A Systematic Literature Review of Adaptive Learning Systems Based on

the Assessment of Collaboration Quality

Nadia Hocine

CSTL Laboratory, University of Mostaganem, Av. Hamadou Hossine, Mostaganem, Algeria

Keywords:

Adaptive Learning Systems, Computer-Supported Collaborative Learning, Collaboration Quality.

Abstract:

Advances in information and communication technologies has led to the development of new data analysis

methods and strategies used to support remote and co-located collaborative learning. These strategies seek to

give meaning to complex data of individual and team interaction with the learning system to inform actionable

insights. However, providing teams and teachers with a substantial amount of data during the collaboration

process can complicate interpretation and hinder decision-making. Adaptive learning systems bear high po-

tential to assist classroom orchestration and support collaborative learning by providing students with adaptive

feedback. This paper systematically reviews existing literature following PRISMA methodology to provide

insights into adaptive collaborative learning systems. It speciﬁcally puts the light on how learning systems

have been adapted by considering the assessment of collaboration quality within teams. The objective is to

present common adaptation approaches, practices, and challenges as well as to discuss opportunities to im-

prove future adaptive learning systems.

1 INTRODUCTION

The ability to work in a group to construct knowl-

edge, negotiate, and meet shared objectives is among

the critical 21st century skills that promote workforce

effectiveness (Lima and de Souza, 2017). However,

teams may also face challenges such as conﬂicts in

views, the lack of social skills, as well as the need for

support and explanation of tasks (Saqr et al., 2018).

Adapting the learning system according to stakehold-

ers’ needs can therefore play an important role in sup-

porting collaborative learning.

Adaptation traditionally refers to the process of

tailoring the learning content, the system or the inter-

face to individual learners (Brusilovsky et al., 2015;

Hocine and Sehaba, 2024). In other studies, it is

viewed as a means to support or guide users dur-

ing the learning process by allowing them to con-

trol their learning and choices (Barria-Pineda et al.,

2023). In computer-supported collaborative learning

(CSCL) studies, adaptation aimed to support collabo-

rative work by recommending or adapting for instance

team composition on the basis of group members’

characteristics (Lykourentzou et al., 2016). Quite re-

cently, research has been concentrated on various con-

https://orcid.org/0000-0001-7875-1064

cerns such as supporting orchestration and transitive

activities between individual and collaborative learn-

ing (Yang et al., 2023), adaptive scaffolding (Splichal

et al., 2018), and dashboard design (Zamecnik et al.,

2022).

This paper reviews research on adaptive learning

systems that assess collaboration quality to support

collaborative learning, addressing two key questions.

First, the review aims to understand how collabo-

ration quality analysis methods and indicators were

used to adapt the learning system and whether this

contributed to improving collaborative learning and

teaching. RQ1: What adaptation approaches have

been used to enhance collaborative learning based

on collaboration quality assessment?. Second, the

review puts the light on collaboration quality analy-

sis methods and indicators. RQ2: What methods

and indicators have been employed to automati-

cally assess the quality of student collaboration for

adapting the learning system?.

Despite numerous literature review studies on as-

sessing collaboration quality through indicators in-

ferred from student interactions with learning sys-

tems (Eryilmaz et al., 2021), there is a lack of studies

that explore how these indicators contribute to system

adaptation. For example, Neumayr and Augstein re-

viewed personalized collaborative systems across var-

Hocine, N.

A Systematic Literature Review of Adaptive Learning Systems Based on the Assessment of Collaboration Quality.

DOI: 10.5220/0013196300003932

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 17th International Conference on Computer Supported Education (CSEDU 2025) - Volume 2, pages 909-916

ISBN: 978-989-758-746-7; ISSN: 2184-5026

909

ious contexts (Neumayr and Augstein, 2020), but did

not examine how collaboration indicators informed

system adaptation. Similarly, Vogel and colleagues

conducted a meta-analysis of socio-cognitive scaf-

folding techniques (Vogel et al., 2017), but focused on

adaptive scaffolding’s impact on learners’ skills, with-

out addressing how collaboration quality was used to

adjust the scaffolds.

The remainder of this paper is organized as fol-

lows: Section 2 describes the methodology of the sys-

tematic literature review. Section 3 presents the re-

sults of the review by answering the two previous re-

search questions. Section 4 discusses the obtained re-

sults, current challenges, and some opportunities to

advance research in adaptive collaborative learning

systems. We conclude this paper by presenting a sum-

mary of the review ﬁndings, its limitations, practical

implications, and perspectives for future work.

2 METHODOLOGY

This systematic literature review follows the preferred

reporting items for systematic reviews and meta-

analyses (PRISMA) protocol (Page et al., 2021). It

is performed in four phases: a comprehensive search

of eligible studies using databases, screening the ti-

tle and abstract of these studies, selection of relevant

papers following exclusion and inclusion criteria, and

reviewing the full texts of the screened studies to ex-

tract data. 27 research papers were included in this

review.

Figure 1: Paper selection ﬂowchart.

