DYNAMIC PROFILING TO ENHANCE LEARNING AND REDUCE

COGNITIVE LOAD ON EACH LEARNER

Keith Maycock, Sujana Jyothi and John Keating

National University of Ireland, Maynooth

Maynooth, Co.Kildare Ireland

Keywords:

SCORM learning object, Multiple Representation Approach, Exploratory Space Control, Cognitive Trait

Model.

Abstract:

This paper proposes extensions to the architecture of any Learning Management System (LMS) that utilizes

the Sharable Content Object Reference Model (SCORM), to incorporate Multiple Representation Approaches

(MRA) and Exploratory Space Control (ESC) features, when interacting with a learner. The learners proﬁle

will consist of the traditional student modeling features such as students goals, preferences and knowledge.

The proﬁle will also incorporate a Cognitive Trait Model (CTM) to measure the learners cognitive abilities.

The LMS provides functionality for dynamic login to reduce the time spent getting to know the learner. If a

course is changed to ensure that the course is MRA compliant to suit the cognitive needs of that learner, the

transformation that course encored is stored in a Learning Experience Repository (LER) for future reference.

In effect, the learners proﬁle becomes the author of educational content throughout the learning experience,

ensuring that the content delivered will suit the cognitive ability of each learner to increase the throughput.

ESC techniques are used throughout the LMS interaction with the learner once a learning experience has

concluded to offer suitable links to other related courses. This paper also discusses various factors that must

be taken into account when developing a LMS, for example, teaching styles, different types of students and

learning styles. The proposed extensions will enhance the learning experience for individual users.

1 INTRODUCTION

Currently, in higher education, there are roughly 70

million students worldwide. This number is expected

to more than double before the year 2025 to over 160

million students (LittleJohn, 2003). The only possi-

ble solution to cater for the expected inﬂux of peo-

ple entering into higher education is to automate the

process of learning. However, this is not an elemen-

tary task. If we look at the results of a number of stud-

ies carried out on the performance of individually tu-

tored students against the performance of an average

student in a typical classroom environment, we ﬁnd

that, the speed with which different students progress

through instructional material varies by a factor of 3

to 7 (Gettinger, 1984). An average student in a typi-

cal classroom environment asks on average 0.1 ques-

tions every hour in contrast to an individually tutored

student asking on average 120 questions every hour

(Graesser and Perso, 1994). Furthermore the achieve-

ment of individually tutored students will exceed that

of classroom students by as much as two standard de-

viations (Bloom, 1984) - an equivalent which is equal

to raising the performance of 50 percentile students

to that of 98 percentile students. These results show

the vast range of differences between the learning ca-

pabilities of each learner. It is very important when

developing an education environment to take into ac-

count the environmental contexts. These contexts in-

clude the nature of the subject discipline and the level

of its learning; the characteristics of the learning ma-

terial and the role of the human teacher (Patel, 1998).

Support should also be available for dealing with a

learner’s learning proﬁle. The proﬁle should consist

of the entire learner’s educational history, learner’s

goals, preferences and cognitive ability.

In November 1997, the Department of Defense

(DoD) and the White House Ofﬁce of Science Tech-

nology Policy (OSTP) launched the ADL initiative.

The mission of the ADL is to provide access to the

highest quality of education and training, tailored to

the individual needs of each user, anytime anywhere

(ADL, 2004). The ADL initiative borrowed from

many different speciﬁcations and standards such as:

287

Maycock K., Jyothi S. and Keating J. (2006).

DYNAMIC PROFILING TO ENHANCE LEARNING AND REDUCE COGNITIVE LOAD ON EACH LEARNER.

