Towards a Dynamic Visualization of Online Collaborative Learning

Malik Koné

1,2

, Madeth May

and Sébastien Iksal

LIUM, Le Mans Université, Le Mans, France

LARIT, INP-HB, Yamoussoukro, Cote d’Ivoire

Keywords:

Collaborative Learning, Visualization, MOOC.

Abstract:

Socio-constructivism and connectivism theories pinpoint the importance of collaboration for learning. Ne-

vertheless, the online social interactions underlying the collaboration processes are still not well understood.

As a result, learning designers have difﬁculties creating effective collaborative activities in Massive Open On-

line Courses (MOOCs). As for online learners, they are often isolated and require a lot of self-regulation to

succeed. The research effort presented in this paper covers a review of visualization techniques supporting

the online collaborative learning process. Our ﬁndings show that some visualizations have the potential to

develop the learners’ reﬂexivity. Therefore, we give an overview of collaboration importance and how it could

be enhanced with such visualizations. Our goal is to identify a new approach to visualize learners’ activities

in MOOCs, while supporting collaboration and self-regulation.

1 INTRODUCTION

For socio-constructivism and connectivism theories,

people learn within a group when they exchange and

confront their ideas (Bandura and McClelland, 1977;

Siemens, 2014). Discussing opinions, arguing, ex-

pressing ideas help learners to evaluate and consoli-

date their knowledge. While doing so, they explore

their Zone of Proximal Development (ZPD) and can

gradually shift it towards new knowledge (Vygotskij

et al., 1985).

Online collaboration analysis is based on observa-

ble electronic interactions between learners. Graphs

are often used to visualize theses interactions. A set

of nodes represents the agents and links between no-

des idealize the interactions, but collaboration is also

a process and it cannot be fully captured with a sta-

tic image. Most studies on collaborative learning give

an overview of the collaboration at the end of the le-

arning session. They are intended for instructors or

researchers, to help them reﬂect on the course design

and to support future pedagogical interventions. In

our case, we also want to provide learners with feed-

back. Yet, if the latter comes at the end of the les-

son, it will be too late. Hence, we need to visualize

the collaboration as the course unfolds. Also, we will

look at visualizations to ensure that temporal infor-

mation, from the collaborative activities, is not lost.

Animations, timelines or interactive visualizations are

among the most effective ways of achieving this. We

refer to them as dynamic visualizations. In the follo-

wing section, we deﬁne collaboration and discuss the

challenges to model collaborative learning. Then we

illustrate the change in the Learning Analytics (LA)

goal, moving from modelling to supporting learning.

In the third section, we will detail the visualization

concept and why it is important in the new LA para-

digm. Finally, a review of dynamical visualizations

techniques, supporting learner’s collaboration, is pro-

posed in the fourth section.

2 COLLABORATIVE LEARNING

Roschelle et Teasley (1995) cited in Dillenbourg

(1999) deﬁne collaboration as “a coordinated syn-

chronous activity born from the persistent will to

share a common perception of a problem”. Dillen-

bourg perfects this deﬁnition by specifying that colla-

boration occurs between people roughly equal in so-

cial status. There is no collaboration if one is an in-

structor and the other his student.

We call online collaborative learning or collabo-

rative learning the work and process of learners in-

teracting online, supporting each other to achieve a

precise common learning goal. For example, in a

MOOC, it could be the grouped activities of online

Koné, M., May, M. and Iksal, S.

Towards a Dynamic Visualization of Online Collaborative Learning.

DOI: 10.5220/0006687202050212

In Proceedings of the 10th International Conference on Computer Supported Education (CSEDU 2018), pages 205-212

ISBN: 978-989-758-291-2

205

learners interacting and coordinating over a forum to

help each other understand the subject matter or to get

good grades at the end of course.

Online collaborative learning is based on net-

works, which we call online learning networks. We

deﬁne an online learning network as the set of indi-

viduals and their interactions in a MOOC. Mathema-

tical tools from the network science provide models

to study online learning networks. Graphs for exam-

ple, can have nodes representing individuals (or so-

cial entities), and links between them can symbolize

the social interactions. But graphs are not enough to

model online collaborative learning. This endeavour

has challenges that we will present now.

