MANAGING INTER-ACTIVITIES IN CSCW
Supporting users emerging needs in the CooLDA platform
Gregory Bourguin, Arnaud Lewandowski
Laboratoire d’Informatique du Littoral (LIL), 50 rue Ferdinand Buisson, 62228 Calais, France
Keywords: CSCW, tailorability, dynamic integration, Activity Theory, Java.
Abstract: The CSCW research domain still tries to find a b
etter way for supporting the users needs. Some groupware
systems propose global and integrated environments supporting collaborative activities, but empirical
studies show that these environments usually omit to support some dimensions of the work. On the other
hand, some groups work with diverse applications that do not know each other. Mainly inspired by results
coming from Social and Human Sciences, we believe that the complete CSCW environment cannot be
defined a priori. However, we also believe that a global CSCW environment is really valuable for users.
Taking our foundations in the Activity Theory, we aim at creating a global but tailorable environment that
supports the dynamic integration of external applications and manages the links between them, i.e. it
manages the inter-activities. This work is concretized in the CooLDA platform.
1 INTRODUCTION
Following the impressive rise of communication
means, the CSCW (Computer Supported
Cooperative Work) holds today a strong place in
computer science. Many communities use
groupware tools while realizing their cooperative
activities. However, even if a real need for
groupware exists, a quick look at the large literature
helps to understand that many questions still exist
about how to realize a good CSCW application. The
main question that remains after years of research is
how to realize an application really fitting the users
needs. As an example, we can cite a recent study
about systems supporting collaborative writing on
the Web (Noël & Robert, 2003). The study presents
some systems that are interesting because they
propose global and integrated environments for the
realization of cooperative editing tasks. However,
the authors underline that none of these systems
totally fulfils the requirements that better fits the
users needs. It happens despite of the large number
of systems supporting this activity. This work
concludes that it is necessary to further develop
existing tools while adding required functionalities,
or to create new and more complete ones. These
critics are characterizing a major problem in the
CSCW research domain highlighted by results
coming from Human and Social sciences. For
example, Activity Theory (Bedny, 1997) teaches us
that human activity is reflective and that the users
needs are emerging from the activity, while they are
realizing it. These results lead to believe that the
complete CSCW system cannot be defined a priori.
The solution that is usually adopted by groups is
t
o use concurrently different groupware applications,
when they need them, each one fulfilling a
dimension of the collaborative activity. The matter
in this situation is that these different tools do not
know each other and the coherence of the
collaborative environment, the global activity
contextualizing their use, is only managed by
humans.
Trying to solve this problem, we are working on
a platform
named CooLDA (Cooperative Layer
supporting Distributed Activities) that provides the
means to create some global and integrated but
tailorable cooperative environments. In this
approach, the environment is not designed to support
some a priori users needs, but we want it to be able
to dynamically integrate groupware tools supporting
their emerging needs.
The first part of this paper describes how we
defi
ne a tailorable global and integrated environment
while characterising what we call the inter-activities
approach. The second part of the paper introduces
the CooLDA platform that proposes a conceptual
and technical solution to these problems. The last
part of the paper briefly underlines some CSCW
research works that are close but different ours.
134
Bourguin G. and Lewandowski A. (2005).
MANAGING INTER-ACTIVITIES IN CSCW - Supporting users emerging needs in the CooLDA platform.
In Proceedings of the Seventh International Conference on Enterprise Information Systems, pages 134-139
DOI: 10.5220/0002529601340139
Copyright
c
SciTePress
2 THE INTER-ACTIVITIES
2.1 A fine-grained integration need
As we introduced before, groups usually use
concurrently different tools while realizing their
cooperative activities. Each tool supports an aspect
of the group’s activity. From our point of view, each
tool itself supports an activity existing in the context
of another more general one. As an example we can
consider the setting that is described in
Figure 1.
Figure 1: The inter-activities
In this situation, an instant messaging application
is available over the Internet. It supports a
synchronous discussion activity. In the same way,
another application supports an asynchronous
document sharing activity. In this example, our
approach aims at providing the support for the paper
writing activity that creates a context for the
discussion and document sharing. We want to
support the inter-activities between these two sub-
activities. Such contextualization supposes that the
global environment is able to start the application,
control it and, in the case of a really fine integration
scheme, receive some events concerning state
changes in the activity it supports.
