COLLABORATIVE LEARNING IN HETEROGENEOUS CLASSES

Towards a Group Formation Methodology

Liana Razmerita

and Armelle Brun

Copenhagen Business School, CBS, ISV, Dalgas Have, 15, DK-2000 Frederiksberg, Denmark

LORIA - Nancy Universit

e, Campus Scientiﬁque, 54506 Vandoeuvre les Nancy, France

Keywords:

Group formation, Group work, Collaborative learning, Web 2.0.

Abstract:

Group work has been adopted as an important tool to support collaborative work in order to enhance learning

processes. There is a wealth of literature related to group performance and the impact of group composition on

group and individual performance. However, very few studies address the issue on how to automatically form

groups. This article proposes a methodology that could be used by professors to form groups automatically

taking into account different criteria as well as the students’ proﬁle. This methodology is based on a pilot

study that analyzes group composition of self-formed student groups. The pilot study ﬁndings suggest that

students tend to form homogeneous group in terms of level of the knowledge. Furthermore, students report

that working on common topics of interests was a decisive factor in forming the groups.

1 INTRODUCTION

Group work and collaborative work are important

pedagogical tools for classroom assignments and have

proved to have a strong impact on individual and co-

operative learning (Johnson and Johnson, 1994). Fur-

thermore Web2.0 has created new possibilities for stu-

dents to engage, interact and collaborate in various

learning tasks that may enhance learning processes

and the overall learning experience. In this context

the didactic challenge for educators is to design and

integrate a new set of tools based on speciﬁc didac-

tic principles associated with a speciﬁc domain of

learning (Mondahl et al., 2009). Consider a professor

who wants to assign group work to students during a

course. One issue he/she has to handle is: How to

form groups? He/she may either let the students form

the groups by themselves or may form student groups

randomly or based on their physical proximity (e.g.

their position in the class). Even though this is a very

easy convenient approach, randomly formed groups

may not be fair or may not be the best approach.

Group work has many variables and different fac-

tors that can inﬂuence group performance. For exam-

ple, literature related to motivation and group learning

shows that performance is not only linked to the inter-

est in the subject to be learned but may also be related

to relations to peers, gender differences, age, individ-

ual differences, cultural backgrounds within the gro-

up, personality traits, etc. (Souren et al., 2003). Fur-

thermore, explaining to students why they form a

group and why they are performing certain tasks may

lead to better performance (Bekele, 2005). Moreover,

a student has to feel comfortable in a group in order

to communicate his/her ideas, to express his point of

view with his/her group-mates. Thus, a professor may

select a special criterion based on which he/she wants

to create the groups. This criterion can be either a sin-

gle criterion or a set of different criteria (e.g. hetero-

geneous in relation with their background). However

forming groups of students, especially for large size

classes, is an intractable and time consuming task for

the professor.

Previous studies related to collaboration and

group work have emphasized the importance of het-

erogeneity for performance, creativity and learning

(Rich, 1997). A heterogeneous group is made of

members that are balanced in terms of diversity based

on some criteria: culture, gender, personality, etc.).

There is a wealth of literature related to group per-

formance and their composition (Slavin, 1995; Sha-

ran, 1999) but very few literature addresses the issue

of how to automatically form groups. Group work

and collaborative learning open a number of challeng-

ing research questions related to group performance

and group composition such as: How heterogeneity

or homogeneity of a group inﬂuence the group per-

formance and students’ learning? Do homogenous

189

Razmerita L. and Brun A..

COLLABORATIVE LEARNING IN HETEROGENEOUS CLASSES - Towards a Group Formation Methodology.

DOI: 10.5220/0003338901890194

In Proceedings of the 3rd International Conference on Computer Supported Education (CSEDU-2011), pages 189-194

ISBN: 978-989-8425-50-8

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

groups perform better than heterogeneous groups?

How students form groups? Literature suggests that

groups should be formed differently according to the

type of assignments. Homogeneous groups are bet-

ter for achieving speciﬁc goals (e.g. short-term and

guided problem solving) while heterogeneous groups

are better for long-term knowledge discovery prob-

lems (Bekele, 2005).