2.1 Challenges in Modelling

Collaborative Learning

Human have a unique talent to learn. We infer efﬁ-

ciently generalizable knowledge from a few speciﬁc

examples (Tenenbaum et al., 2006). Artiﬁcial Intelli-

gences (AIs) based on neural networks, despite their

biological inspiration and success – driving cars, win-

ning world class go players– cannot explain how hu-

mans learn. Lake et al. (2016) argue that human le-

arning models should be causal models of the world

able to support explanation and understanding. These

models should learn-to-learn and be reﬂexive. Baye-

sian models of cognition may go in that direction but

they are not yet able to implement the subtle and intui-

tive psycho-sociological phenomena described by the

theory of mind and at play in collaborative learning.

On the other hand, traditional sociological appro-

aches of collaboration base themselves on ﬁeld obser-

vations and questionnaires, but contingencies in time

and human resources usually limit the experiments to

small sets of individuals. In contrast, online collabo-

rative learning studies are cheap. It is an opportunity

to gather data about behaviours of humans learning. It

could help elaborate a model of online collaborative

learning. But, with this abundance of information,

data scientists need to carefully ground their approa-

ches in learning theories. The exploration of correla-

tions and predictive models may dominate the search

for causal relationships and explanations. Kirschner

(2016) alerts that the latter should be the essence ad-

vancing learning sciences.

Learning Analytics, as the practice of informing

learning with statistical analysis (Kirschner, 2016),

may not yet be able to produce stand alone collabo-

rative learning models but it could effectively support

it, says Baker (2016).

2.2 Supporting Collaborative Learning

with LA

In distant education, learners and instructors are iso-

lated from the cohort. They hardly know each other.

Collaborating is therefore difﬁcult.

Collaboration does not spur on its own says Dil-

lenbourg (2002). Grouping learners and letting them

loose will likely fail to induce a successful collabora-

tion. Agreeing on a common goal, coordinating acti-

ons during a substantial time need a coercive force

that is usually the role of the instructor. In MOOCs,

the number of learners is overwhelming. They are of-

ten geographically distant, may live in different time

zones and have different learning objectives. In this

context, it is not possible for an instructor or even a

team of instructors to manage the collaborative lear-

ning efﬁciently.

One option is to help automate the collaboration

process with scripts. Scripts are sets of instructions,

which in this context tell learners how to collabo-

rate. They may deﬁne criteria for team building, spe-

cify the role of each member and the means offered

for collaboration. Scripts have been LA scientists’

attempts to model collaboration, but over-scripting

sometimes lead learners in clumsy situation (Dillen-

bourg, 2002). This is partly why Baker (2016) argues

for a paradigm change. Instead of seeking to create

“intelligent tutors” (or AI), which fail to grasp the

subtle psychological complexity of human learning,

he says provocatively that the community should aim

for “stupid systems”... but stupid systems expert in

intelligent human support. Following his logic, let’s

review some LA tools supporting collaborative lear-

ning.

2.2.1 LA Tools

Soller et al. (2005) compile a list of articles evaluating

the tools effect on group learning and categorize the

different approaches relative to the degree of control

devoted to the Learning Management System (LMS).

Mirroring Tools: display gathered data almost in

raw format. For example, connections’ numbers,

posts’ percentage by such or such user.

Meta-cognitive Tools: present elaborated measures

of higher order cognitive functions. Motivation,

mastery level are often compound measures built

on top of lower level indicators, such as grades,

post, # of friend.

Guiding Tools: propose actions to take. This can be

a learner to help, a keyword to deﬁne a post, the

next learning resource to access.

CSEDU 2018 - 10th International Conference on Computer Supported Education

206

Table 1: We classify collaborative learning articles in two groups : visualization and indicator based studies. Visualization

based studies, focus on the impact of visualizations in online collaborative setting. Indicator based studies focus on ways to

measure online collaboration and its relation with performance. Different social contexts are covered, small groups (2 or 3

persons), medium (a few tens), big (a few hundreds) and massive (a few thousands).

