SOFTWARE AGENTS FOR SUPPORTING STUDENT TEAM
PROJECT WORK
Janice E. Whatley
Information Systems Institute, University of Salford, Manchester, M5 4WT,
K
eywords: Software Agents, Teamwork, E-Learning,
Abstract: In this paper an agent system is described, which has been designed to support students undertaking team projects
as part of their studies on campus or online. Team projects form an important part of the learning process
for campus based students, but are not easily incorporated into the learning activities for online students.
The particular problems of working on projects in teams are explored, and an agent system was designed to
support some of the maintenance tasks of team working. Agent technology is suggested because of the ease
of communication between software agents and their autonomy in operation. The agent system has been
tested on student teams working on campus, and the results indicate that this type of support agent may be
helpful to students. The modified version of the agent system was successfully implemented, and the trial
suggests that it may be scaled up to use over the Internet to support online student teams.
1 INTRODUCTION
Online learners rely on Internet connections to
communicate with institutions, tutors and other
learners, and there is often a sense of isolation from
the support of others (Hill and Raven, 2000).
Campus based learners are beginning to rely more
on the Internet to support their studies, such as to
enable them to access material outside of lecture
times, to work at more convenient times and
wherever they choose and to supplement their face
to face contact with other students.
Working in teams is particularly problematical,
both for campus based students and for online
students, but team working, and virtual team
working, especially, is becoming an essential skill
for our graduates. Groupware and virtual learning
environments help team members to communicate
and share files, but do not support the maintenance
needs of team working, which are necessary for
successful team operation.
Artificial intelligence has been used to develop
tutoring systems for individualised learning, and
agent technology is being harnessed for Internet
based communicating systems. The analogy between
multi-agent systems and student teams has pointed
to the possibility of agent technology as a solution to
some of the difficulties of working in teams, by
combining the benefits of intelligent tutoring, advice
from an agent and communication.
2 STUDENTS WORKING ON
TEAM PROJECTS
Traditional undergraduate campus-based courses
incorporate a team project element, as an essential
means of “learning by doing”. The learning cycle by
Kolb (Kolb, 1984) summarises the stages of
experiential learning as:
concrete experience;
reflective observation;
conceptualisation;
active experimentation.
These stages give a starting point for thinking about
how we approach the design of learning activities to
achieve the learning outcomes. The main feature is
that students do not learn by simply being told facts.
They need to be able to practice using the facts, and
reflect on the way they are used in order to form
connections in the brain, which can be regarded as
knowledge. Further experimentation, experience and
reflection leads to intelligence or expertise in a
subject. If the students are able to talk about this
information, then they can be said to have
knowledge of the subject, and intelligence shows in
190
E. Whatley J. (2004).
SOFTWARE AGENTS FOR SUPPORTING STUDENT TEAM PROJECT WORK.
In Proceedings of the Sixth International Conference on Enterprise Information Systems, pages 190-196
DOI: 10.5220/0002633201900196
Copyright
c
SciTePress
their ability to apply the knowledge in a variety of
situations. Collaborative learning may range from a
pair of students working together to a large class of
students learning together, but as Dillenbourg says,
ther is no agreed definition of collaborative learning
(Dillenbourg 1999). Team projects give students an
opportunity to discuss their understanding of the
subject with their peers, as they apply the theory to
practice (Sharan, 1990). Students may be working at
times collaboratively and at times cooperatively,
coordinating their efforts to achieve a project
outcome, and learning about themsolves and about
the subject in the process (Dillenbourg, Baker et al.
1996). When campus based students work on team
projects they experience difficulties in organising
their work together, such as arranging meetings,
deciding who should carry out which part of the
work, and coping with non participation from
members. If we are to provide students undertaking
online courses with a similar opportunity to
experience team working, we need to provide some
means of helping them to organise their work
together. Where face to face contact is not possible,
technologies may be able to provide some additional
resources to help make the online team experience
comparible to the campus based team experience.
