively used for online collaborative learning process

analysis, and the test model results fit well. The entropy index of the test model took a

value of about 0.85, and about less than 10% of the individuals were assigned to the

wrong profile. During the test, the participation of participants gradually increased from

5% to about 25%, and the participation effect improved by about 80%. This indicates

the strong applicability value of the computer-supported collaborative learning process

analysis model under the multimodal data analysis perspective.

KEYWORDS

multimodality; computer-supported collaborative learning; visualization; process

analysis model; online classroom

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

PAPER INDEX

ABSTRACT

KEYWORDS

1. INTRODUCTION

2. ROLE-BASED AND CSCL MODELS FOR COLLABORATIVE

WEB-BASED LEARNING

2.1. Application of role mechanism

2.2. Model framework

2.3. Description of the collaboration process

3. VISUAL COLLABORATIVE LEARNING ANALYTICS MODEL

CONSTRUCTION

3.1. CSCL-KBS learning analysis model

3.2. Analytical model-based tool design and practice

3.2.1. Participation Analysis

3.2.2. Visualization design and results presentation

3.2.3. Potential profile analysis

4. DISCUSSION OF PRACTICE RESULTS

4.1. Perceptiveness of the collaborative activity process

4.2. Impact of visual presentation on the pattern of collaborative

activities

4.3. Evidence support for process evaluation

5. CONCLUSION AND OUTLOOK

REFERENCES

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

1. INTRODUCTION

Computer-supported collaborative learning (CSCL) is the theory and practice of

learners supported by computer network technology for the purpose of improving

learning performance. It enables collaborative cognition, exchange of emotions, and

development of collaborative skills in shared activities and interactions. In online

collaborative learning supported by technology, learners are able to ask questions to a

greater extent. They are able to express ideas clearly, exchange ideas with each

other, and share information. They can negotiate meaning. Ultimately, learners are

able to improve their collaborative learning skills. They can promote the development

of their cognitive skills and critical thinking [1]. Studies have shown that good

collaborative learning will have a positive effect on learning outcomes [2]. However,

the most central issue of collaborative learning in current research is how to visualize

models and thus motivate learners. Moreover, deep learning evaluation is a new

direction formed by the intersection of multiple domains. It can collect and build a

deep learning database and create a deep learning evaluation analysis model.

Ultimately, it can achieve the purpose of optimizing educational evaluation.

In the field of research and practice, there is a growing interest in computer-

supported collaborative learning (CSCL). An educational practice in which students

form learning groups and learn through social interaction via computers or the Internet

[3].CSCL can take place in classroom or online learning environments and can be

synchronous or asynchronous [4]. However, there are still many substantial problems

with collaborative learning [5]. For example, learners have uneven participation in the

learning process, lack of deep interaction, and biased support tools. Since

collaborative learning is a complex social process. Learners in CSCL are different

individuals. They have unique personality, cognitive, and affective characteristics.

They do not actively and voluntarily collaborate with other members [6]. Individual

differences and diversity, as well as the complexity of the learning environment, may

negatively affect cognition, emotion, and motivation. During collaboration, it is difficult

for learners to collaborate on complex problems or concepts through high-quality

cognitive mapping, active interaction, and sharing [7]. sotani, Mizoguchi, and Jaques

et al [8-9] argued that to improve collaboration in CSCL settings, students'

engagement needs to be increased to increase their interaction rate. This implies that

issues such as the allocation of responsibilities and resources and the mode of

interaction need to be addressed. There is variation in learners' interactive

engagement in CSCL, but it is not clear what causes this variation [10]. It has been

suggested that differences in engagement during interaction may stem from students'

motivation to participate in CSCL [11]. Therefore, a new generation of researchers has

begun to seek to identify the causes and mechanisms hidden behind the positive

collaborative outcomes. They focus on the process of collaborative interaction among

members and try to analyze the collaborative learning process in depth. They

understand the internal mechanisms by which effective collaboration occurs and build

long-lasting analytical models. In recent years, due to the main properties of deep

learning, it is increasingly used to solve several 3D visual problems [12-16], and these

collaborative learning analytical models are based on different theoretical

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

perspectives such as cognition or metacognition, knowledge construction, and critical

thinking. The models construct various analytical frameworks oriented towards

artifacts, contexts, interactions, and knowledge development. Based on different

theoretical perspectives, the analytical frameworks also cover elements such as

participants, interaction behaviors, cognitive and metacognitive, affect, learning

output, social support, and topic space. Models have focused on different aspects of

collaborative learning. Some models emphasize the importance of social interaction

for collaboration. Some models focus on elements of collaboration related to

interaction such as engagement, affect, and structural features of interaction (size,

density, intensity). For example, the models proposed by Fahy [17] and Veldhuis [18].

Some models emphasize the important role of cognition in collaborative analysis.

Some models focus on factors related to cognitive involvement, considering whether

new ideas are presented in the discussion, whether the problem space is clarified,

etc., such as Henri [19] and Newman [20]. There are also models that emphasize the

importance of conversational behavior and focus on elements such as questions,

answers, arguments, and comments, such as the analytical model constructed by Zhu

[21]. However, knowledge construction in groups in collaborative learning is a process

in which multiple factors are organically combined and interact with each other.

Collaborative learning has the limitation of narrow perspective in examining the

process of collaborative learning from a single side. From the analysis of the literature,

though, researchers have tried to construct as comprehensive a dimension as

possible to analyze the collaborative learning process. There are also researchers

who have enhanced their understanding of the internal mechanisms of collaboration

and the process of knowledge construction. For example, Li Yanyan's [22] model uses

knowledge construction as the theoretical basis and messages as the minimum unit of

analysis. The model quantitatively analyzes the collaborative learning process from

three aspects: topic space, topic intention, and social network. the model of Paul [23],

on the other hand, is based on knowledge construction theory and explores the

collaborative learning process based on content analysis from three aspects:

cognitive, social, and motivational. However, in general, the model construction and

analysis of multidimensional perspectives are still incomplete and scarce. Moreover,

along with the continuous development of the online collaborative learning

environment, the model needs to be constantly revised and improved in new

scenarios. Models to fit the analytical requirements of online collaborative learning. In

this paper, after comparing numerous existing collaborative learning models such as:

Web-based 1CAI model [24], Intermet-based intelligent teaching system ITS, and

intelligent agent-based model of online collaborative environment [25], it is found that

each of these virtual learning environment models is constructed with defects. Some

models are limited to only two roles, teacher and student, and students are in a

passive state. Some models emphasize the active role of students, but ignore the

collaboration between teachers and the role of teachers as learners. The models are

poorly interactive, lack intelligence, and do not allow for good collaborative learning.

