2012 45th Hawaii International Conference on System Sciences
Cognitive Load in Collaboration - Convergence
Gwendolyn L. Kolfschoten
Delft University of Technology
Faculty of Technology, Policy and Management
[email protected]
Frances Brazier
Delft University of Technology
Faculty of Technology, Policy and Management
[email protected]
Abstract
because it is the key to resolve paradoxes in research
findings with respect to the use of collaboration support
[6]. To gain an understanding of the effect of
interventions in collaborative effort (made though
collaboration support) and to resolve the conflicting
findings with respect to the effects of collaboration
support, interventions in collaboration need to be studied
in more detail, with a focus on cognitive effects of these
interventions [6, 7].
Collaboration is inherent to complex
participatory multi-actor and multi-agent social
technical systems. Supporting collaboration is
challenging. One of the key production factors in
collaboration is cognitive effort. Understanding
cognitive load involved in collaborative tasks is
therefore important to the design of collaboration
support. This paper focuses on cognitive load related to
convergence, a very complex collaborative task, that is
much less studied than the often preceding, divergence
or brainstorming task. On the basis of an overview of
convergence techniques, and literature on convergence
this paper presents a framework for the assessment of
cognitive load during collaboration processes, and
strategies to deal with cognitive load in convergence.
The paper ends with a reflection on the use and
implications of the framework.
To design interventions that improve cognitive
efficiency and effectiveness and reduce the demand on
the central executive in collaborative tasks requires an
understanding of cognitive activities and processes in
collaboration. This paper presents first steps in
developing a framework for assessment of cognitive
load in a collaboration context. Based on literature and
examples of existing collaboration support techniques
and tools the role and design implications of cognitive
load in the convergence phase of collaborative problem
solving tasks are explored. Initial validation of the
framework is offered though evaluation by four expert
facilitators.
1. Introduction
Collaboration is a sine qua non for innovation
and productivity of organizations[1]. Knowledge
intensive collaborative tasks often require high cognitive
effort. Collaboration implies a team to perform a task
jointly, thus requiring interaction and coordination of
cognitive effort[2]. Coordination and support of
collaboration can be offered by a group member
(leader), or by an external facilitator or mediator.
Furthermore, tools and training can offer guidance in
collaborative activities [3]. Cognitive load is typically
higher for collaborative tasks then for individual tasks
[2], and therefore groups often benefit from tools and
facilitation to structure their cognitive effort[4].
Examples of such tools are social software, workshops,
facilitated (electronic) meetings or discussions, and
Group Support Systems.
2. Cognitive Load
Cognitive load can be defined as the cognitive
effort made by a person to understand and perform
his/her task [8]. Cognitive load has both a task-based
dimension (mental load) and a person-based dimension
(mental effort) [5, 9]. The task based load, further has a
perceptual and a cognitive dimension, related to the
amount of information presented, and the amount of
information that needs to be processed in the working
memory [10]. The concept of cognitive load in cognitive
load theory is associated with the seminal assumption
that our short-term or working memory, also called
central executive is limited [11-13]. The central
executive in this theoretical model is a simplification of
the various parts of the brain that are involved in
cognitive processing of information. Problem solving
tasks are mainly associated with the prefrontal cortex,
which is also used to recall things from memory and to
inhibit distraction [14]. While cognitive- and
neuropsychologists offer ample debate on the different
ways in which cognitive tasks are performed in the
brain, they agree that our capacity to process
information is limited, and that these limitations are
actively experienced in problem solving tasks [14].
Furthermore, research in instructional design has shown
Cognitive load is the cognitive effort made by a person,
required to perform a task [5]. Cognitive load theory
(CLT) distinguishes various design principles to
efficiently and effectively use cognitive capacity [5] in
the context of learning. However, in the context of
collaboration support, less research is devoted to
understanding the cognitive implications of process and
technology design.
Recently scholars have indicated that the cognitive
perspective on collaboration requires further research,
978-0-7695-4525-7/12 $26.00 © 2012 IEEE
DOI 10.1109/HICSS.2012.156
129
ample evidence that the way in which information is
offered and structured has significant effects on
performance in problem solving [8].
[5]. The same holds for intrinsic cognitive load that is
directly related to task complexity. Pollock et al [20],
propose that the intrinsic cognitive load can be reduced
for by pre-structuring complex information. By prestructuring the information, it can be absorbed in parts
(requiring less cognitive effort) that are meaningfully
connected, which facilitates schema building [16]. Also
research has reported initial results on stimulating
learners explicitly to construct and automate schema,
increasing productive germane load [21]. CLT has
provided new insights in instructional design improving
learning efficiency and effectiveness.
Besides short-term memory, the theory also
assumes people have a long-term memory, which stores
large quantities of information, in so called schemata
[8]. Information in long term memory is related,
networked or associated in these schemata [14].
Through a process called automation, and also called
chunking [11, 12], larger schema can be used in the
working memory as a single component, allowing the
processing of more complex information. Problem
solving is a complex task, and can benefit from support.
Early research from Simon et al already indicated that
we can learn a lot about problem solving and how to
support it, when looking at it from a cognitive
perspective [15]. Cognitive overload causes effects such
as impaired performance and decision making, stress,
difficulty to retrieve knowledge, impeding creativity,
and difficulty to analyse and organize knowledge,
impeding schema building and learning [16]. Solutions
to cognitive overload are amongst others to plan and
regulate information flow, to structure information and
to use intelligent agents to analyse and filter information
[16]. Cognitive overload is ineffective, but too low
cognitive load can also be ineffective as it leaves
cognitive capacity for distraction, which can then take
attention away from the task. The cognitive load of a
task differs for each individual, depending on their
experience in the domain and skill in the type of
problem solving task [17, 18]. Furthermore, people can
get distracted while performing a task, or use an
ineffective way to process the information due to fatigue
or lack of skill [14].