Indicator based studies

Papers Social

context

Audience What is measured Expected output

Duque et al. (2015) Small Instructors Interactions Collective work

Rehm et al. (2015) Medium Researchers Hierarchical position Activity & Performance

Hommes et al. (2012) Big Researchers Social network structure,

learners’ motivation &

prio-performance

Performance

Wang et al. (2016) Massive Researchers Interactions Performance

Tempelaar et al. (2015) – Researchers Online activity & Survey Performance

Visualization based studies

Papers Social

context

Audience What is visualized Expected output

Anaya et al. (2016) Small Instructors

& learners

Decision Tree & Collabo-

ration

Learners’ activity

van Leeuwen et al. (2014)

Small Instructors Agreement level & Con-

tributions

Teacher’ activity

Medina et al. (2016) Medium Learners Activity, Grades & Usabi-

lity

Users’ satisfaction

Yousuf and Conlan (2015) – Learners Engagement Learner’s satisfaction

Lonn et al. (2015) – Instructors

& Learners

Activity Motivation & Perfor-

mance impact

AI Tools: act on behalf of a human. They may au-

tomatically group users depending on a predeﬁ-

ned model. The AI model could use a learner’s

past performances to automatically adapt quiz dif-

ﬁculty.

Meta-cognitive and guiding tools seem the most

promising tools to help user gain insight on their col-

laborative behaviours, without been overwhelmed by

data or proned to over-scripting.

Duque et al. (2015) apply a multidimensional ge-

neralization of the median mathematical concept to

identify central and peripheral learners based on the

activity. The authors use dimensions such as the level

of work, interaction, coordination, speed, syntactic

correctness and quality to group learners heterogene-

ously, resting on Wang et al. (2007)’s suggestion that

heterogeneous groups induce more collaboration than

homogeneous ones. The authors emphases the need

to give more control to the instructors over the hete-

rogeneity degree used to group the learners.

In our review (Table 1) we notice that few LA re-

searches focus consider the learners as the ﬁnal tool

users. The reason is that learners may not have suf-

ﬁcient knowledge to effectively interpret recommen-

dations made by the LA tools. But Anaya et al.

(2016) investigate the students’ reactions to a semi-

automated recommendation about their collaborative

behaviour. The authors provide a decision tree visu-

alization, giving students insight on the recommen-

dation machinery (Figure. 1). The recommendation

is based on six meta-cognitive indicators : reputa-

tion, leadership, initiative (I-level, I-regularity), acti-

vity (A-level, A-regularity). Thus students better un-

derstand why they receive the recommendation, this

makes it more effective. The decision tree also rein-

forces the learners’ meta-cognitive abilities by hel-

ping the learners gain insight on the way to improve

their collaboration.

These studies points to visualizations as promi-

sing tools to support collaborative learning from an

instructor’s point of view but also from the learners’.

Towards a Dynamic Visualization of Online Collaborative Learning

207

A-regularity

A-level

Low

A-level

Medium

Yes

High

Yes

Low

Medium

High

Yes

Low

Medium

High

Figure 1: Excerpt with two indicators from Anaya et al.

(2016)’s visualization explanatory decision tree for a stu-

dent. The “yes” meaning the systems suggest to send a re-

commendation.

3 VISUALIZATIONS

Kosslyn, cited in (Twissell, 2014), break down the vi-

sualization concept in two parts, which he terms vi-

sual perception and visual mental imagery. Visual

perception points to the bio-physical process. We per-

ceive an object when it reﬂects light and stimulate the

retina’s cones cells. This process is an efﬁcient high

bandwidth way to receive information about the en-

vironment. It can provide rapid and precise feedback

about a speciﬁc situation and it is therefore a useful

awareness mechanisms. Visual mental imagery iden-

tiﬁes the process of representing, manipulating and

transforming mental images. It is the cognitive pro-

cess, the persistent image one can have of a diagram,

an animation or a photo when the external stimuli ce-

ase. This is what we call visualization.

Visualizations enable creative and holistic thin-

king. They improve the ability to make effective in-

ferences while translating or making visual analogies

reinforces conceptual development. They impact cog-

nition, help sense making and understanding (Twis-

sell, 2014; Klerkx et al., 2014). The designer’s in-

genuity is his ability to convey a speciﬁc information

or an emotion by arranging judiciously symbolic ele-

ments (Figure 2).