Considering that the roles of teachers and students participating in collaborative

learning are dynamic and changing in CSCL, we introduced the role mechanism.

Therefore, according to the current research base and research questions. In this

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

paper, we collect, label, and analyze data based on five modalities: brain, behavior,

cognition, environment, and technology, based on a deep learning database. In this

paper, deep learning evaluation based on multimodal data is implemented and

improved in terms of automating data collection, integrating predictive models,

deepening educational applications, unifying mechanisms, and enhancing decision

wisdom. This paper realizes real-time knowledge sharing and promotes interactive

mutual assistance of learners through computer-supported collaborative learning

(CSCL) technology. This paper establishes a model for analyzing the process of

computer-supported collaborative learning in the context of multimodal data analysis.

Based on this model, an interactive visualization tool is designed and developed to

support online collaborative learning process analysis, and a practical study is

conducted in an online classroom.

In this paper, we propose and build an intelligent network collaboration model

RICLU based on role and CSCL, which focuses on role-based interaction,

collaboration and negotiation mechanisms among multiple agents in a collaborative

learning environment. The intelligent Agent takes on a role in learning on behalf of the

client user and interacts with other users and the management Agent on the server

side. This is a dynamic and open virtual learning environment, which better reflects

the characteristics of autonomy, interactivity, collaboration and distribution

transparency. This paper provides a good model for online education. In particular, the

introduction of multi-role mechanism better reflects the personalization of user

learning while promoting extensive cooperation among users.

2. ROLE-BASED AND CSCL MODELS FOR

COLLABORATIVE WEB-BASED LEARNING

2.1. APPLICATION OF ROLE MECHANISM

A role is a unity of responsibilities and rights and has four attributes:

responsibilities, rights, activities and agreements. As a reasonable criterion for

classifying things, roles are abstracted by grouping participants according to their

skills, abilities and other elements of the activity. A single participant may fill multiple

roles. A class of roles can also be filled by multiple users. A collaborative organization

can be considered as a collection of roles. There are specific relationships between

roles. In the collaborative process, a role is an active, relatively independent

abstraction unit. A role has a certain goal and can perform a series of operations in a

sequential manner. At different moments, roles can be in different states. A role R is

usually defined as a mapping f: (O,Ts)→action. O is the object on which the role acts.

Ts is the task to be performed. Action is the action of the object. The role-based

collaboration process is defined as a binary: P:=<Role, Relation>. Role denotes the

set of role spaces, and Relation denotes the collaborative relationship between roles.

The CSCL-based web-based learning environment is a distributed web-based

system. Users are located on the client side. Intelligent Agents represent users in their

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

learning roles. Interacts with other users and the server-side Management Agent.

Teachers and students are the most basic roles. Group managers, system managers,

resource managers, and message carriers assist in learning as secondary roles to

achieve better interactivity. All these roles are performed by the Smart Agent. Each

role has different authority according to its task: Management Agent controls and

manages all collaborative activities, shared resources, network communication, etc. in

the virtual environment; Collaboration Group Agent manages the activities of all

members of the group; Resource Management Agent manages the resource

repository; Routing Agent is responsible for inter-member messaging. In the role-

based and CSCL collaborative learning model, the following relationships exist among

the roles: active relationship (equal relationship). The relationship between the

collaborative agents is equal, and there is no controlling party and controlled party.

They can engage in free learning and mutual learning. For example, the interactions

between Student Agent and Student Agent and between Teacher Agent and Teacher

Agent during group learning and free discussion are equal. They all have equal

access to resources and privileges. One of the collaborators is the controlling party,

who is responsible for management and supervision. The other party is the controlled

party, whose actions are constrained by the control of one of the subjects. For

example, the interaction between the Teacher Agent and the Student Agent reflects

the relationship between teaching and learning. The Teacher supervises the students

and guides them in their learning. The Environment Management Agent is the master

when interacting with other subjects, and the other subjects are the controlled parties.

For example, the Student Agent requests services from it (registration, access to a

group, access to a repository, exit). The passive relationship is also manifested in the

collaborative learning between the Group Agent and the Member Agent, where the

latter is controlled by the former.

2.2. MODEL FRAMEWORK

From the perspective of application, collaborative learning can be divided into 3

layers: resource layer, functional layer, and management layer. The resource layer

provides a large amount of basic resource data for building the learning environment,

including text, audio and video, and WWW. These resources form the basic

databases such as the book database, audio and video database, and test bank. The

functional layer provides a friendly user interface to interact directly with learners and

realize specific application functions. Such as electronic forums, online groups, e-mail,

real-time video playback and evaluation of students' learning effects. The

management implements effective monitoring of resource data in the resource layer

and ensures data security. It performs daily maintenance of the functions in the

learning environment and manages the basic information of registered students and

teachers.

These management functions can be implemented through software and hardware.

Four types of Agents in collaborative learning can be defined based on three

characteristics of Agents: Autonomy, Cooperation, and Learning: Cooperative Agent,

Learning Agent, Interface Agent, and Decision Agent. It establishes negotiations with

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

other Agents and performs limited role learning. The Learning Agent emphasizes

autonomy and learning. It observes the user's behavior, learns from the patterns it

finds, and takes actions based on the user's preferences. The Interface Agent

retrieves information intelligently. It finds data automatically and quickly. Decision

Agent automatically performs tasks using intelligent mechanisms. It helps the user to

learn.