The literature review search was conducted

using: ACM, IEEE, SpringerLink, and Google

Scholar databases. The following general research

query was used: (personaliz OR adapt* OR cus-

tomiz*) AND (”computer-supported collaborative

learning” OR ”CSCL” OR ”collaborative learning”

OR ”technology-enhanced learning”) AND (”Collab-

oration analytics” OR ”collaboration assessment” OR

”collaboration quality”). The search was limited to

peer-reviewed studies from 2015 to June 2024 writ-

ten in english.

Figure 1. summarizes the main steps of the search

and analysis process. In the eligibility phase of

search, the full texts of the 68 articles were reviewed.

Given that the review objective is to understand how

collaboration quality assessment has contributed to

the adaptation of collaborative learning systems, two

inclusion criteria of articles were set out. First, the

study has to be deﬁned in the context of remote or co-

located collaborative learning. Second, the collabo-

ration quality assessment has to be considered to per-

sonalize, guide, or adapt learning or teaching. Studies

that were limited to the recommendation or the visual

representation of educational data and collaboration

indicators or models without explainability, collabo-

ration assistance or guidance were excluded. Finally,

adaptable systems that are based on manual system

conﬁguration by the human, or that deal with human-

agent collaboration were excluded. The ﬁnal list of

relevant publications consisted of 27 articles. We an-

alyzed the full text of these articles to address our re-

search questions, RQ1 and RQ2.

3 RESULTS

The results show that research in adaptive collabora-

tive learning systems based on the assessment of col-

laboration quality has increased over years. A grow-

ing interest has been particularly devoted to providing

a helping hand to stakeholders in co-located (or face-

to-face) collaboration.

3.1 Adaptation Approaches

The following adaptation approaches were suggested

to support collaborative learning: adaptive feedback

and scaffolding (30% of papers), personalized rec-

ommendation (25% of papers), adaptive dashboards

(21% of papers), AI assistants (14% of papers), as

well as adaptive visualizations of educational data

and collaboration indicators (10% of papers). Table

1. presents a brief description of the adaptation ap-

proaches used in the reviewed studies, while Table 2.

summarizes their main outcomes.

3.1.1 Adaptive Feedback and Scaffolding

Studies suggest that adaptive scaffolding can improve

students’ reﬂection on their learning and collabora-

CSEDU 2025 - 17th International Conference on Computer Supported Education

910

Table 1: Adaptation approaches of collaborative learning systems based on the assessment of collaboration quality.

Approach Description paper ID

Adaptive

feedback and

scaffolding

Adaptive scaffolding to support students’ reﬂection on their regulatory actions and prompts for the expert to support

group monitoring, and the evaluation of individual reﬂections.

(Splichal et al., 2018) 1

Adaptive generated textual process feedback for CSCL based on discourse indicators (Menzel et al., 2023) 2

Real-time adaptation of the system to support collaboration and generate appropriate action when a group is struggling

(Evans et al., 2019) 3

An intelligent tutoring system to support learning through step-by-step guidance, adaptive hints and feedback

(Yang et al., 2023) 4

Adaptive support for group formation and real-time feedback for reﬂection (Liang et al., 2023) 5

Adaptive collaboration script based on prompts and hints (Rau et al., 2017) 6

Orchestration scripts that detects everyday work situations in a workplace and suggest strategies

(collaborative activities) to attempt during workers’ interactions with each other

(Garg et al., 2023) 7

Collaboration script to regulate the distribution of participation (Strauss et al., 2023) 8

Personalized

recommendation

A teacher-facing orchestration tool based on recommendation of pairing students to work collaboratively

(Yang et al., 2023) 4

Recommendation of students who needs support and a visual explanation of collaboration indicators using a logical tree

(Anaya et al., 2016) 9

Recommendation of collaborative activities based on learning style using neural network model (Troussas et al., 2023) 10

Personalized recommendation of forum posts to promote collaboration (Echeverria et al., 2017) 11

Personalized recommendation of posts based on collaborative ﬁltering and k-nearest neighbors (Kasepalu et al., 2022) 12

Personalized recommendation of learning resources using a rule-based approach and a deep neural network model (Zheng et al., 2024) 13

Recommendation of roles based on the prediction of student group’s collaboration quality using

deep-learning models

(Som et al., 2021) 14

Adaptive Adaptive multimodal analytics dashboard with adaptive visualization of collaboration quality using a network graph (Praharaj et al., 2022) 15

dashboards Learning analytics dashboard with adaptive visual feedback for collaborative argumentation according to students’ needs (Han et al., 2021) 16

Multimodal analytics dashboard with real-time notiﬁcations of collaboration issues (Serrano Iglesias et al., 2021) 17

Multimodal analytics dashboard with adaptive visualizations (graphs) and automatic feedback about epistemic and social

aspects of collaboration

(Chen and Demmans, 2020) 18

Multimodal analytics dashboard with adaptive visualizations that summarize group indicators (Martinez-Maldonado et al., 2015) 19

Learning analytics dashboard to guide orchestration based on Epistemic Network Analysis and an alerting mechanism that

fagged critical moments in collaboration

(Amarasinghe et al., 2021) 20

AI assistant A chatbot with adaptive guidance and feedback on submitted assignments (Burkhard et al., 2022) 21

