The Inception of an Intelligent Modular Education Learning Assistant
Elaine McGovern, Rem Collier and Eleni Mangina
School of Computer Science & Informatics, College of Engineering Mathematical and Physical Sciences University College
Dublin, Belfield, Dublin 4, Ireland
Keywords: Mobile learning, modular education, multi-agent systems, recommender systems, ubiquitous computing.
Abstract: University College Dublin has made an unprecedented transition from its once traditional educational
metaphor to a modularised education framework, the first of its kind in Ireland. This paper questions
whether students, who are unfamiliar with the concepts of modularisation, can successfully make informed
decisions, ensuring success from specifically tailored module combinations. It presents IUMELA, an
intelligent modularised-education learning assistant for use in a mobile context to aid in this career-making
decision making process. Using multi-agent systems and their associated expert systems, it recommends
appropriate modules based on the students own requirements. Proclaimed as a new departure, the ubiquitous
multi agent systems technologies are capable of transcending the boundaries of the traditional into a world
of innovative intelligent mobile learning, where software components embedded with intelligent learning
assistants can enhance the third level students academic career.
“Just as technology can be used to strengthen
different forms of intelligence, so too can it target
different learning styles.” (Snowman et al. 2006)
IUMELA is an acronym for Intelligent Ubiquitous
Modularised Education Learning Assistant. It uses
multi-agent systems (MAS) technologies to create
an intelligent learning assistant that can support
students in their choice of modules based on their
learning preferences, academic abilities and personal
preferences. The learning assistant uses expert
systems analysis functionality to recommend and
predict potential outcomes through the investigation
of the students’ learning styles and comparative
analysis of similar past student’s achievements. A
wireless device can support the functionality of a
MLE and IUMELA has been designed for use on the
XDA Mini S. It was chosen because studies have
shown that third level students tend to purchase a
vast array of mobile phone technologies and upgrade
these devices more frequently than other consumer
groups within the market. Mobile devices enable
student’s opportunities to connect their academic
queries to data in a real time setting thus reducing
many administrative overheads (McGovern, 2005).
The remainder of this paper is structured as follows:
Section 2 provides a description of current research
in the areas of mobile intelligent learning
applications and learning styles theory. Section 3
presents the internal structure of the IUMELA
application. Section 4 presents the results of the
software’s feasibility study. Finally, the conclusions
and future developments are discussed in Section 5.
2.1 The Mobile Device
A wireless device is any form of networked
hardware that can communicate with other devices
without being physically attached to them
(Tarasewich, 2002). IUMELA has been designed to
run using the smart phone technologies available on
the XDA Mini S. This smart phone combines
integrated personal information management
facilities as well as mobile phone capabilities
(Ballagas, et al. 2006). IUMELA’s preliminary
findings suggest that providing third level students
with access to modularised education learning
assistant via smart phone technologies, as well as via
traditional means, ensures that the resources are
being provided in an on-demand, an anytime,
McGovern E., Collier R. and Mangina E. (2007).
IUMELA - The Inception of an Intelligent Modular Education Learning Assistant.
In Proceedings of the Third International Conference on Web Information Systems and Technologies - Society, e-Business and e-Government /
e-Learning, pages 372-377
DOI: 10.5220/0001267403720377
anywhere manner. Previous studies have shown that
the XDA class of mobile computing devices
provides flexibility, connectivity, pro-activity, cost-
efficiency and multimedia capabilities that its users
have expressed as essential to the successful
completion of their computing activities (Doe,
Microsoft Windows Mobile 5.0 drives the XDA
Mini. This is a scaled down version of Microsoft
Windows that has been specifically designed for
PDA’s and smart phone technologies (Doe, 2006).
The operating system provides a familiar platform to
students ensuring its ease of use and simplistic
integration with the Managed Learning Environment
(MLE) installed at the University. Active Sync
technologies enable the XDA to synchronise with
the university’s MLE, ensuring that IUMELA’s
student case base is kept current and relevant.
In past research, it was evident that students were
using mobile device as a graphing calculator, word
processor, database, test prep tool, and as a means of
accessing resource. These devices have afforded
students with "opportunities to connect questions
and investigations to the data in a real time setting
that enhances "systematic investigations, critical
thinking and cooperation" (Staudt, 1999) Additional
research suggests that these facilitate group work,
the immediate analysis of data particularly during
laboratory exercises or when conducting scientific
investigations in the field (Belanger, 2000).