The ability to start the integrated tool seems an
evidence: a user that connects to the environment
and enters in the paper writing activity should have
access to the instant messaging and document
sharing tools. This corresponds to the minimal
integration level. Such integration scheme can be
found in web sites providing simple links towards
different applications.
In the case of a finer integration, the role of a
user in an activity should influence its roles in the
linked activities. For example, in a distance learning
activity like a course, the professor role may imply
the presenter role in an activity supported by a
WYSIWIS (What You See Is What I See) document
sharing application and the speaker role in an
associated audio conference. To do this, the platform
that supports the course has to control the tools
implied in this global activity. We can also imagine
that an action that is triggered in the global activity
has direct repercussions on the sub-activities, i.e. on
the integrated tools. These repercussions may be
different according to the different roles played by
the users in the global activity. For example, after its
presentation, the professor may decide to start an
exercise where students cooperatively annotate a
shared document while discussing altogether. This
professor action in the global course activity triggers
a state change in the audio-conference activity, thus
allowing students to speak when they want, and
another related state change in the document sharing
application allowing students to annotate the shared
document. Those are simple examples, but they
show why it is important that the global integrated
environment is able to pilot tools supporting sub-
activities. If this is not the case, the teacher has to
explicitly change the state of each tool he uses for its
course, and this, each time he switches from a
presentation to an exercise.
Finally, in the case of a maximum integration
scheme, the action of a user in a sub-activity may
also have repercussions on the state of the other sub-
activities. As an example, we can imagine a global
debate activity using a vote tool and a forum.
Proposing a motion in the vote sub-activity may
engender a state change in the global debate activity
that closes the discussions in the associated forum
sub-activity. The platform should be able to receive
some events that are generated by a tool, to give
them a sense in the global activity according to the
role of the actor, and eventually to trigger a global
action having repercussions in the linked sub-
activities.
2.2 A dynamic integration need
We have talked about the mechanisms that are
necessary to support the links weaved between an
external tool and our environment. One can notice
that these mechanisms may be found in certain full-
integrated environment like complex web portals.
However, it is important to remember that the main
difference here is that we want to support the
emerging user needs. In our approach of
tailorability, the users should be able to integrate
tools they need and when they need them. Thus, our
platform must also propose the means for
dynamically creating and evolving those links. This
supposes that the environment is able to support the
MANAGING INTER-ACTIVITIES IN CSCW: Supporting users emerging needs in the CooLDA platform
135
dynamic integration of diverse tools as linked
activities, and the dynamic (re)definition of the links
created between them.
Finally, we would like to underline that in our
approach, this dynamicity should be offered not only
to computer scientists or system administrators, but
also to domain specialists, i.e. the end-users. In fact,
if a teacher finds integrable tools that are interesting
for its activity, we want to let him integrate them
himself, without stopping the system, i.e. during its
activity. This is what the CooLDA platform is
designed for.
3 THE COOLDA PLATFORM
The CooLDA platform takes benefits from our
experience in the CSCW research domain (Bourguin
& Derycke, 2001). This work mainly takes its
foundation in the Activity Theory (AT) (Bedny,
1997), a conceptual framework that has a wide
audience in CSCW (Nardi, 1996). Presenting all of
our previous work in the domain with the AT is out
of the scope of this paper. More details can be found
in (Bourguin & Derycke, 2001). However, we will
start this presentation of CooLDA by defining some
concepts necessary for understanding our
demonstration.
3.1 A generic model for activity
CooLDA has to support the inter-activities. For this
purpose, we have defined a generic model of
activity-support. This model is presented in Figure 2.
Figure 2: The activity-support model
A user is an actor in a particular activity-support.
Each activity-support is linked to a resource,
namely, a tool that is integrated in the platform.
Each resource proposes some operations that can be
triggered. An actor plays a role allowing the user to
perform some actions according to the activity state.
An action is realized through a chain of operations.
An activity can be linked to another (sub-)activity.
In this case, a role in the activity may imply another
role for the user being an actor in the other activity.