Animations alleviate visual complexity by mat-

ching the temporal dimension with that of the con-

cept to analyse. “A meta-study of algorithm visualiza-

tion effectiveness” (Hundhausen et al., 2002) descri-

bes how animations grounded in cognitive constructi-

vism are an efﬁcient ways to teach algorithms. In-

deed, since algorithms are processes, animations faci-

litate their exploration and understanding.

Navigation with interactive visualizations can op-

timize topic exploration. INSIGHT, a web applica-

tion uses Latent Dirichlet Allocation (LDA) to au-

tomatically extract major topics from the stackex-

change.com network and displays them using bubbles

of different sizes and colours. The bubbles’ attributes

depend on the semantic proximity of the posts’ key-

word and the searched concept.

In the following subsection, we look at other vi-

sualizations speciﬁcally supporting collaborative le-

arning.

3.1 Supporting Collaborative Learning

with Visualizations

Many studies proved visualizations useful from the

instructors’ point of view. Theses researches are clo-

sely related to the ﬁeld of Computer Supported Colla-

borative Work (CSCW). There, users are not necessa-

rily learners. They can be experts in their ﬁeld and

therefore have less difﬁculties visualizing correctly

the information. Gilbert and Karahalios (2009) give

another example of a study supporting collaborative

users with, CodeSaw, which let users infer knowledge

from the logged data generated while working on an

open source project. The authors advocate the use of

simple visualizations, leaving most of the reasoning

on the users’ side.

In Computer Supported Collaborative Learning

(CSCL), collaboration visualizations mostly deal with

supporting exploration of the user’s social context

(Heer and Boyd, 2005; Klamma et al., 2006), or to

help them collaborate by enhancing their reﬂexivity.

The difﬁculty prevails with visualizations aimed at in-

forming the learners but a few examples exist.

3.1.1 Visualizations for Learners

In cognitive science, reﬂexivity is the meta-cognitive

ability to visualize one’s own cognitive abilities. It

is, for example, the ability to answer the following

questions : do I think that I am able to play tennis?

How should I change my behaviour to succeed in that

online course? It is closely related to Zimmerman

(1990) Self-Regulated Learning concept. Reﬂexivity

Figure 2: Dmitry Mendeleev’s periodic table. The clever

organization of information provides a deep understanding

and facilitates memorization of the chemical elements’ pro-

perties. . It illustrate “how a diagram is (sometimes) worth

ten thousand words” (Larkin and Simon, 1987).

CSEDU 2018 - 10th International Conference on Computer Supported Education

208

helps learners build a model of their own learning be-

haviour, which they can confront to the learner model

held by the LMS. Learner Models (LMs) can be as

simple as an ID number. It can also have thousands

of dimensions trying to model a learner’s motivation,

knowledge level. Accessing and visualizing this in-

formation is capital for the learner say Bull and Kay

(2016). Therefore, this makes it an important element

for learners to visualize.

Ideally, learners should also be able to interact

with an Inspectable learner model. Govaerts et al.

(2012) use learning data to generate feedback for the

learners, creating activity awareness and fostering re-

ﬂexivity. More recently Yousuf and Conlan (2015)

found that it indeed helped their learning. But as we

mentioned in Section 2.2.1 on page 2, visualizations

may not be easily interpreted by learners.

3.2 Visualizations Limits and Bias

Klerkx et al. (2014) review visualizations applied to

learning and ﬁnd that they have a positive impact on

cognitive abilities but Twissell (2014) identiﬁes the

following limits: a) different learning styles, natural

differences in learners have a signiﬁcant impact on

the way diagrams are perceived, visualized and under-

stood b) visualizations do not equally affect all types

of learning activities.

Kahneman (2011) and Dehaene (2007) neuro-

psychological ﬁndings, pinpoint to our visual and

cognitive biases, an anchoring mechanism, which

succinctly means that we cannot refrain ourselves

from using our personal history to infer meanings.

Therefore, there is no objective way of representing

an information. Even a “simple” tasks such as evalua-

ting a length or an area is bias. We are more precise at

evaluating vertical length than oblique ones because

statistically we encounter the ﬁrst in our environment

more often than the latter and developed more neu-

rons to recognize it (Girshick et al., 2011).