Computer mediated communication (CMC)
tools, such as conferencing, email and discussion
forums support the communication needs for the
task roles of team projects, examples of their use are
given in (English and Yazdani, 1998) and (Hendson,
1997). The facilities included in Virtual Learning
Environments (VLE) give students the capability to
communicate with each other and the tutors, and are
based to a large extent on the facilities incorporated
in Groupware products, which in turn have been
developed as a result of research into Computer
Supported Cooperative Working (CSCW)
(Connolly, 1994). The VLE’s provide a structure to
enable communication, but little help in the process
of communication to help the students form
workable learning networks (Lawther and Walker,
2001). Opie used the term “knowledge-based
teamwork” to describe the sort of interaction
between team members who are all bringing to the
case in hand their own interpretation of the situation,
through their own knowledge or expertise. Her work
is specifically related to health care, but this is a
typical domain in which teamwork is essential for
achieving outcomes (Opie, 2000).
Successful team working requires that the
maintenance roles as well as the task roles of the
team are given attention (Hartley, 1997). Group
dynamics play an important role in determining how
successful the outcome of the project is, i.e. the
ways in which the members interact with each other
and how this changes with time as the team develops
(Bion, 1961), (Gibbs, 1994), (Jaques, 1984). Gilly
Salmon (Salmon, 2000) suggested ways in which
tutors can help students to interact socially online, in
order to develop team cohesion. Student support
using commercial groupware products enables
communication between team members and
instructors (Tiwari and Holtham, 1998), also BSCW
(Basic Support for Cooperative Working) is an
example of a tool that has been used as support for
team projects and was found useful for information
sharing, offering greater flexibility in students’ face
to face communication, but it offers limited support
for the maintenance roles of teamwork (Vliem,
1998). In previous work, students’ perceptions of the
manner in which their team worked together
confirmed that teams were more likely to be
successful in their projects if they pay attention to
some of the maintenance factors (Whatley et al.,
1999a).
The essence of learning how to work in a team is an
important aspect of team projects, because
organisations make much use of team working,
whether face to face, or, increasingly, in a virtual
team. The experience provided in Higher Education
is important, but concentrates on face to face teams,
whereas there is an increasing need to offer the
opportunity to learn to work virtually as well.
3 ONLINE TUTORING WITH
AGENTS
The Internet is providing possibilities for learners to
access their course materials in a variety of ways.
Some may prefer to use traditional face to face
means of learning, whereas others may prefer to
learn from home or the workplace, taking advantage
of online access. The new breed of “blended
learners” expect to be able to choose when and
where to access their lecarning, and require support
to enable them to learn effectively.
Intelligent tutoring using artificial intelligence
(AI) concepts has been associated with distance
learning, providing interfaces that respond to an
individual user’s needs (Farr and Psotka, 1992),
though very much aimed at individuals using
programmed learning packages. Agent technology is
a relatively new field of applying AI to practical
areas, e.g. knowledge management (Ferneley and
Berney, 1999) and Internet searchbots (Lieberman,
1997). Knowledge management aims to enable
SOFTWARE AGENTS FOR SUPPORTING STUDENT TEAM PROJECT WORK
191
collaboration between individuals online, notably for
problem solving (Corkill, 2003). Virtual
communities in the workplace are becoming more
common, but collaborating globally requires
different skills from those used for face to face
collaboration (Lipnack and Stamps, 2000).
The concept of an agent originates from human
agents that provide services, such as estate agents
and travel agents. These agents have specialist skills,
access to relevant information, contacts for obtaining
information and are focused on a particular task. In
the same way software agents are autonomous
systems that work on behalf of a user (Bradshaw,
1997). They exhibit the ability to recognise what the
user needs to accomplish and reacts to the user’s
input. A more formal definition is:
An agent is a self-contained, concurrently executing
software process, which encapsulates the current
state in terms of knowledge, and is able to
communicate with other agents through message
passing (Wooldridge, 1995).
In the field of e-learning software agents have
the potential to help online learners in several ways.