In collaboration, cognitive load has many sources. It can
originate from the information shared among
participants through various communication channels,
from constructing and thinking up new information,
from explaining or arguing positions, from assessing
value, implications and effects of decisions, from
various procedures, and from distractions. In previous
work we explored cognitive load involved in
brainstorming [22]. In the next section cognitive load in
a difficult cognitive task: convergence will be explored.
3. Convergence
When groups collaborate they often go through a goal
oriented problem solving or design process with roughly
three phases; they diverge to gather, share or brainstorm
information. Second, they analyze and converge the
information available to create meaning and shared
understanding. Third, they make decisions based on the
information analyzed. This paper focuses on
convergence; the process of analyzing and organizing
the information shared in a group. Divergence (also
called generation or brainstorming) [23] often produces
a large volume of content of varying relevance, across
multiple levels of abstraction and of varying granularity.
This knowledge, shared and created by a group, needs to
be converged to a manageable size, to create an
overview of its content, to be used for further analysis,
evaluation or decision making. There are many ways to
approach convergence, and there are significant
differences in the cognitive effort they require, and the
rigor with which the convergence is achieved. Typically,
the goal of convergence is not yet to choose among
alternatives, rather it is to create a parsimonious
overview of alternatives. However, we acknowledge that
some judgment or evaluation is likely to happen as ‘out
of scope’ alternatives could be removed, similar
alternatives can be merged, and strategic behavior can
occur in this process. Furthermore, convergence
activities can give rise to new ideas, and consequently
causing ideation and divergence. The exclusive focus in
this paper on convergence is to scope the paper. For a
full collaboration process design, cognitive activities in
brainstorming and evaluation should also be considered.
Cognitive load theory (CLT) in the context of learning
explains how cognitive capacity is used to construct
schemata and use them for problem solving tasks. Three
types of cognitive load [8] are distinguished:
• Intrinsic cognitive load is the cognitive load that is
inherent to the task, defined by the intrinsic task
complexity.
• Extraneous cognitive load is the cognitive load
caused by the presentation and transition method of
the information. Extraneous load should be reduced
as much as possible, as it is ineffective. However, it
cannot be completely eliminated.
• Germane cognitive load is the cognitive load
instrumental to building schemata and storing them
in the long term memory. For learning, germane
load should be stimulated.
Limiting unproductive mental activity is a
critical challenge for information systems [19],
particularly those that support collaboration and
collaborative effort. CLT provides different methods to
reduce extraneous cognitive load such as providing
parsimonious information elements, avoiding split
attention by disintegrated or unstructured information
We decompose the convergence process in six phases.
First there is a preparation phase in which the group
130
pool of information considered by the group (cherry
picking). In this case, information that meets e.g. the
criteria of fit and consistent are selected.
familiarizes itself with the convergence task, next the
group analyses the set of contributions for convergence.
After analysis contributions are actually converged;
clarified, integrated or related. The next phase involves
reflection on the implications of individual convergence
activities, possibly leading to further convergence or
adjustment of convergence. Furthermore there is overall
reflection. Once the converged set of information has
taken shape, the total set is re- considered for overall
quality. Again this can lead back to convergence and
analysis activities. Finally the group can be distracted in
their activity, which is a ‘phase’ parallel to the other
phases.
A second approach to reduction is a form of
summarization. Rather than filtering, the aim of this
approach is not to select part of the information, but
rather to capture the essence of information with fewer
information elements. Redundant information is
removed; but no choices are made with regard to which
remaining information is preferred over other
information.
There are several approaches to summarizing. First,
summarizing can be accomplished by selecting only
unique information. Second, similar contributions can be
merged, to keep only unique information. Third, an
instance of similar pieces of information can be selected
to represent multiple instances. Last, extraction of the
most important information can also be used as a
summarizing method [30].
Convergence has three key aspects: (1) creating shared
understanding, (2) eliminating redundancy, similarity
and overlap, and (3) creating overview and structure in a
set of contributions by identifying relations among
contributions [24, 25]. Not all convergence processes
focus on all three aspects, teams might initially have a
good shared understanding, or they might have
structured their brainstorm session effectively,
diminishing the need for organizing in the convergence
phase.
A third approach to reduction is to reduce information
through abstraction. The purpose of abstraction is to
make content more intellectually manageable by
allowing group members to pay attention to relevant
information and to ignore other details. Smith et al. [31]
describe two approaches for abstraction: generalization
and aggregation. Generalization refers to an abstraction
in which a set of similar objects is regarded to be of a
specific generic type/object, aggregation refers to an
abstraction in a hierarchical relationship between
objects.
The difference between summarizing and abstraction is
that in summarizing the same information is represented
with fewer information elements while in abstraction
information is characterized by higher level concepts
that encompass relevant information in the original set.
3.1 Clarify for shared understanding
To create shared understanding, entails; creating shared
meaning of language symbols and labels, resolving
asymmetry of information and resolving differences in
mental models [26]. Mulder, Swaak, & Kessels [27]
explain the notion of shared understanding as mutual
knowledge, mutual beliefs, and mutual assumptions.
Groups achieve shared understanding when they come
to a common understanding of concepts and words that
are related to the task at hand.
