Metacognitive Agent's Contribution to the Learner in a Technology

Enhanced Learning

Hanane Elbasri

, Adil Haddi

, Hakim Allali

, Othmane El Meslouhi

LAVETE Laboratory, FST, Hassan 1st University,Settat, Morocco, North Africa

Ibn Zohr University, Agadir, Morocco, North Africa

Keywords: Technology Enhanced Learning (TEL), metacognition, metacognitive agent, multi-agent system.

Abstract: Online learning is the field of action of the article. We are, therefore, in full "Technology Enhanced

Learning” (TEL).In the absence of the teacher in the TEL and faced with the difficulties encountered,

learners have a high probability of being demotivated and can, therefore, give up learning very early.To

remedy these drawbacks, the article proposes an intelligent agent that helps the learner to build a reading

plan for the course, to choose a learning strategy and readjust it according to his progress towards his

objectives. This agent is the metacognitive agent that, by design, is precisely there to encourage the learner

to self-evaluate his learning.The article shows the interactions between the proposed agent and the learner

through a number of criteria that allow the agent to determine when and how to react.These criteria take into

account the progression and speed of learning, the performance of the learner as well as the result of the

formative evaluation throughout the learning process.The agent is modeled by distributing the different

stages of metacognition and designing the interaction between the learner and the metacognitive agent.

1 INTRODUCTION

A number of studies have been conducted to

improve the metacognitive competence of

programmers (Bernard and Bachu, 2015). Other

studies have focused on how learners provide the

necessary knowledge (Amine and al., 2017) while

other works deal with metacognitive support to

accelerate computer-assisted learning (Mohd and

Ismail, 2017).

In our work, we start from the observation that

the absence of the teacher in online learning requires

the learner to cope with the learning process without

a guide or guidance. To support, guide learners to

succeed in learning, in distance education, and avoid

the high dropout rate observed (Clément, 2014), we

propose to integrate an agent who encourages the

learner to define his goal, to make good planning,

and determine the appropriate learning strategy. In

case of failure, our agent will trigger a series of

incentives before, during and after reading a course.

These incentives are done automatically and at

predefined times.

The aim of our study is the definition of the

interactions between the learner and the

metacognitive agent, interactions that go through the

integration, in the system, of intelligent agents for

the validation of the distribution rules of the

metacognitive incentives in a "Technology

Enhanced Learning "(TEL).

TEL is an environment that has human users

(learners, teachers) and a computer system with

which users interact. This interaction is intended to

facilitate the training and enrichment of learners.

This interaction is intended to facilitate the training

and enrichment of learners.Online learning has

evolved significantly in recent years. Today, learners

have the opportunity to learn lessons through a

series of direct interactions with TEL. The problem

is that the learner must learn without follow-up and

without assistance throughout the learning process.

Our agent can intervene in the learning processes

at predefined times, and communicate with other

agents to carry out its work. It helps the learner to

focus on his goal - a predefined goal - before, during

and after reading the lesson.

The main contribution of this work is the

modeling of interactions between our metacognitive

agent and the learner.

The article is organized as follows: Section II

describes the work related to our study. Section III

Elbasri, H., Haddi, A., Allali, H. and El Meslouhi, O.

Metacognitive Agent’s Contribution to the Learner in a Technology Enhanced Learning.

DOI: 10.5220/0009770900550061

In Proceedings of the 1st International Conference of Computer Science and Renewable Energies (ICCSRE 2018), pages 55-61

ISBN: 978-989-758-431-2

briefly discusses artificial intelligence and

metacognition, and section IV describes the

methodological approach. In section V, we present

the modeling of our metacognitive agent in

interaction with the learner. Finally, in section VI,

we present our main conclusions.

2 RELATED STUDIES

Learning can take many forms. In face-to-face

learning, some works have focused on metacognitive

integration initiated by the teacher from time to time.

These works show that this kind of metacognitive

intervention helps the learner as LilianePortelance

does in 2002 (Liliane, 2002).

In classrooms, and in e-learning, some researches

have been done to identify tools for improving

metacognitive skills in learners. This is the case, for

example, of Bernard's team in 2015 (Bernard and

Bachu, 2015). Other works discuss the

characteristics of a metacognitive support system.

The works of MohdRum and others in 2017 (Mohd

and Ismail, 2017) go in this direction.

Other studies have focused their research on

improving platforms in distance education to help

the learner follow his studies. The problem with

these platforms is that, in all the works, the focus has

been on the integration of agents at the cognitive

level of the learner even if we observe the

abandonment of the continuation of learning.