With this model, we can describe how CooLDA
conceptually manages the inter-activities. Going
back to our example with the professor and students
in a course, professor and student are the roles in the
course activity. The course activity is linked to the
document sharing and the audio-conference
activities. The professor role implies a presenter role
and a speaker role in the respective two sub-
activities. The document-sharing resource proposes
an enable/disable drawing operation that will be
mapped in a corresponding action in the document
sharing activity. This action will automatically be
triggered for a student when the professor performs
the start exercise action in the course activity.
Several other points should be described in this
scenario. It is not the purpose of this paper to
describe them all here, but those we have explicated
let understand how our generic model enables the
specification of the inter-activities managed by
CooLDA. We now would like to describe how we
realize the dynamic inter-activities management
from the technical viewpoint.
3.2 A simple dynamic integration
environment
One particularity in our project is that we do not
want to develop ourselves the tools that may be
integrated in the environment and we would like to
let users integrate themselves tools available over
the Internet. The generic properties needed for
integration that we have already enunciated in part 2
exist in diverse distributed environments like
CORBA. The same type of elements can be found in
the more recent Web Services technology (Kleijnen
& Raju, 2003). These elements support the dynamic
discovery and invocation of objects (or services) and
methods over a network.
Another particularity of our project is that we are
actually working with teachers involved in the
development of distance learning situations. These
teachers use concurrently different tools supporting
their distributed collaborative teaching. They are
interested in using CooLDA as a global integrated
environment supporting scenarios related to their
teaching activities. The tools they use are mostly
ICEIS 2005 - HUMAN-COMPUTER INTERACTION
136
simple Java applications they find over the Internet
and, for some, Java applications developed by
students at university..
According to these constraints, we have
developed the current version of CooLDA using a
very simple integration environment : the standard
Java Virtual Machine (JVM). The description of
CooLDA in the rest of this paper will be linked to
this choice. However, the principles we are going to
describe are transposable in other technologies.
3.3 Supporting inter-activities
Java enables to download and dynamically create
instances of any Java application whose classes have
been made available over the Internet in a simple
Web server. An URL allows downloading the client
part of a groupware application. Thus, the CooLDA
client is itself a Java application which instantiates
client tools in its own virtual machine.
As the instance of the object-application is
created inside CooLDA, we benefit from the Java
message sending mechanisms for supporting the
interactions between activities. These mechanisms
are synthesized in Figure 3.
Figure 3: A debate inter-activities scenario.
This figure takes back to the debate example.
The action (1) of a user in its vote client by
proposing a motion warns (2) the vote server. This
corresponds to the “normal/external” functioning of
this distributed application. The CooLDA client,
which is in the same JVM, can receive a message or
event (3) that is generated by the vote client. This
message is thrown (4) to the CooLDA server
(actually using CORBA) that determines the sense
of this local action in the global debate activity. This
may imply a call-back (5) on the CooLDA clients
that will then control the linked sub-activities, i.e. it
will cause needed method calls on the other client
applications. In this example, the CooLDA client
will ask to the Forum client to change its activity
(sending a message to its server) towards a state
where discussions are closed.
It is important to notice that each separate
application is free to choose its communication
protocol (CORBA, SOAP, RMI, JSDT, JXTA, etc.)
to manage its own activity. On the other hand and
for a fine integration, it is necessary that the
application designer has defined methods on the
client side enabling the communication with a tierce
environment. In our example, the Forum client
should propose a method like stopDiscussion() that
changes the Forum activity state managed by the
Forum server. This server will itself warn the other
Forum clients interested in this Forum activity. Any
integration approach needs the creation of such
interface. However, our approach does not constrain
the application designer to implement a particular
extra communication protocol. Only the basic
message sending mechanism, available in any JVM,
is solicited. Moreover, a part (if not all) of this
needed interface certainly already exists in the
client. It is indeed probable that the application
designer has created a menu or a button calling the
stopDiscussion() method. In fact, such a method
belongs to the Forum activity abstraction level.
3.4 Inter activities dynamic
specification
3.4.1 Using introspection
We want to let users themselves weave the links
between their activities. A particular link will for
example define which methods of the integrated tool
will be called by CooLDA to create a configuration
according to the role of a user in a global activity. In
the same way and as in the debate example, a state
change in the global activity will use these links to
be transmitted to the associated activities.
For this dynamic (re)definition of the links, we
use the Java reflective mechanisms, particularly the
introspection. When an application-object has been
downloaded from a Web site, it is possible to
dynamically get a list of its methods. These
mechanisms are similar to those used in Sun’s
BeanBox that proposes to graphically link
JavaBeans. The main difference with CooLDA is
that we use them to create links between high
abstraction level components like a chat tool.