Thus far, the choice for the collaborative lear-

ning’s representations aimed to learners, should care-

fully be thought and rest on works from Bertin (1983).

Visualizations can easily be misunderstood, as in

this study from Lonn et al. (2015), where they ﬁnd

a modiﬁcation of the students’ goals related to how

often the students visualize their progress. Mastery

goal orientated students, which are those seeking a

deep understanding, became competency based stu-

dents (Pintrich and De Groot, 1990), only caring to

show external proof of performance.

In the last section we present visualizations which

believe could support online collaborative learning

from a student point of view.

4 DYNAMIC GRAPHS

We propose to build a visualization considering the

two pedagogical theories: socio-constructivism and

cognitivism. Socio-constructivism because the visua-

lization shows the learners’ social online interactions

in the MOOC. It emphasizes the collaboration rather

than the competition. Learners will not be able to de-

rive each others exact participation from the visuali-

zation. The self-regulated learning or cognitivism as-

pect comes from the feedback provided through the

visualization to each learner about his behavioural

pattern.

Our visualization should support cognitive com-

petence, such as improved performance, and goal

achievement. It should provide some means for com-

parison but only with cohort of similar students or

with one’s future goals and past progression.

We envision that a dynamic network visualiza-

tion could satisfy those characteristics. The network

should be content focused. Nodes would represent

forums and not users. Two forums would be linked if

the same user posted in both. Also, the size of the

forum-nodes, the links’ width would be a function

of the user count. The keywords extracted from the

forum would help build a conceptual map from the

users’ interactions.

Another reason not to represent learners as nodes

is a readability issue. There are far more learners than

forums discussions. Therefore a network showing fo-

rums as nodes should be more comprehensible that

one showing learners. Meanwhile, it would still be

possible for instructors to transform a forum-network

in its dual, that is the a learner-network.

Boroujeni et al. (2017) propose an integrated ap-

proach to analyse the dynamic of MOOCs discussion

forums. They describe the interplay of temporal pat-

terns, discussion content and social structure emer-

ging from the learners’ interactions. Our objective

is to ﬁnd a visualization carrying this information to

learners. In the following subsection we summarize

the different visualization techniques for dynamical

graphs.

4.1 Dynamic Graph Visualizations

A collaborative learning visualization could build

upon dynamic graph visualizations. Directed graphs

(whose links have directions) can be represented in

four ways (see Figure 3 on the next page). 1) Force-

directed, strong links are the shortest and nodes are

correspondingly placed. 2) Orthogonal, nodes are

placed on an orthogonal grid. 3) Hierarchical, links

deﬁne node hierarchies. They tend to always go in

Towards a Dynamic Visualization of Online Collaborative Learning

209

the same direction, for example from top to bottom.

Nodes can therefore be grouped based on there ver-

tical position. 4) Matrix, nodes are represented by

the matrix’s lines and columns. The line-column in-

tersections correspond to the interaction between two

nodes. A "1" value at an intersection code for the ex-

istence of a link, a "0 " for its absence. Intermediaries

values can quantify the link’s strength.

Beck et al. (2017) further identify two major cate-

gories of dynamic visualizations: animations and ti-

melines. By dynamic visualization we consider all vi-

sualizations evolving in time, and because collabora-

tive learning is a process, therefore evolving in time,

we believe that a dynamic visualization will reduce

the learners cognitive load.

Animations integrate time changes in unique

moving representations. They are general-purpose,

special-purpose or matrix based representations.

General purpose methods are further subdivided

in online and ofﬂine approaches. In ofﬂine approa-

ches, the full sequence of the interactions is known.

For example in a playback situation the user sees how

the graph evolved up to present. In that case, the graph

layout can be optimized to smooth transitions from

the ﬁrst to the last layout. For online, or real-time,

approaches. The full set of interactions is not know.

The graph layouts are arranged without knowing the

future. The transition problem, here, is salient. How

should nodes and links’ addition deletion be displayed

to keep a visually comprehensible graph ? One of the

approaches proposed by Beck et al. (2017) is that of

Friedrich and Eades (2002): Marey. It consists of four

phases. 1) nodes and links are removed ; 2) the whole

graph is translated to a new optimal position ; 3) each

node takes its new position ; 4) new nodes and edges

are placed.