One such way is improving the effectiveness of
searching or enable the sharing of resources between
students who have similar interests (Ferneley and
Berney, 1999). Another aims to bring together
students with similar interests or needs into a
discussion area where they can receive help on
particular problems (Vassileva and Deters, 2001).
There are agents for guiding students in completing
work, by offering tutorial help using a character
(Nijholt, 2001). Finally, software agents may be
used to help teach learners, for example using virtual
environments to portray an example scenario
(Aylett, 2001). Soller suggests an architecture for
multi-agents to support online group learning,
concentrating on knowledge sharing between
students (Soller and Busetta). Software agents can
be made to work actively and adapt to users, which
means they can simulate some of the roles of tutors.
Pedagogical agents can monitor progress, give
instruction when needed, help organise students’
work and provide feedback for tutors (Baggetun et
al., 2002). These agent systems continuously operate
in the background on a student’s workstation and act
autonomously to suggest ways in which the learner
might improve performance.
A software agent may operate in isolation,
working on behalf of an individual, similar to
personalised intelligent tutors, but their power
derives from an ability to communicate with other
agents to fulfil tasks they would be unable to
complete alone. Typically a multi-agent system may
consist of several agents, each capable of performing
a different task autonomously. A network of agent
systems, communicating over a wide area network
(WAN) or a local area network (LAN), will make
use of Internet connectivity to pass messages
between each other. These multi-agent systems are
the main thrust of current research, and have arisen
as a result of the massive global infrastructure of
networks now available.
We now turn to the notion that multi-agents may
be applied to supporting collaborating members of a
team, and in particular teams of learners. In the next
section an application of a software agent for
supporting students working on team projects is
described.
4 DESIGNING A SOFTWARE
AGENT SYSTEM FOR TEAM
WORKING
The support needed by students for teamwork differs
from that which might be appropriate for an
individual working alone, as the dynamics of team
working also need to be considered. The advantage
of using software agents for supporting online
students is that agents can bridge the divide between
time and place. Students may be dispersed and
working at times to suit themselves, so the agents
can keep track of the students’ progress on the work,
and enable all the students to be aware of the status
of the project. Similarly, campus based students may
benefit from such a software agent system, as these
students demand the flexibility to work at different
times and places.
A preliminary version of an agent system
prototype has been developed, performing a limited
set of functions to help students to get started on
their teamwork, and the results of a trial carried out
using teams working on projects on campus are
discussed. These results have informed our further
design, leading to a second version of the agent
system, which has similarly been tried on student
teams working on campus.
An action research approach was adopted for this
study, because a more user-centred design may be
achieved by active user involvement in the
development process. Over several iterations of a
prototyping method, further functions may be added
and refined, by considering feedback from students
in the form of questionnaires, interviews and focus
groups. Although each successive cycle will not
involve the same individuals, a broadly similar range
ICEIS 2004 - SOFTWARE AGENTS AND INTERNET COMPUTING
192
of students have participated in the design process,
and the final product should be acceptable to a
generic type of students.
5 FUNCTIONS OF THE AGENT
SYSTEM
There are limited examples of teamwork being used
for learning activities with online learners, because
there is a belief that face to face contact is essential
for successful teamwork (Lewis, 2002). Previous
surveys have been carried out to determine the
nature and extent of the difficulties experienced by
students working face to face, the intention being to
design an agent system to alleviate these problems
initially (Whatley et al., 1999b). Three main stages
of a team project have been identified, each with its
own associated problems, and these are summarised
in Table 1.
These stages of a team project do not correlate
directly with the stages of team development
originally defined by Bruce Tuckman (Tuckman,
1965), but represent stages of the tasks with which
students will identify (O'Sullivan et al., 1996). The
identified factors “introductions” and “setting the
ground rules” are processes that contribute towards
the maintenance roles of team projects. It was
decided that the initial work on developing a
software agent to support students should be targeted
at these functions, forming part of the planning stage
of a project.