2.2 Modular Education at UCD
UCD Horizons is the flagship of modularised
education in Ireland. Modular education in UCD has
provided a structured modular and credit-based
taught degree programme. It has been designed to be
more flexible than its traditional counterpart and
enables students to individualise their academic
career. They are required to undertake some core
modules and have the opportunity to elect some
optional and free choice modules also. This, in
theory, enables them to adapt their degree
programme based on their own study preferences
and strengths (Nolan, 2006).
A primary motivation behind developing
IUMELA was that, although there is enhanced
freedom of choice in a modularised education,
students entering third level education are often
poorly equipped to deal with such freedom. They
subsequently make misinformed module choices,
frequently resorting to poor decision-making
2.3 Learning Styles & Teaching
Psychologists agree that intelligence is an ability.
Significant resources have gone into developing an
understanding of how students use these abilities for
the purpose of education, otherwise known as
learning styles theory. Learning styles are
considered to be preferences for dealing with
intellectual tasks (Snowman et al, 2006). It is
possible to adopt different learning styles as the need
arises. Kagan found that some students seem
characteristically impulsive, while others are
reflective (Morgan, 1997). Witken (Witken et al.
1977) theorised that individuals can be influenced by
their surrounding context and that there are two
groups of learners: field dependent and field
independent. Sternberg’s (Sternberg, 2001) styles of
mental self –government theory describes thirteen
styles that fall into one of five categories. There are
functions, forms, levels, scope and learning. This
concept supports the belief that IUMELA would
assist students by suggesting appropriate modules
based on their preferred learning styles.
Educators often use various instructional
methodologies to engage any number of styles of
learning at one time or another. They are required to
use various test formats to measure accurately what
various students have learned. IUMELA measures
those classes in which students consistently
participate well, through the inclusion of an expert
agent. IUMELA’s expert agent defines each teachers
style based on one of several well-documented
behavioural approaches; constructivist, humanistic
and social (Bloom et al. 1956), (Krathwohl et al.
1964), (Simpson, 1972), (Adams & Engelmann,
1996). IUMELA, as a learner-centric intelligent
assistant, can link concepts to every day experiences,
guide students in their problem solving processes
and encourage learners to think analytically when
reasoning in a humanistic manner.
Historically, assessment involves measuring how
much knowledge and skills a student has and its
acceptability based on the teachers’ eventual goals.
The summative and formative techniques are two
popular methods of evaluation. Teachers use a
variety of means to evaluate, either summarily or
formatively, a student’s knowledge or skills level. A
vast array of methodologies - including written
assignments, short answer tests and e-portfolios - are
frequently used by lectures in UCD and so have
been incorporated into IUMELA expert agents
reasoning abilities and knowledge base
Software systems are becoming increasingly
more complex and online information spaces are
growing exponentially. Kay (Kay, 1990) highlights
IUMELA - The Inception of an Intelligent Modular Education Learning Assistant
how the use of MAS has resulted in a transition from
the traditional direct manipulation of a system to
indirect human-intelligent agent interactions. Agents
have enabled the delegation of the mundane and
tedious. Shneiderman (Shneiderman, 1979)
considered user confidence and agent autonomy as
areas that required consideration. MAS should have
the overall effect of reducing the users workload
(Collier, 2001).
IUMELA conforms to FIPA specifications (FIPA,
2001). The multi-agent system (MAS) was
developed using Agent Factory (Collier, 2001)
toolkit, using Java as the programming language. In
particular, the Assistant Agent runs on a XDA Mini
S. The high-level communication protocols have
been implemented using ACL messages, whose
content refers to the IUMELA ontology. The GAIA
methodology was used to identify the agent
structures, roles and interactions within IUMELA
(FIPA, 2001) and can be identified in figure 1.
Figure 1:The Agent Architecture.
IUMELA uses a FIPA compliant MAS
architecture, displayed in figure 1, to fulfil the task
of an intelligent application capable of autonomous
human computer interaction for communication,
event monitoring and the performance of higher
order cognitive tasks. IUMELA consists of a
community of five agent types: assistant, moderator,
learning agent, expert agent and analysis agent.
