7
Chapter 1. Collaborative Learning
Collaborative learning is en vogue. Recent advances in educational psychology as
well as new theories in organizational science have placed big expectations into
group learning and team work. The goal of this first chapter is to define more
precisely where collaborative learning comes from, to give a short overview of
the processes that are held responsible for learning in groups, as well as point out
difficulties encountered by practitioners of collaborative learning, with a special
The concept of collaborative learning covers a wide range of situations that differ
with regard to the size of the group, the definition of learning, the modalities of
• First, the size of the group may vary from a dyad (2 subjects) to a
small group (3-5 subjects), a class (20-30 subjects), and a community
(a few hundred to a thousand people) or even to society (several
• Second, ‘learning’ may refer among others to following a course,
studying course material, or performing learning activities like
• Third, the modalities of joint learning may encompass face to face or
distance interaction, simultaneous or deferred activity, joint activity
or organization of activity through some division of labor,
• Finally, activities labeled as collaborative learning may last from a
Dillenbourg (1999) points out the limits that the diversity of scales imposes
upon the generalizability of results from one particular study (e.g. a dyad
working together for one hour) to another (e.g. a class working on a project for
one semester). However, concepts developed within studies that cover one
particular timeframe or one particular group size are sometimes exported to
other contexts. For example, the concept of culture which stems from disciplines
like sociology and anthropology and is used to describe a phenomenon at the
level of society is sometimes used to describe the results of interaction among the
members of a dyad. Starting at the other end of the scale, the concept of common
ground (Clark & Brennan, 1991) which is initially used to describe at the micro
level how two persons manage to build and maintain shared understanding
during dialogue, is also used to describe the rules and collective knowledge held
Part I: Literature Review - Collaborative Learning
8
In the experimental part of this thesis, we chose to focus on dyads who work
together for one hour on a problem solving task.
A first way to define collaboration consists of differentiating it from
cooperation by using the degree of division of labor as a criterion. “In cooperation,
partners split the work, solve sub-tasks individually and then assemble the
partial results into the final output. In collaboration, partners do the work
'together'” (Dillenbourg, 1999, p. 8). However, there are situations where partners
adopt roles while still working ‘together’, thus collaborating. One partner might
take over low-level aspects of the task (e.g. manipulating the parameters of a
simulation) while the other might be in charge of strategic aspects of the task (e.g.
diagnosing the situation, planning the next actions). Dillenbourg (1999)
designates this way of splitting the task as ‘horizontal’ division of labor.
Cooperation on the other hand, consists in a ‘vertical’ division of labor, where the
task is split into sub-tasks that are accomplished independently. Also, in
cooperation, the vertical division of labor is set at the beginning of the interaction
and stays constant throughout the duration of the interaction whereas in
collaboration, the horizontal division of labor might constantly be renegotiated,
which results for instance in partners switching roles. To sum up, what
characterizes collaborative learning with regard to division of labor is the
flexibility of the division of labor, not the absence of any division of labor. Also, in
collaboration, division of labor is rather horizontal, in terms of roles, than vertical
in terms of subtasks accomplished independently.
A collaborative situation is further characterized by symmetry of action,
symmetry of knowledge as well as the existence of a common goal that is shared
by the participants. “… [A] situation is termed 'collaborative' if peers are more or
less at the same level, can perform the same actions, have a common goal and
work together” (Dillenbourg, 1999, p. 7). Symmetry of action means that all the
learners have access to the same repertoire of actions. Symmetry of knowledge
means that all the learners have a more or less equal level of knowledge in the
domain they are working in, which is not the case in tutoring situations.
Symmetry of participation, i.e. all the participants are active (not necessarily in
the same way) and contribute more or less equally to the learning activity, is
another characteristic of collaboration. We give more details about participation
problems in collaborative situations in Section 1.3. Finally, the existence of a
common goal differentiates collaboration from competition where subjects have
antagonist goals.
Sometimes the rules of symmetry are broken in order to structure interaction.
The symmetry of knowledge is one criterion used by Damon and Phelps (1989) to
distinguish between peer tutoring, where one student (the tutor) knows more
than the other (the tutee); cooperative learning where groups are heterogeneous
with regard to competence; and peer collaboration, where relative novices learn
together. In cooperative learning it is common to use tasks that require the
contribution of several specialists, each bringing a different kind of knowledge to
the group. This specialization of roles and uneven distribution of knowledge is
typically imposed through a jigsaw scenario for research purposes (e.g. Soller,
2002) or for instructional purposes (Aronson, 1978).
Part I: Literature Review - Collaborative Learning
9
Section 1.2. Research Overview
Research on collaborative learning includes numerous studies which differ in
terms of the population studied, the tasks that are used and the
operationalization of outcomes that they propose. Among the disciplines
involved in the study of collaborative learning, psychology studies the
mechanisms of learning (in the individual), social psychology studies the social
norms and representations that exist in any group (usually larger than two
members), educational science studies collaborative learning as a teaching
method.
Dillenbourg, Baker, O’Malley and Blaye (1995) propose a synthesis of the
evolution of research on collaborative learning by distinguishing three
paradigms that we employ here to present a short overview. The first paradigm
(the “effects” paradigm) which we won’t detail much, studies whether group
learning leads to better results than individual learning. These studies use the
term cooperative rather than collaborative to refer to “classroom techniques in
which students work on learning activities in small groups and receive rewards
or recognition based on their group’s performance” (Slavin, 1980, p. 315). Most of
the studies lead to an advantage of cooperative learning over individual or
competitive learning (Slavin, 1995; Johnson, Johnson & Stanne, 2000).