6 DESIGN OF THE FIRST
PROTOTYPE AGENT SYSTEM
The initial prototype for the agent system was
developed in LPA Prolog, using their Agent
Development Kit (Logic Programming Associates,
2000). This tool enabled the developer to code the
interfacing aspects of the agent without worrying
about the technicalities of the agent communication,
which is dealt with by the tool. The declarative
features of Prolog were used for handling facts and
rules, which can be passed between each student’s
agent and the server agent. The first prototype
considered the allocation of tasks to the team
members.
In designing the prototype agent system, we were
interested in these main features:
whether such an agent system would be
acceptable to students;
whether the agent system would be of any
help to the students;
whether communicating agents could be
implemented successfully within a typical
intranet environment.
Figure 1: Structure of Guardian Agent System
In the chosen system structure, each individual
student communicates with the agent system by
means of their individual agent (Figure 1). Each
agent will have a similar structure when the team
project begins, with interfacing capabilities for
communicating with its student, reasoning
capabilities for monitoring and analysing the current
situation, a knowledge base personal to its student
and communication capabilities for communicating
with other students’ agents. All communications
between agents is through a server agent, allowing
for a knowledge base to be built up for the particular
project the students are working on.
Table 1: Stages of the team project
Project stage Factors identified as
problematical
Planning Introductions
Setting ground rules
Produce a project plan
Allocate tasks
Doing the
project
Check the time schedule
Ensure all members
contribute
Identify lack of skills
Discuss each others’
contributions
Completing Collating the individual parts
Preparing a report
Appraising the team’s
performance
Student
Student
Student
Agent
Agent
Agent
Server
Agent
SOFTWARE AGENTS FOR SUPPORTING STUDENT TEAM PROJECT WORK
193
Online students may be working on the
team project at different times of the day, so there is
limited possibility of discussing the allocation of the
tasks for the project. Even campus based students
may not all be present at the same time for formal
meetings. Comparing individual abilities would take
a considerable length of time using a discussion
forum, or other CMC means. Thus the agent system
aims to reduce the time spent on a mundane task,
and inform the students of the status of this task.
The process of allocating tasks begins with the
agent asking its student to enter details of their
abilities and preferences. The agent system will
obtain its own student’s abilities and preferences and
post these to the server agent so that all of the
students’ agents can access them. Once all of the
students in the team have posted their abilities and
preferences the agent system can apply a set of rules
to the facts, in order to determine which tasks of the
project could be allocated to each student. The agent
system will maintain a record of the suggested
allocations on the server agent, necessary later when
the agent system will be able to monitor student
activity against the work plan. As each student
returns to the team project, the agent will present the
allocations, so that all of the students can consider
and discuss them with the other students on the
project. Any allocations proposed by the agent
system may be subject to negotiation between the
students, the allocations are simply suggestions.
The agent system has been programmed to work
with three levels of allocation, using the following
rules:
Allocation1 -
If studentA likes X and is able at X
Then studentA should do X.
Allocation2 –
If studentB is good at X and has not expressed a
dislike of X
Then studentB could do X
Allocation of tutoring -
If studentC likes X, but is unable at X
Then studentC could be offered tutoring in X
7 RESULTS FROM THE FIRST
TRIAL OF THE AGENT
SYSTEM
The agent system was tested with seven teams
working on projects in systems development as part
of their undergraduate programme. The teams
consisted of between 6 and 10 second and final year
members, working on campus, and they were asked
to use the allocation of tasks function as they began
their projects. Each team project is slightly different,
so the tasks were specific to each team. After some
brief instructions for using the agent system, each
student in the teams used the agent to input their
details over a period of four weeks. As not all
students were present for each session, they did not
all use the system on the same occasion, which
matched the way in which the agent might be used
online.
Afterwards the students were asked to complete
questionnaires and were invited to a focus group so
that we could obtain feedback on the usefulness of
the system.
The interface was generally acceptable, but some
students suggested improvements, which we have
incorporated into the second prototype. About half
of the students said that the output from the
allocation of tasks function was useful, these were
mainly team leaders, who compared the output with
the ways in which they would have normally made
task allocations. A majority of the students thought
that such an agent system would be useful to
students working online as well as for campus based
students. Just over half of the students said that they
personally would like to use such an agent.