3.1 The Assistant Agent
The assistant agent resides on the student’s client
device and is responsible for the seamless interaction
between IUMELA’s MAS and the student. The
effectiveness of interface agent technologies as a
means of facilitating human-computer interactions
ensures that they remain a prevalent research area
within the fields of artificial intelligence (AI),
human-computer interaction (HCI), and user
modelling. It is considered a mechanism by which
the mundane and tedious can be altered or delegated.
Each of these research domains reflects upon a
unique facet of the agents’ capabilities, rating its
effectiveness in terms of their own requirements.
AI research has afforded the assistive interface
agent with knowledge representation skills, learning
heuristics, and reasoning abilities. HCI research in
interface agents has ensured attentiveness to the
user, focusing on that which the user requires. In
order to effectively display personalized information
to the student, a user modelling functionality is
required (Harrington, et al. 1996). Deploying
IUMELA on a mobile device ensures that the
manner and location of the student-agent interaction
will differ significantly from that which occurs when
taking place via the traditional desktop metaphor.
Enabling the student to interact with IUMELA in a
ubiquitous manner ensures that their learning
experience can be transformed into a larger context,
incorporating it into every aspect of their third level
academic career. It is the task of the assistant agent
to interact with the student and the other agents
within the MAS in order to provide appropriate
assistance based on context. Adaptive
personalization was considered as the mechanism
that would best assist students.
To correctly aid the student, the assistant is
required to be aware of several fundamental
variables: their degree program, level, current stage,
preferred learning style, academic history, and
preferred teaching strategy. The application is
capable of using a multitude of web-based
technologies ensuring that the student’s information
could be displayed in an appropriate and discernable
3.2 The Moderator Agent
The mediator agent family is composed of three
basic agent patterns: the broker, the matchmaker,
and the mediator. They act as intermediaries
between any number of other agent types. Similar to
the broker and mediator agents the moderator
WEBIST 2007 - International Conference on Web Information Systems and Technologies
arbitrates interactions between the other agent types.
In addition to this, it also maintains an acquaintance
model based on past interactions. The moderator can
interpret the requests received and, based on a
combined analysis of the stored acquaintance model
and current context it acts accordingly. However,
further to this the moderator can also seamlessly
interact with resources external to the IUMELA
MAS, this functionality was necessary in order to
enhance its communicative capacity. Assistant and
expert agents inform each other of actions preformed
via the moderator agent.
3.3 Expert Agent Technologies
The student agent enables all other agents in
IUMELA to connect to the student case base and to
access the administrative data of the analysed
student. It provides a single, generic method by
which other agents can interact with the student data
while, simultanelosuy, ensuring student
confidentiality. It has an ability to examine incoming
messages and retain and return information based on
the senders clearance level. It can then match action
requests to the appropriate agent role controlling the
student case base.
IUMELA aims to help students to achieve their
ultimate academic goals by assisting them in
devising competent and obtainable academic goals
while traversing through a specially tailored module
schema. The student agent enables students to
envision a potential overview of their academic
journey based on the student’s current profile and
previous academic achievements.
The fundamental role of the expert agent is to
accurately depict the teaching strategies of the
module. So too, is it the task of this agent to retrieve
all potential evaluation techniques for each module,
ensuring that any prediction or recommendation
made is based on 100% current information. Within
IUMELA, the expert agent maintains a knowledge
base of all possible teaching strategies used within
the university. It is then linked to a list of all
potential learning strategies within each module
offered. It is the task of the expert agent to maintain
this directory of all available modules, the lecturer
directing it, and their preferred teaching style and
examination technique.
The analysis agent maintains a knowledge base
that predicts all plausible academic outcomes based
on the information it receives from the student and
expert agents. Although it maintains several
potential recommendation algorithms, it will
proactively choose an appropriate reasoning model
based on the students prior knowledge, their
academic history, and their chosen degree program
and current level. This is achieved through the
knowledge-reasoning centre that enables it to
autonomously choose an appropriate reasoning
algorithm. This agent type is capable of adapting its
current reasoning strategies and assimilates new and
improved ones, therefore ensuring that IUMELA’s
recommendations are constantly improving,
becoming more accurate.
3.4 The Agent Interactions within
Figure 2:Sample HCI with IUMELA.
Upon logging on to IUMELA for the first time, the
student is required to complete an initial survey. To
do so is necessary to enable the multi-agent system
(MAS) to assist the new student in choosing their
preferred learning style, teaching strategy and
examination procedures. To ensure that IUMELA is
useful from the onset, a student will be unable to
navigate through the application without having
completed the survey beforehand.