Sense making is done to prepare a group to act in a
principled, and informed manner [28]. Sense-making
tasks often involve searching for contributions from
others in a group that are relevant for the purpose at
hand and then extracting and reformulating the
information so that it can be used [29]. Note that sensemaking requires more than shared understanding of
concepts and terms; it also requires a common
understanding of a shared problem, its context, and
possible solutions.
3.3 Organizing or structuring
Finally, a group can choose to identify and document
relations among contributions in order to structure the
information, to make it easier to store in memory, and to
create an overview of the knowledge for analysis and
decision making. Creating such structure can be done by
relating information. The most common relations are
abstraction to identify hierarchical relations and
sequencing to create temporal relations [32]. Further,
groups can create causal relations to describe cause and
effect. Categorization, sometimes referred to as
classification, is the most common form of organize
[33], and is often used as a step after brainstorming. A
similar activity is performed in the creation of a shared
ontology [34], in which shared meaning is created of
concepts and their relations.
3.2 Reduction, abstraction or filtering
The next phase in convergence is to reduce the actual
amount of information considered, the knowledge
shared and created in a group. A group can choose to
select key concepts. Two approaches to selection are
distinguished. First a group can filter information; each
piece of information is considered and kept, adjusted or
discarded, to reduce redundancy and increase
consistency. Second, a group can choose an approach
where key concepts are identified and selected from the
Sequencing is mainly found in scheduling, planning,
workflow and project management [35, 36]. Causal
modeling in collaborative setting is researched in Group
Model Building literature [37, 38] and in soft systems
131
methodology (strategic options development and
analysis) [39]. Causal relations are often more complex
for participants to understand, and many of these
approaches prescribe the role of a modeler to support the
group in capturing relations and guarding consistency.
Another approach is to use shared task and decision
modeling [40].
which might be more or less required depending on the
task, group and experience of group members. We can
classify cognitive activities to a convergence goal, based
on which the extent to which they are effective (on-task
load) or ineffective (off-task load). Also we distinguish
cognitive effort that is required for the process, but not
directly effective, similar to the ‘extraneous cognitive
load’ in instructional design.
In convergence tasks, especially for organizing
visualization is intuitively used to reduce cognitive load.
Complex concepts and contributions are often visualized
to explain and convey them to others. Visualization can
be any style of capturing information for all to see, from
writing/typing to drawing to modelling. This is
especially the case in modelling, where often nodes and
their relations are visualized in a specific language.
Research has indicated that visual attention is
competitive; on can only hold on to one or a few mental
images given our limited cognitive capacity. Therefore,
it is very difficult for people to understand someone
else’s visualization. This makes collaborative modelling
particularly challenging, as it will require people to be
open minded to visualizations of others, and to switch
between their own visualization and those of others.
The overview of the cognitive processes involved in
convergence below is organized in the six phases of
convergence describe above. Cognitive activities and
their cognitive implication are listed in the table below,
and numbered in the text with numbers between
brackets ().
4.1 Cognitive activities in Convergence
Phase1: Preparation. In the preparation phase a group
receives (listens, reads) a specific convergence task (1)
and process it for understanding (2). Further, they
receive instruction about the tools and methods that will
be used for the purpose of convergence (3), observe,
study or try it (4) and understand it to infer their
personal task (5). Finally, a group needs to make the
transition from preparation to the actual convergence
activity (6).
4. Cognitive Load of Convergence
To create an overview of cognitive activities in
convergence this paper focuses on ten convergence
ThinkLets. ThinkLets capture best practices in patterns
of collaboration [41-43]. The thinkLets used are:
CheckMark, StakeholdersPoll, Relate-it, Evolution,
FastFocus, BucketWalk, PopcornSort, Concentration,
Goldminer and ThemeSeeker, as documented in [44].
One of the authors has facilitated over a hundred
workshops in education and industry settings based on
thinkLets, including the listed convergence thinkLets,
and thus has experience in conducting these techniques.
ThinkLets are documented as scripts and prescribe how
to facilitate a collaborative activity. For each thinkLet
this paper identifies the cognitive activities required to
perform the thinkLet, and lists these in a table. Overlap
of cognitive activities was then identified and removed.
Next, cognitive activities are grouped in the
convergence phases to create a complete list. The
resulting set of cognitive activities are compared to the
convergence aspects found in the literature described
above, to verify completeness of the convergence
activities. This completeness check also considers basic
cognitive activities; understand, decide, recall (from
memory), memorize (store in memory) and inhibit (to
push away distracting thoughts or stimuli) [14]. In this
way an overview of the cognitive activities in
collaborative convergence is created, which we will then
classify based on their effectiveness. .
Phase 2: Analysis. In this phase contributions are
verified and marked if they are to be processed in a next
step. Contributions are analyzed for fit to the scope (8)
to identify similarity (9), to identify relations (10), to
verify clarity (11), to verify fit to the class in which it is
classified (12), and to assess its importance with respect
to personal stakes (13). Next those contributions that
don’t fit the scope (14), are inconsistent relations or
classifications (15), are unclear (16), or are similar
(17), are marked. Markings can be made with pens,
colors, though re-location, or verbally. Visual marks
help identification for the purpose of retrieval.
Phase 3: Core Convergence. In the convergence
process firstly, contributions can be marked to be
included in the converged list (18). Next, similar
contributions can be merged (19), or different
contributions, can be rephrased to elicit their uniqueness
(20). Contributions can be related. For the purpose of
simplifications relations (causal, sequential or other)
(21) are distinguished from abstract classes (22).