Most web-based open source learning

management systems, such as GANESHA,

MOODLE and BLACKBOARD, are widely used,

and successfully, in distance learning. These systems

offer a variety of functions to support the learner to

understand his or her courses. Despite this, currently

such environments offer very little intelligent

support for learners.

The software agent technologies are based on:

 Cognitive agents (S.Pestyand al., 2003):

knowledge and reasoning related to

applications,

 Rational Agents: justification of

decisions and illustration of results

according to rules,

 Intentional agents: choice of the task

according to the means of specific

assignment. One example is the BDI

agent (Belief-Desire-Intention) (Karl,

2014).

The indirect monitoring of the learner, that is to

say the notions of "metacognition" and "intelligent"

will be developed in our modeling.

3 ARTIFICIAL INTELLIGENCE

AND METACOGNITION

3.1 Artificial Intelligence

Artificial intelligence is recognized as a computer

discipline that aims to model so-called "intelligent"

human behaviors such as perception, decision-

making, understanding, learning.

The intelligent agent is a physical or virtual

entity that operates automatically and autonomously.

Indeed, he is able to communicate directly with

other agents and to perceive his environment. In

addition, he is able to learn from experience and

perform activities in a flexible and intelligent way.

An intelligent agent is, quite simply, a simple

informationretrieval system in an automatic manner

that is to say without the intervention of the user. It

is characterized by interactivity, autonomy and

intelligence.

3.2 Metacognition

Metacognition is about having a mental activity on

one's own mental processes, that is to say, what an

individual knows about his or her way of knowing.

And more precisely

1- to know that we know,

2- to know that one is able to memorize.

Metacognition is thus a factor facilitating

learning and contributing to the development of the

learner through a better knowledge of oneself and

one's possibilities. In a socio-constructive approach,

the learner is an actor of his own learning.

In our work, we modeled:

 the role of the planning phase: so that

the learner is able to organize the way in

which he will use the information, that is

to say to define his objectives, to ask

himself questions before reading a text,

etc.;

 the role of the control phase: so that the

learner can make the decisions that aim to

manage the understanding, that is to say,

to concentrate his attention, to test

himself during the reading, etc. ;

 the role of the self-regulation phase: so

that the learner is aware of the activities

that are strongly related to control, that is

to say, reduce the speed of reading to

adjust to the difficulty of the text, etc.

ICCSRE 2018 - International Conference of Computer Science and Renewable Energies

3.3 The Characteristics of the

Intelligent Agent

An intelligent agent is a physical entity like a robot,

and a virtual entity like a software. The role of the

intelligent agent is to assist you, to help you find the

information.

An intelligent agent has characteristics

 Autonomy: it can make autonomous

decisions;

 Interactivity: it can interact with its

environment and other agents to

accomplish its task;

 Intelligence: it knows how to reason and

learn from the information it collects..

There are different categories of agents and

among these agents we find:

 the reactive agent: it uses the stimulus-

reaction type capacity;

 the cognitive agent: it is based on

knowledge;

 the intentional agent: it drafts plans and

makes them possible;

 the rational agent: it is based on logic.

Depending on the rules adopted or the

methods applied, he justifies it decisions

and illustrates the results;

 the adaptive agent: it is able to learn

according to the constraints of the

environment;

 the communicating agent: it has the

ability to communicate;

 the metacognitive agent: it is able to

assist the learner when needed, that is to

say to guide the learner to adopt a

learning strategy (what ?, when ?, how?).

It communicates with other agents

(Hamid and al., 2015) (Lotfiand al.,

2015). This agent is used here.

4 THE METHODOLOGICAL

APPROACH USED

Figure 1: Tree illustrating the operations and the

integration of the agents according to the state of the

learners (H.Elbasriand al., 2018)

Given the results obtained in the test we conducted

at the Faculty of Science and Technology of Settat

(Morocco, North Africa) in July 2017 with students

in a computer course (H. Elbasri and al, 2018), the

following is the method of integrating metacognitive

incentives:

 the metacognitive agent focuses, first, on

the detection of the state (disoriented,

confused, unwise, ...) of the learner as

shown in the decision tree (see Figure 1),

follows the learner, state by state, and

triggers, if necessary, the appropriate

incentives according to the rules

determined by the decision tree ;

 our study focuses, first, on the

management of the dialogue between the

agent and the learners. The metacognitive

agent acts by sending messages,

questions, metacognitive incentives in an

implicit manner, the goal being to follow

the metacognitive level of the learner;

 the work of our agent is as follows:

if the learner answers the implicitly asked

questionnaire, then

 the results are calculated to identify the

state of the learner (disoriented, unwise,

confused, failure) ;

 the metacognitive agent is used to

"encourage" the learner according to his

needs, given his condition.

