An Approach to Assist Learners to Build Their Own Curriculum in Personal Learning Environment Context, Based on the AI Concepts

Abstract

With the emergence of information technologies and the proliferation and diversification of learning tools, educational institutions have diversified their approaches and resources in order to improve learners’ performance and education quality. Thus, they have invested in the learning environments which offer learners’ personalization learning possibilities. But this option depends generally on institutional constraints. Considering this limit, personal learning environments (PLEs) have come to allow learners to individually develop their learning environment by selecting the right curriculum, resources, and appropriate activities. This concept is in vogue, especially in the lifelong learning context. Mostly, the setting up of such an environment is not based on educational concepts (choice of objectives, selection of appropriate programs and activities). As a result, we are faced with undesirable situations: learning is not aligned with learner prerequisites; training layout does not align with the content and learner expectations. The question arises: how can learners define their learning objectives, set up their own activities, and follow their training? In this context, we propose an approach to assist learners to build their own curriculum, which is supported by an assisted PLE developed on the basis of artificial intelligence (AI) concepts and using a dynamic questionnaire.

1. Introduction

Our contribution consists of defining an approach to assist engineering learners to build their own curriculum without pedagogy prior knowledge and to take autonomous control of their learning. It promotes lifelong learning, which aligns perfectly with the goals of sustainable development (SDGs). Three phases characterize our approach: the first concerns assistance in the identification of learning objectives and the recommending appropriate curriculum, the second will be assistance in the search for appropriate activities to the objectives already set, and the last will be responsible for maintaining and managing the learner’s profile. In this paper, we will only focus on the first phase; the purpose is to build a learner’s own curriculum. To do this, a consistent environment must be able to provide assistance to engineering learners to identify and choose their learning objectives. It is based on a dynamic questionnaire that takes into consideration the profiles and feedback learners’ information. An educational learning objective according to IMS-LD standard [1] is represented by a couple formed by a concept (C) and a learning level (N): C being a concept belonging to a domain ontology θ and N is one among the taxonomic levels in pedagogy [2]. Once the couple is identified, our environment will be able to offer one or more curricula from a curricula corpus issue from different sources: educational and training institutions and other resources (Cloud, OER, Moocs). A classification strategy using unsupervised machine learning algorithms (ascending hierarchical classification also called clustering) will then be applied to recommend appropriate curricula. Before going into the details, it would be appropriate to review the basic PLE theoretical foundations and the related work carried out in this area. We will first start by explaining the PLE concept, the PLE-related work, and then we will present our assistance approach, and subsequently our recommender system. We will end by presenting our experiment, and the results are obtained.

2. Theoretical foundations of a PLE

Personal learning environment can be considered as a concept related to the use of learning technologies emphasizing learner ownership of tools and resources. The questions discussed are how does the learner use technology to manage his/her learning? How is individual activity captured? What are the distinctive characteristics of personal learning environment? This study supports the idea that PLE can be considered as a complex activity system using the activity theory (AT) framework [3].

2.1 Activity theory (AT) as an integrated framework

The PLE concept emphasizes the appropriation of tools and resources by learners. The view of learning as a mediation tool or collective activity is the basic principle of activity [4, 5]. Activity theory (AT) has been used as a framework for exploring pedagogical innovations and as a conceptual framework for analyzing and designing support systems for collaborative learning [6, 7, 8, 9, 10], mobile learning [4], and learning technologies evaluation [11, 12].

The study conducted by Buchem on a wide range of PLE publications supports the idea that a PLE can be considered as a complex system of activities and can be analyzed using the framework of activity theory in order to describe its main components [3, 9] (Figure 1).

Figure 1.

Summary of the PLE elements and its main dimensions.

2.2 PLE definitions

Several works have been carried out to bring a certain maturity to the PLE concept and above all a definition; we mainly cite the works of Buchem [3, 13, 14, 15]. From this review of the literature, we have identified several definitions of the term PLE: some have an educational vocation and others a technological vocation (Tables 1 and 2).