A conversational agent to provide adaptive scaffold for students based on APT(Academically Productive Talk) and an

orchestration support for teachers using a concept map and a classiﬁcation of students activities

(Tegos et al., 2015) 22

AI assistant for classroom orchestration based on collaboration problems modeling (Eryilmaz et al., 2021) 23

AI assistant for collaborative learning using various explainable ML methods (Tomic et al., 2023) 24

Adaptive A feedback tool based on visual analytics using storytelling of the learner model and a rule-based system (Martinez-Maldonado et al., 2020) 25

visualizations Generated visualizations of of critical sub-processes in teams’ activity using a rule-based system (Venegas-Reynoso et al., 2018) 26

Visual analytics using social network analysis to support learning, ﬁnd aspects in need of improvement, and guide

an informed intervention

(Saqr et al., 2018) 27

tion processes, while metacognitive prompts help ex-

perts monitor group and individual progress. Adap-

tive feedback and scaffolding strategies were gener-

ally suggested in different learning contexts to en-

hance knowledge acquisition (Menzel et al., 2023),

literacy (Liang et al., 2023), and communication skills

(Strauss et al., 2023). The assessment of collabora-

tion quality has been primarily used to reﬁne collabo-

rative scripts, support decision-making for orchestra-

tion and tutoring, and identify appropriate prompts to

enhance learning outcomes and increase group aware-

ness (Rau et al., 2017). By providing adaptive feed-

back and guiding students through scripts, hints, and

prompts, these strategies signiﬁcantly improved indi-

vidual learning outcomes (Splichal et al., 2018), sat-

isfaction (Strauss et al., 2023), collaboration skills

(Menzel et al., 2023; Garg et al., 2023), perceived

collaboration usefulness (Rau et al., 2017), and group

awareness (Evans et al., 2019).

However, despite the positive impact of adap-

tive feedback and scaffolding strategies on individual

learning and collaborative skills, a certain amount of

concerns have to be addressed. This includes for in-

stance the limited metrics and features considered to

assess collaboration quality (Splichal et al., 2018), the

dependence of the adaptation strategy on a particular

learning context, and the difﬁculty to considering stu-

dents’ soft skills such as self-regulation and creative

thinking (Yang et al., 2023). Some authors also high-

lighted the need for studies that identify teacher inter-

action patterns to lend a hand to classroom orchestra-

tion (Garg et al., 2023).

3.1.2 Personalized Recommendation

Personalized recommendation systems have been

proposed generally in the context of collaborative

problem solving (Zheng et al., 2024; Som et al.,

2021), project-based learning (Christos Troussas and

Voyiatzis, 2023), as well as creative and reﬂective

learning (Eryilmaz et al., 2021). They were based

on recommending collaborative activities (Chris-

tos Troussas and Voyiatzis, 2023), discussion fo-

rum posts (Kasepalu et al., 2022), learning resources

(Zheng et al., 2024), and roles (Som et al., 2021).

Personalized recommendation has also been deployed

in classroom orchestration to assist group formation

(Yang et al., 2023) and suggest assistance to stu-

dents who face collaboration and learning difﬁcul-

ties (Anaya et al., 2016). Research studies showed

the usefulness of personalized recommendation sys-

tems in increasing for instance students’ participation

in collaborative activities (Echeverria et al., 2017),

creativity, group awareness (Kasepalu et al., 2022;

Anaya et al., 2016), as well as individual learning out-

comes (Christos Troussas and Voyiatzis, 2023).

However, studies raised different concerns about

the consideration of individual students’ skills such

A Systematic Literature Review of Adaptive Learning Systems Based on the Assessment of Collaboration Quality

911

Table 2: Summary of samples, research designs, and main studies outcome following the adaptation approaches, (n) the

number of students (m) the number of teachers.

Approach Sample Research design Study ﬁndings ID

Adaptive feedback

and scaffolding

n=48 Post analysis of a questionnaire and posts Improvement of students who augmented their scripts 1

m=5

n=408 Experimental group (generated feedback) vs. control group (simple feedback) Generated personalized feedback were helpful 2

n=11 Experimental group (adaptation) vs. control groups (without adaptation) The effectiveness of the detection of collaboration issues 3

n=199 Post analysis of the learning system traces and interviews Utility of dynamic transitions between activities in personalizing learning 4

n=25 Post study analysis of the learning system traces The grouping strategy can predict group performance 5

n=69 Experimental group (adaptive script) vs. a control group (without adaptation) Improvement of learning outcomes and perceived collaboration 6

n=17 Pre-post tests and post-study interviews Situated scripts usefulness in supporting students 7

n=150 Experimental group (adaptive scripts) vs. an awareness tool vs. control group Positive impact on students’ satisfaction 8

Personalized n=23 A survey to evaluate the user experience The explanation enhanced collaboration issues perception 21

recommendation n=80 Experimental group (personalized recommendation) vs. control group A high degree of pedagogical affordance and the positive impact 10

(simple recommendation) on learning

n=57 Experimental group (personalized recommendation) vs. control group The effectiveness of predicting students preferences and increasing 11

(other recommendation algorithms) participation

n=70 Experimental group (personalized recommendation) vs. control group Increase the number of messages, cultivated a sense of collective 12