The student enters their UCD email address and
student number, which will suffice as their unique
student identifiers within IUMELA. The assistant
agent submits them to the moderator agent via the
getSurvey() message, which adheres to the FIPA
compliant IUMELA ontology. The moderator
receives the message and, using its internal
reasoning abilities, forwards it to the appropriate
student agent. The student agent, responsible for
maintaining the knowledge base for all students,
creates a new entry for the student, retrieves the
question to be posed and returns it to the assistant
agent for display, see figure 2. When there are no
more questions to be answered, the assistant agent
IUMELA - The Inception of an Intelligent Modular Education Learning Assistant
affords the student the opportunity to review the
survey, and update any answers he so wishes.
The task of the expert agent is similar to that of
the student agent in that it maintains a registry of the
user information. Instead of maintaining that of the
student, however, it stores all information relating to
the modules available at the university. Including,
but not limited to, the current lecturer, their
preferred teaching strategies and examination
procedures. It is the task of the analysis agent to
analyse all information received. It combines these
with past student histories to predict all plausible
academic outcomes based on the possible module
combinations currently available.
The purpose of the initial IUMELA user trial was
two fold; the first was to determine if the agent
architecture was robust enough to display returned
stored student information, to retrieve information
entered by the student via the assistant agent and to
store this information in a meaningful and cohesive
manner for later reuse. The second purpose of these
initial user trials was to obtain an initial knowledge
base from the current degree candidates and to offer
them module opportunities not previously
This trial took the form of a survey to be
undertaken, via the XDA Mini S, by the general
student body at University College Dublin. By
enlisting a broad range of students across the
university’s campus it was ensured that a wide
variety of students with varying interests and
academic backgrounds as accounted for. The
participation levels varied from faculty to faculty;
21% of the participants were from the Faculty of
Arts and Humanities, 25% were from the Faculty of
Engineering, 17% were from Law and 28% were
from the Faculty of Science.
The First set of questions required the
students to submit information pertaining to their
academic history via the assistant agents interface.
This information was retrieved by the assistant agent
sent, via the appropriate ACL, for unbundling by the
moderator agent. The moderator agent analysed the
received data and determined, using its internal
reasoning capabilities, to which agent the messages
should be sent. Results showed that the moderator
agent was safely abstracting necessary information
and accurately identifying its content for sending to
either the student or expert agent 100% of the time.
The information received by the student agent
was specific to the students learning styles and their
perceived understanding of modules currently being
undertaken. The students were required to enter a
difficulty rating and relevance rating for each
module that they were attending. The results
displayed that 83 % of the students believed that one
or more of the modules taken was out of their
difficulty comfort zone. Furthermore, 42% of these
students indicated that one or more of these modules
were not relevant to their degree course.
The aim of the final section was to store, in the
student case base, the students personal
characteristics; interests, abilities, values, personality
and motivations. Its aim was also to indicate how
students were incorporating their personal
characteristics into their module decision-making.
The survey indicated that only 31% of the students
surveyed considered extra curricular activities they
enjoy to partake in, with 62% of these indicating that
the influence of television and media played a major
Regarding the student’s academic abilities, 26%
of them felt that their academic scores were higher
than their classmates and 17% felt that their scores
were significantly lower. Of the latter, 88% had
previously indicated that one or more of the modules
taken had a high difficulty rating. This indicated to
the student agent that these students had poor to fair
module decision-making strategies.
Furthermore, 37% of the participants indicated
that their moral values and principles had played a
significant role in deciding upon the modules they
were currently undertaking. And finally, an
overwhelming 87% had indicated that their
personality had played a considerable role in their
decision to undertake a particular degree
programme, but only 45% of these had used the
same belief set in choosing their modules. In many
cases students were empowering others to indicate
modules that should be undertaken; 37% received
assistance from their family members, 33% from
their peers and 22% from academic authorities, the
remainder had chosen their academic courses alone.
With regards the feasibility of constructing an
intelligent learning assistant to aid modular students
in their academic decision making in a mobile
context, 82% of those surveyed considered the
applications structure and navigation functions
enabled them to access and utilize IUMELA’s
survey function and vicariously enabled them to
consider modules in a way not previously done.
