Using Intelligent Agent-managers to Build Personal Learning

Environments in the e-Learning System

Nadiia B. Pasko

1 a

, Oleksandr B. Viunenko

1 b

, Svitlana V. Agadzhanova

1 c

and

Karen H. Ahadzhanov-Honsales

1 d

Sumy National Agrarian University, 160 Herasyma Kondratieva Str., Sumy, 40000, Ukraine

Keywords:

e-Learning, Distance Learning, Personal Learning Environment, Intelligent Agent-Manager.

Abstract:

The article focuses on the issues of developing the structure of a multi-agent environment for e-learning sys-

tems and proposes a computer technology to ensure student activities in e-learning modular systems. The

relevance of the research topic is due to the low level of modern e-learning systems adaptation to the individ-

ual characteristics of the student, the lack of ability to predict learning outcomes. The technology enables to

take into consideration the factors affecting the students’ learning outcomes and to form an individual trajec-

tory of the learning session from a holistic perspective.

1 INTRODUCTION

In modern e-learning systems, it is important to de-

liver dynamic learning materials, as well as manage

the training course system in a prompt manner, that

is, the e-learning system should provide the user with

optimal content and encourage working in groups. An

intelligent agent-manager should refer students to the

most relevant community or knowledge communities,

examining the materials that other community mem-

bers look through, and connect students and experts

(Al-Sakran, 2006).

The introduction of e-learning systems has also

accelerated the evolution and the learning process in

higher education institutions, given the constraints of

non-adaptive systems, resulting in the introduction of

new open intelligent systems that are used simulta-

neously with web technology. This is critical to the

e-learning technology being implemented across the

globe (Arif and Hussain, 2016).

Tutor agents and support systems play an impor-

tant role in improving learning outcomes, as they pro-

vide continuous assistance to students in the learn-

ing process. Some of the existing learning support

systems are used at the organizational level and in-

tegrated into the current organizational structure of

https://orcid.org/0000-0002-9943-3775

https://orcid.org/0000-0002-8835-0704

https://orcid.org/0000-0002-0486-3511

https://orcid.org/0000-0002-1409-4648

the educational institution (Chen et al., 2003). Such

learning support systems enable to connect existing

users, share important information, improve the train-

ing of technical personnel, and improve organiza-

tional processes, making them more efﬁcient. How-

ever, most existing learning support systems operate

with a small number of functions that do not con-

tribute to the development of the e-learning environ-

ment required for groups and students to achieve their

learning goals in the corresponding ﬁelds (Hung and

Nichani, 2001).

The main disadvantage of present-day learning

management systems is the failure to provide stu-

dents with assistance in the distance learning process,

and therefore they are unable to replace the physi-

cal presence of a tutor, who generates the students’

work progress. In fact, it is proposed to integrate

for each student a metacognitive agent that would en-

sure metacognition assistance and reveal defects in

the learning process and strategies. The goal is to en-

courage students to improve their learning outcomes

measured against the learning goals and reﬁne the

learning method. The results show that there are re-

lationships between different metacognitive attributes

and student’s academic excellence, that is, there is

a dependence of metacognitive inﬂuence on learning

outcomes, reﬂecting the degree of student’s under-

standing of a particular training unit (Elbasri et al.,

2018). There are certain difﬁculties associated with a

large number of micro-modules and the need to form

a learning trajectory tailored to the student’s needs.

292

Pasko, N., Viunenko, O., Agadzhanova, S. and Ahadzhanov-Honsales, K.

Using Intelligent Agent-managers to Build Personal Learning Environments in the E-Learning System.

DOI: 10.5220/0010931000003364

In Proceedings of the 1st Symposium on Advances in Educational Technology (AET 2020) - Volume 2, pages 292-299

ISBN: 978-989-758-558-6

One of the ways to overcome these obstacles may be

the use of adaptation technology (Kla

snja-Mili

cevi

et al., 2017).