(without recommendation) agency and creativity

n=135 Between-subject pre-post test design and interviews Improved socially shared regulated behaviors 13

NA Post study analysis of the performance of representations of students roles The effectiveness and accuracy of the model of collaboration quality 14

Adaptive n=14 Post analysis of team discourses and log data Positive impact of inﬂuential role-role interactions on collaboration 15

dashboards n=22 A within-subject design to evaluate two conditions: personalized dashboard The improvement of participation and argumentation 16

m=88 and without dashboard

NA A use case of the system integration The adoption of MD for the smart learning environment 17

n=15 Post-study questionnaire and interviews The dashboard enhanced students’ post-hoc reﬂection about their 18

m= 1 Post-hoc activity: writing reﬂection activity

n=150 Post-study interviews and questionnaires Presented the teachers perspectives and issues to orchestrate 19

m=4 a multi-tabletop classroom

m=6 A within-subject design to evaluate 3 conditions: guidance, mirroring, and The guidance enabled teachers to perform more, orchestration 20

control condition actions, interactions, and announcements

AI assistant n=11 Post evaluation of the chatbot through structured interviews The perceived usefulness of the guidance 9

n=43 Experimental group (with agent) vs. control group (without agent) Improved engagement in a productive dialogue, reasoning, and 22

argumentation

m=20 Wizard-of-Oz protocol founded on interviews The utility of the virtual assistant in co-regulation understanding 23

n=252 Post study interviews and analysis of methods performance and explainability Fuzzy rules and decision trees combined with neural networks 24

m=6 make best performance and explainability

Adaptive n=44 Qualitative studies based on interviews Assistance of student reﬂection on their activity, stress management, 25

visualizations m=8 and errors made

n=60 Post study structured interviews The meaningfulness of the generated visualizations

n=164 A repeated measurement design (pre-intervention vs. Post-intervention) Enhancement of student-student and teacher-student interactions 27

as critical thinking and social skills (Zheng et al.,

2024; Kasepalu et al., 2022). There is also still a

need for studies to evaluate the recommendation qual-

ity (Som et al., 2021) and its effect on collabora-

tive learning and teaching. Several opportunities can

be discussed, including the consideration of multi-

ple aspects and metrics to assess collaboration quality

(Christos Troussas and Voyiatzis, 2023; Som et al.,

2021), as well as improving the explainability of rec-

ommendation methods (Som et al., 2021), especially

in co-location collaboration. In fact, due to the signif-

icant amount of real-time interaction data, both teach-

ers and students may have difﬁculty understanding the

system’s recommendations and how they can inﬂu-

ence the collaboration process and students’ perfor-

mance.

3.1.3 Adaptive Dashboards

Adaptive dashboards have been used to support stu-

dents in different contexts, including collaborative

problem solving, co-located project-based learning

(Echeverria et al., 2017; Praharaj et al., 2022), and

argumentation (Han et al., 2021). They were also

used to guide classroom orchestration by altering col-

laboration issues (Amarasinghe et al., 2021). More-

over, adaptive multimodal analytics dashboards were

used to reduce the cognitive load of analyzing col-

laboration quality indicators that relied on multiple

data modalities (Praharaj et al., 2022; Chen and Dem-

mans, 2020). Studies showed the effectiveness of

adaptive dashboards in improving learners’ participa-

tion and argumentation skills (Han et al., 2021), re-

ﬂective learning (Chen and Demmans, 2020), collab-

oration skills (Praharaj et al., 2022) as well as teach-

ers’ orchestration actions (Amarasinghe et al., 2021;

Serrano Iglesias et al., 2021)

However, adaptive dashboards often depend on

particular structures of learning activities and sce-

narios (Amarasinghe et al., 2021; Han et al., 2021).

Moreover, as reported by some studies, mining tem-

poral interaction patterns for real-time explanation of

indicators (Martinez-Maldonado, 2019) and consid-

ering social and epistemic aspects of collaboration to

adapt the dashboard (Praharaj et al., 2022) is still chal-

lenging. Research opportunities also include the eval-

uation of the effect of dashboards in different learn-

ing contexts (Amarasinghe et al., 2021; Han et al.,

2021) and how they can be adapted to improve self-

regulation skills.

3.1.4 AI Assistant

Adaptive virtual agents and chatbots were generally

used in the context of collaborative project-based

learning (Tomic et al., 2023), argumentation (Tegos

CSEDU 2025 - 17th International Conference on Computer Supported Education

912

et al., 2015), and literacy. The studies showed the use-

fulness of this approach in improving students’ col-

laboration (Anaya et al., 2016; Eryilmaz et al., 2021),

engagement, and learning (Tegos et al., 2015). Adap-

tive guidance using virtual agents and chatbots mostly

focused on individual support of students that stands

on a particular learning context and activities. There

is also a lack of studies that consider socio-emotional

indicators of collaboration to adapt the guidance in

the learning system (Liang et al., 2023).