In Proceedings of WEBIST 2006 - Second International Conference on Web Information Systems and Technologies - Society, e-Business and

e-Government / e-Learning, pages 287-292

DOI: 10.5220/0001257702870292

 SciTePress

AICC (AICC, ), ARIADNE (ARIADNE, ), IEEE

LTSC (LTSC, ) and IMS (IMS, ) when develop-

ing the Sharable Content Object Reference Model

(SCORM). SCORM is used to produce and deploy

courses that can be tracked and delivered to a stu-

dent by a Learning Management System (LMS) in

a standardized way. An LMS is software that auto-

mates training event administration through a stan-

dard set of services that launch learning content, keep

track of the learner’s progress and sequence learning

content. SCORM courses are fully deﬁned within

a SCORM content package. The SCORM manifest

is inside the content package. The manifest con-

sists of metadata, organisations, resources and sub-

manifests. The metadata is used to describe in full the

version of SCORM and type of course. The organisa-

tions section details the sequencing information of the

various learning objects that are encapsulated within

the content package. The resources section is fully

described using XML metadata elements to describe

the content that is being delivered. Sub-Manifests can

also be used to create structured courses with different

layers of dept.

One of the problems that we perceive with most e-

learning educational systems is that authors of educa-

tional material are likely to have different ideas on the

best teaching practices which, can hence hinder the

development of a learner’s learning experience. De-

veloping an educational system around the SCORM

would easily be able to overcome the problem of the

teacher being in full control of the learning experience

as the hierarchical learning activities and the corre-

sponding sequencing information are fully described

within an activity tree (Maycock and Keating, 2005).

The activity tree is not a static structure and is free to

change with the requirements of the author of educa-

tional media. Once the learning experience has initi-

ated the learner’s proﬁle becomes the author for the

duration of the learning experience and is capable of

changing the educational media to adapt the speciﬁc

learners needs immediately. After the learning experi-

ence has concluded, the learner’s proﬁle is returned to

the learner. Enabling learners to store their own pro-

ﬁle locally, enhances the learning experience, as the

learners would be free to utilise any LMS and imme-

diately initiate a learning experience, based on their

learning proﬁle. The next section details a learning

proﬁle that is suited to automatically control a learn-

ing environment utilising SCORM.

2 DYNAMIC PROFILING TO

ENHANCE LEARNING

Most of the existing student models are focused on the

speciﬁc domains with which they interact with, for

example, the domain concepts competence and do-

main skills required. Such student models, are called

performance based student models and include the

student competence state models (Staff, 2001) and

process state models (Martin, 1999). To create a

truly adaptive learning environment across multiple

domains the cognitive traits of a learner should be

catered for. The cognitive traits that are associated

with learning are working memory capacity, induc-

tive reasoning ability, information processing speed

and associative learning skill.

Working memory capacity also known as Short-

Term Store (STS) facilitates temporal storage of re-

cently perceived information, allows active retention

a limited amount of information, (7 +/- 2 items), for a

short period of time (Miller, 1956). The range of in-

formation perceived to be active is almost double and

should be catered for by a learning system. Induc-

tive reasoning ability is the ability that allows us to

construct concepts from examples (Kinshuk and Mc-

Nab, 2005). Inductive reasoning is seen as one of the

important characteristics of human intelligence. It is

strongly recognised that inductive reasoning ability

can be extracted from most aptitude tests and is the

best predictor for academic performance. Information

processing speed determines how fast learners can

acquire new information correctly. Adapting to the

information processing speed would enable a learn-

ing environment to reduce the possibility of cognitive

overload. Associative learning skill is the skill to link

new knowledge to existing knowledge. Students with

high associate learning skill should be given content

that has been adapted to existing relevant information

already encountered in past learning experiences.

Our proposed proﬁle consists of two distinct tiers.

The ﬁrst tier consists of information entered by the

learner, detailing personal information to allow a

LMS to personalize content. The second tier con-

sists of the learner’s cognitive traits and educational

history. This latter tier of the proﬁle is automati-

cally updated by the LMS after learning experiences.

The proﬁle will be contained within an XML ﬁle and

stored in a repository of personal proﬁles. Storing the

personal proﬁle as an XML ﬁle enables the LMS easy

access to the proﬁle and is easily incorporated into

a SCORM learning object at the start of a learning

experience, as discussed in (Maycock and Keating,

2005). Each of the cognitive traits that are being mon-

itored will be assigned a numeric value in the range of

-1 to 1. After a learning experience has concluded, a

graphical representation of the current learning object

WEBIST 2006 - E-LEARNING

288

is generated, mirroring the structure of the SCORM

learning objects activity tree with associated meta-

data. This graph is stored in the educational history

section of the proﬁle under past experiences.