The model is represented by a seven-tuple: F=(A,R,T,Task,Source,D,K). Where A is

the set of Agents involved in the environment. A is represented by their internal

identifier Aid. r = {Teacher, Student, Manager, Facilitator}, is the set of roles. t =

{T1,T2,...Tn} is the set of collaborative groups Ti(1≤i≤

n). Task is a set of collaborative

tasks, indicating the tasks that each role in the group completes together. source is a

collection of system resources, including multimedia database, courseware library,

test bank, etc. D represents the database, describing system information such as

student management information, resource management information, teacher

information, etc. K is the knowledge base, which stores the collaboration rules and

guides the collaborative learning activities of the collaborative groups. Based on the

role analysis, the model defines Interface Agent (Student Agent and Teacher Agent),

Routing Agent, Group Agent, Management Agent, and CORBA-based Object

Requirements Agent, where Interface Agent is the functional layer. The Group Agent

and the Management Agent belong to the management layer. CORBA-based Object

Requirements Agents belong to the resource layer. Agents collaborate with each other

over a network (Internet, Intranet or small local area network). Learners can take on

the role of teachers or students. Common Object Request Broker Architecture CORBA

can provide security services, naming services, lifetime services and external

services. This facilitates distributed computing applications in a network environment

and effectively describes the dynamic nature of the Agent. This is a better

representation of object-oriented features. Combining the functions and roles of each

Agent, a unified model is used to describe the basic framework and internal structure

of the Agent in the network environment. The intelligent Agent in the model is defined

as a nine-tuple: Ag=(M,A,R,B,I, D,V,K,T). M - describes the activities such as

methods, executable behaviors and processes that the Agent has. A - describes the

type of Agent, the intent to perform the activity and the status information of the cohort

collaborators. R - Describes the role of the Agent in a collaborative activity. B -

describes the Agent's personal workspace. It is the equivalent of a network

blackboard and stores interaction information. I - Reasoning and problem processing

system, which controls the behavior of the Agent. It is responsible for the

interpretation and execution of domain knowledge, pattern matching, interaction

information processing, and result evaluation. D - The basic elements and data sets of

the problem solving domain. V - describes the domain knowledge (models, rules) and

collaborative interaction communication behavior. K - a knowledge system consisting

of domain-specific knowledge. It includes algorithms, models, generative rules and

semantic networks, etc. T - the communication mechanism with other Agents. Each

Agent performs the following functions in the model:

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

(1) Interface Agent: Interacts with other Agents on behalf of learners. It exchanges

requests, goals, resources, commitments, etc. to achieve the purpose of collaborative

learning. It includes user interface layer, semantic understanding layer, operation layer

and interaction layer. The operational layer includes behavior control, goal or rule

base, information retrieval and reasoning engine. The operational layer corresponds

to the belief set and knowledge base of the Agent. It learns the mindset of the user

(teacher or student) and understands their preferences. It automatically perceives the

learning environment and makes requests to other Agents (e.g., to the administrative

Agent to register, exit, enter a group, request a resource, etc.). The Teacher Agent has

the same functions as the Student Agent, and in the role of the Teacher, it is

responsible for making decisions about teaching and learning issues, making inquiries

about teaching and learning situations, and controlling and monitoring the Student

Agent.

(2) Group Agent: A special kind of interface agent who supervises the activities of

the members of the collaborative group and coordinates their activities as necessary.

Tutor all members during group instruction. Responsible for the distribution of

speaking rights during free discussion. It is responsible for assigning and coordinating

tasks when learning together. It can be elected by the Interface Agent or assigned by

the Management Agent. It is created when a group is created and disappears when a

task is completed. When intergroup learning occurs, it acts as a representative of the

group and negotiates with other group Agents about joint intentions.

(3) Management Agent: It is responsible for coordinating and supervising the

activities of members and the allocation of resources in the entire dynamic

environment, and any request for resources must be approved by it. It is the super

user of the learning environment, and any Agent can communicate with it directly. It is

connected to the Resource Module for data storage and retrieval. It can assign a

member as a group leader and assign teachers to individual and group instructional

activities. It can also record relevant information (including user joins, logins,

processing interactions, collaboration information, student information, teacher

information).

(4) Routing Agent: Responsible for communication between Interface Agent,

between Group Agent and Member Agent, and between Group Agents. He is

responsible for passing resource requests, task requests, goals, negotiation requests,

information feedback, etc. It can also communicate directly with the Management

Agent. It has mobility and is a mobile Agent.

(5) CORBA-based Object Requirements Agent: It follows the CORBA specification

and provides CORBA-based public request services, and is connected with resource

repositories (multimedia repository, courseware repository, test repository, answer

repository), databases (student management information, teacher information,

collaborative activity information), and knowledge repositories (collaborative rule

repository and goal planning repository). The Management Agent accesses the

repository by making resource access requests to it. It can also read, write, and

update the database and knowledge base.

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

2.3. DESCRIPTION OF THE COLLABORATION PROCESS

Collaborative learning includes individual instruction, group instruction, free

discussion, and joint learning (collaborative lesson preparation and collaborative

practice). In group instruction, the Teacher Agent is the controlling party, supervising

and guiding the learning activities of each Agent. Free discussion uses the group

Agent's web board as the workspace. The group uses a voice mechanism for

collaboration. All members have equal relationships and are learners. However, each

member is supervised and managed by the Group Agent. In joint learning, the

Teacher Agent breaks down the learning task into subtasks. These subtasks are

performed by a number of group members, with each task corresponding to a role.

The assignment of tasks is based on a combination of assignment and voluntariness.

Each member chooses whether to accept the task according to his or her ability and

willingness.

The mechanisms of co-operative learning are conflict and competition, self-

explanation, internalization, apprenticeship, shared cognitive tasking, and shared

rules. We take co-learning as an example. Combining the above mechanisms and

role mechanisms, a formal description of collaborative learning is given. The language

system V uses predicates to represent collaborative interaction activities, and defines

the interaction activities in task assignment as follows:

State (Ai): indicates the state of a member Agent Ai in the collaborative group.