Special purpose techniques are particularly useful

for visualizing subcommunities, if we know for ex-

ample that the graph has a hierarchical structure.

In general, node-link animations are well suited

for medium sized graphs (with a few hundred nodes).

Bigger graphs rapidly become unreadable even for ex-

perts. A common option, is the to use colours to mark

c d

e e

a b c d e

force-directed orthogonal hierarchical matrix

Figure 3: Different representations of the same network

(Beck et al., 2017).

different communities. But structural changes may

always be difﬁcult to track on big scales, and if we

were to represent learners as nodes we would need

other techniques to provide a collaborative learning

overview.

Other techniques include timeline, matrix and

hybrid visualizations.

Timelines map different timesteps in space. Each

timestep requires a graphical representation. There-

fore changes in the interactions give rise to several

graphs combined in one visualization. Such techni-

ques shrinks each node-link graphs, but it facilita-

tes comparison. Timelines make it easier to analyse

structural changes when nodes are at ﬁxed position

(see Figure. 4)

Matrix visualizations are found both with anima-

tions and timeline techniques. Due to its concise size,

matrices scale better than other approaches. It can

conveniently represent data with several thousand no-

des.

A few hybrid methods exist. They combine time-

lines and animations. For example, a selection of ti-

mesteps could animated on request by the user. This

brings up our discussion on the interactions.

As Yi et al. (2007) recall, information visualiza-

tion has two parts : representation and interaction.

4.2 Interactions to Support Visual

Analytics

Heer and Shneiderman (2012) emphasize the impor-

tance of interaction in information visualization. With

Yi et al. (2007), they provide two overlapping user

centred taxonomies listing the tools to support a the

ﬂuent use of visualizations. a) Data & View speciﬁ-

cation tools ﬁlter out, sort and reconﬁgure the data

Timestep ∆

Figure 4: Three timesteps representation of the same graph.

In ∆

, some users posting Forum 3 (F

) also posted in F

Some from F

also posted in F

. Some users posted in F

but did not post elsewhere during this time interval. In ∆

a new forum (F

) and a new link (F

− F

) are created. So

people are still posting in the previous forums. Finally in

∆

since nobody posted in F

it is removed.

CSEDU 2018 - 10th International Conference on Computer Supported Education

210

to give it meaning, b) View manipulation tools se-

lect items, navigate, explore, abstract/elaborate, en-

code the information, c) Process & Provenance tools

record, annotate, share visual conﬁgurations.

We propose to integrate the following interacti-

ons in our collaborative learning visualization : a) ﬁl-

ter forums based on keywords, time stamps, pos-

ting/reading users, b) ﬁlter links based on quantitative

threshold, time stamps, users’ characteristics, c) sort

the nodes’ dispositions based on forum creation date,

most recent post, topics, or a hierarchy derived from

the users’ interactions, d) get detailed information

about a selected forum, e) see the path from forums

to forums for a selected user or cohort, f) zoom, pan,

scroll the forum network map, g) enable an overlay

instructor based concept map, h) annotate, comment

and share visualizations.

We understand that, as shown by Jivet et al.

(2017), awareness is not enough. With our inte-

ractive visualization, we hypothesize that the users

will engage in collaborative activities to ﬁnd topics

of interest and collaborative partners, supporting their

learning goals without fostering competition. They

develop self-regulated competences.

5 CONCLUSION

In this paper we presented the challenges related to

modelling learning and collaborative learning. We

explained why the LA community focuses on desig-

ning and evaluating tools supporting and not repla-

cing human learning. We argued that visualizations

can be used to support some parts of learning if they

are well grounded in learning theories. We found that

static graph images of collaboration are not suitable

to support learners because they are only available at

the end of a course. Instead, we proposed a dyna-

mic visualization to support collaborative learning in

a self-regulated learning and socio-constructivist fra-

mework. Hence, our future challenge is to represent

in a real-time and intuitive way to the learners, the col-

laboration dynamics. We plan to achieve this using a

dynamic and interactive graph, which learners could

explore and derive insights from, about their learning

behaviour.