We propose adapting the Cognitive Trait Model

(CTM) to supplement a traditional performance based

student model in our proposed learning environment.

CTM was developed by Taiyu Li, Kinshuk and Ashok

Patel (Taiyu Lin and Ashok, 2003) at Massey Univer-

sity in New Zealand, and facilitates the construction

of a long-term student model, based on invariant cog-

nitive traits. Once the CTM has successfully identi-

ﬁed the characteristics of a learner no more training is

required.

2.1 Cognitive Trait Model(CTM)

Learner modeling motivates the creation of systems

that are adaptive to each of the learner’s interests,

preferences and background knowledge in order to

provide personalised instruction to a particular learner

The Cognitive Trait Model (Taiyu Lin and Ashok,

2003) may be adopted to produce a student proﬁle that

is easily incorporated into the SCORM content pack-

age as discussed in (Maycock and Keating, 2005).

The CTM is composed of a user Interface module, a

Learner-Behavior model, Interaction model, Learner-

Performance model and Trait Analyser as seen in 1.

Figure 1: Cognitive Trait Model.

The Interface Module depicts the system-learner

interaction. It is an important component of this

model and acts as a communication medium, and as

an external representation of all the system’s models.

This acts as a mediator between the learner and the

system. The Learner-Behavior Model records learner

interaction associated with different cognitive traits

such as the working memory capacity, inductive rea-

soning ability, information processing speed and as-

sociative learning. It generates a graphical represen-

tation of the current SCROM learning object after the

learning experience has concluded. The objective of

the Interaction Model is twofold: to deliver an adap-

tive unit of learning for each learner and to observe

the behavior of the learner while the learner interacts

with the learning material. Every student interaction

recorded in the behavior model has one interaction

entry. This model produces an XML ﬁle that is di-

rectly associated with the behavior of each student.

The Learner-Performance Model is used to identify

the weight of each of the cognitive trait affected by

the learning experience.

Figure 2: Components of a Trait Analyser Derived from -

Taiyu Lin, Kinshuk & Ashok, P.(2003) (Taiyu Lin and Ashok,

2003).

The inputs of the Trait Analyser are from the three

components of the CTM namely the Learner behavior

model, the Interaction model and the Learner perfor-

mance model. The main criterion of the Analyser is

to analyse the learner based on the cognitive traits and

the other information known about the learner. The

trait analyser is composed of three different compo-

nents: the pattern detector, the individualised tem-

perament network and the CTM updater as shown

in 2. The pattern detector examines the records of

the student’s current actions that are produced by the

LMS to recognise patterns that are associated with

different cognitive abilities. There are many different

types of manifestations that symbolize different traits,

for example navigational linearity, reverse navigation,

excursions, simultaneous tasks, retrieval of informa-

tion from long-term memory, and long sequences of

calculation or procedures. The individualised tem-

perament network is used to adjust the CTM accord-

ing to the results obtained from the pattern detector

and the CTM updater is the communication channel

that communicates with the CTM. After an analy-

DYNAMIC PROFILING TO ENHANCE LEARNING AND REDUCE COGNITIVE LOAD ON EACH LEARNER

289

sis is carried out the CTM is updated and communi-

cates back to the interface module to incorporate some

adaptive functionality. A combination of the adaptive

techniques in neural network will be used to track the

navigation patterns of the learners and their cognitive

traits. A modiﬁcation to the cognitive trait model is

possible during the implementation of the techniques.

3 ENHANCING EACH LEARNING

EXPERIENCES TO REDUCE

LEARNER COGNITIVE LOAD

Multiple Representation Approach (MRA) (Kinshuk

and Kashihara, 1999) and Exploratory Space Control

(ESC) (Kinshuk and Lin, 2003) are used to ﬁne tune

learning experiences. The following section details

the advantages of both techniques and illustrates how

these techniques are incorporated into our proposed

learning environment architecture.