Relations can also be visualized (23). This can be done
e.g. by linking contributions, marking the contributions
in the same way, locating them together, or coloring
them.
Relations, visualizations and mental models can be
inconsistent, causing the revision of relations, or reclassification of contributions (24). When summarizing
similar or related contributions, a new contribution is
created that encompasses the content of these
contributions (25). When contributions are unclear, they
Effectiveness of convergence depends on the context of
the problem solving task; it can create reduced
complexity, shared understanding, and overview
(interrelatedness, system) of the shared information,
132
can be rephrased to make them clear (26), or the can be
explained to the group, to create shared understanding
(27). Finally, contributions that don’t fit the scope of
the convergence task can be rephrased to ensure the fit
(28), or they can be discarded (29). Once contributions
are converged, they become clearer, and they form a
structure. Participants often have personal stakes in the
structure. For instance the structure can reflect one
perspective better than another perspective, or the
structure can have implications for future tasks or
implied quality or relevance of the contributions.
Therefore, convergence activities are designed to
consider implications. When implications are considered
to be negative, they will cause iteration to the
convergence steps to for instance identify different
relations, different clarifications or classifications.
reflection determines which information should be
reconsidered and thus put back in working memory.
As discussed above convergence can also create synergy
when the result of analysis is of better quality than its
components, or when it inspires new contributions,
which are now classified under distraction, as there is an
entire set of cognitive processes that are then triggered
[22]. Synergy and inspiration thus cause a loop back to
the divergence phase of collaboration. While this loop is
not part of the focus of this study, its importance for
further research is acknowledged.
Trust is a determining factor that needs to be considered
in convergence processes. For instance, if a group is
split up into sub-groups, and each sub-group assigned
different parts of a convergence task, sub-groups need to
trust each other. Group members have to trust other
group members not to discard (29) valuable
contributions, or manipulate a set of contributions to
focus on a specific perspective (21,22,26,28). If such
trust does not exist, all group members have to verify all
convergence activities for inclusiveness. This can be
time consuming, but critical if the converged list is input
for (democratic) decision making. Further, the group can
consider reviewing each other’s work jointly doing the
reflection (30-34) and overall reflection tasks (35-39),
while distributing the analysis and convergence tasks
[45, 46].
Phase 4: Reflection. Convergence activities can have
implications for personal stakes when contributions are
discarded (30), when the label for the abstract class is
chosen (31), when a relation is elicited (32), when a
contribution is clarified or rephrased (33) or when a
contribution is selected to be included in the converged
list (34).
Phase 5: Overall reflection. Once the set of converged
contributions is taking shape, it can also be evaluated as
a whole. This can be based on its size (35) for sufficient
complexity reduction, based on its fit to personal stakes
and perspectives (36), based on its completeness (37), its
consistency (38) and its overall quality (39). Overall
quality reflection can also lead to reflection on the
generation or brainstorming phase. As the converged list
of contributions improves, providing more overview of
the content of the set, the next step is to evaluate the
quality of the whole set on criteria such as e.g.
usefulness, innovativeness or feasibility.
To further validate the overview of cognitive activities
and their (in)effectiveness, we asked four expert
facilitators to evaluate the overview of tasks. The
facilitators had an average of 7 years of facilitation
experience, and worked in academia and industry. Their
average age is 40. We asked them to consider a
convergence task, and to evaluate based on their
experience if the cognitive tasks listed will typically
occur in this task. Further we asked them to classify the
cognitive activity in one or more categories of
effectiveness, and to report any missing tasks. We
repeated this exercise twice with a clarification round of
differences to reach more consensus among the experts.
There were no missing tasks reported, but two tasks
were merged as they were similar.
Phase 6: Distraction. Distraction can exist of a loop
back to ideation (40) (see below), personal thoughts
(e.g. other tasks, personal considerations) (41) and
external distractions such as a siren, or a picture on the
wall (42)
In general many organized convergence activities are
about identifying relations, similarities or inconsistency
in a set of contributions. Contributions are pieces of
information provided by individual participants such as
for instance ideas from a brainstorming session. It is
often not possible for a single individual to consider an
entire set of contributions at once, as it is larger than the
capacity of any one individual's working memory.
Therefore an underlying coping strategy is to add new
contributions to an existing set of contributions under
consideration in the working memory (7) and to discard
contributions that are no longer relevant (29) or that are
marked for a next step (14,15,16,17,18) and thus can be
retrieved when required. This makes it possible for the
set for consideration to be maintained within the
capacity limits of the working memory. Overall
Table 1 lists the numbered cognitive activities
distinguished from the 10 Convergence ThinkLets, and
our experience with these ThinkLets in practice, and
indicate if they are ineffective, extraneous (needed for
the process, but not contributing to outcomes) or
effective for reduction, shared understanding or
overview of the set. The total scores of the experts are
expressed in a percentage. Some tasks were required,
but not for the process, and neither for directly for any
of the outcomes, as they are required for the cognitive
process of convergence, to deal with it’s complexity.