This article is a preliminary study that focuses on

how students form groups and proposes a method to

form groups automatically, taking into account dif-

ferent criteria as well as the students’ proﬁle. This

paper relies on a preliminary study case investigat-

ing the way students form groups, in the framework

of a course where they have to conduct research and

present project related ideas and ﬁndings in group.

The article is structured in ﬁve sections. The sec-

ond section presents an overview of literature on col-

laborative learning, collaborative work, group compo-

sition and group formation. The third section presents

a pilot study of group work within a heterogeneous

classroom supported by a Web 2.0-enabled learning

environment. The fourth section proposes a method-

ology that enables professors in very different teach-

ing areas to form groups automatically.

2 LITERATURE REVIEW

2.1 Web 2.0 and e-Learning 2.0

Web 2.0 and the associated technologies have

changed the way the web is used; web is a dynamic

social space where participation, collaboration, on-

line interaction are core elements. Web 2.0 or social

software may be approached from different perspec-

tives: as a new social media tool, a facilitator of new

forms of interaction and knowledge sharing (Kirchner

et al., 2008), an enabler of personal information and

knowledge management tools (Razmerita and Kirch-

ner, 2009) and new didactic tools that facilitate in-

teraction and social processes. Web 2.0 has a large

inﬂuence on learning approaches (e.g. using wikis,

blogs, micro blogs) and it offers new means to inter-

act, socialize on-line, ﬁnd information, and commu-

nicate using a wide range of new collaborative ser-

vices. Learning is not anymore viewed as a unidirec-

tional process, where teachers are in the same place

at the same moment with the learners, and knowledge

is transferred from teachers to learners. Learners are

now participants in the learning process thanks to the

tools that enable and encourage them to participate,

interact and collaborate more easily with other learn-

ers, teachers or peers, etc.

2.2 Collaborative Learning and

Cooperative Learning

Collaborative learning can be described as a situation

in which two or more people learn or attempt to learn

something together. Collaborative learning is differ-

ent from cooperative learning. In cooperative learning

peers split the work in tasks and tend to solve these

tasks individually and then assemble their results into

the ﬁnal output. While in collaborative learning learn-

ers interact and do work together in order to complete

their assignments. Despite the difference, many ar-

ticles use these two terms interchangeable. For ex-

ample, cooperative learning (CL) is deﬁned as an in-

structional strategy in which students work actively

and purposefully together in small groups to enhance

both their own and their teammates’ learning (Abrami

et al., 2004). Literature emphasizes that collaborative

learning is one of the most successful techniques to

enhance student performance. Several studies report

that group work and cooperative learning enhance the

learning of an individual compared to when he/she

learns alone (Dansereau and Johnson, 1994; Slavin,

1983). Several approaches have been proposed to

support collaborative learning with the aim of facil-

itating information and resource sharing between stu-

dents (Florea, 1999; Krejins et al., 2002). Group

work can be performed either by students physically

present at the same place and at the same time (syn-

chronous work) or remotely through asynchronous

work (Souren et al., 2003).

2.3 Group Composition

The quality of the learning process in the context

of collaborative work highly depends on the char-

acteristics of the group. Previous studies suggest

that groups should be formed randomly or by stu-

dents themselves and groups should have a small size,

made up to four-members of different level of knowl-

edge (Slavin, 1987). Related work emphasized the

importance of personality attributes, gender, school

background, ethnic background, motivation (Bradley

and Herbert, 1997; Alfonseca et al., 2006) in group

performance. The learning style is also an impor-

tant criterion in group composition (Martin and Pare-

des, 2004; Wang et al., 2007). It has been observed

that the quality of learning in groups is inﬂuenced

by their diversity. Heterogeneous groups may out-

perform homogeneous groups (Nijstad and Carsten,

2002). Some studies emphasized that heterogeneous

groups may be more creative and innovative (Paulus

and Nijstad, 2003) and furthermore they may be more

effective for individual learning. One of the ﬁrst het-

CSEDU 2011 - 3rd International Conference on Computer Supported Education

190

erogeneity criteria was in relation with the level of

knowledge and skills (Slavin, 1987; Webb, 1989).