There are three kinds of states: idle (idle), waiting (waiting), and busy (working).

Ask (T, Ai, Ti): Teacher Agent T asks if Ai can complete the task Ti.

Cando (T, Ai, Ti):Ai tells T that it can do the task Ti alone.

Notcand (o T, Ai, Ti): Ai tells T that it cannot complete the task Ti alone.

Assig (n T, Ai, Ti): T assigns the task Ti to Ai.

Needhelp (T, Ai, Aj, Ti): Ai can complete the task assigned to it by T only with

the help of Aj.

Askhelp (T, Ai, Aj, Ti): Ai asks Aj for help in completing the task assigned to it by

Help (T, Ai, Aj, Ti): Aj is willing to help Ai to complete the task assigned to it by T.

Refusehelp (T, Ai, Aj, Ti):Aj refuses to help Ai to complete the task assigned to it

by T.

Do (Aj, Aj, Ti): Aj and Aj work together to complete the task Ti.

Report (Ai,T,result): Ai submits the execution result to T.

For a given task Ti, the interactions in the task assignment process are described

by the following algorithm:

FOR each team member Agent Ai

IF State (Ai)=idle

{Ask (T, Ai,Ti);

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

IF Cando (T, Ai,Ti)

{Assign (T, Ai, Ti); Stat (e Ai)= busy; return; }

ELSE

IF Needhelp (T, Ai, Aj, Ti)

{State(Ai)=waiting;

REPEAT

Askhelp (T, Ai, Aj, Ti);

UNTIL (find an Aj, satisfying: Help (T, Ai, Aj, Ti), or ask all Aj);

IF find Aj Do (Aj, Aj, Ti) that satisfy the condition;

ELSE {State (Ai)=idle; return;}}}

3. VISUAL COLLABORATIVE LEARNING ANALYTICS

MODEL CONSTRUCTION

3.1. CSCL-KBS LEARNING ANALYSIS MODEL

Although there are many models on collaborative learning analysis, however,

existing research still lacks a systematic and global perspective to overview the

dimensions of collaborative process analysis [26]. A review of the current literature on

collaborative learning process analysis. We were able to identify some new research

perspectives that are gradually gaining attention in the study of collaborative learning

process analysis. One of the important aspects is the research on knowledge

processing in the collaborative learning process. Knowledge processing plays an

important role in the collaborative process [27]. Knowledge processing is concerned

with the process of knowledge creation and generation in collaborative learning. This

process allows learners to organize knowledge into coherent structures and to

generate new knowledge using existing knowledge. Studies have shown that the

measurement of knowledge processing can measure whether a cluster is successfully

engaged in collaborative problem solving [28]. Another important aspect is the

research on social relationships in collaborative learning. Numerous studies point out

that active online participation is a key factor in the success of student learning. In

online collaborative learning, individuals in a group interact effectively for the common

learning goals of the group. Social relationships among members can influence the

process and quality of knowledge construction [29]. In addition, the analysis of

behavioral patterns of collaborative processes is an important topic of current CSCL

research. Group members accomplish activities with specific goals through

interaction. This can be seen as consisting of a series of intentional interaction

behaviors. Abstracting the sequence of interactions of these behaviors can lead to

different behavioral patterns. The different behavioral patterns reflect the collaborative

interaction strategies embodied by the collaborative group during the interaction

activities.

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

In order to provide a comprehensive portrayal of learning analysis in CSCL. On the

basis of the findings of Li et al. [22]. This paper proposes an improved

multidimensional analysis model. The model explores student knowledge construction

in collaborative learning discussion activities. The model contains three analytical

dimensions: knowledge processing (K), behavioral patterns (B), and social

relationships (S). The model is designed for three different levels of study: individual,

group, and community. The model is named as KBS model.

3.2. ANALYTICAL MODEL-BASED TOOL DESIGN AND

PRACTICE

3.2.1. PARTICIPATION ANALYSIS

It is quite obvious that the basic activity in collaborative learning is to participate in

discussions [30-31]. This in a way implies that the participants externalize and share

the sense of information or knowledge. This is one of the most important

manifestations of active interdependence of individuals. Social construction theory

suggests that our knowledge and experience is not objectively "discovered". It is

discussed, negotiated, and constructed by participants in group interaction. From the

perspective of social construction theory, the process of understanding is not driven

by natural forces. It is the result of the active, collaborative work of people in certain

relationships. Therefore, participation in a collaborative group is the most basic

requirement and behavior.

Researchers have argued that a participant's engagement can be measured by his/

her interaction with peers or the teacher. Previous research has shown that participant

engagement is a positive predictor of actual learning, individual retention in

continuous learning, and learning satisfaction. In general, individual engagement in

computer-supported collaborative learning refers to the number of individual

perspectives the length of posts in the online environment or whether the perspectives

are social rather than focusing on content creativity. Researchers have argued that

the number of participants' perspectives is a better indicator of how engaged

participants are in the computer-supported collaborative learning process. In this

study a viewpoint is primarily a sentence of an individual.

Let C denote the sequence of viewpoints and Cr

denote the tth viewpoint in the

sequence. n denotes the length of the sequence of viewpoints. Since views vary over

time, the variable 1 will be used to index individual views, also called "time" (the value

of 1 ranges from 1 to n).

(1)

Let P be a set of individuals. The variables a and b will be used to refer to any

member (individual) of this set. To determine the initiator (or individual) of each

viewpoint, we define the following participation function as shown in Equation (2):

1t n≤ ≤

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

(2)

is denoted as 1 if participant a ∈

P contributes view c, and 0 otherwise. The

participation function of any participant (a), can be defined as a sequence:

(3)

where n is the same length as the sequence of viewpoints C. It takes the value 1

when participant a initiates the corresponding viewpoint in C, and 0 otherwise. Using

this participation function, several useful descriptive measures of participation in the

discussion can be defined in a relatively simple way. The number of points of view of

any participant is:

(4)

In addition, during the discussion, participants may have "pandering" opinions.