133
15 mark inconsistently
related contribution
16 mark unclear
contribution
17 mark similar
contributions
convergence
18 mark contribution for
inclusion in converged
list
19 merge similar
contributions
20 rephrase similar
contribution to articulate
uniqueness
21 label relation of
contributions
ineffective
classify
100
0
50
0
75
0
100
0
50
0
100
0
100
0
0
0
75
0
100
0
0
0
75
0
100
0
50
0
100
0
100
0
25
25
100
0
100
Reduction
process
1 receive (listen, read)
the convergence task
2 understanding the
convergence task
3 listening to the
explanation of tool &
method
4 study/observe/try out
the tool & method
5 understanding the
convergence tool &
method
6 cognitive effort to
make the transition from
understanding the
convergence task to
performing it
analysis
7 store contributions in
memory to consider set
of contributions
8 consider if the
contribution fits the
scope
9 consider set of
contributions to identify
similarity
10 consider set of
contributions to identify
relation
11 consider contribution
to verify clarity
12 consider contribution
to verify fit to class
13 consider contribution
to assess importance
with respect to personal
stakes
14 mark contribution that
is out of task scope
overview/organization
Preparation
EXPERIENCE Yes/No
cognitive activity
shared understanding
22 label abstraction of
contribution(s) as class
25
0
50
25
25
100
50
25
0
100
50
100
100
25
0
100
25
100
100
25
0
100
0
100
25
25
0
100
25
75
50
25
0
0
25
0
0
32 consider implications
of creating relation
33 consider implications
of clarification and
rephrasing
34 consider implications
of selection for
converged set
overall reflection
35 reflect on size of
converged list
36 reflect on inclusion of
stake/perspective in
converged list
37 reflect on
completeness of
converged list
38 reflect on consistency
of converged list
39 reflect on quality of
converge list
distraction
40 identify new
contribution
41 personal distraction
0
100
100
23 visualize relation
between contributions
24 resolve inconsistency
in relation
25 phrase summarized
contribution
26 rephrase unclear
contribution
27 explain contribution
to the group
28 rephrase out of
scope contribution to fit
scope
29 discard out of scope
contribution
reflection
30 consider implications
of discarding
contribution
31 consider implications
of class choice
50
42 external distraction
100
75
50
75
25
0
75
25
25
50
25
0
100
0
100
0
25
0
100
75
50
75
25
0
100
75
25
25
25
0
100
100
75
50
50
0
100
25
100
0
25
0
75
0
50
50
25
0
100
25
50
75
50
0
75
25
25
100
0
0
100
0
50
100
25
0
100
50
75
50
25
0
100
0
100
25
25
0
100
0
100
0
25
25
75
0
25
25
0
25
100
75
50
50
25
0
100
25
25
25
25
0
100
0
25
25
25
0
100
25
50
25
25
0
100
0
75
0
25
0
100
25
25
0
25
0
100
0
0
25
0
50
100
0
50
0
25
0
100
0
50
25
0
25
100
25
50
75
25
0
100
25
50
25
25
0
100
0
50
25
0
75
100
0
25
25
25
75
100
0
25
25
25
75
The results are not yet conclusive on all aspects. But
some first tentative conclusions can be drawn. Firstly,
experts by majority agreed that all cognitive tasks
occurred in the convergence process described. With
regards to effectiveness of the tasks, experts noticed a
reinforcing effect in the outcomes; when the set of ideas
is reduced, it becomes easier to build shared
understanding and to organize, and vice versa. Therefore
it was often difficult to distinguish effectiveness for one
of the three outcomes. As expected, the experts also
noticed that some cognitive activities involve other
phases of the problem solving process such as
conceiving new ideas (brainstorming phase) and
reflecting on implications of ideas (evaluation and
decision making phase). We will now discuss the
effectiveness as rated by the experts.
134
For the preparation phase, experts found the activities
were mainly procedural. Cognitive load of these tasks
can be phased before the convergence task to reduce
cognitive load, and therefore a clear task and explicit
transition is critical. Some experts indicated that the
preparation activities can also foster shared
understanding.
which can then be used as a basis to learn the complete
information with the interaction to the initial framework
[20]. This is called the isolated interacting elements
approach. Associated is also the pre-training effect;
better transfer when the learner recognize the concepts
as part of a structure [47]. This effect is recreated when
relations are identified to structure and organize the
converged set of contributions. While it takes cognitive
effort to setup the structure at first, in the later phase of
the convergence activity, over time the structure
supports the activity making it easier to recognize
relations and to create consistency. A side effect is that
contributions are grouped or related. This relates to the
segmentation effect [47]; better transfer if material is
presented in segments, that can be controlled by the
learner, then as continuous whole. Thus, classifying
contributions in categories should reduce cognitive load
of creating an overview and holistic understanding of
the set of contributions as a whole. Parsimoniousness
of the set of contributions implies that the set of
contributions contains all information required to
understand the problem and solution, yet no interesting
but extraneous information [47]. Redundant information
increases cognitive load and has no added value to
performance or learning [48, 49]. This is called the
Coherence effect [47] or the redundancy effect [47].
Some convergence approaches have higher levels of
parsimoniousness of the output than others because of
the structure they offer, depending on the criteria they
focus on (scope, overlap, clarity, and inconsistency in
structure[50]).
Analysis tasks were generally not supporting reduction,
except for marking the ideas that are out of scope or
similar. Shared understanding and overview were
affected by the analysis tasks. Task 7, to hold ideas in
memory for comparison, was difficult to classify, as it is
instrumental to the other analysis tasks. Task 13, to
reflect on implications of ideas, is considered by some
as ineffective. Depending on a more rational or a more
social process of decision making following the process
this can be more or less important.
The convergence phase shows a more distinct consensus
of experts on the effectiveness of activities. Tasks that
reduce the list are distinguished from tasks that focus on
identifying relations for organizing and tasks focused on
rephrasing and resolving unclear ideas. Remarkable is
that for many activities that support organizing and
reduction, shared understanding is also listed by several
experts as effect. Task 21, labeling the relation between
ideas and task 28, rephrasing out of scope ideas where
difficult to attribute to a specific success factor, although
experts acknowledge the activities take place.