Year	Author(s)	Definition
2006	Van Harmelen[16]	“A PLE is a system that helps learners to take charge of their own learning: set their own learning goals, manage their learning (content and process), communicate with others and thus achieve their learning goals.”
2006	Lubensky[17]	“A PLE is the ease for an individual to access, aggregate, configure and manipulate digital objects during his/her learning.”
2006	Milligan, Johnson, Sharples, Wilson, Liber [18]	“A PLE facilitates choice and control for a learner and enables the selection and combination of formal and informal learning opportunities issue from various sources.”
2007	Attwell [19]	“A PLE should not be considered a software application, but a new approach in the use of new technologies in learning.”
2007	Chen, Huang, Li [20]	“A PLE is a learning-oriented strategy that promotes each learner’s learning ability in a web-based learning environment.”
2007	Downes[13]	“A PLE is recognition that the “one size fits all” approach which characterizes LMSs will no longer be sufficient to meet the varied needs of the learners. It is not a software application in itself, but rather a characterization of an e-Learning approach.”
2008	Johnson, Liber[21]	“A PLE is the desire to create learning centered on the learner; who is seen as the provider of his/her learning and that personal learning is fundamentally a learner-driven model of education where the traditional role centered on institutions is contested.”
2008	Wilson, Liber, Johnson, Beauvoir,; Sharples, Milligan[22]	“A PLE can be considered as a concept related to the use of technology for learning with an emphasis on learner appropriation of tools and resources.”
2008	Schaffert,; Hilzensauer[23]	“A PLE is the idea of a user-centered approach to learning, using social tools and software.”
2008	Aviram, Ronen,; Somekh, Winer, Sarid [24]	“A PLE is an educational model used for the self-regulated development of learners who are able to make informed, considered and strategic choices and plan their own learning as well as adapt the learning process according to their own needs, interests and their preferences.”
2010	Buchem[3]	“A PLE can be considered as a concept related to the use of technology for learning with an emphasis on learner appropriation of tools and resources.”
2010	Mcloughlin, Lee [25]	“A PLE is a concept adopting web 2.0 to promote continuous learning, informal learning and self-directed learning. It is an approach and not an application that allows the learner to take control of own learning environment.”
2011	Chatti, Jarke, Specht, Schroeder[14]	“A PLE is a model for learner-centered learning that takes small, loosely-glued instructional chunks, characterized by their free-form use by a set of learner-controlled learning tools.”
2022	Sarah, Serhan, Noraffandy [26]	“A PLE can be identified as a lifelong learning environment.”

Table 1.

PLE definitions from an educational point of view.

Year	Author(s)	Definition
2001	Olivier, Liber [27]	“A PLE is a single-user e-Learning system that provides access to a variety of learning resources, and which may provide access to learners and teachers who use other PLEs and/or other VLE (Virtual Learning Environments).”
2007	Wilson, Liber, Johnson, Beauvoir, Sharples, Milligan [22]	“A PLE is not a piece of software. It is an environment where people, tools, communities and resources interact flexibly.”
2007	Anderson[28]	“A PLE is a web interface in the owners’ digital environment.”
2007	Siemens [29]	“A PLE is a collection of tools brought together under the conceptual notion of openness, interoperability and learner control.”
2008	Attwell, Costa [19]	“A PLE provides both the framework and technologies to integrate personal learning and the work.”
2009	Educause [30]	“A PLE describes the tools, communities and services that constitute the different educational platforms learners use to direct their own learning and to achieve educational goals.”
2010	Martindale, Dowdy[31]	“A PLE is a specific tool or collection of defined tools used by the learner to organize and control their own learning.”
2011	Chatti, Jarke, Specht, Schroeder[14]	“A PLE is a way of creating new web applications by combining (by aggregation and/or integration) existing data and services from different sources.”
2022	Sarah, Serhan, Noraffandy [26]	“A PLE is open access online learning with learner-based guidance, self-direction and self-regulation.”

Table 2.

PLE definitions from a technological point of view.

This diversification in the definitions led us to analyze them more closely and to propose our own definition which has a techno-pedagogical vocation: “A PLE is a learner-centered approach, based on web technologies and allowing support, control and appropriation of learning independently of technical and institutional constraints” [15].

2.3 PLE characteristics

Based on the views of Chattiand et al. [14], Martindale et al. [31], Drachsler et al.[32], Jafari et al. [33], Johnson et al. [21], Lubensky [17], and Guettat et al. [15], PLEs should have the following characteristics:

• PLEs are open systems controlled by learners independent of the educational establishment.

• PLEs are customizable by learners.

• PLEs concentrate all the tools useful for the learner in a single environment.

• PLEs promote informal learning and lifelong learning.

As a result, it becomes clear that PLEs represent a turning point, from a model where learners simply consume information to one where learners become autonomous and create connections with a variety of resources that they select and curate themselves.

2.4 PLE objectives

Although some of the fundamental needs of users of PLEs have not yet been clearly defined, two major objectives have nonetheless emerged in the literature: a PLE must be centered on learners and should enable lifelong learning [3]. These two goals align with the Sustainable Development Goals (SDGs).

2.5 PLE needs

With the trend in PLE being clearly visible today, questions continue to arise [14]: Who would need a PLE? Is there any feedback on PLEs? We need PLEs for lifelong learning. The need is also the response to approaches requiring that the learning environment be under the control of the learner and that activities can be carried out offline. These needs have encouraged certain educational establishments to raise awareness among learners and teachers of the interest and contribution of PLEs, by instilling in them this PLENK 2010 culture [13].

2.6 PLE statistics

A study carried out by Sarah et al. [26] and published in 2022 in the “International Journal of Information and Education Technology,” presented statistics and graphs showing the keen interest shown in this concept by researchers (Figure 2).

Figure 2.

Number of PLE article publications per year.

2.7 PLE limits

Following this brief overview of PLEs, we noted the lack of consensus on the terminology and even on the definition, hence the need for a more stable theoretical framework. In addition, the tools making up a PLE are heterogeneous; the need to orchestrate them is therefore desired. We also noted the limited number of methods and approaches that respond to the various issues raised [15]. We believe that, with the emergence of information technologies, PLEs work will progress further to compensate for the scientific deficiencies observed.