3.1 Multiple Representation

Approach (MRA)

MRA is used to change the presentation of domain

knowledge concepts, in terms of the complexity and

granularity, to suit the learner’s cognitive ability and

progress through a learning experience. It enhances

the educational system’s design to suit the learner’s

perspective. There are various types of multime-

dia objects, each stimulating different cognitive re-

sponses. Audio stimulates imagination, video clips

stimulate action information, text conveys details and

diagrams convey ideas. Generating MRA compliant

learning objects in a learning environment can reduce

the cognitive load by using similar multimedia objects

to convey domain concepts. If any media objects are

omitted during the MRA process they must be avail-

able to a user on speciﬁc request, reducing the possi-

bility of losing any relevant information.

There are three different types of ﬁltering used in

MRA: restriction, extension and approximation. Re-

striction is used when a learning object contains an

excessive number of media objects, thereby causing

cognitive overload. A subset of these media objects

maybe selected to produce an MRA compliant learn-

ing object conveying the current domain concept. If

several different MRA compliant learning objects are

available then the combination of media objects offer-

ing the best learning experience suited to that learner’s

cognitive ability maybe selected. When the number of

media objects is insufﬁcient to produce an MRA com-

pliant learning object extension maybe used. An LMS

will search learning object repositories to ﬁnd suitable

learning objects that will enhance that learning ob-

ject and make it MRA compliant. When the structure

or content of a SCORM learning object is changed,

a transformation occurs within that learning object’s

activity tree, as discussed in (Maycock and Keating,

2005). MRA extensions are seen also as transforma-

tions and are saved in a Learning Experience Repos-

itory (LER) for further search and discovery during

different learning experiences. If a learning object

was poorly designed, and the complete learning ob-

ject cannot be made MRA compliant, the largest mul-

timedia rich subset is selected. The process of exten-

sion is then carried out on the reduced learning object.

The number of learning object repositories are in-

creasing. Many of these repositories are freely avail-

able online (Repositories, ). Several organizations

that have generated learning objects and host there

own repository or have provided guidelines, tem-

plates, or frameworks for objects that are stored in

their repository. SCORM is fast becoming the stan-

dard for producing learning objects and has been

adopted by many different LMSs. SCORM is cen-

tered around developing small granular learning ob-

jects with minimal content associated with the learn-

ing material, ensuring maximum reusability. With ex-

pansive metadata our learning environment will ef-

fectively be able to generate new content to suit the

cognitive needs of each user by manipulating the

graphical structures that have been generated mirror-

ing each learning object’s activity tree.

3.2 Exploratory Space Control

(ESC)

ESC limits the learning space to reduce the cognitive

load of each learner and to make sure that learners do

not get lost in hyperspace (Kinshuk and Lin, 2003). In

our proposed system, ESC is used in the exploration

of further reading once a learning experience has con-

cluded. The exploration elements catered for are the

learning content and navigational paths. When deal-

ing with the learning content, the ability of the student

to interpret the content exactly as the content devel-

oper expected, is a very complex task and depends on

the learner’s cognitive ability. There have been many

studies carried out on how learners perceive instruc-

tional material, in particular, Phenomenography (Lau-

rillard, 2002) (Marton and Booth, 1997) (Ramsden,

1988) (Ramsden, 1998). This is successful at illu-

minating how students deal with structure and mean-

ing. These studies have led to the identiﬁcation of two

contrasting approaches to studying content, i.e. an

atomistic approach and a holistic approach. Learners

utilizing a holistic approach interpreting some con-

tent retain the concepts that are trying to be conveyed

but may suffer some cognitive overload. Learners uti-

lizing an atomistic approach lose the structure of the

WEBIST 2006 - E-LEARNING

290

content being delivered, hence, may have a different

interpretation to the actual meaning. The ESC server

uses MRA techniques when searching for a suitable

course for any given learner. As SCORM is focused

around producing reusable learning objects the struc-

ture of learning objects can be adapted to suit the indi-

vidual cognitive needs of a user and ensure that only

suitable courses are offered. These elements are fully

controlled by the learner’s learning proﬁle and the ef-

fects of the cognitive abilities can be seen in (Kinshuk

and Lin, 2003).

Figure 3: Proposed outline architecture for LMS model.

Figure 3 represents the outline architecture of our

LMS with MRA and ESC servers incorporated. A

learner logs into the LMS from a remote platform and

the learner’s proﬁle is uploaded to the LMS server.