3.1.5 Adaptive Visualizations

Many studies have reported issues with the high

cognitive load required to interpret real-time indica-

tors, particularly in co-located collaboration (Gavse-

vic et al., 2015b). To address this, visual analytics

have been adopted to help stakeholders gain insights

through data ﬁltering and predictive modeling (Va-

trapu et al., 2011). However, despite aiding data ex-

ploration, these methods do not always ensure mean-

ingfulness or efﬁciency in decision-making (Barria-

Pineda et al., 2023). The analysis of collaboration

processes may require complex datasets, while ana-

lytics methods, especially black-box machine learn-

ing models, often lack explainability.

Some studies dealt with tailoring visual represen-

tation of educational data and collaboration indica-

tors (Yang et al., 2023; Saqr et al., 2018). Studies

showed the positive impact of this approach in frost-

ing students’ reﬂection on their activity (Martinez-

Maldonado et al., 2020), communication (Saqr et al.,

2018), and group awareness (Venegas-Reynoso et al.,

2018). These studies often relied on rule-based sys-

tems, however, the generation of rules has been seen

as challenging to adapt and to capture the intent of

team members (Venegas-Reynoso et al., 2018).

3.2 Collaboration Quality Assessment

Methods and Indicators

Research on adaptive collaborative learning systems

has extensively explored various methods for assess-

ing collaboration quality, leveraging learning analyt-

ics (LA, 44%), multimodal analytics (MA, 33%),

text analysis (TA, 26%), and social network analy-

sis (SNA, 18%). Collaboration quality assessment

using LA was generally used to adapt feedback and

personalize recommendations. Common methods in-

clude statistical analysis (Menzel et al., 2023), su-

pervised learning (Eryilmaz et al., 2021), neural net-

works (Christos Troussas and Voyiatzis, 2023) and

deep learning (Zheng et al., 2024) using using logs

and chat data.

Numerous studies also focused on TA methods

based on students’ forum posts, questions, and tran-

scribed speech to tailor feedback and scaffolding.

Techniques such as thematic similarity (Chen and

Demmans, 2020) and question-answer analysis (Rau

et al., 2017) enable a detailed examination of com-

munication patterns, while interaction analysis helps

reveal the nuances of group discourse (Menzel et al.,

2023).

Some studies used SNA to model group interac-

tions and assess collaboration (Eryilmaz et al., 2021)

to support personalized recommendations (Chen and

Demmans, 2020) and adapt visualizations of learning

and collaboration indicators (Saqr et al., 2018).

Finally, recent studies extend the analysis by in-

tegrating diverse data types such as logs, audio, ges-

tures, and tabletops touch actions using MA methods.

They generally relied on rule-based systems (Evans

et al., 2019), predictive analyses of interactions (Ser-

rano Iglesias et al., 2021), and deep learning models

(Som et al., 2021). By capturing multiple modalities,

MA provides richer insights into both individual and

group dynamics.

Across these analytics approaches, research stud-

ies identiﬁed multiple indicators of collaboration

quality, especially knowledge contribution (52%),

task participation (30%), collaboration quality mod-

els (26%), and group performance (19%).

Knowledge contribution has been generally as-

sessed through the quality of messages, posts, and an-

swers, with metrics such as keyword usage and activ-

ity types (Serrano Iglesias et al., 2021). This analy-

sis ties into task participation, where the frequency of

actions, speech inputs, and forum interactions high-

lights the level of engagement (Martinez-Maldonado,

2019). Some studies suggested a collaboration model

that depends on connection ratios, readability scores,

and machine learning models such as support vector

machines and decision trees (Kasepalu et al., 2022).

Other studies were limited to the evaluation of group

performance as an indicator of collaboration qual-

ity. It has been measured through computational

models of problem-solving, task completion metrics,

and tutor evaluations (Rau et al., 2017). Finally,

some studies incorporate individual performance met-

rics, learning styles, and emotional responses, utiliz-

ing psycholinguistic attributes and multimodal data

to capture the affective dimensions of collaboration

(Martinez-Maldonado et al., 2020).

To sum up, research studies on adaptive learn-

ing systems relied on different analytics methods to

capture and assess collaboration quality under differ-

ent group settings. However, despite some similar

collaboration indicators revealed in several studies,

A Systematic Literature Review of Adaptive Learning Systems Based on the Assessment of Collaboration Quality

913

their low-level metrics were often different and de-

pend on multiple factors, including the learning con-

text of studies and the collaboration scenario. In addi-

tion, some metrics were interpreted differently by re-

search studies as they have been used to compute dif-

ferent collaboration indicators. For instance, the qual-

ity of learners’ messages and posts measured often

by the frequency of keywords and messages length

has been used to assess the students’ contribution to

knowledge construction (Strauss et al., 2023; Praharaj

et al., 2022) as well as to evaluate participation and

collaboration quality (Eryilmaz et al., 2021).Finally,

despite the use of collaboration quality to adapt learn-

ing systems, there is a lack of studies identifying and

evaluating the most relevant collaboration indicators

that can positively inﬂuence stakeholders’ decision-

making.