From our initial results it is clear that
students would find an agent system acceptable,
particularly for online students (though the term
“online” was not defined). It is difficult to establish
how helpful the agent system might be to either
online or campus based students, but comments
made during the focus group session did suggest that
the agent system performed a helpful function.
Table 2 – Questionnaire findings
Questions to students after
completing the Guardian Agent
trial
% of total
responses
Did you find the function useful? 56
Did you find the system easy to
use?
89
Was it self explanatory? 78
Do you think it would be useful for
students online?
81
Do you think it would be useful for
students on campus?
64
Do you like the concept of agent
help for working online?
75
Do you like the concept of agent
help for working on campus?
61
Would you personally like to use
this sort of agent?
56
ICEIS 2004 - SOFTWARE AGENTS AND INTERNET COMPUTING
194
Most importantly, we did find that the agent system
implementation, using a server agent, was
successful, though a limited number of students
were involved in the study. However, implementing
the system using Prolog proved difficult, owing to
the reluctance of the code to compile correctly.
8 DESIGN FOR THE SECOND
TRIAL OF THE AGENT
SYSTEM
Taking into consideration the feedback from
students and issues of portability, we built the
second prototype in Java. The programs for the
agents were produced as Java executable files,
incorporated into web pages, and an improved user
interface was designed for the system. The server
agent was replaced by a file server, running a
MySQL database, in which the facts are stored.
Figure 2 shows the screen to obtain a student’s
preferences.
Figure 2: example of the interface to obtain preferences.
The issue of agreeing ground rules for team working
has been little explored, and problems cited included
difficulties getting students to attend meetings,
inform the team leader if they cannot attend and
complete their assigned work on time (Hill and
Raven, 2000). So this additional functionality was
included, to help the students to agree ground rules
for the way they intend to work together. Figure 3
shows the interface for asking about ground rules.
The second prototype was tried on 25 teams of
between 10 and 15 students, working on campus,
and took place over five weeks, when theses teams
were establishing their individual task areas for their
projects.
Once again questionnaires and focus groups were
used to capture the students’ opinions, together with
interviews carried out with the team leaders to
ascertain their views on the differences the agent
system made to their team project.
9 RESULTS FROM THE SECOND
TRIAL OF THE AGENT
SYSTEM
Initial analysis of the findings from the second trial,
indicates some satisfaction with the functions of the
agent system, several team leaders said that the
allocation of tasks was a useful function. Some
limitations with the interface were identified and the
pre-programmed task list did give rise to some
reluctance to use the agent system, as one team
leader said that the tasks included were not relevant
to their particular project.
Technically, the agent system was a
success, as the MySQL database was able to cope
adequately with the number of students using the
system. In spite of the fact that not all of the
computers in the laboratory were equipped to run
Java programs, sufficient machines were available to
satisfy the demand for using the agent system at any
one time. The pattern of usage for the campus based
students probably matched the expected pattern for
online teams.
Further development work will take place to
improve the interface and add more functionality to
the agent system.
10 CONCLUSIONS
In this paper we have described two phases in the
development of an agent system for supporting
Figure 3: interface to ask about ground rules.
SOFTWARE AGENTS FOR SUPPORTING STUDENT TEAM PROJECT WORK
195
students working on team projects. At present the
system has only been tried with campus based
students, but feedback from these students indicates
that the results are likely to be applicable to online
teams.
We have developed an agent system to help
student teams to allocate the tasks between the
members of the team, and to help the team members
to agree to a set of ground rules. Although these are
two relatively simple functions, and only a part of
the planning stage of a project, our findings indicate
that an agent system that can support students
through their team project would be acceptable.
Future investigations will try to establish the
extent to which student teams may be helped by an
agent system to perform more successfully.
The mode of implementation of the second
prototype agent system perhaps deviates from the
true definition of an agent, as each instance of the
agent does not maintain its own knowledge base.