Another difference is that we want to help end
users to integrate themselves the tools they need in
the environment. The matter with introspection is
that it presents many methods that should not be
used for specifying integration. Moreover, the name
MANAGING INTER-ACTIVITIES IN CSCW: Supporting users emerging needs in the CooLDA platform
137
of the methods and the parameters types are close to
the underlying language abstraction level, and then,
the description is difficult to read and understand for
a non-programmer. This is why, we need a
technique to filter and better describe the elements
presented to the user integrating a tool component.
3.4.2 The Activity Descriptor
Filtering can be realized by different means and
some examples exist in classical software design.
For example, the JavaBeans recommendation uses
prefixes for filtering the methods related to
properties. Another technique is to create a specific
separated interface description. This is the role of
IDL with CORBA and WSDL in the Web Services
technology. However, classical interface description
languages use semantics designed to inform only
programmers. Java proposes another interesting
mean with the BeanInfo technique. A BeanInfo
allows the component programmer to create a
specific interface for integration by filtering the
methods and, if needed, by proposing a high
abstraction level user interface that is instantiated to
manipulate the component at integration time. This
technique is interesting but it requires over-
development.
Inspired by all these techniques but still thinking
to our end users, we have decided to create a tool,
named Activity Descriptor that helps to create an
XML description of the activity supported by a tool
component. The XML file contains a separated
higher abstraction level description. When a user
needs to integrate a tool in CooLDA, the platform
uses Java introspection coupled with the tool’s XML
description. If the description is not found, only the
standard introspection mechanism is used.
The Activity Descriptor also uses introspection.
This enables to document a component whose
source code is not available. Activity Descriptor
presents all of the component’s constructors and
methods. Selecting an element uses its structure to
propose a form for creating a description. For
example, selecting the ChatApp(…) constructor in a
Chat application allows to create an operation named
Start the chat with its particular description
according to the required parameters. This
description will be used in CooLDA to help a user to
define an action while hiding the implementation
level : the user will be able to specify a specific role
with specific actions realized by operations. Even if
there is a direct mapping between the Start the chat
operation and the ChapApp(…) constructor at run-
time, the underlying programming language
abstraction level is partially hidden. The Activity
Descriptor tool still has to be improved, but the
current version helps us to develop our approach.
We have described how it is possible to
dynamically create specifications to start and pilot
integrated tools in CooLDA. Even if this is not the
maximum integration level, these functionalities are
helpful : it allows a professor to perform a start
exercise action in a course activity, this
automatically triggering operations on the tools
involved in the realisation of sub-actions performed
by the sub-roles in the sub-activities.
3.4.3 Activity events
However, the last point is related to the maximum
integration means, i.e. receiving events from an
integrated tool. We already noticed that CooLDA
could technically receive events from other
components because they are running in the same
JVM. The question is, what to listen to ? With the
hierarchical approach defined in Java for creating
user interfaces, it would be possible to register
CooLDA for receiving events from each widget
composing the tool. The matter is to give a sense to
this event in the global activity. This is why we
develop a different approach. We propose a simple
framework, based on the publish/subscribe
paradigm, which can be used by tools developers to
extend the functionalities of their applications. This
extension provides an activity event generator and a
simple activity event model. Using introspection,
CooLDA can dynamically discover if a tool
proposes an activity event generator and listen to
receive activity events. With the same mechanisms,
the Activity Descriptor tool helps in creating a high
abstraction level description of the events that may
be generated by a tool. When describing inter-
activities, a user can then know which event like a
vote has been proposed can be generated by a
particular tool, and specify which actions should
automatically be triggered in the global activity. The
main drawback of this approach is that it requires
that the integrable tool use our simple framework for
activity event generation. Then, this type of
integration is constraining for the tool developer :
the component has to be adapted for CooLDA. This
seems to be the prise for a maximum integration
scheme.
4 RELATED WORK
Different works have been realized by CSCW
researchers using a component approach for
tailorability and it would not be possible to
summarize them all here. There are several
differences between these propositions and ours.