Thus, the state of elaboration of this problem and

current trends in the development of management sys-

tems for educational environments for e-learning are

indicative of its theoretical and practical signiﬁcance,

and determine the urgency of the chosen theme. The

goal of the research is to develop a functional archi-

tecture that supports the above goals of e-learning us-

ing mobile agent technology.

The introduction of multi-agent systems is one of

the most promising areas for building virtual educa-

tional environments for distance education systems.

The goal of this article is the possibility of illustrat-

ing the advantages of using intelligent agents to op-

timize the location and conﬁguration of appropriate

resources for distance learning courses and organiz-

ing collective collaboration in the e-learning environ-

ment.

The main objectives of the research are to develop

the structure of the training service based on the use of

a personal learning environment and intelligent agent-

managers, which may be used to ensure individual

learning. It uses a set of agents that may personal-

ize learning based on previous requests from students

(or groups of students), and improve learning and col-

laboration based on previous knowledge and learning

styles.

2 RESULTS

As of today, the SCORM (Shareable Content Object

Reference Model) standard that is a standard for shar-

ing learning materials based on the IEEE 1484.12.1

standard model (IEEE, 2020) has been developed, and

is currently being used. SCORM has been devel-

oped to ensure the multiple use of learning materi-

als, support for and adaptation of training courses, in-

troduction of information of individual training ma-

terials into training courses or disciplines in accor-

dance with individual user requests. In June 2006, the

United States Department of Defense established that

all developments in the ﬁeld of e-learning should meet

the SCORM requirements. A promising direction for

e-learning standardization has become the successor

of SCORM – Tin Can API model (Romero, 2015),

which enables to consider the types of learning activi-

ties that are not available in SCORM: mobile learning,

simulations, informal learning, games; tracks events

without using the Internet, and has a reliable system

for maintaining the required level of security and user

authentication.

When creating complicated and distributed sys-

tems, multi-agent systems (MAS) can offer a vari-

ety of solutions, especially in the ﬁeld of distance

learning. The combining of agent technology with

other methods such as the Educational Data Mining

(EDM) and Case-Based Reasoning (CBR), which in

turn are based on cloud technology, is important in

taking the learning process to the next level. The

three-level multi-agent management architecture for

distance learning in the e-learning system, which con-

tains the following set of intelligent agents, is pro-

posed to meet the above functional requirements (ﬁg-

ure 1):

• Tutor Agent is a set of tools for creating rules

that enable tutors to adapt the selection of learning

material, deﬁne appropriate search terms for ﬁnd-

ing learning materials based on certain learning

styles, and to communicate with other agents for

collaboration and establish interaction between

tutors and students in a distance learning system.

• Lesson Planning Agent is designed to collect in-

formation and complicated reasoning required for

deﬁning and developing a curriculum (Woolf and

Eliot, 2005).

• Learner Agents are required to organize the ef-

fective interaction of students with the e-learning

environment, and enable to unite various learning

resources into a single whole and constantly mon-

itor learning outcomes.

• Personalization Agents are responsible for cus-

tomizing training materials based on the preferred

learning style of each individual student or work-

group (Wilson, 2000).

The greatest interest for implementing LMS is

represented by learning agents, which in some liter-

ature are also referred to as autonomous intelligent

agents that determines their independence and ability

to learn. Figure 2 shows the ﬂow of work of an agent-

manager as part of LMS, which meets the following

requirements: to work in real-time mode; learn based

on a large amount of data; analyze oneself in terms

of behavior, mistakes and success; contain a database

of examples with the possibility of replenishing it, as

well as learn and develop in the process of interaction

with the environment.

The objective of formalized description of modu-

lar e-learning systems to ensure the ergonomic qual-

ity of human-machine interaction has been solved. As

a result, a complex of component and morphological

models, which is the basis for the formation of infor-

mation support to adaptive e-learning as the “man –

technology – environment” classical systems and con-

tribute to the search for ergonomic reserves of com-

Using Intelligent Agent-managers to Build Personal Learning Environments in the E-Learning System

293

Figure 1: Architecture of distance learning multi-agent management in e-learning systems.

puter human dialogue interaction has been obtained

(Lavrov et al., 2017b) (ﬁgure 1). The set of models

is given by the scheme shown in ﬁgure 3, and is de-

scribed by structural formula (1). The description of

the designations accepted in the formula is given in

ﬁgure 3.