The reflection of implications of ideas, and implications
of the converged set were difficult to attribute. One
expert explained that these tasks can motivate the group
to improve convergence, and to detect deficiencies, but
do not directly contribute to the outcome. Some
considered these activities procedural, but others did not
see a coordinating effect of these tasks.
To reduce extraneous cognitive load, a first technique is
integrating information to reduce split attention. When
information is offered in separate components (e.g.
picture and separate text) and these components are not
self-explanatory, both need to be held in working
memory to process the information. Integrating text in a
picture will therefore require less cognitive capacity
[48], which is named special contiguity effect [47]. This
effect is very closely related to the way in which the set
of contributions and the converged set are visualized; for
instance, when collaborative modeling approaches are
used, their visualization can help to reduce split
attention. When converged contributions are clearly
marked, this can support the process, as marked items
can be quickly retrieved, reducing the need to analyze
the set again to identify the specific contribution.
However, too many different markings can increase the
complexity of the total set. Also, markings can help to
distinguish those contributions that have been
converged, from those that still need to be converged.
For verification purposes, and to avoid permanent loss
of discarded contributions, it is useful to keep a copy of
the original set of contributions. However, the
representation of the set to modify should not contain
both sets at the same time to avoid split attention.
Talking about the convergence process can facilitate the
use of both the visual and the audio channel for the task.
Finally, distractions were classified by a majority as
ineffective. However, a change of perspective can give
insight that leads to more shared understanding or a
different way to organize ideas. Further, experts noted
that a new idea is not really a distraction, but rather a
loop to the brainstorming phase.
5. Reducing cognitive load in convergence
Given the classification of the cognitive load
for each of the activities distinguished above, this
section explores means to reduce ineffective and
extraneous cognitive load in convergence, and to
facilitate groups to focus cognitive effort on those
cognitive activities that are most important for a specific
convergence task.
To this purpose the techniques used in educational
design to reduce cognitive load of learning and problem
solving tasks were evaluated for their applicability in
convergence. Pollock et al [20] suggest that in tasks with
high cognitive load information can best be taught in
two steps. First the learner is offered a basic framework
135
Based on these insights we offer five guidelines for the
design of convergence tasks:
1. First create a structure, and then identify relations
or classifications. When an initial structure is
created, the convergence task becomes easier. For
instance first identify key categories, then cluster
contributions, and identify categories for the
contributions that remain. In this way the structure
can facilitate the convergence activity.
2. Separate reduction from choice. When convergence
is combined with choice or preference selection, it
becomes much more complex. Keeping a
contribution on the converged list makes it larger,
but discussing inclusion of, or discarding the
contribution can take much more time, than the
extra effort of considering the extra contribution in
the next step.
3. Phase convergence in a ‘rough’ and ‘more refined’
step. In the first step contributions are converged
intuitively, in the next step they are verified more
specifically for fit in the scope, overlap, clarity and
consistency. This leaves fewer contributions for the
more specific analysis.
4. Phase clarification after reduction. Like the rough
and refined step, this reduces the amount of
contributions that need to be verified and when
unclear rephrased or explained, reducing the size of
the task.
5. Consider the need for collaborative convergence.
Convergence takes much effort from the group
when done collaboratively, while it is not always
required to reach consensus over the converged set
of contributions. Groups can successfully make
choices and decisions based on a fuzzy set of
contributions, especially when there is less conflict
and misunderstanding in the group. However, a lack
of participation in convergence can be an argument
to discard the outcome of a later decision phase.
Therefore, participation in convergence can be
critical. However, in tasks with less implication or
in groups that are more cohesive, rough
convergence, individual convergence or no
convergence can also function as a basis for
decision making.
6. Discussion and conclusions
This paper provides an overview of cognitive
processes involved in collaborative convergence, and a
framework within which the implications of approaches
for convergence processes are analyzed. For instance,
when focusing on shared understanding, the clarification
task is most critical, however, structure and reduction
can support this task by reducing it in size and offering a
mental framework that can serve as a basis for a shared
mental model. This framework can inspire designers of
collaboration support systems to create their systems in
a way that supports groups to benefit from the gains of
collaboration while reducing the need for ineffective or
less effective cognitive effort.
Further research is required to further evaluate the
framework for completeness and to verify the existence
of these cognitive activities in various convergence
tasks. An approach for this would be to ask participants
in a convergence process to evaluate the framework
directly after the task, ensuring better recall of cognitive
processes [51]. A similar framework has been devised
for the cognitive tasks in brainstorming [22]. Next,
additional frameworks need to be devised to understand
evaluation, decision making and consensus building.
Furthermore, interdependencies between ideation and
decision making and convergence tasks need to be
further examined. During analysis and re-structuring of
shared information synergy can be created; merging
ideas gives rise to better ideas, concepts or information
elements. While this effect is highly relevant, it seems to
be a secondary effect, not directly related to a single
cognitive task, but rather to a combination. Finally, three
styles or aspects of convergence are presented. They can
be combined or phased as activities. It is important to
further understand the interplaying effects of these
convergence activities on cognitive load.
References
[1]
[2]
Besides these design guidelines above, it is important to
offer tools (manual or electronic tools) for convergence
that enable visualization of relations between
contributions, and help the user keeping track of original
contributions and converged contributions. Further,
tools need to ensure that large contribution sets are
structured, paced or split over ‘views’ of smaller sets so
the amount of contributions that need to be considered
remains limited to keep cognitive load of reading and
reflecting low enough to leave capacity for convergence
activities.