These are single-word opinion sentences such as "um", "ah", "yes", and "yes".

Treating these as equivalent to longer sentences may result in higher participation by

participants who are not seriously engaged in the discussion. This would affect the

accuracy of the later analysis. Therefore, it is also necessary to calculate the length of

the opinions expressed by the participants. The length

of any participant (a)

expressing an opinion can be considered as:

(5)

where

is denoted as the length of opinions published by participant a at time

t. and the total opinion length W is:

(6)

where k is the number of individuals in the group. The participation of participants

can be estimated by the sum of the relative proportions of their participation to the

total number of participants and the relative proportions of the total number of words

(its variation with rounds is shown in Figure 1):

(7)

( ) 0,

P t ⎧

=⎨

⎩

{ } { }

, ,

( ) (1), (2), (3) , ( )

a a a a a a

P P t P P P P n

…

= =

( )

Pa P t

=∑

( )

a a

W w t

=∑

Wa(t)

W W=∑

a a

P W

n W

⎛ ⎞

⎜ ⎟

⎝ ⎠

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

Figure 1 Changes in participant engagement over time

3.2.2. VISUALIZATION DESIGN AND RESULTS PRESENTATION

In the context of knowledge processing, knowledge and the connections between

knowledge are important assessment criteria. Students form new knowledge

structures by integrating and linking knowledge, thus facilitating knowledge

processing. Therefore, it is important to be able to monitor in real time the

development of key knowledge points during student discussions. This is important for

teachers to keep track of the development of students' cognitive engagement and to

effectively monitor the teaching process. In the operationalization analysis of

knowledge processing, natural language syllabification techniques were used. The

Chinese word-sorting system of CAS was used to keyword-sort the discussion texts

during the collaborative process and identify the keywords during the students'

discussions. This paper matches with the knowledge concept map provided by

experts about the collaborative discussion problem. Meanwhile, this paper measures

the cognitive involvement in the collaborative process from the perspective of cluster

or student knowledge structure formation. As shown in Figure 2, the knowledge point

development change map presents teachers with the development change and

distribution pattern of cluster knowledge points from the time dimension. When the

mouse hovers over a knowledge point, it also automatically shows back the content of

the post where the knowledge point is located, the poster and the time of the posting.

Using this visual information, teachers can help discover in-depth information about

the process of group discussion. For example, how the group knowledge points were

generated over time, whether any group had problems with off-topic or stagnant

discussions, and whether relationships between knowledge points were established.

This makes the development process of cognitive engagement easier to monitor.

However, knowledge processing can only focus on cognitive engagement during

collaborative discussions. This lacks a clear indication of the behavioral interactions of

the cluster. The developmental changes in cognitive processes are influenced by the

behavioral aspects of the cluster interactions themselves. Therefore, further, the visual

presentation of behavioral patterns can be used to explore the strategies and patterns

of students' behavioral interactions during collaborative knowledge construction.

0 10 20 30 40 50

0.00

0.05

0.10

0.15

0.20

0.25

Degree of involvement

Sequence of sentence

speaker

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

100

(a) Group 1

(b) Group 5

Figure 2 Development of knowledge points

In terms of behavioral patterns, to dig deeper into the influence of group behavioral

patterns on students' collaborative learning. This paper investigates the characteristics

of discussion-based online collaborative learning. In this paper, collaborative

behaviors are coded into five first-level categories: presentation (C1), negotiation

(C2), questioning (C3), management (C4), and emotional communication (C5), and

each category is further refined into 14 second-level categories (C11: gives ideas/

options; C12: further explains ideas; C13: revises ideas/options; C14: summarizes

ideas/options. C21: agrees; C22: Agree, give evidence/reference; C23: Disagree;

C24: Disagree, give evidence.C31: Ask questions; C32: Ask follow-up questions. C41:

Organize/assign tasks; C42: Coordinate management/reminders). Embed these

codes in the posting area of the Moodle platform. The selection of behavioral

categories can be made when students submit postings. This can support the

automated processing of analytics tools. Finally, the association rule approach to data

mining in learning analytics is used through the analytics system. This method

calculates the probability that each behavior will be accompanied by the next behavior

and the intensity of the next behavior, extracts behavior transition pairs that occur at

high frequencies, and finally forms behavior sequence transition patterns. These

behavior patterns characterize the different behavior patterns of the collaborative

group in the collaborative interaction.

In terms of social relationship analysis, the analysis of social interactions will help

teachers to better understand who are the central participants in the knowledge

stack

queue

under

shelves

Order stack

Chain queue

Fig. 2 (a) Group 1

14:10 14:20 14:30 14:40 14:50 15:00 15:10

Chain queue

Order stack

shelves

under

queue

stack

14:15 14:20 14:30 14:35 14:40 14:45 14:50

Fig. 2 (b) Group 5

14:25 14:55 14:58

https://doi.org/10.17993/3cemp.2023.120151.87-109

101

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

construction dialogue. It can see if there are some undesirable social relationships

that can have an impact on the motivation for collaborative learning. The visualization

diagram based on the interactions can graphically represent the characteristics of the

interaction network structure. It can effectively support teachers in qualitatively

analyzing the attributes of the interactive network structure and discovering whether

there are distinct central, peripheral, and isolated figures in the network.

3.2.3. POTENTIAL PROFILE ANALYSIS

For potential profile models, the most important issue is to determine the number of

their profiles. Currently, researchers still determine the model mainly based on its fit

indices. These fit indices include the Akaike Information Criterion (AIC), Bayesian

Information Criterion (BIC), and likelihood ratio test. The better the model fit is, the

smaller the values of these indices are. In addition, the entropy index has a value

range from 0 to 1. This can be used to measure the accuracy of the model in

classifying profiles (classes). The higher the value is, the more accurate the

classification is. For example, when it is 0.6, about 20% of the individuals may be

classified into the wrong profiles (potential classes). While when Entropy=0.8, about

less than 10% of the individuals were classified into the wrong profiles (potential

classes). As shown in Figure 3, the model results fit well (when choosing a model, it is

important to consider not only the statistical indicators but also the substantive

significance of each class).