MMS =< EE, OT, PO, MODUL, KPKT, SPF,

SV P, SV FS, SGOT, SMT, EREM, KvPEE,

KvN pEE,KvOT, KvPKT, KvPT, MKvHEE,

MFSEE, MKvMODUL, MKvMOD,ProgPPR,

MDV, MUT >

(1)

Here are the structures of some models.

Component model of elements of module. It de-

scribes the structure of educational module.

MODUL =< [idmod

, [PO

, [tema

k j

j ∈ [1, 2, ...,KT

]|k ∈ [1, 2, ..., KPO],

[PMod

, [Srmod

iln

]||n ∈ [1, 2, ..., KSr

[Sdmodilk]|z ∈ [1, 2, ..., KSd

Pruk

]|l ∈ [1, 2, ..., KPmod

] >

(2)

where idmod

is the identiﬁcation of the i-th module;

is the k-th subject area;

tema

k j

is the j-th theme of the k-th subject area;

is the number of themes of the k-th subject

area;

PMod

is the ﬁrst sub-module of the i-th module;

Srmod

iln

is the n-th self-control of the ﬁrst sub-

module of the i-th module;

KSr

is the number of variants of self-control of

the ﬁrst sub-module of the i-th module;

Sdmod

ilk

is the z-th means of “ﬁnishing” of addi-

tional learning (in terms of (Adamenko et al., 1993) –

“ﬁnishing”) of the ﬁrst sub-module of the i-th module;

KSd

is the number of means of “additional learn-

ing” of the ﬁrst sub-module of the i-th module;

KPmod

is the number of sub-modules of the i-th

module;

Pruk

is a sign of existence of means of control-

ling the quality level (provides a possibility of chang-

ing learning technologies depending on the current

level of the learning quality) of the ﬁrst sub-module

of the i-th module, Pruk

[0,1].

Component model of the means of revealing

motivation levels. The model gives enumeration of

means for revealing motivation levels of EE.

SMT =< [INSmt

, NameSmt

, [PSmt

i j

]

| j ∈ [1, 2,..., KPMT

]|i ∈ [1, 2, ..., KMT ]] >

(3)

where INSmt

is the identiﬁer of the i-th means of

deﬁning motivation of EE;

NameSmt

is the name of the i-th means of deﬁn-

ing motivation of EE;

AET 2020 - Symposium on Advances in Educational Technology

294

Figure 2: Flow of work of an agent-manager as part of LMS.

PSmt

i j

is the j-th indicator for the i-th means,

PSmt

i j

∈ MMT ;

KPMT

is the number of all indicators of motiva-

tion for the i-th means;

KMT is the number of means of deﬁning the mo-

tivation level of EE.

Component model of means of revealing pref-

erences of EE. The model describes the means for

revealing preferences and indicators of EE and pref-

erence indicators of the EE, revealed by this means.

SV P =< INSvp

, NameV p

, [PSvp

i j

]

| j ∈ [1, 2,..., KPV P

]|i ∈ [1, 2, ..., KV P] >

(4)

where INSV p

is the identiﬁer of the i-th means of

revealing the EE preferences;

NameV p

is the name of the i-th means of reveal-

ing the EE preferences;

PSvp

i j

is the j-th indicator for the i-th means,

PSvp

i j

∈ [PMOD];

KPV P

is the number of all indicators of the EE

preferences, revealed by the i-th means;

KV P is the number of means of revealing the EE

preferences.

Component-qualitative model of non-

pragmatic indicators of EE. The model deﬁnes

the composition of the EE characteristics, which

are revealed for deﬁning individual preferences,

psycho-physiological characteristics, functional state,

motivation and level of readiness for learning.