[3]
[4]
136
Frost and Sullivan, "Meetings Around the World:
The Impact of Collaboration on Business
Performance," 2007.
Dillenbourg, P. and Betrancourt, M., (2006),
"Collaboration Load," in Handling Complexity in
Learning Environments: Research and Theory, J.
Elen and R. E. Clark, Eds., ed: Pergamon, pp.
142-163.
Dennis,
A.R.
and
Wixom,
B.H.,
(2002)"Investigating the Moderators of the Group
Support Systems Use with Meta-Analysis,"
Journal of Management Information Systems,
vol. 18, pp. 235-257,
Niederman, F., Vreede, G.J. de, Briggs, R.O., and
Kolfschoten, G.L., (2008)"Extending the
Contextual and Organizational Elements of
Adaptive Structuration Theory in GSS
Research.," Journal of the Association for
Information Systems, vol. 9,
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
Sweller, J., Merrienboer, J.G. van, and Paas,
F.G.W.C., (1998)"Cognitive Architecture and
Instructional Design," Educational Psychology
Review, vol. 10, pp. 251-296,
Heninger, W.G., Dennis, A.R., and Hilmer, K.,
(2006)"Individual Cognition and Dual-Task
Interference in Group Support Systems,"
Information Systems Research, vol. 17, pp. 415424,
Nijstad, B.A. and Stroebe, W., (2006)"How the
Group Affects the Mind: A Cognitive Model of
Idea Generation in Groups," Personality and
Social Psychology Review, vol. 10, pp. 186-213,
Sweller, J., (1988)"Cognitive load during
problem solving: Effects on learning," Cognitive
Science, vol. 12, pp. 257-285,
Paas, F.G.W.C. and Merrienboer, J.G. van,
(1994)"Variability of Worked Examples and
Transfer of Geometrical Problem-Solving Sills:
A Cognitive Load Approach," Journal of
Educational Psychology, vol. 86, pp. 122-133,
Fitousi, D and Wenger, M.J., (2011)"Processing
Capacity Under Perceptual and Cognitive Load:
A Closer Look at Load Theory," Journal of
Experimental Psychology: Human Perception
and Performance, vol. 37, pp. 781-798,
Miller, G.A., (1956)"The magical number seven
plus or minus two: some limits on our capacity
for processing information," Psychological
Review, vol. 63, pp. 81-97,
Simon, H.A., (1974)"How Big is a Chunk?,"
Science, vol. 183, pp. 482-488,
Farrington, J., (2011)"From the Research: Myths
Worth Dispelling: Seven Plus or Minus Two,"
Performance Improvement Quarterly, vol. 23, pp.
113-116,
Rock, D., (2009) Your Brain at Work. New York:
HarperCollins Publishers.
Newell, A., and Simon, H.A., (1972) Human
problem solving the state of theory in 1970
Englewood Cliffs: Prentice-Hall.
Eppler, J.M. and Mengis, J. , (2004)"The
Concept of Information Overload: a Review of
Literature
from
Organizational
Science,
Accounting, Marketing, MIS and Related
Diciplines," The Information Society, vol. 20, pp.
325-344,
Kolfschoten, G.L., Lukosch, S.G., Verbraeck, A.,
Valentin,
E.,
and
Vreede,
G.J.de.,
(2010)"Cognitive Learning Efficiency Through
the Use of Design Patterns in Teaching,"
Computers and Education, vol. 54, pp. 652-660,
Kalyuga, S., Ayres, P., Chandler, P., and Sweller,
J., (2003)"The Expertise Reversal Effect,"
Educational Psychologist, vol. 38, pp. 23-31,
Bray, D.A., (2008),"Information Pollution,
Knowledge Overload, Limited Attention Spans,
and Our Responsibilities as IS Professionals," in
GITMA World Conference Atlanta.
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
137
Pollock, E., Chandler, P., and Sweller, J.,
(2002)"Assimilating Complex Information,"
Learning and Instruction, vol. 12, pp. 61-86,
Paas, F., Renkl, A. , and Sweller, J.,
(2004)"Cognitive Load Theory: Instructional
Implications of the Interaction between
Information
Structures
and
Cognitive
Architecture," Instructional Science, vol. 32, pp.
1-8,
Kolfschoten, G.L., (2011) "Cognitive Load in
Collaboration - Brainstorming," presented at the
Hawaii International Conference on System
Science, Kauai, HI.
Briggs, R.O., Kolfschoten, G.L., Vreede, G.J. de.,
Albrecht, C.C., Lukosch, S.G., and Dean, D.L.,
(2009),"A Seven-Layer Model of Collaboration:
Separation of Concerns For Designers of
Collaboration Systems,"
in International
Conference on Information Systems.
Briggs, R.O., Kolfschoten, G.L., Vreede, G.J. de,
and Dean, D.L., (2006),"Defining Key Concepts
for Collaboration Engineering," in Americas
Conference on Information Systems, Acapulco,
Mexico.
Kolfschoten, G.L., Lowry, P.B., Dean, D. L.,
Vreede, G.J. de, and Briggs, R.O., (working
paper), "Patterns in Collaboration," in Advances
in Collaboration Science, J. J. Nunamaker, et al.,
Eds., ed: to appear.