4 DISCUSSION

This review reveals that research studies proposed

various adaptation approaches based on collabora-

tion quality assessment to support collaborative learn-

ing and teaching, including adaptive feedback and

scaffolds (AF), personalized recommendations (PR),

adaptive dashboards (AD), adaptive visualizations

(AV), and AI assistants (AIA). Despite their signif-

icant positive impact on improving students’ learn-

ing outcomes and skills, as revealed in the reviewed

studies, these approaches were developed in varied

learning contexts and utilized different collaboration

quality assessment methods and indicators. Some ap-

proaches, such as AF, were suggested across different

learning contexts. In contrast, other approaches such

as AD and PR were generally used for collaborative

problem solving and co-located project-based learn-

ing.

All proposed adaptation techniques have demon-

strated their usefulness in enhancing group aware-

ness and individual learning outcomes. Studies have

shown the effectiveness of PR, AD, and AIA in boost-

ing student participation in collaborative activities,

while AF, AD, and AV were found to improve stu-

dents’ collaboration and communication skills. AD

and AV were particularly effective in fostering stu-

dents’ reﬂection on their activities. Additionally, PR

was found effective in enhancing students’ creativity,

and AD in improving teachers’ orchestration actions.

Although this review can help identify and com-

pare relevant adaptation strategies in certain learn-

ing contexts, it does not allow for deﬁnitive conclu-

sions about when adaptation techniques are valuable

in other learning contexts or under further collabora-

tion indicators. In fact, despite the extensive literature

on collaboration quality assessment, there is a notable

lack of well-established frameworks for assessing the

collaboration process using standardized metrics and

measures. Additionally, there is a shortage of studies

that identify the most relevant collaboration indicators

inﬂuencing stakeholders and the decision-making of

adaptive systems.

In addition, AF has been seen in the literature

as instructional strategies that aim to support col-

laborative learning (Gavsevic et al., 2015a). How-

ever, there is a lack of research on how to effec-

tively adjust scaffolds by addressing collaboration is-

sues. In addition, to improve students’ reﬂection on

their actions and collaboration process, other stud-

ies have proposed PR to reduce the cognitive load

of processing collaboration indicators in real-time.

In recent years, research opportunities have been

geared towards the explainability of recommenda-

tions (Martinez-Maldonado et al., 2020). In the case

of co-located collaboration, the stakeholders can be

overloaded by the different recommendations of the

system and need assistance to evaluate the impact of

their actions and decisions on the system decisions.

Moreover, there is a lack of studies that consider how

the assessment of collaboration quality can improve

the decisions of the recommendation system. Finally,

despite the potential of dashboards in promoting col-

laborative learning and teaching, they still lack adapt-

ability by taking into account stakeholders’ needs and

differences (Han et al., 2021) as well as teacher or-

chestration strategies (Tomic et al., 2023).

Many research opportunities can be suggested to

promote collaborative learning, including the devel-

opment of new infrastructures of collaborative learn-

ing systems that consider multiple learning contexts

and collaboration scenarios. There is also a need to

develop standardized instruments, measures, indica-

tors and models to assess collaboration quality dimen-

sions while improving their explainability and mean-

ingfulness. This can not only help inform actionable

insights but also improve adaptation strategies. Con-

sidering individual students’ differences can also con-

tribute to improving their skills acquisition, including

regulation, social, creative, and critical thinking skills

(Splichal et al., 2018; Yang et al., 2023).

5 CONCLUSION

The review highlights the important role of adaptive

collaborative learning systems, based on collabora-

tion quality assessment, in improving students’ learn-

ing outcomes and skills. Despite increasing research

CSEDU 2025 - 17th International Conference on Computer Supported Education

914

in this area, there is a need for standardized tools and

infrastructures to evaluate collaboration quality across

different contexts. Adapting learning systems by con-

sidering students’ collaboration quality, differences,

and soft skills can boost engagement and productiv-

ity. The review identiﬁes opportunities for improv-

ing the explainability of collaboration models and tai-

loring learning systems to students’ needs. However,

limitations include restricted database inclusion due

to the large number of papers and some access limi-

tations. It also focused exclusively on studies in the

context of education conducted after 2015.

Finally, this review provides insights about adap-

tation in remote and co-located collaborative learn-

ing. Practical implications of this review analysis for

educators and system designers include the considera-

tion of different collaboration quality metrics that are

independent from a particular collaboration scenario

and learning activities structures. This also includes

examining the impact of collaboration indicators on

stakeholders’ decision-making to determine adapta-

tion strategies that better address the needs of learners

and teachers, as well as targeted competencies.

REFERENCES

Amarasinghe, I., Hernandez-Leo, D., and Ulrich Hoppe, H.

(2021). Deconstructing orchestration load: comparing

teacher support through mirroring and guiding. Inter-

national Journal of Computer-Supported Collabora-

tive Learning, 16(3):307–338.

Anaya, A. R., Luque, M., and Peinado, M. (2016). A visual

recommender tool in a collaborative learning experi-

ence. Expert Systems with Applications, 45:248–259.

Barria-Pineda, J., Akhuseyinoglu, K., and Brusilovsky, P.

(2023). Adaptive navigational support and explain-

able recommendations in a personalized programming

practice system. In Proceedings of the 34th ACM Con-

ference on Hypertext and Social Media, pages 1–9.

Association for Computing Machinery.