Figure 3 Fitting index gravel plot for potential profile analysis

For the model with six categories of potential profile analysis, the results showed

that engagement differed significantly on participant categories (F(6,164) = 74.22, p <

0.01). Social influence differed significantly across participant categories (F(6,164) =

76.80, p < 0.01). Overall response rate was significantly different on participant

categories (F(6,164) = 97.89, p < 0.01). Intrinsic correlation was significantly different

across participant categories (F(6,164)= 32.85, p<0.01). Communication density

234567

14000

14200

14400

14600

14800

15000

15200

15400

15600

AIC&BIC Vaule

Porfile Number

BIC

AIC

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

102

differed significantly across participant categories (F(6,164)= 86.89, p< 0.01).

Response rate was significantly different across participant categories (F(6,164) =

86.89, p < 0.01).

4. DISCUSSION OF PRACTICE RESULTS

Through the analysis of the three important dimensions in the model and the

visualization of the results based on tool support, it can be seen that the use of

learning analytics improves the subjective drawbacks of the mainstream manual

coding-based analysis of the original collaborative learning process. The technique

overcomes the shortcomings of manual analysis, which is time-consuming and

laborious and can only be used for post-collaboration analysis, and provides

implementation feedback on the collaborative process. The technique enhances the

evaluation, feedback, perception and adaptation of collaborative learning. At the same

time, the presentation of visualizations based on tool analysis transforms the data

generated by the collaborative process into a friendly visual form. This brings to the

fore some important features, patterns, and anomalies. Thus, visual presentation

supported by analytical tools can be a key feature. It is used to gain insight into the

learning process as well as to provide basic support for monitoring, feedback, and

evaluation. It is important for teachers to monitor their teaching, researchers to

uncover large-scale teaching patterns, and process evaluation.

4.1. PERCEPTIVENESS OF THE COLLABORATIVE ACTIVITY

PROCESS

Model-based visual presentation can improve teachers' perception of the process

of collaborative activities. When multiple groups are discussing online at the same

time, it can be difficult for teachers to monitor problems with group collaboration in real

time without the help of tools. With visual information, it is easier for teachers to

identify problems such as digressions and stagnation in the discussion. Teachers can

gain a deeper understanding of the discussion process. As can be seen from the

comparative display of the two groups in Figure 2, Group 1 had more discussion on

the six knowledge points selected by the teacher during the discussion time. There

was no deviation or stagnation in the middle of Group 1. In contrast, the discussion in

Group 5 was fragmented and disorganized, without establishing relationships among

related knowledge points. Further, when the teacher hovers over a particular bullet

point, the tool automatically displays more detailed information about that knowledge

point. For example, which student mentioned the point at that point in time, and the

original text content of their discussion of the point. With this information, teachers can

more easily find out at what point the group entered into the discussion of a particular

issue. It is also possible to discover how the group gradually builds knowledge-to-

knowledge connections in the discussion that facilitate problem solving. In other

words, presenting the distribution patterns of knowledge over time makes the

traditional "black box" collaborative process of knowledge processing visible. This will

provide teachers with sufficient information to better observe the discussion process.

https://doi.org/10.17993/3cemp.2023.120151.87-109

103

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

It can provide additional evidence for researchers to conduct ongoing inquiry into the

internal mechanisms of collaborative processes.

4.2. IMPACT OF VISUAL PRESENTATION ON THE PATTERN OF

COLLABORATIVE ACTIVITIES

Model-based visual presentation can provide powerful support for exploring the

patterns of collaborative activities. In large-scale online education scenarios, the

visual presentation of behavioral sequence patterns in behavioral models can be used

to flexibly explore the behavioral transition patterns of online collaborative activities.

The visual presentation helps teachers gain deeper insight into the internal patterns of

group interaction behaviors. In Group 1, the self-loop of C11→

C11 during the

collaborative discussion to complete the task shows that each member can

continuously put forward his or her own point of view. Members actively think about

the problem and express their suggestions. In addition, the transitions from

C11→C32, C11→C23, and C11→

C13 show that the members of Group 1 presented

their ideas accompanied by further follow-up questions from other members,

questioning with evidence, and revision to improve their ideas. This indicates that the

members of the group were able to argue the issue sufficiently to keep moving the

task forward to completion. Moreover, the transformation of C32→

C12 shows that

when a member pursues a point of view, he or she is given a more detailed

explanation by other members. This indicates that the group is very interactive. In

contrast, after a member of group 5 raises a viewpoint, other members give questions

(C11→

C21), but the questions are not followed by corresponding explanations.

c31→C11 and C32→C11 show that members of the group do not give explanations or

revise their views after facing questions or follow-up questions, but continue to raise

new views. From the later analysis of the content based on knowledge processing and

the synthesis of the interaction structure, it is clear that Group 5 did not reach the

pattern of deep interaction of questioning-pursuing-questioning. This is related to its

group members' lack of attention to other people's viewpoints and the fact that group

members' discussions are more about posting only rather than engaging in dialogue.

Extraction results using real-time behavioral sequence transformation provided by

the tool. Teachers or researchers can conduct the mining of online collaborative

learning behavioral patterns in various scenarios. For example, in terms of the

characteristics of group behavior patterns, the behavioral characteristics of

collaborative groups with regular patterns in the process of collaborative knowledge

construction can be found. Their effects on knowledge construction can also be found.