Briggs, R.O., Kolfschoten, G.L., and Vreede, G.J.
de, (2005),"Toward a Theoretical Model of
Consensus Building," in Americas Conference on
Information Systems, Omaha.
Mulder, I., Swaak, J., and Kessels, J.,
(2002)"Assessing
learning
and
shared
understanding
in
technology-mediated
interaction," Educational Technology & Society,
vol. 5, pp. 35-47,
Ntuen, C.A., Leedom, D., and Schmeisser, E.,
(2005)"Performance of Individuals and Groups in
Sensemaking of Information from Computer
Decision Aiding Systems,"
Weick,
K.E.,
(1995)
Sensemaking
in
Organizations. Thousand Oaks, CA: Sage
Publications Inc.
Barzilay, R., McKeown, K.R., and Elhadad, M.,
(1999),"Information Fusion in the Context of
Multi-Document Summarization," in Annual
meeting of the Association for Computational
Linguistics on Computational Linguistics,
College Park, Maryland.
Smith, J.M., Smith, J.M., and Smith, D.C.P.,
(1977)"Database Abstractions: Aggregation and
Generalization," ACM Transactions on Database
Systems, vol. 2, pp. 105-133,
Brazier, F.M.T., Lange, P.H.G. van, and Treur,
J., (1998)"Strategic knowledge in design: a
compositional approach," Knowledge-Based
Systems, vol. 11, pp. 405-416,
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
Grisé, M. and Gallupe, R. B., (2000)"Information
Overload: Addressing the Productivity Paradox
in Face-to-Face Electronic Meetings," Journal of
Management Information Systems, vol. 16, pp.
157-185,
Stuckenschmidt, H. and Timm, I.J., (2002)
"Adapting Communication Vocabularies using
Shared Ontologies," presented at the Second
International Workshop on Ontologies in Agent
Systems, Bologna, Italy.
Chen, F., Romano, N.C. Jr., and Nunamaker, J.
Jr., (2003),"An Overview of a Collaborative
Project Management Approach and Supporting
Software," in AMCIS.
Deokar, A.V., Kolfschoten, G.L., and Vreede,
G.J. de, (2007)"Prescriptive Workflow Design
for Collaboration Intensive Processes Using the
Collaboration Engineering Approach," Global
Journal of Flexible Systems Management, vol. 9,
pp. 13-24,
Ackermann, F. and Eden, C., (2005)"Using
Causal Mapping with Group Support Systems to
Elicit an Understanding of Failure in Complex
Projects: Some Implications for Organizational
Research," Group Decision and Negotiation, vol.
14, pp. 355-376,
Rouwette, E.A.J.A., Vennix, J.A.M. , and
Mullekom, T. van, (2002)"Group Model Building
Effectiveness: a Review of Assessment Studies,"
System Dynamics Review, vol. 18, pp. 5-45,
Checkland, P.B., (1981) Systems Thinking,
Systems Practice. Chichester: John Wiley &
Sons.
Brazier, F., Treur, J., and Wijngaards, N., (1996)
"The Acquisition of a Shared Task Model,"
presented at the European Knowledge
Acquisition Workshop on Advances in
Knowledge Acquisition
Briggs, R.O., Vreede, G.J. de, and Nunamaker,
J.F. jr, (2003)"Collaboration Engineering with
ThinkLets to Pursue Sustained Success with
Group
Support
Systems,"
Journal
of
Management Information Systems, vol. 19, pp.
31-63,
Kolfschoten, G.L., Briggs, R.O., Vreede, G.J., de,
Jacobs,
P.H.M.,
and
Appelman,
J.H.,
(2006)"Conceptual Foundation of the ThinkLet
Concept for Collaboration Engineering,"
International Journal of Human Computer
Science, vol. 64, pp. 611-621,
Vreede, G.J. de, Briggs, R.O., and Kolfschoten,
G.L., (2006)"ThinkLets: A Pattern Language for
Facilitated and Practitioner-Guided Collaboration
Processes," International Journal of Computer
Applications in Technology, vol. 25, pp. 140-154,
Briggs, R.O. and Vreede, G.J., de, (2001)
ThinkLets, Building Blocks for Concerted
Collaboration. Delft: Delft University of
Technology.
[46]
[47]
[48]
[49]
[50]
[51]
138
Helquist, J.H., Kruse, J., and Adkins, M.,
(2006),"Group support systems for very large
groups: A peer review process to filter
brainstorming input," in Americas Conference on
Information Systems, Acapulco, Mexico.
Helquist, J.H., Kruse, J., and Adkins, M., (2008)
"Participant-Driven Collaborative Convergence,"
presented at the Hawaii International Conference
on System Science, Waikoloa, HI.
Mayer, R.E. and Moreno, R., (2003)"Nine ways
to reduce Cognitive Load in Multimedia
Learning," Educational Psychologist, vol. 28, pp.
43-52,
Sweller, J., Merrienboer, J.G. van, and Paas,
F.G.W.C., (1998)"Cognitvie Architecture and
Instructional Design," Educational Psychology
Review, vol. 10, pp. 251-296,
Mayer, R.E. and Moreno, R., (2002)"Aids to
Computer-based Multimedia Learning," Learning
and Instruction, vol. 12, pp. 107-119,
Kolfschoten, G.L., Grünbacher, P., and Briggs,
R.O., (2011)"Modifiers for Quality Assurance in
Group Facilitation," Group Decision and
Negotiation, vol. 20, pp. 685–705,
Ericsson, K.A. and Simon, H.A., (1993) Protocol
Analysis: Verbal Reports as Data. Cambridge:
MIT Press.