Brusilovsky, P., Somyurek, S., Guerra, J., Hosseini, R.,

Zadorozhny, V., and Durlach, P. J. (2015). Open

social student modeling for personalized learning.

IEEE Transactions on Emerging Topics in Computing,

4(3):450–461.

Burkhard, M., Seufert, S., Cetto, M., and Handschuh, S.

(2022). Educational chatbots for collaborative learing:

Results of a design experiment in a middle school. In

19th International Conference on Cognition and Ex-

ploratory Learning in Digital Age (CELDA 2022). In-

ternational Association for Development of the Infor-

mation Society, ERIC.

Chen, Z. and Demmans, C. (2020). Csclrec: Personal-

ized recommendation of forum posts to support socio-

collaborative learning. In Proceedings of The 13th In-

ternational Conference on Educational Data Mining

(EDM 2020), July 10 - 13, EDM2020, page 364–373.

International Association for Development of the In-

formation Society, ERIC.

Christos Troussas, Filippos Giannakas, C. S. and Voyiatzis,

I. (2023). Collaborative activities recommendation

based on students’ collaborative learning styles us-

ing ann and wsm. Interactive Learning Environments,

31(1):54–67.

Echeverria, V., Martinez-Maldonado, R., Chiluiza, K., and

Buckingham Shum, S. (2017). Dbcollab: Automated

feedback for face-to-face group database design. In

for Computers in Education, A.-P. S., editor, Proceed-

ings of the 25th International Conference on Comput-

ers in Education, ICCE 2017, pages 56–165. Associ-

ation for Computing Machinery.

Eryilmaz, E., Thoms, B., Ahmed, Z., and Lee, K.-H. (2021).

Effects of recommendations on message quality and

community formation in online conversations. Edu-

cation and information technologies, 26:49–68.

Evans, A., Davis, K., and Wobbrock, J. (2019). Adap-

tive support for collaboration on tabletop computers.

In 13th International Conference on Computer Sup-

ported Collaborative Learning (CSCL) 2019. Interna-

tional Society of the Learning Sciences.

Garg, K., Gergle, D., and Zhang, H. (2023). Orchestra-

tion scripts: A system for encoding an organization’s

ways of working to support situated work. In Proceed-

ings of the 2023 CHI Conference on Human Factors

in Computing Systems, CHI ’23, New York, NY, USA.

Association for Computing Machinery.

Gavsevic, D., Adesope, O., Joksimovic, S., and Kovanovic,

V. (2015a). Externally-facilitated regulation scaffold-

ing and role assignment to develop cognitive presence

in asynchronous online discussions. The internet and

higher education, 24:53–65.

Gavsevic, D., Dawson, S., and Siemens, G. (2015b). Let’s

not forget: Learning analytics are about learning.

TechTrends, 59:64–71.

Han, J., Kim, K. H., Rhee, W., and Cho, Y. H. (2021).

Learning analytics dashboards for adaptive support in

face-to-face collaborative argumentation. Computers

and Education, 163:104041.

Hocine, N. and Sehaba, K. (2024). A systematic review

of online personalized systems for the autonomous

learning of people with cognitive disabilities. Human–

Computer Interaction, 39(3-4):174–205.

Kasepalu, R., Prieto, L. P., Ley, T., and Chejara, P.

(2022). Teacher artiﬁcial intelligence-supported peda-

gogical actions in collaborative learning coregulation:

A wizard-of-oz study. Frontiers in Education, 7:15.

Liang, C., Horikoshi, I., Majumdar, R., Flanagan, B., and

Ogata, H. (2023). Towards predictable process and

consequence attributes of data-driven group work. Ed-

ucational Technology and Society, 26(4):90–103.

Lima, Y. O. and de Souza, J. M. (2017). The future of

work: Insights for cscw. In 2017 IEEE 21st Inter-

national Conference on Computer Supported Cooper-

ative Work in Design (CSCWD), pages 42–47. IEEE,

IEEE.

A Systematic Literature Review of Adaptive Learning Systems Based on the Assessment of Collaboration Quality

915

Lykourentzou, I., Antoniou, A., Naudet, Y., and Dow,

S. P. (2016). Personality matters: Balancing for

personality types leads to better outcomes for crowd

teams. In Proceedings of the 19th ACM Conference

on Computer-Supported Cooperative Work and Social

Computing, pages 260–273. Association for Comput-

ing Machinery.

Martinez-Maldonado, R. (2019). A handheld classroom

dashboard: Teachers’ perspectives on the use of real-

time collaborative learning analytics. International

Journal of Computer-Supported Collaborative Learn-

ing, 14:383–411.

Martinez-Maldonado, R., Echeverria, V., Fernandez Nieto,

G., and Buckingham Shum, S. (2020). From data to

insights: A layered storytelling approach for multi-

modal learning analytics. In Proceedings of the 2020

CHI Conference on Human Factors in Computing Sys-

tems, page 1–15, New York, NY, USA. Association

for Computing Machinery.

Martinez-Maldonado, R., Yacef, K., and Kay, J. (2015).

Tscl: A conceptual model to inform understanding

of collaborative learning processes at interactive table-

tops. International Journal of Human-Computer Stud-

ies, 83:62–82.