We are in the process of building a prototype to

evaluate how this will scale up to our ﬁnal objective,

supporting MOOCs’ collaborative learning.

REFERENCES

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

sual recommender tool in a collaborative learning ex-

perience. Expert Systems with Applications, 45:248–

259.

Baker, R. S. (2016). Stupid Tutoring Systems, Intelligent

Humans. International Journal of Artiﬁcial Intelli-

gence in Education, 26(2):600–614.

Bandura, A. and McClelland, D. C. (1977). Social learning

theory. 35324.

Beck, F., Burch, M., Diehl, S., and Weiskopf, D. (2017). A

taxonomy and survey of dynamic graph visualization.

In Computer Graphics Forum, volume 36, pages 133–

159. Wiley Online Library.

Bertin, J. (1983). Semiology of Graphics: Diagrams, Net-

works, Maps. University of Wisconsin Press.

Boroujeni, M. S., Hecking, T., Hoppe, H. U., and Dillen-

bourg, P. (2017). Dynamics of MOOC discussion fo-

rums. In LAK, pages 128–137.

Bull, S. and Kay, J. (2016). SMILI: a framework for inter-

faces to learning data in open learner models, learning

analytics and related ﬁelds. International Journal of

Artiﬁcial Intelligence in Education, 26(1):293–331.

Dehaene, S. (2007). Neurones de la lecture (Les): La nou-

velle science de la lecture et de son apprentissage.

Odile Jacob.

Dillenbourg, P. (1999). What do you mean by collabora-

tive learning. Collaborative-learning: Cognitive and

computational approaches, 1:1–15. 02876.

Dillenbourg, P. (2002). Over-scripting CSCL: The risks of

blending collaborative learning with instructional de-

sign. Heerlen, Open Universiteit Nederland.

Duque, R., Gómez-Pérez, D., Nieto-Reyes, A., and Bravo,

C. (2015). Analyzing collaboration and interaction in

learning environments to form learner groups. Com-

puters in Human Behavior, 47:42–49.

Friedrich, C. and Eades, P. (2002). Graph drawing in mo-

tion. J. Graph Algorithms Appl., 6(3):353–370.

Gilbert, E. and Karahalios, K. (2009). Using social visuali-

zation to motivate social production. IEEE Transacti-

ons on Multimedia, 11(3):413–421.

Girshick, A. R., Landy, M. S., and Simoncelli, E. P. (2011).

Cardinal rules: visual orientation perception reﬂects

knowledge of environmental statistics. Nature Neu-

roscience, 14(7):926–932.

Govaerts, S., Verbert, K., Duval, E., and Pardo, A. (2012).

The student activity meter for awareness and self-

reﬂection. In CHI’12 Extended Abstracts on Human

Factors in Computing Systems, pages 869–884. ACM.

Heer, J. and Boyd, D. (2005). Vizster: Visualizing online

social networks. In Information Visualization, 2005.

INFOVIS 2005. IEEE Symposium on, pages 32–39.

IEEE.

Heer, J. and Shneiderman, B. (2012). Interactive dynamics

for visual analysis. Queue, 10(2):30.

Hommes, J., Rienties, B., de Grave, W., Bos, G., Schuwirth,

L., and Scherpbier, A. (2012). Visualising the invisi-

ble: a network approach to reveal the informal social

side of student learning. Advances in Health Sciences

Education, 17(5):743–757.

Towards a Dynamic Visualization of Online Collaborative Learning

211

Hundhausen, C. D., Douglas, S. A., and Stasko, J. T.

(2002). A meta-study of algorithm visualization ef-

fectiveness. Journal of Visual Languages & Compu-

ting, 13(3):259–290.

Jivet, I., Scheffel, M., Drachsler, H., and Specht, M. (2017).

Awareness Is Not Enough: Pitfalls of Learning Ana-

lytics Dashboards in the Educational Practice. In

Data Driven Approaches in Digital Education, Lec-

ture Notes in Computer Science, pages 82–96. Sprin-

ger, Cham.

Kahneman, D. (2011). Thinking, Fast and Slow. Farra,

Straus and Giroux.

Kirschner, P. A. (2016). Keynote: Learning Analytics: Uto-

pia or Dystopia. City.

Klamma, R., Spaniol, M., Cao, Y., and Jarke, M. (2006).