ICEIS 2005 - HUMAN-COMPUTER INTERACTION
138
Systems like in (Grundy & Hosking, 2000) only
allow arranging predefined plugins components at
the user interface level. We develop an approach
allowing end-users to finely integrate themselves the
components they need in the platform.
In (Stiemerling & Cremers, 2000) the authors
develop a component model named Flexibeans
designed for creating components that are finely and
dynamically integrable by end-users in the Evolve
platform. This approach is different because it
proposes a completely new component model for
composition as we try to directly use standard
mechanisms. Evolve also differs from our platform
because CooLDA takes directly its roots in the
Activity Theory and then our view of what is a
component is a little different. For example, in
CooLDA, a tool that has been documented and/or
adapted while its integration becomes a high
abstraction level component that may be easily
integrated in someone else’s activity. The scenarios
for inter-activities that are defined or adapted by
users are also reusable CooLDA components for
other users. This way and thanks to our component
approach, CooLDA aims at presenting an important
property inspired from the Activity Theory : the
crystallization of the users experience inside the
computer artefacts supporting their activities. In
other words, reusing a CooLDA component is taking
benefits from the experience of the actors that used
and developed it during their own activities.
5 CONCLUSION
In this paper, we have shown that despite of the
strong need towards CSCW, problems still exist in
proposing the needed groupware applications to
their potential users. We have underlined that
existing global and integrated groupware
environments are interesting but, as they are usually
designed for fulfilling a priori users needs, they are
not completely satisfying. Users usually solve this
problem by concurrently using different groupware
applications without links between them. This
solution is also not satisfying because it does not
provide a computer support for the global group
activity that creates the context for the use of these
tools. This is why we aim at creating the CooLDA
platform proposing a global, integrated and
tailorable CSCW environment.
CooLDA is designed to support what we call the
inter-activities, i.e. the global activity giving sense to
the links existing between different sub-activities.
We have proposed a generic recursive model for
activity that enables the specification of scenarios
for inter-activities computer support. We also have
identified the different properties that should be
proposed by the platform for enabling the dynamic
integration of tools supporting diverse activities.
Finally, we have introduced an approach for end-
users tailorability that proposes means for dynamic
tool integration by end-users while elevating the
components abstraction level. These concepts and
mechanisms have been exemplified in the current
version of CooLDA that has been designed for
teachers involved in distance learning activities. This
version uses the standard Java Virtual Machine as a
simple technical integration environment.
The next step of our work will focus on usability
tests and improvements. For example, the Activity
Descriptor should evolve for creating tools
descriptions closer to the end-users abstraction level.
CooLDA is going to be used in real long-term
distance learning activities to see how its users
effectively evolve the environment. We hope that
CooLDA will grow from its own use, the users
experience crystallizing in the components they
integrate or evolve.
REFERENCES
Bedny G., Meister D., 1997, The Russian theory of
activity, Current Applications to Design and
Learning,. Lawrence Erlbaum Associates, Publishers.
Bourguin G., Derycke A., Tarby J.C., Beyond the
Interface: Co-evolution inside Interactive Systems - A
proposal founded on Activity Theory, in Proceedings
of IHM-HCI 2001, Lille, France, 10-14 September
2001, Springer Verlag, pp. 297-310.
Grundy J., Hosking J., 2000, Developing Adaptable User
Interfaces for Component-Based Systems, in
Proceedings of the First Australasian User Interface
Conference, pp 17-25.
Hummes J., Kohrs A., Merialdo B., 1998, Software Com-
ponents for Co-operation: a Solution for the “Get
help” Problem, in Proceeding of COOP’98
conference, INRIA.
Kleijnen S., Raju S., 2003, An Open Web Services
Architecture, QUEUE Volume 1 , Issue 1 (March
2003), pp. 38-46.
Nardi, B. A., 1996, Context and consciousness : activity
theory and Human-Computer Interaction., Cambridge,
Ma : MIT Press.
Noël, S., Robert, J-M., 2003, How the Web is used to
support collaborative writing, Behaviour &
Information Technology, 22 (4), pp. 245-262
Stiemerling, O., Cremers, A. B., 2000, The EVOLVE
Project: Component-Based Tailorability for CSCW
Applications, AI & Society (2000), 14, pp. 120-141.
MANAGING INTER-ACTIVITIES IN CSCW: Supporting users emerging needs in the CooLDA platform
139