Diversity in background, opinions (Siemens, 2010),

ideas, personality, gender, are also criteria that can

be considered in the heterogeneity (Slavin, 1995).

Groups may be automatically formed based on think-

ing styles (Wang et al., 2007), competence, learn-

ing style and interactions (Ho et al., 2009), learning

achievements (Chan et al., 2010). The clustering ap-

proaches proposed in the literature use genetic algo-

rithms (Graf and Bekele, 2006; Ho et al., 2009; Chan

et al., 2010) or ant colony optimization (Bekele, 2005;

Graf and Bekele, 2006).

3 THE PILOT STUDY

3.1 The StudyBook Context

This pilot study aims to investigate the use of a Web

2.0 collaborative platform, StudyBook, to support

new ways of teaching and learning. The underly-

ing hypothesis of the overall study is that collabora-

tive groupwork creates natural opportunities for the

learners to articulate their understanding, reﬂect on

and justify actions and these activities may improve

learning. Furthermore new Web2.0/Web3.0 technolo-

gies integrated in a learning platform may support

group work, collaboration and learning processes in

new ways that are more natural for new generation of

learners. In particular, the StudyBook project aims

to investigate the use of social-collaborative services

for collaborative learning and groupwork (Mondahl

et al., 2009). A screenshot of StudyBook is pre-

sented in the Figure 1. This study has been conducted

within the Web Interaction Design and Communica-

tion course, an elective course for bachelor students

at the Copenhagen Business School (CBS) in Den-

mark. This course enrolled 46 bachelor students from

seventeen different countries and from different study

programs (Business Administration, European stud-

ies, International Business, Marketing, HRM, etc).

Within this course, a project has to be conducted col-

laboratively, partly in group and partly individually.

The students had to select a topic of interest, to form

groups and work in group on this topic. Within their

projects, the students deﬁne their own topic to work

on in relation with their core areas of interest and the

main course topics. The project work is divided in

two steps. 1)Work in group: The group has to work

on a common topic of interest. The StudyBook plat-

form facilitates students’ interaction and group col-

laboration. Students are not in the same study pro-

grams and do not have common free time-slots, their

Figure 1: Screenshot of the StudyBook platform.

project work had to be constructed collaboratively us-

ing wikis and afterwards had to make a presentation.

2) Work individually: In the second step, students

have to continue to work on the selected topic indi-

vidually in order to write a ten page report that will

be graded.

3.2 Data Collection: The Questionnaire

Two questionnaires have been designed with two ob-

jectives in mind: to assess the student’s perception

of StudyBook as a learning platform and collect data

about students and the way they form groups. Within

the following section of the article we focus on the

ﬁndings related to the analysis of the groups that have

been formed. In the questionnaire, data about stu-

dents include their background (study programs), cul-

ture (the country they come from), topics of interests

and level of knowledge in relation with the course.

Students had to answer about both their level of

knowledge (beginner, intermediate, advanced) and re-

search interests (not interested, interested, very inter-

ested). In addition, students were asked about the way

they have formed their groups.

3.3 Data Analysis: Evaluation of

Groups

This section presents an analysis about the way

groups have been formed in the heterogeneous classe.

Among the 46 students enrolled in the course, 29 stu-

dents have answered the questionnaire. Eight groups

in total have answered the questionnaire. First of all,

students declare they have enjoyed working in groups

and have perceived positively their workgroup collab-

oration. A ﬁrst analysis of the student provided an-

swers shows that the students who answered the ques-

tionnaire are from ﬁfteen different countries all over

the world. The most represented countries are Den-

mark, Spain and France (4 students from each coun-

COLLABORATIVE LEARNING IN HETEROGENEOUS CLASSES - Towards a Group Formation Methodology

191

Table 1: Average distances within groups.