KvN pEE =< [PMOD, PFH, [PFS,VP f s],

[MMT,V Mmt], [InU Got,VU Got]] >

(5)

where PMOD is the set of characteristics of preferable

modalities of the EE;

Using Intelligent Agent-managers to Build Personal Learning Environments in the E-Learning System

295

Figure 3: Structure of complex of models of systems ergonomic analysis.

PFH is the set of psycho-physiological character-

istics of the EE;

PFS is the indicator of functional state;

V P f s is the range of values of functional state;

MMT is the level of the EE motivations;

V Mmt is the range of values of motivation level;

InU Got is the integral level of professional readi-

ness for learning of EE;

VUGot is the range of values of the level of pro-

fessional readiness for learning of EE.

AET 2020 - Symposium on Advances in Educational Technology

296

The set of characteristics of preferable modalities

of the EE are determined by formula:

PMOD =< [Pmod

,VPmod

]| j ∈ [1, 2, 3, 4] >

where Pmod

is the name of the j-th characteristic of

preferable modalities of EE;

V Pmod

is the range of values of the j-th charac-

teristic of preferable modalities of the EE.

The set of psycho-physiological characteristics of

the EE is determined by formula:

PFH =< [N p f h

,V p f h

]| j ∈ [1, 2, ..., K p f h] >

where N p f h

is the name of the j-th psycho-

physiological characteristic of the EE;

V p f h

is the range of values of the j-th psycho-

physiological characteristic of the EE;

K p f h is the number of psycho-physiological char-

acteristics of the EE.

Component-qualitative model of implements of

labor. The model describes the characteristics of im-

plements of labor, used in the system.

KvOT =< [idOt

, NameOt

, TipOt

, [Pk

i j

,Val

i j

]

| j ∈ [1, 2,..., KPK

]|i ∈ [1, 2, ..., KOT ]] >

(6)

where idOt

is the identiﬁer of the i-th implement of

labor;

NameOt

is the name of the i-th implement of la-

bor;

TipOt

is the type of the i-th implement of labor;

i j

is the j-th characteristic (quality indicator of

the i-th implement of labor);

Val

i j

is the value of the j-th characteristic of the

i-th implement of labor;

KPK

is the number of all quality indicators of the

i-th implement of labor;

KOT is the number of implements of labor.

Morphological-qualitative model of electronic

learning module. The model contains the values of

the results of ergonomic assessment of learning mod-

ule quality.

MKvMODU L =< idMod;[PO

;[tema

k j

j ∈ [1, 2, ...,KT

];[px

]|i = [1, 2, 3];

[py

]| = [1, 2];[pz

]|i = [1, 2];[pv

]|i = [1, 2, 3];

[pm

];[mod

]|i = [1, 2, 3, 4];e

|i ∈ [1, 2, 3]] >

(7)

where idMod is the identiﬁer of a module;

is the k-th subject area;

tema

k j

is the j-th theme of the k-th subject area;

is the i − th indicator of the interface assess-

ment;

is the i − th indicator of assessment of slide’s

parameters;

is the i −th indicator of test assessment;

is the i − th indicator of assessment of visual

environment;

mod

is the i − th indicator of information modal-

ity;

is the result of assessment (resolution on corre-

spondence of a module to ergonomic requirements).

The developed models deﬁned the concept of

building databases and knowledge of the learning

management system in the software package “Agent –

Manager for e-learning” (Lavrov et al., 2017a). Each

module can also be divided into parts (submodules),

depending on the levels of complexity of the train-

ing material. The individual learning trajectory is

a sequence of e-learning modules(ELM) and self-

monitoring procedures. Individuality of the trajectory

of learning is achieved through the use of different

types of self-control procedures. Self-monitoring is

a test procedure performed by a student after study-

ing part of the module. UML-diagram of options for

using the software package ”Agent - Manager for e-

learning” is shown in ﬁgure 4.

To study the effectiveness of the developed mod-

els and computer technology, the experiments were

conducted on the basis of Sumy National Agrarian

University. The quality expertise and evaluation of

the parameters of electronic training modules “Infor-

matics” for ﬁrst-year students of the specialty “Agron-

omy” of the Bachelor’s educational level were carried

out.