However, as the functionality of the agent system is
enlarged in the future, it is suggested that a self-
contained knowledge base might be an essential part
of each agent for monitoring its student against a
project plan.
Although this trial was carried out on
campus, within a department’s intranet, the results
suggest that the Java agent can be distributed over
the Internet. Issues of security and firewalls may
need addressing, in preparation for it to be
implemented for real online teams.
REFERENCES
Aylett, R. (2001) In IVA2001, Vol. LNAI 2190 Springer
Verlag, London.
Baggetun, R., Dolonen, J. and Dragsnes, S. (2002).
Bion, W. (1961) Experiences in groups, Tavistock,
London.
Bradshaw, J. (1997) Software agents, MIT Press, London.
Connolly, J. (1994) CSCW and artificial intelligence,
Springer Verlag, London.
Corkill, D. (2003) In International Lisp Conference.
Dillenbourg, P. (1999). Collaborative learning: cognitive
and computational approaches. Oxford, Elsevier.
Dillenbourg, P., M. Baker, et al. (1996). The evolution of
research on collaborative learning. Learning in
humans and machine: towards an interdisciplinary
learning science. E. Spada and P. Reiman. Oxford,
Elsevier: 189-211.
English, S. and Yazdani, M. (1998) Virtual University,
Internet.
Farr, M. and Psotka, J. (1992) Intelligent instruction by
computer: theory and practice, Taylor & francis, New
York.
Ferneley, E. H. and Berney, B. (1999) In 3rd International
ACM Conference on Autonomous Agents (Agents
'99)Seattle, WA, USA.
Gibbs, G. (1994) Improving student learning:through
assessment and evaluation, Oxford Staff Dev, Oxford.
Hartley, P. (1997) Group communication, Routledge,
London.
Hendson, B. (1997) Br J of educational technology, 28, 4,
257 - 270.
Hill, J. R. and Raven, A. (2000), Vol. 2002 ITForum.
Jaques, D. (1984) Learning in Groups, Kogan Page,
London.
Kolb, D. (1984) Experiential learning, Prentice Hall,
London.
Lawther, P. and Walker, D. (2001) Education and
Training, 43, 105-116.
Lewis, R. (2002) International Review of Research in
Open and Distance Learning, 2.
Lieberman, H. (1997) In Proceedings of the ACM
Conference on Computers and Human Interface, CHI-
97Atlanta, Georgia.
Lipnack, J. and Stamps, J. (2000) Virtual teams: people
working across boundaries with technology, John
Wiley, New York.
Logic Programming Associates (2000) LPA, London.
Nijholt, A. (2001) In First International Workshop on
Agents and Internet LearningInternational Conference
on Autonomous Agents, Montreal.
Opie, A. (2000) Thinking teams/ thinking clients:
knowledge-based teamwork, Columbia University
Press, New York.
O'Sullivan, T., Rice, J. and Saunders, C. (1996) Successful
group work, Kogan Page, London.
Salmon, G. (2000) E-moderating, Kogan Page, London.
Sharan, S. (1990) Collaborative learning: theory and
research, Praeger Publishers, New York.
Soller, A. and P. Busetta "An intelligent agent architecture
for faciltating knowledge sharing communication.”
Tiwari and Holtham, C. (1998) In ITiCSEACM.
Tuckman, B. W. (1965) Psychological Bulletin, 63, 384-
399.
Vassileva, J. and Deters, R. (2001) In First International
Workshop on Agents and Internet
learningInternational Conference on Autonomous
Agents, Montreal.
Vliem, M. (1998) University of Twente., Twente.
Whatley, J., Beer, M., Staniford, G. and Scown, P. (1999a)
In HCI International 1999Lawrence Erlbaum,
London., Munich, Germany.
Whatley, J., Staniford, G., Beer, M. and Scown, P.
(1999b) Journal of Interactive Learning Research, 10,
361-373.
Wooldridge, M. J., N (1995) Towards a theory of
cooperative problem solving.
ICEIS 2004 - SOFTWARE AGENTS AND INTERNET COMPUTING
196