1.2.1. Conditions for success
The second paradigm (the “conditions” paradigm) tries to determine the initial
conditions that guarantee successful collaboration. Independent variables are the
size of the group, the heterogeneity of group members (in terms for example of
gender, academic achievement, or ethnicity) and finally the task features (for
example seat work or open-ended problem-solving).
Many different methods were used from the mid sixties to the late eighties to
structure group work. Most of the studies reviewed in recent meta-analyses
(Slavin, 1995; Johnson, Johnson & Stanne, 2000) show an advantage of
cooperative learning over competitive or individual learning. We will not give
much detail about each of these methods but rather discuss briefly what their
potential is and give extensive details later about structuring interaction through
computer tools (Chapter 4). According to Cohen (1994), structuring methods
range from simple task instructions about the activity (e.g. discuss to solve the
problem, make sure everyone understands), to detailed procedures that specify
what has to be discussed (e.g. establish a plan, evaluate results, etc).
• STAD (Student Teams-Achievement Divisions) is the most studied
methods for cooperative learning in classrooms. Students are
assigned to heterogeneous four-member teams in terms of ability,
gender and ethnicity. The teams’ task is to review the teacher’s lesson
so that all team members have mastered the content. Students then
take individual quizzes which are rated with regard to every
student’s relative progress. The individual scores are then added up
for each team and rewards are distributed to the best teams. Slavin
(1995) review 29 studies of STAD across all grades and possible
Part I: Literature Review - Collaborative Learning
10
subject matters and finds a positive median effect size of +.32 on all
tests and +.21 on standardized measures.
• The Team-Games-Tournaments (TGT) method is a variation of STAD
where individual quizzes are replaced by a tournament where
students of equal past academic performance compete in front of the
class to answer questions. The “fun” of the tournaments leads many
practitioners to use it in their classes. In the twelve studies reviewed
by Slavin (1995) TGT produced an effect size of +.38 on seven studies
and +.40 on four studies with standardized measures.
Three important features of these cooperative learning methods are that all
students must have equal opportunities for success regardless of their level of
ability, that individuals should be held accountable for their contribution and
that situation induces the team members’ interdependence (Slavin, 1995). The
notion of interdependence was applied to rewards, goals, and resources. Positive
reward interdependence refers to the distribution of rewards for the group as a
whole. Reward interdependence does not guarantee that all the team members
participate in the group effort: some participants might not do their fair share by
free-riding on the group’s effort. To foster fair and equitable distribution of labor
or effort, individuals are held accountable for their performance by adding an
independent evaluation of learning for each group member to the group reward
(Hymel, Zinck & Ditner, 1993 cited by McGrenere, 1996; see Section 1.3 for more
details about measures to counter participation inequalities). Goal
interdependence differentiates cooperation from competition. When goals are
interdependent individuals can achieve their goals only if other individuals from
their group also achieve their goals. Finally, resource interdependence refers to a
situation where resources are distributed in such a way that individuals can only
achieve their goals if their partners provide them with missing resources.
The methods that we present now rely on some division of labor and the
assignment of roles, either based on a partition of the content to be learned, or on
the performance of a particular cognitive function.
• With Group Investigation (Sharan & Sharan, 1992), students form
groups by themselves and work in parallel on parts of the curriculum
using cooperative inquiry, group discussion and cooperative
planning and projects. Group Investigation works best for openended
problems where there is not one single solution to be found
but where a domain has to be explored. The studies reported by
Slavin (1995) produced rather small effects with the exception of one
study that produced an effect size of +1.43. Slavin reports that this
study took place after teachers used Group Investigation for several
months thereby enabling the teachers to be fully competent. Another
explanation for this effect would be that the students became
competent collaborators and were able after several months to fully
take advantage from the method. Later in his book, Slavin (1995)
points out that the prior acquisition of group skills is critical for the
method’s success.
Part I: Literature Review - Collaborative Learning
11
• Following the Jigsaw method (Aronson, 1978), students also work in
heterogeneous groups of four. Each team member is assigned the role
of being an expert in one aspect of the task. For example, imagine that
the problem at hand is security policy about earthquakes in the San
Francisco Bay area (Dillenbourg, personal communication). One
student would adopt the role of mayor of San Francisco, another
would play the insurance agent, still another would be the geologist
and the last one would play the head of police department. After
reading the material, all the team members who play the same role
would meet together, to discuss their point of view, and then rejoin
their groups to teach their team members about their point of view on
the question. Slavin (1995) reports that the outcomes from Jigsaw
methods are highly variable.
• Working in a group allows for work to be divided through horizontal
division of labor. Horizontal division of labor allows for the
distribution of cognitive functions over participants and potentially
decreases the individual work load. One participant would for
instance be in charge of strategic aspects of the task while another
takes over more operational aspects. This might explain why
members of a group get better at regulating each other, and
ultimately oneself. Work by Brown and Palincsar (1989) on reciprocal
teaching of reading shows for example that the main benefit of
interacting with other learners is above all enhanced metacognitive
capabilities. Another example is King’s (1998) method which consists
of assigning roles (explainer, questioner) to the students and training
them to ask each other “thought-provoking” questions (see Section
3.5, p. 49 for more details).
In Group Investigation and the Jigsaw methods, individual accountability is
replaced by the “unicity” of each group’s contribution which motivates students
to contribute to the team effort (see also 1.3.2, p.18). According to the principle of
unicity, students are motivated to participate because they believe that their
contribution to the group is unique and important. Slavin (1995) points out
students may not learn about all the aspects of the task when work is divided like
in Jigsaw or Group Investigation.
According to Cohen (1994) the only stable result with regard to group
composition is that low achievers benefit from being in a heterogeneous group as
compared to a homogeneous group of low-achievers. The difference of level
between low and high achievers should however not be too large because low
achievers might become passive when collaborating with high achievers
(Salomon & Globerson, 1989). Cohen also points out the negative effects of status