Menzel, L., Gombert, S., Weidlich, J., Fink, A., Frey,

A., and Drachsler, H. (2023). Why you should give

your students automatic process feedback on their

collaboration: Evidence from a randomized experi-

ment. In European Conference on Technology En-

hanced Learning ECTEL, pages 198–212. Springer.

Neumayr, T. and Augstein, M. (2020). A systematic re-

view of personalized collaborative systems. Frontiers

in Computer Science, 2:562679.

Page, M., McKenzie, J., Bossuyt, P., Boutron, I., and Hoff-

mann, T. (2021). The prisma 2020 statement: an up-

dated guideline for reporting systematic reviews. Sys-

tematic reviews, 10(1):1–11.

Praharaj, S., Scheffel, M., Schmitz, M., Specht, M., and

Drachsler, H. (2022). Towards collaborative conver-

gence: Quantifying collaboration quality with auto-

mated co-located collaboration analytics. In 12th In-

ternational Learning Analytics and Knowledge Con-

ference, LAK22, page 358–369, New York, NY, USA.

Association for Computing Machinery.

Rau, M. A., Bowman, H. E., and Moore, J. W. (2017). An

adaptive collaboration script for learning with multi-

ple visual representations in chemistry. Computers

and Education, 109:38–55.

Saqr, M., Fors, U., Tedre, M., and Nouri, J. (2018). How

social network analysis can be used to monitor online

collaborative learning and guide an informed interven-

tion. PLOS ONE, 13(3):1–22.

Serrano Iglesias, S., Spikol, D., Bote Lorenzo, M. L., and

Ouhaichi, H. (2021). Adaptable smart learning en-

vironments supported by multimodal learning analyt-

ics. In Proceedings of the LA4SLE 2021 Workshop:

Learning Analytics for Smart Learning Environments,

pages 24–30.

Som, A., Kim, S., Lopez-Prado, B., Dhamija, S., Alozie, N.,

and Tamrakar, A. (2021). Automated student group

collaboration assessment and recommendation system

using individual role and behavioral cues. Frontiers in

Computer Science, 3:16.

Splichal, J. M., Oshima, J., and Oshima, R. (2018). Regu-

lation of collaboration in project-based learning medi-

ated by cscl scripting reﬂection. Computers and Edu-

cation, 125:132–145.

Strauss, S., Tunnigkeit, I., Eberle, J., vom Bovert, L.,

Avdullahu, A., Schmittchen, M., and Rummel, N.

(2023). Differential effects of a script and a group

awareness tool on the acquisition of collaboration

skills. In Proceedings of the 16th International Con-

ference on Computer-Supported Collaborative Learn-

ing CSCL, pages 75–82. International Society of the

Learning Sciences.

Tegos, S., Demetriadis, S., and Karakostas, A. (2015). Pro-

moting academically productive talk with conversa-

tional agent interventions in collaborative learning set-

tings. Computers and Education, 87:309–325.

Tomic, B. B., Kijevcanin, A. D., sevarac, Z. V., and Jo-

vanovic, J. M. (2023). An ai-based approach for grad-

ing students’ collaboration. IEEE Transactions on

Learning Technologies, 16(3):292–305.

Vatrapu, R., Teplovs, C., Fujita, N., and Bull, S. (2011). To-

wards visual analytics for teachers’ dynamic diagnos-

tic pedagogical decision-making. In Proceedings of

the 1st international conference on learning analytics

and knowledge, pages 93–98. Association for Com-

puting Machinery.

Venegas-Reynoso, A., Gaytan-Lugo, L. S., and Martinez-

Maldonado, R. (2018). From touches to teamwork

constructs: Towards automatically visualising collab-

oration processes. In Proceedings of the 7th Mexican

Conference on Human-Computer Interaction, New

York, NY, USA. Association for Computing Machin-

ery.

Vogel, F., Wecker, C., Kollar, I., and Fischer, F. (2017).

Socio-cognitive scaffolding with computer-supported

collaboration scripts: A meta-analysis. Educational

Psychology Review, 29:477–511.

Yang, K. B., Echeverria, V., Lu, Z., Mao, H., Holstein, K.,

Rummel, N., and Aleven, V. (2023). Pair-up: Pro-

totyping human-ai co-orchestration of dynamic tran-

sitions between individual and collaborative learning

in the classroom. In Proceedings of the 2023 CHI

Conference on Human Factors in Computing Systems,

number 453 in CHI ’23, New York, NY, USA. Asso-

ciation for Computing Machinery.

Zamecnik, A., Kovanovic, V., Grossmann, G., Joksimovic,

S., Jolliffe, G., Gibson, D., and Pardo, A. (2022).

Team interactions with learning analytics dashboards.

Computers and Education, 185:104514.

Zheng, L., Fan, Y., Gao, L., Huang, Z., Chen, B., and Long,

M. (2024). Using ai-empowered assessments and per-

sonalized recommendations to promote online collab-

orative learning performance. Journal of Research on

Technology in Education, pages 1–27.

CSEDU 2025 - 17th International Conference on Computer Supported Education

916