Group level of knowledge topics of interest

Group1 0.02 0.27

Group2 0.48 0.10

Group3 0.24 0.23

Group4 0.16 0.15

Group5 0.22 0.20

Group6 0.02 0.05

Group7 0.05 0.18

Group8 0.24 0.22

All Gr 0.18 0.18

Class 0.23 0.18

try) and Hong Kong (3 students). Thus, the formed

groups are probably heterogeneous in terms of cul-

tural background. In our overall study the heterogene-

ity is related to students’ background (different study

programs), cultural (different countries), topics of in-

terests and the level of knowledge. In this preliminary

study, we speciﬁcally focus on level of knowledge and

topics of interest. In order to study homogeneity and

heterogeneity of groups we compute a metric distance

between students taking into account their declared

level of knowledge or/and topics of interests.

3.3.1 Group Heterogeneity in Relation with the

Level of Knowledge

Studies suggest that if students organize themselves

in groups, they usually tend to form homogeneous

groups (Souren et al., 2003). The ﬁrst column of Ta-

ble 1 represents the average distance in terms of level

of knowledge within groups and within the class. The

average distance between all students in the class is

0.23 and standard deviation is 0.17. Less than 3%

of the pairs of students have a distance greater than

0.8. This means that few students are very different

in terms of their declared level of knowledge. Among

the eight groups, the average distance is 0.18, the most

homogeneous group has an average distance of 0.02

and the less homogeneous has an average distance

of 0.48. Three groups have an average distance less

than 0.05. The average distance between students in

the groups is smaller than the average distance in the

whole matrix. Thus groups formed by students tend

to be similar in terms of knowledge. The ﬁndings of

our study conﬁrm the fact that students tend to form

homogeneous groups.

3.3.2 Group Heterogeneity in Relation with

their Topics of Interests

The second column of Table 1 represents the average

distance in terms of topics of interest within groups

and within the class. The average distance between

all students in the class is 0.18 and standard deviation

is 0.12. Less than 0.2% have a distance greater than

0.7. Despite their heterogeneity, students appear to be

quite homogeneous in relation with their level of in-

terest. Among the eight groups, the average distance

is also 0.18, which is similar to the average distance

among groups in terms of level of knowledge. The

most homogeneous group has an average distance of

0.05 and the less homogeneous one has an average

distance of 0.27. Five groups out of the eight have

similar average interest distance and average knowl-

edge distance. Within the class, the average distance

between students in terms of topics of interest (0.18)

is lower than the one in terms of level of knowledge

(0.23); students are more similar in terms of topics of

interest than in level of knowledge. However, the av-

erage distance within groups is similar to the average

distance within the class; thus we can deduce that stu-

dents do not tend to form similar groups in terms of

topics of interest.

3.3.3 How Students Have Formed Groups?

Based on the assignment requirements described in

section 3.1, the results of the study in relation with

how students form groups are presented in Figure 2.

Based on the questionnaire’s answers, a large major-

ity of students have declared forming groups based on

identiﬁed common topics of interests (68,97%), fol-

lowed by afﬁnity with the others members (10, 34%)

and different backgrounds and culture (10, 34%).

10,71%&

67,86%&

7,14%&

10,71%&

3,57%&

0,00%&

You&know&&or&you&like&the&others&

Common&topics&of&interests&

Compa@ble&working&periods&

Diﬀerent&background&and&culture&

Similar&background&and&culture&

You&had&to&form&groups&

Figure 2: Reported criteria on groups formed.

CSEDU 2011 - 3rd International Conference on Computer Supported Education

192

As a conclusion, the groups formed by students

tend to be homogeneous in terms of the level of

knowledge. However, these groups are not really

homogeneous in terms of the topics of interests,

whereas the students answered in the questionnaire

they formed homogeneous groups in terms of com-

mon topics of interests. Thus, when leaving students

to form groups by themselves, the resulting groups

may not have the expected characteristics. Thus a tool

that automatically forms groups may be highly useful

to ensure that groups have given characteristics and

lead to the expected type of learning and performance.