End if

Metacognitive Agent’s Contribution to the Learner in a Technology Enhanced Learning

Once the state of the learner has been identified, the

metacognitive agent sends explicit incentives to

trigger and ask the relevant questions.

If the state of the learner = disoriented, then

 our agent focuses its work on planning

(H. Elbasri and al., 2018) : at each shift

from one concept to another, at the time

of reading, the agent sends out incentives

that help the learner to orient him

correctly. The agent abstains when the

learner, voluntarily and indirectly, has the

capacity to plan without help, that is to

say when the learner shows autonomy.

End if

If the state of the learner = confused, then

 our agent focuses its work on self-

assessment and self-regulation strategy : at

each shift from one concept to another, at

the time of reading, the agent sends out

incentives that help the learner to self-

evaluate and self-regulate. The agent

abstains when the learner, voluntarily and

indirectly, has confidence in himself, that is

to say when the learner shows autonomy

and confidence.

End if

If the state of the learner = unwise, then

 our agent focuses its work on the

connection between the concepts, that is to

say on the routing of a course: at each

passage from one concept to another, at the

moment of reading, the agent sends

incentives that help the learner find the

links between the concepts. The agent

abstains when the learner, voluntarily and

indirectly, has the ability to find the links

between the concepts when moving from

one concept to another.

End if

The aim of the system is to provide the teacher

with indicators that classify learners according to

their states (motivated, confused, disoriented, etc.)

and this, taking into account the difficulties

encountered during the training process. The system

has 3 agents and 3 databases (see Figure 2).

Figure 2: Architecture based on the multi-agent system

The 3 agents developed for automated

monitoring are:

- the Metacognitive Agent : this agent

assists the learner in his learning process. It

intervenes at predefined times. It sends

message-alerts, metacognitive incentives to

help the learner plan his learning, maintain

his motivation and self-confidence ;

- the Time Agent : this agent records the

consultation time of each teaching unit. It

also calculates the speed or time elapsed

since the execution of the previous task and

the progress of the mission ;

- The Verification Agent: this agent

compares the minimum consultation time

and the effective consultation time of the

learner and the average mark and the mark

obtained by the learner.

The three databases are DB1, DB2 and DB3 and

have been designed to record learner traces (see

Table 1)..

Table 1: Database contents

DB1

1-The feedback of the learners

2-The comments

3-The questions

4-The answers to the

uestionnaires

DB2

1-The answers to predefined questions

2-The answers to predefined messages

3-The minimum consultation times

DB3

1- The final results of each learner

2-The consultation times

3-The objectives

4- The strategies adopted

5- The references to ambiguities,

difficulties

6- The links between the concepts

ICCSRE 2018 - International Conference of Computer Science and Renewable Energies

5 MODELING AND DESIGN OF

THE METACOGNITIVE

AGENT

5.1 Design Methods for "multi-agent"

Systems

In Table 2, below, are presented some

methodologies for designing "multi-agent" systems.

Table 2: Some design methodologies for multi-agent

Methodology Description

AAII (Australian

Artificial Intelligence

Institute (Kinnyand al.,

2015

)

This method defines the

specifications of the

BDI agents

GAIA (Wooldridge and

al., 2000)

It uses organizational

paradigms. It is

relatively limited to

describe systems

“multi-agent”

AUML (Agent Unified

Modeling Language)

notation

(S.Lynch and

Rajendran, 2004)

It is designed to fit the

ratings

"UML" to the writing

of the "agent" oriented

modeling.

In this article, we are interested only in the

communication between the metacognitive agent

and the learner, communication which is done by

sending messages whose content is a series of

incentives helping the learner to improve his

learning and triggering his autonomy. So, to

communicate, agentsneed communication protocols

to ensure the continuity of relations and exchanges

between them. Communication is, therefore, an

essential characteristic of our agent because

communication is the basis of the social behavior of

agents.

To model our agent, we used the "AUML" as a

modeling language.

5.2 Agent Design

The following diagram allows us to define the

functioning of our metacognitive agent and also

allows us to define its "capacity", that is to say it

specifies what an agent can do and under what

conditions it can do the task.