WEBIST 2007 - International Conference on Web Information Systems and Technologies
It was the intention of the IUMELA feasibility study
to ascertain whether there was a requirement for an
intelligent learning assistant to aid students in their
modularised education decision making strategies as
UCD transitions from a traditional to a modularised
educational framework. Research into educational
psychology determined that, not only could students
learning styles be categorised into discreet forms,
but so too could lecturers teaching strategies be
identified. By assimilating these and the university’s
current examination procedures into MAS based
application, an intelligent agent capable of
autonomous human-computer interactions could be
developed. The results from the initial user trial
indicated that students at University College Dublin
are unfamiliar with modular education and the
required decision-making strategies required in
choosing third level academic modules. Even
student’s, who had demonstrated significant ability
at choosing their degree programs had failed to do so
based on their learning styles, preferred teaching
strategies, evaluation procedures and personal
considerations. In such a myriad of academic
modules, students were not averse to receiving
modular course recommendations from a multi-
agent systems based learning assistant via a mobile
device. In future work, the role of the analysis agent
will be examined and a comparative study of
reasoning algorithms used will be undertaken.
Adams, G.L & Engelmann, S. Research on Direct
Instruction: 25 Years beyond DISTAR [WA]
Educational Achievement system 1996
Ballagas, R et al The Smart Phone: A Ubiquitous Input
Device [NJ] Merrill Prentice Hall 2006
Belanger, Y. Laptop Computers in the K-12 Classroom
(2000, May) Retrieved 11 September 2006 from ERIC
DIGEST: Notebooksystems.com
Bloom, B.S. et al. Taxonomy of Educational Objectives.
The Classification of Educational Goals Handbook I,
Cognitive Domain, 1956
Collier, R. The Realisation of Agent Factory: An
Environment for the Rapid Prototyping of Intelligent
Agents UMIST, Manchester, UK (M.Phil. Thesis),
Doe, J Microsoft Release Windows Mobile 5.0 (2006,
June) Retrieved 8 October 2006 from
Foundation for Intelligent Physical Agents, FIPA 97
Specification Part 2: Agent Communication Language,
Retrieved 18 October 2005 from
http://citeseer.ist.psu.edu/411222.html 1997.
Harrington, R.A. Banks, S. and Santos Jr. E. Development
of an Intelligent User Interface for a Generic Expert
System. Retrieved 1 October 2006 from
http://citeseer.ist.psu.edu/context/107458/0 In
Michael Gasser, editor, Online Proceedings of the
Seventh Midwest Artificial Intelligence and Cognitive
Science Conference, 1996.
Kay, A. User Interfaces: A Personal View. The art of
Human Computer Interface Design. Addison-Wesley
Krathwohl et al. Taxonomy of Educational Objectives
Handbook II: Affective Domain [NY] McKay 1964
McGovern, E., A Comparative Study of Multi-Agent and
Object Oriented Methodologies for Mobile Managed
Learning Environments. MSc. Thesis, Department of
Computer Science, University College Dublin. 2005
Morgan, H. Cognitive Styles & Classroom Learning [CT]
Praeger Publishers. Mowrer, R.R. and Klein, S.B. (eds)
Nolan, P. (2006, July). UCD Horizons University College
Dublin, 2(2). Retrieved from www.ucd.ie/horizons/
Shneiderman, S. Human Factors Experiments in
Designing Interactive Systems. Computer 12 (12)
Simpson, E.J. The Classification of Educational
Objectives: Psychomotor Domain Urbana. [IL]
University of Illinois Press 1972
Snowman et al., Psychology Applied to Teaching
Houghton Mifflin Company, Boston 2006
Staudt, C. Probing untested ground: Young Students Learn
to Use Handheld computers. (1999 September)
Retrieved 13 November 2006 from Http://concord.org
Sternberg R.J. and Grigorienko E.L. A capsule history of
theory and research on styles. In R.J. Sternberg and
L.F. Zhang, editors, Perspectives on Thinking,
Learning and Cognitive Styles.
Lawrence Erlbaum,
Mahwah, NJ, London, 2001 pp 1-21
Tarasewich, P. Designing Mobile Commerce
Applications, Communications of the ACM, 46, 12,
57-60. 2003
Witkin, H. Moore, C. Goodenough, D. Cox, P. Field-
dependent and field-independent cognitive styles and
their educational implications, Review of Educational
Research, Vol. 47, No. 1 (Winter, 1977), pp. 1-64
IUMELA - The Inception of an Intelligent Modular Education Learning Assistant