Also, the similarities and differences in behavioral patterns presented by high and low

quality groups can be examined to help teachers explore important positive influences

in high quality discussions as well as potential limitations present in low quality

groups. This will provide a valuable reference for teachers to design better online

collaborative activities and teaching strategies. Using real-time process information

from the visualization tool, it can also be explored to obtain comparisons of

differences in behavior patterns at different stages. By analyzing the different

behavioral patterns of collaborative groups at the beginning, unfolding, and concluding

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

104

stages of collaboration, the changes of behavioral development in the collaborative

knowledge construction process can be explored. The results can provide a basis for

exploring the internal mechanism of the knowledge construction process.

4.3. EVIDENCE SUPPORT FOR PROCESS EVALUATION

Model-based visual representations can provide comprehensive evidence to

support process evaluation. Collaborative assignments within online classrooms often

involve a large amount of participation and contribution. This makes monitoring and

evaluation by tutors time-consuming, tedious, and error-prone. It is nearly impossible

for tutors to manually process the hundreds of sequences of contributions in a

discussion topic and the relationships between these contributions. As a direct

consequence, most online learning environments use simple metrics based on the

number of posts, reads, topics created, and average length of statements. These

measures are very useful in capturing the dynamics of online collaborative activity.

However, they ignore the essential feature of the need to continuously consider the

process of knowledge construction. This does not serve the core of process-based

evaluation of clusters. Thus, using the three dimensions of the multidimensional

model to complement and explain each other will help teachers to comprehensively

evaluate the group collaboration process in terms of multiple dimensions, including

cognitive engagement, interactive behaviors, and social relationships.

By observing the behavioral transition pattern, it can be found that Group 5 mainly

reflected more behavioral strategies of questioning or pursuing in the interaction

pattern of behavior. However, group 5 did not show more meaningful negotiation

processes such as arguing in the process of question reaching. Further, a deeper

examination of the content revealed by the mouse locating knowledge points shows

that Group 5 lacked sufficient motivation in the content of their statements to explain

their views or to discuss alternative options. Group 5 prefers to seek ultimate help,

such as being told the answer directly. In other words, the knowledge processing

dimension was further combined. The analysis of the content provides an

understanding of the micro-level of online collaborative discussions. Exploring the

relationship between the influence of social network structure on knowledge

construction can explore the internal causes affecting the effectiveness of

collaborative learning from multiple dimensions. This can provide more robust

evidence to support process evaluation.

5. CONCLUSION AND OUTLOOK

In this paper, we use computer-supported collaborative learning (CSCL) technology

to share knowledge in real time and promote interactive learning. In this paper, we

collect, annotate and analyze data based on five modalities: brain, behavior, cognition,

environment and technology, and establish a model of computer-supported

collaborative learning process analysis in the context of multimodal data analysis. The

model is based on the intelligent network collaboration of roles and CSCL, and an

interactive visualization tool is designed and developed to support the analysis of

https://doi.org/10.17993/3cemp.2023.120151.87-109

105

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

online collaborative learning process. In addition, this paper conducts a practical study

in an online classroom. The study shows that the model and tool can be effectively

used for online collaborative learning process analysis. It can help teachers monitor

the discussion process in real time, identify problems in collaboration, and provide

timely intervention and guidance. The following conclusions were obtained:

(1) In this paper, we elaborate the application of role theory in online learning

environment, and propose a new intelligent CSCL model by combining the three-layer

structure of application perspective. The model is personalized and intelligent, with

good practicality, and well reflects the dynamic distribution of multiple roles in

collaborative learning. Especially, the multiple roles of teachers in the environment

better realize the unity of teaching and learning.

(1) The intelligent network collaboration model based on roles and CSCL is

proposed in this paper. The model combines the theory of CSCL, intelligent agent

technology and role theory, and has good application value. The model results fit well

according to the Akaike Information Criterion (AIC), Bayesian Information Criterion

(BIC) and likelihood ratio test. When the entropy index takes a value of about 0.85,

about less than 10% of the individuals are assigned to the wrong profile.

(1) Model-based visualization enhances teachers' perception of the process of

collaborative activities. It provides powerful support for exploring the patterns of

collaborative activities and provides comprehensive evidence to support process

evaluation. Visual presentation improves the effectiveness of monitoring the

discussion process in real time. During the test, the participation of participants

gradually increased from 5% to 25%, and the participation effect increased by about

80%.

In this paper, we study the interaction, cooperation and coordination among role-

based multi-intelligent Agents from the perspective of multiple roles. This paper

proposes and describes the collaboration and coordination mechanism among role-

based intelligent agents. In particular, the role mechanism is studied in the application

of open and dynamic network environment. The article greatly enriches the multi-

agent system (MAS) theory. The mechanism, if combined with adaptive knowledge

mining algorithms, will greatly contribute to the progress of distributed data mining

research. The next step is to apply the role-based cooperation and coordination

negotiation mechanism among intelligent agents and adaptive knowledge mining of

intelligent agents to distributed data mining systems with knowledge orientation. This

has good application prospects and is the direction we need to study in the future.

REFERENCES

(1)

LI Y, DONG M, HUANG R. Toward a semantie forum for active collaborative

learning [J]. Educational technology & society, 2009, 12(4):71-86.

(2)

ALAVI M, DUFNER D. Technology-mediated collaborative learning: a research

perspective [M/]/Learning together online: research on asynchronous learning

networks. Mahwah, NJ: Lawrence Erlbaum Associates, 2005 :191-213.

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

106

(3) Soualah-Alila, F., Nicolle, C., & Mendes, F. (2015).

Towards a methodology for

semantic and context-aware mobile learning [M/EB].[2019-4-26]

(4) Stahl, G., Koschmann, T. D., & Suthers, D. D. (2006).

Computer-supported

collaborative learning: An historical perspective [C]

R. K. Sawyer(Ed.).

Cambridge handbook of the learning sciences. Cambridge, UK: Cambridge

University Press:409-426.

(5) Heimbuch, S., Ollesch, L., & Bodemer, D. ( 2018).

Comparing effects of two

collaboration scripts on learning activities for wiki-based environments [J]

International Journal of Computer-Supported Collaborative Learning

13(3):331-357.

(6) Kreijns, K, Kirschner, P. A., & Jochems, W. (2003).