In the following section, we propose a methodology

to form groups automatically according to students’

proﬁle and speciﬁed citeria.

4 METHODOLOGY FOR

FORMING GROUPS

AUTOMATICALLY:

CLUSTERING STUDENTS

The proposed methodology comprises four main

steps: collecting data about students, initialize

the vectors representing the students’ characteris-

tics, clustering students, evaluating the group perfor-

mance.

Step 1. Collect data about students. Questionnaires

provide an effective way to collect data about the

students. However data can be provided by stu-

dents or by the administration (e.g., the study pro-

gram, the background) or by the professor (pos-

sible topic, learning concepts, learning objectives,

students’ skills, heterogeneous versus homogeneous

groups). As described previously, in our pilot study

the following data (topics of interest, level of knowl-

edge, country of origin, study program) was collected

with the purpose of automatic group formation and

other purposes. The collected data needs to be pre-

processed in order to be used by the algorithm. For

example qualitative data needs to be transformed into

quantitative data.

Step 2. Initialize the input vectors. Each student is

represented by a vector with features/components that

are made up of the attributes values associated with

the student, initialized from the questionnaire.

Step 3. Select and run the clustering algorithm (e.g.

K-means, hierarchical clustering, etc) in order to gen-

erate the groups; depending on the algorithm select

the number of clusters, the size of the cluster or the

quality criterion. One distance measure that can be

used to compute similarity between students in groups

is for example the Euclidean distance:

d(x, y) =

∑

− y

|, (1)

where x = [X

] i = 1..n represents one student’s proﬁle

and y = i = 1..n represents another student’s proﬁle.

Once the grouping task is achieved, the students can

work on their assignments in groups formed.

Step 4. Evaluate the group performance in relation

with the selected criterion or criteria. Depending on

the assigned task and the learning objectives, the pro-

fessor might decide to evaluate the performance of the

groups and may decide to change or keep the type of

clustering method for the following assignments. The

performance may be evaluated at either group level or

individual level.

5 CONCLUSIONS

This paper investigates group formation in the con-

text of a collaborative learning platform. This pa-

per proposes a methodology to form groups of stu-

dents which relies on a preliminary study on how

students formed groups within a course in a hetero-

geneous class. A pilot study has been conducted to

study the way students form groups in the context

of heterogeneous classes. According to the ques-

tionnaires ﬁndings, students have enjoyed working in

groups and have perceived positively their workgroup

collaboration. However not all students in the class

have formed a group. Furthermore, the analysis of

the groups reveals that students tend to form homoge-

neous groups in terms of level of knowledge, which

is in line with what other previous studies have sug-

gested (Souren et al., 2003). In relation with the top-

ics of interests, according to the students’ answers,

the groups were formed based on topics of interests

but according to the analysis of groups as presented

in section 3.3.2 groups are not as homogeneous as

students declared. The proposed methodology using

clustering algorithm tools for group formation is a

useful tool to help professors form groups automat-

ically using a certain criteria. We plan to further test

the methodology and the performance of automati-

cally formed groups within different type of courses

and assignments. In further studies we will assess

the performance of automatically formed groups us-

ing different clustering methods.

COLLABORATIVE LEARNING IN HETEROGENEOUS CLASSES - Towards a Group Formation Methodology

193

REFERENCES

Abrami, P. C., Poulsen, C., and Chambers, B. (2004).

Teacher motivation to implement an educational inno-

vation: Factors differentiating users and non-users of

cooperative learning. Educational Psychology, pages

201–216.

Alfonseca, E., Carro, R., Martn, E., Ortigosa, A., and Pare-

des, P. (2006). The impact of learning styles on stu-

dent grouping for collaborative learning: A case study.

User Modeling and User-Adapted Interaction Jour-

nal, pages 377–401.

Bekele, R. (2005). Computer-Assisted Learner Group For-

mation Based on Personality Traits. PhD thesis, Ham-

burg University.

Bradley, J. H. and Herbert, F. J. (1997). The effect of per-

sonality type on team performance. Journal of Man-

agement Development, pages 337–353.