Figure 3: Capacity diagram of the metacognitive agent

Table 3: Capacities of the metacognitive agent

Capacity

Input Output Goals

Planning Replies

to the

question

naire

Information

that can be

used to set

goals

Establish

learning

plan and

set goals

Orientin

g the

action

Determi

ne the

strategie

s to

reach its

objective

Control Replies

to the

question

naire

Clic

streamin

Side

anal

sis

Information

that can be

used to

regulate

learning

Monitori

learning

Regulati

Informat

ion

provide

d by the

"control

" tas

Rules Correct

the

learning

process

Metacognitive Agent’s Contribution to the Learner in a Technology Enhanced Learning

5.4 Interactions between Agent and

Learner in Proposed Architecture

The diagram in Figure 4 shows, by way of example,

the interactions between the agents if the learner is

in a state of "disoriented".

Figure 4: Interactions between learner and agents

6 CONCLUSIONS

In this article, we are interested in defining the

interaction between the learner and the

metacognitive agent in its role and in its interest for

the learning system. The proposed agent can help the

learner to trigger self-reliance, improve critical

thinking and, finally, build one's own active

metacognitive level.

Our agent allows the learner to develop

metacognitive skills such as:

- to think about your own action,

- to become aware of the routing that led to

the obtained results,

- to find, himself, his mistakes and seek

appropriate solutions.

The intervention of our agent is done according

to predetermined rules and according to the state of

the learner.

To make our system dynamic, these rules will be

automatically discovered. This will be the subject of

future team work.

REFERENCES

Amine Belahbibandal., 2017. “Serious Games Adaptation

According to the Learner’sPerformances”,

International Journal of Electrical and Computer

Engineering (IJECE), vol/issue:7(1), pp. 451~459.

Bernard, and Bachu, 2015.“Enhancing the metacognitive

skill of novice programmers through collaborative

learning” in “Metacognition: Fundaments,

applications, and trends”, volume 76,pp. 277-

298..doi:10.1007/978-3-319-11062-2_11.

Clément Dussarps,2017. « L’abandon en formation à

distance», Distances et médiations des savoirs [en

ligne], 10 | 2015, mis en ligne le 17 juin 2015, consulté

le 12 Septembre. URL :http://dms.revues.org/1039.

H. Elbasri, A.Haddi, H.Allali, 2018.” Improving E-

learning by Integrating a Metacognitive Agent”, article

to be published in October, 2018 by International

Journal of Electrical and Computer Engineering

(IJECE).

Hamid B., Mohammad Ali T. and Zhaleh G., 2013.

“Architectural Approaches for Self-Healing Systems

Based on Multi Agent Technologies”, International

Journal of Electrical and Computer Engineering

(IJECE),vol/issue: 3(6), pp. 779-783.

Karl Devooght,2007.article"Modélisation de la capacité

d’un agent intentionnel en fonction de ses activités".

Kinny D. Georgeff M. Rao A.1996.

«AMethodologyandModellingTechniqueforSystems of

BDI agents»,Van de Velde W.,Perram J.

W.(dir.),Agents Breaking Away:Proceedings of the

Seventh European Workshop on Modelling

Autonomous Agents in aMultiAgent World,vol.1038

de LNAI,Springer,p.51-71.

Liliane Portelance, 2002.“Intégrer la métacognition dans

l’ensemble de ses interventions pédagogiques” (1 réf.)

No 122, p. 20-23.

Lotfi E., Amine B., Mohammed B., 2015. “Towards a

System of Guidance, Assistance and

LearningAnalytics Based on Multi Agent System

Applied on SeriousGames”, International Journal of

Electrical and Computer Engineering

(IJECE),vol/issue: 5(2), pp. 344-354.

ICCSRE 2018 - International Conference of Computer Science and Renewable Energies

Mohd Rum, S. N., & Ismail, M. A., 2017.

“Metocognitive Support Accelerates Computer

Assisted Learning for Novice Programmers”,

EducationalTechnology&Society,vol/issue:20(3),pp:1

70–181.

S. Lynch, K. Rajendran, 2004. “Design Diagrams for

Multi-Agent Systems”, pp: 66~ 78,

S. Pesty, C. Webber, and N. Balacheff, 2003.« Baghera:

une architecture multi-agents pour l'apprentissage

humain” Cognitique, (P. Aniorte, S. Gouardereseds),

CepadeusEdition,Toulouse.

Wooldridge M., Jennings N.R, and Kinny D. ,2000.

«The Gaia Methodology For Agent-Oriented

Analysis and Design », Journal of Autonomous

Agents and Multi-Agent Systems, 3 (3),pp 285-312.

Metacognitive Agent’s Contribution to the Learner in a Technology Enhanced Learning