Identifying the pitfalls for

social interaction in computer-supported collaborative learning

environments: A review of the research [J]. Computers in Human Behavior

19(3):35-353.

(7) Ludvigsen, S., Cress, U, Law, N., Stahl, G., & Rose, C. P. (2017).

Future

direction for the CSCL field: Methodologies and eight controversies [J]

International Journal of Computer-Supported Collaborative Learning

12(4):337-341.

(8)

Isotani, S., Mizoguchi, R., Isotani, S., Capeli, 0. M., Isotani, N., Albuquerque, A.

R., & Jaques, P. (2013).

A semantic web- based authoring tool to facilitate

the planning of collaborative learning scenarios compliant with learning

theories [J]. Computers & Education, 63(2):267-284.

(9)

Reis, R. C. D., Isotani, S., Rodriguez, C. L., Lyra, K. T, Jaques, P. A., &

Bittencourt, I. I. (2018).

Affective states in computer-supported collaborative

learning: Studying the past to drive the future [J]. Computers & Education

(120): 29-50.

(10)

Rienties, B., Tempelaar, D., Van den Bossche, P, Gijselaers, W., & Segers, M.

(2009).

The role of academic motivation in computer-supported

collaborative learning [J]. Computers in Human Behavior, 25(6):1195-1206.

(11) Koops, W., & Van der Vleuten, C. (2018).

A computer-supported

collaborative learning environment in medical education: The importance

for educators to consider medical students’ motivation [J]. J

ournal of

Contemporary Medical Education, 8(1): 10-17.

(12) Xiang, W. A. , et al.

Multi-view stereo in the Deep Learning Era: A

Comprehensive Review. (2021).

(13) W. Cai, D. Liu, X. Ning, et al.,

Voxel-based Three-view Hybrid Parallel

Network for 3D Object Classification, Displays 69 (1) (2021).

(14) Bai X, Zhou J, Ning X, et al. 3D data computation and visualization. Displays

2022: 102169.

(15) X. Ning, P. Duan, W. Li, and S. Zhang,

Real-time 3D face alignment using an

encoder-decoder network with an efficient deconvolution layer,

IEEE Signal

Processing Letters, vol. 27, pp. 1944–1948, 2020.

(16) Wang C, Zhou J, Xiao B, et al.

Uncertainty Estimation for Stereo Matching

Based on Evidential Deep Learning. Pattern Recognition, 2021.

https://doi.org/10.17993/3cemp.2023.120151.87-109

107

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

(17) Fahy P J , Gail C , Mohamed A .

Patterns of Interaction in a Computer

Conference Transcript[J].

International Review of Research in Open and

Distance Learning, 2001, 2(1).

(18) VELDHUIS-DIERMANSE A E.

CSC Learning: participation, learning,

activities and knowledge construction in computer-supported

collaborative learning in higher education[D].

Netherlands: Wageningen

University, 2002.

(19) HENRI F. Computer conferencing and content analysis [M],

Collaborative

learning through computer conferencing. Springer BerlinHeidelberg,

1992:117-136.

(20) NEWMAN D R, OTHERS A.

A content analysis method to measure critical

thinking in face-to-face and computer supported group learming[J],

Interpersonal computing & technology, 1995, 3(2) :56-77.

(21) ZHU E.

Meaning knowledge construction and mentoring in a distance

leaming course [C],

National convention of the association for educational

communications and technology, Indianopolis, 1996; 821-844.

(22) LI Y Y, LIAO J,WANG J, et al.

CSCL interaction analysis for assessing

knowledge building outcomes: method and tool [C],

Proceedings of the 8th

iternational conference on Computer supported collaborative learning.

Intemational Society of the Learning Sciences, 2007 :431-440.

(23) PAUL A, ERKENS G. Toward a framework for CSCL research[J].

Educational

Psychologist, 2013, 48(1):1-8.

(24) LI. Yanda et al.

Distance Learming.Proceedings of the international

Conference on Distance Education,Distance Learning and 215t Century

Education Development. April 12~15,1999,Beijing,China.

(25) Ayala, G and Yano, Y.

Intelligent Agents to Support the Effective

Collaboration in a CSCL Environment.

Proceedings of the ED-TELECOM 96

World Conference on Educational Communications, Boston, Mass. AACE,

Patricia Carlson and Fillia Makedon (eds.), pp.19-24, June,1996.

(26) KIM M, LEE E.

A multidimensional analysis tool for visualizing online

interactions [J]. Educational technology & society, 2012, 15 (3):89-102.

(27) ZHENG L, HUANG R, HWANG G J, et al.

Measuring knowledge elaboration

based on a computer -assisted knowledge map analytical approach to

collaborative learning[J]. Educational technology & society

, 2015,

18(1):321-336.

(28) Eryilmaz E , Pol J , Ryan T , et al.

Enhancing student knowledge acquisition

from online learning conversations[J]. International Journal of Computer-

supported Collaborative Learning, 2013, 8(1):113-144.

(29) TIRADO R, ANGEL HERNANDO, AGUADED J I.

The effect of centralization

and cohesion on the social construction of knowledge in discussion

forums Interactive learning environments, 2015, volume 23(3):1-24.

(30) Julca A. B., Tapia, C. D., Hilario, F. M., y Corpus, C. M. (2021).

Qualitative

benchmarking study of software for switch performance evaluation.

Tecnología. Glosas de innovación aplicadas a la pyme, 10(4), 35-49. https://

doi.org/10.17993/3ctecno/2021.v10n4e40.35-49.

https://doi.org/10.17993/3cemp.2023.120151.87-109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023

108

(31) Li Zhengjian & Li Lifeng (2021).

Mathematical statistics algorithm in the

bending performance test of corroded reinforced concrete beams under

fatigue load. Applied Mathematics and Nonlinear Sciences (1). https://doi.org/

10.2478/AMNS.2021.2.00142.

https://doi.org/10.17993/3cemp.2023.120151.87-109

109

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 51 Iss.12 N.1 January - March, 2023