The developed technology makes it possible to

take into account the factors affecting the students’

learning outcomes from a holistic perspective and

form an individual trajectory of the learning session.

3 CONCLUSION

The proposed architecture of the training service

based on the use of a personal learning environment

and intelligent agent-managers provides users with

the opportunity to collect, analyze, distribute and use

knowledge in the e-learning system from various in-

dependent sources.

The computer technology that enables to autom-

atize the processes of organizing high-quality human

computer interaction in e-learning systems has been

developed:

• ensuring a focus on comprehensive accounting of

factors affecting the students learning outcomes;

• automatic selection of an individual training ses-

sion trajectory.

The direction for future research:

Using Intelligent Agent-managers to Build Personal Learning Environments in the E-Learning System

297

Figure 4: UML Use Case Diagram of agent-manager for e-learning.

• development of intelligent agent models based on

dynamic data extraction rules and interaction with

LMS;

• formation of intelligent agent operation algo-

rithms that automatically detect the student’s sta-

tus, proﬁle, and agent response in real time.

REFERENCES

Adamenko, A. N., Asherov, A. T., and Berdnikov, I. L.

(1993). Informacionno-upravljajushhie cheloveko-

mashinnye sistemy: Issledovanie, proektirovanie,

ispitanija. Spravochnik Information controlling

human-machine systems: research, design, testing.

Mashinostroenie, Moscow.

Al-Sakran, H. (2006). An Agent-based Architecture for De-

veloping E-learning System. Information Technology

Journal, 5(1):121–127.

Arif, M. and Hussain, M. (2016). Intelligent Agent Based

Architectures for E-Learning System: Survey. Inter-

national Journal of u- and e- Service, Science and

Technology, 8(6):9–24.

Chen, J. Q., Lee, T. E., Zhang, R., and Zhang, Y. J.

(2003). Systems requirements for organizational

learning. Commun. ACM, 46(12):73–78.

Elbasri, H., Haddi, A., and Allali, H. (2018). Improving

E-Learning by Integrating a Metacognitive Agent. In-

ternational Journal of Electrical and Computer Engi-

neering, 8(5):3359–3367.

Hung, D. and Nichani, M. (2001). Constructivism and e-

learning: Balancing between the individual and so-

cial levels of cognition. Educational Technology,

41(2):40–44. http://www.jstor.org/stable/44428658.

IEEE (2020). IEEE Standard for Learning Object Metadata.

IEEE Std 1484.12.1-2020, pages 1–50.

AET 2020 - Symposium on Advances in Educational Technology

298

Kla

snja-Mili

cevi

c, A., Vesin, B., Ivanovi

c, M., Budimac,

Z., and Jain, L. C. (2017). Personalization and Adap-

tation in E-Learning Systems, pages 21–25. Springer

International Publishing, Cham.

Lavrov, E., Barchenko, N., and Pasko, N. (2017a). Com-

puter program “Modeling qualimetric complex of di-

alog interaction in the system student-computer”.

Lavrov, E., Barchenko, N., Pasko, N., and Borozenec,

I. (2017b). Development of models for the for-

malized description of modular e-learning systems

for the problems on providing ergonomic quality

of human-computer interaction. Eastern-European

Journal of Enterprise Technologies, 2(2 (86)):4–13.

http://journals.uran.ua/eejet/article/view/97718.

Romero, D. (2015). What Is the Tin Can

API? An Intro to Tin Can Compliance.

https://www.unboxedtechnology.com/blog/

what-is-the-tin-can-api/.

Wilson, E. V. (2000). Student characteristics and computer-

mediated communication. Computers & Education,

34(2):67–76.

Woolf, B. and Eliot, C. (2005). Customizing the in-

structional grid. Applied Artiﬁcial Intelligence, 19(9-

10):825–844.

Using Intelligent Agent-managers to Build Personal Learning Environments in the E-Learning System

299