Chan, T., Chen, C. M., Wu, Y. L., Jong, B. S., Hsia, Y. T.,

and Lin, T. W. (2010). Applying the genetic encoded

conceptual graph to grouping learning. Expert Sys-

tems with Applications, 37:4103–4118.

Dansereau, D. and Johnson, D. (1994). Learning, Remem-

bering, Believing: Enhancing Human Performance,

chapter Cooperative learning, pages 83–111. Druck-

man, D. and Bjork, R.A.

Florea, A. M. (1999). An agent-based collaborative learning

system: Advanced research in computers and commu-

nications in education. In Proc. of the Int. Conf. on

Computers in Education.

Graf, S. and Bekele, R. (2006). Forming heterogeneous

groups for intelligent collaborative learning systems

with ant colony optimization. In Intelligent Tutoring

System Conference.

Ho, T. F., Shyu, S. J., Wang, F. H., and Li, C. T. (2009).

Composing high-heterogeneous and high-interaction

groups in collaborative learning with particle swarm

optimization. In World Congress on Computer Sci-

ence and Information Engineer-ing.

Johnson, D. W. and Johnson, R. T. (1994). Perspectives on

Small group Learning: Theory and practice, chapter

What is cooperative Learning? Rubicon Publishing

Inc.

Kirchner, K., Razmerita, L., and Sudzina, F. (2008). Web

2.0: The Business Model, chapter New Forms of Inter-

action and Knowledge Sharing on Web 2.0. Springer

Science and Business Media.

Krejins, K., Kirschner, P. A., and Jochems, W. (2002). The

sociability of computer-supported collaborative learn-

ing environ-ments. Educational Technology and Soci-

ety, pages 26–37.

Martin, E. and Paredes, P. (2004). Using learning styles

for dynamic group formation in adaptive collaborative

hypermedia systems. In Proceeding of the conference

AHCWS.

Mondahl, M., Rasmussen, J., and Razmerita, L. (2009).

Web 2.0 applications, collaboration and cognitive pro-

cesses in case-based foreign language learning. In 2nd

World Summit On The Knowledge Society.

Nijstad, B. and Carsten, K. (2002). Creativity and group

innovation. Applied Psychology, 51(3).

Paulus, P. B. and Nijstad, B. A. (2003). Group Creativity:

Innovation through Collaboration. Oxford University

Press.

Razmerita, L. and Kirchner, K. (2009). Personal knowledge

management: The role of web 2.0 tools for manag-

ing knowledge at individual and organisational levels.

Online Information Review, 33(6):1021–1039.

Rich, M. (1997). A learning community on the internet: An

exercise with masters students. In Proc. of Americas

Conf. on Information Systems.

Sharan, S. (1999). Handbook of cooperative learning meth-

ods. Westport.

Siemens, G. (2010). Connectivism: A learning theory

for digital age. http://www.elearnspace.org/ Arti-

cles/connectivism.htm retrieved the 16th of May.

Slavin, R. E. (1983). When does cooperative learning in-

crease achievement? Psychological Bulletin.

Slavin, R. E. (1987). Developmental and motivational per-

spectives on cooperative learning: A reconciliation.

Child Development, pages 1161–1167.

Slavin, R. E. (1995). Cooperative learning: Theory, re-

search, and Practice. Allyn & Bacon.

Souren, P., Seetharaman, P., Samarah, I., and Mykytyn,

P. (2003). Impact of heterogeneity and collaborative

conﬂict management style on the performance of syn-

chronous global virtual teams. Information and Man-

agement, pages 303–321.

Wang, D. Y., Lin, S. J., and Sun, C. T. (2007). Diana:

A computer-supported heterogeneous grouping sys-

tem for teachers to conduct successful small learning

groups. Computers in Human Behavior.

Webb, N. (1989). Peer interaction and learning in small

groups. Intern. Journal of Educational Research,

pages 21–39.

CSEDU 2011 - 3rd International Conference on Computer Supported Education

194