Pattern-based cross media social network analysis for

technology enhanced learning in Europe. In European

Conference on Technology Enhanced Learning, pages

242–256. Springer.

Klerkx, J., Verbert, K., and Duval, E. (2014). Enhancing

Learning with Visualization Techniques. In Spector,

J. M., Merrill, M. D., Elen, J., and Bishop, M. J., edi-

tors, Handbook of Research on Educational Commu-

nications and Technology, pages 791–807. Springer

New York. DOI: 10.1007/978-1-4614-3185-5_64.

Lake, B. M., Ullman, T. D., Tenenbaum, J. B., and Gersh-

man, S. J. (2016). Building Machines That Learn and

Think Like People. Behavioral and Brain Sciences,

pages 1–101.

Larkin, J. H. and Simon, H. A. (1987). Why a diagram

is (sometimes) worth ten thousand words. Cognitive

science, 11(1):65–100.

Lonn, S., Aguilar, S. J., and Teasley, S. D. (2015). In-

vestigating student motivation in the context of a le-

arning analytics intervention during a summer bridge

program. Computers in Human Behavior, 47:90–97.

Medina, E., Meseguer, R., Ochoa, S. F., and Medina, H.

(2016). Providing Behaviour Awareness in Collabora-

tive Project Courses. Journal of Universal Computer

Science, 22(10):1319–1338.

Pintrich, P. R. and De Groot, E. V. (1990). Motivational

and self-regulated learning components of classroom

academic performance. Journal of educational psy-

chology, 82(1):33.

Rehm, M., Gijselaers, W., and Segers, M. (2015). The im-

pact of hierarchical positions on communities of le-

arning. International Journal of Computer-Supported

Collaborative Learning, 10(2):117–138.

Siemens, G. (2014). Connectivism: A learning theory for

the digital age. 03556.

Soller, A., Martínez, A., Jermann, P., and Muehlenbrock,

M. (2005). From mirroring to guiding: A review of

state of the art technology for supporting collabora-

tive learning. International Journal of Artiﬁcial Intel-

ligence in Education, 15(4):261–290.

Tempelaar, D. T., Rienties, B., and Giesbers, B. (2015).

In search for the most informative data for feedback

generation: Learning analytics in a data-rich context.

Computers in Human Behavior, 47:157–167.

Tenenbaum, J. B., Grifﬁths, T. L., and Kemp, C. (2006).

Theory-based Bayesian models of inductive lear-

ning and reasoning. Trends in Cognitive Sciences,

10(7):309–318.

Twissell, A. (2014). Visualisation in applied learning con-

texts: a review. Journal of Educational Technology &

Society, 17(3).

van Leeuwen, A., Janssen, J., Erkens, G., and Brekelmans,

M. (2014). Supporting teachers in guiding collabora-

ting students: Effects of learning analytics in CSCL.

Computers & Education, 79:28–39.

Vygotskij, L. S., Sève, F., and Clot, Y. (1985). Pensée et

langage. Messidor.

Wang, D.-Y., Lin, S. S., and Sun, C.-T. (2007). DIANA:

A computer-supported heterogeneous grouping sy-

stem for teachers to conduct successful small learning

groups. Computers in Human Behavior, 23(4):1997–

2010.

Wang, X., Wen, M., and Rosé, C. P. (2016). Towards trig-

gering higher-order thinking behaviors in MOOCs.

In Proceedings of the Sixth International Conference

on Learning Analytics & Knowledge, pages 398–407.

ACM.

Yi, J. S., ah Kang, Y., and Stasko, J. (2007). Toward a dee-

per understanding of the role of interaction in informa-

tion visualization. IEEE transactions on visualization

and computer graphics, 13(6):1224–1231.

Yousuf, B. and Conlan, O. (2015). VisEN: Motivating Le-

arner Engagement Through Explorable Visual Narra-

tives. In Design for Teaching and Learning in a Net-

worked World, pages 367–380. Springer.

Zimmerman, B. J. (1990). Self-regulated learning and aca-

demic achievement: An overview. Educational psy-

chologist, 25(1):3–17.

CSEDU 2018 - 10th International Conference on Computer Supported Education

212