Once the learner selects a course it is fetched from the

learning object repository and is passed to the MRA

server to ensure that the course is MRA compliant,

thereby suiting the cognitive ability of that learner. If

the course does not suit the individual cognitive needs

of that learner, suitable ﬁltering operations are carried

out. The transformations that occur during ﬁltering

operations are stored in the learning experience repos-

itories for further reuse. Once the learning experience

has concluded the LMS consults the learner’s proﬁle

and the ESC server to offer suitable further reading.

4 CONCLUSION

Combining a CTM with traditional performance stu-

dent models and taking advantage of MRA and ESC

techniques enables our proposed extended LMS func-

tionality to allow a learner’s proﬁle to become the

author of education material during learning experi-

ences. We believe that extending the SCORM man-

ifest and building learning experience repositories is

the next step in achieving an automated one-to-one

tutoring experience.

REFERENCES

ADL (2004). Sharable content object reference model 2nd

edition overview. Available at ADLNet.org.

AICC. Aviation industry cbt committee.

http://www.aicc.org.

ARIADNE. Alliance of remote instructional authoring and

distribution networks for europe. http://www.ariadne-

eu.org.

Bloom, B. (1984). The two sigma problem: The search for

methods of group instruction as effective as one-to-

one tutoring. Educational Researcher, 13, 4-16.

Gettinger, M. (1984). Individual differences in time needed

for learning: A review of the literature. Educational

Psychologist, 19, 15 - 29.

Graesser, A. and Perso, N. (1994). Question asking during

tutoring. American Research Journal, 31, 104 - 137.

IMS. Ims global learning consortium, inc.

http://www.imsglobal.org.

Kinshuk and Lin, T. (2003). User exploration based adap-

tion in adaptive learning systems. International Jour-

nal of Information Systems in Education.

Kinshuk, R. Oppermann, A. P. and Kashihara, A. (1999).

Multiple representation approach in multimedia-based

intelligent educational systems. Artiﬁcial Intelligence

in Education, S.P. Lajoie and M. Vivet (eds.), IOS

Press, Amsterdam.

Kinshuk, T. L. and McNab, P. (2005). Supporting inductive

reasoning in adaptive virtual learning environment.

Proceedings, Web-based Education 2005.

Laurillard, D. (2002). Rethinking university teaching: a

framework for the effective use of learning technolo-

gies. London: RoutledgeFalmer.

LittleJohn, A. (2003). Reusing online resources: a sustain-

able approach to e-learning. Kogan page Limited.

LTSC, I. Institute of electrical and electronics engi-

neers and learning technology standards committee.

http://ltsc.ieee.org.

Martin, B. (1999). Contraint-based modelling: Represent-

ing student knowledge. New Zealand Journal of Com-

puting, 7(2), pp 30 - 38.

DYNAMIC PROFILING TO ENHANCE LEARNING AND REDUCE COGNITIVE LOAD ON EACH LEARNER

291

Marton, F. and Booth, S. (1997). Learning and awareness.

Marwah, NJ: Lawrence Erlbaum Associates.

Maycock, K. and Keating, J. (2005). Building intelli-

gent learning management systems to implement the

teacher student dialogue. IEEE LTTC Learning Tech-

nology Newsletter, vol 7, Issue 1, pages 16 - 20.

Miller, G. (1956). The magic number seven, plus or minus

two: Some limit of our capacity for processing infor-

mation. Psychology Review, vol.63(2), pp 81 - 96.

Patel, A. (1998). Discipline attributes and teaching styles:

Environmental contexts of an its design for deter-

mining multimedia and virtual reality representations.

Knowledge Transfer - An International Journal, 1(1),

107 - 113.

Ramsden, P. (1988). In Improving Learning: New Perspec-

tives. London: Kogan Page.

Ramsden, P. (1998). Learning to lead in higher education.

London: Routledge.

Repositories, L. O. http://elearning.utsa.edu/guides/LO-

repositories.htm.

Staff, C. (2001). Hypercontext: A framework for adaptive

and adaptable hypertext. Phd Thesis, University of

Sussex.

Taiyu Lin, K. and Ashok, P. (2003). Cognitive trait model

for perisitent student modelling. EdMedia 2003 Con-

ference Proceedings, Norfolk, USA,.

WEBIST 2006 - E-LEARNING

292