2.8 Synthesis

After analyzing publications made on PLEs, and reviewing the studies and work carried out by many researchers [3, 15, 26], we deduce the learner could obtain his/her educational and technological independence toward institution, by having the possibility of building his/her own learning environment and choosing his/her own resources and activities in order to achieve learning and personal objectives. However, several questions arise: how will the learner identify his/her objectives and therefore his/her learning curriculum?

Two answers can be considered: either the teacher guides the learner in his/her choices, in that case, we return to institutional learning environments with tutors and curricula dependency, or the learner, through the use of an intelligent system, could compose his/her personal curriculum. This second alternative, which has not been the subject of previous work, will be the subject of our research [15].

3. Lifelong learning concepts

Faced with the new landscape of educational technologies, learners continually face challenges in their learning. The speed of change as well as the growth of needs motivate learners to maintain the direction and extent of their lifelong learning. PLEs can be the appropriate solution to these situations. These environments give learners the freedom to learn beyond course boundaries and institutional constraints and customize their own learning environments before and during training. Additionally, e-portfolios used by learners as a tool to trace their learning provide future employers an overview of the individual’s learning history and results, skills, and achievements. With PLEs, they allow learners to demonstrate their professional abilities in a continuous learning framework [34].

3.1 Lifelong learning vision

Lifelong learning is the “Ongoing, voluntary, and self-motivated pursuit of knowledge for either personal or professional reasons. Therefore, it not only enhances social inclusion, active citizenship, and personal development, but also competitiveness and employability” [13, 35, 36]. The diffusing of the lifelong learning vision signals the need for more personal, social, and participatory approaches that support learners in becoming active users and coproducers of his/her learning resources [35, 37]. The emphasis on the shift from formal to informal e-learning through knowledge management and sharing has been placed, with particular attention on the PLE as a learner-centered space. Nevertheless, the investigations are motivated by the many educational theories, implications, and challenges that PLE concept has posed [25].

3.2 Learner-centered learning

In a landscape marked by the evolution and emergence of educational technologies, and innovation in learning modes, models, and methods, the learner is obliged to assume his/her tool choices to use and contributions intended to make in learning. Therefore, we need a learning model centered on learner, adaptable, flexible, and specific, depending on the context, such that the learner will be able to control his/her individual choices in terms of the technologies to use by aligning them with his/her personal needs, interests, learning style, preferences, and context. In this way, learners will know how to build and manage a personal and self-reflective learning environment rather than operating an environment constructed, managed, and imposed by the teacher and/or institution [19].

3.3 PLEs roles in the lifelong learning

The PLEs give students the freedom to learn beyond course boundaries and to personalize their own learning environment. They allow learners to learn anytime and anywhere. E-portfolios are currently used by learners in many education institutions as a tool to document and reflect on their learning. They provide future employers with a snapshot of the learner’s learning history, learning achievements, and reflective practice [38].

3.4 Our critical analysis

Today’s learning systems should break away from traditional learning methods because they can no longer satisfy everyone, especially with the perpetual evolution of technology. Other measures should be found to motivate learners to learn not only when they are in academic training but also when they are independent. In our context, we are interested in lifelong engineering learners whose appropriation of learning can constitute a challenge for them. The solution that seems to be most appropriate is PLE. However, putting up personal learning environments requires solving a number of problems: how can the learner build his own personal curriculum? How does the learner profile will be maintained?

4. Our assistance approach

4.1 Overview

As part of our research in the field of PLE started in 2008, we have developed an innovative approach [15, 36], allowing learners to build their personal learning environment, by building their own curriculums and choosing their appropriate learning activities. Such an approach will promote lifelong learning. To do this, we defined architecture with three components: the “Curriculum builder,” the “Learning activities recommender,” and the “Profile manager” (Figure 3).

Figure 3.

Overview of our assistance approach.

4.2 Modeling learner

Nowadays, there are several specifications that aim to describe the learner in learning environments: IEEE PAPI [39], IMS-GLC-LIP [40], IMS-LTI [41], IMS-GLC-RDCEO [42], and IMS-GLC-LIS [43]. However, no study has been conducted to assess whether any of those specifications are appropriate to the PLEs.

So, we are concerned with finding a specification useful in the case of PLE in general and for our approach in particular. We have identified a requirement set of the learner model:

• Personal information (used to identify the learner and to interact with tools and learning environments),

• Previous knowledge (used to build his personal curriculum),

• Learning traces (used to manage the learning process),

• Learning objectives (used to store personal curriculum), and

• Preferences of the learner (used to select appropriate learning activities).

Based on our study, we demonstrate how the IEEE PAPI is suitable for the case of our approach and in general for the PLEs (Table 3) [44].

Table 3.

Adequacy of IEEE-PAPI for PLE.

4.3 Assistance for identifying learning objectives

This component helps learners to choose their learning objectives. We start by offering them a list of concepts so they can choose one, for example, mechanics, computer science, management, mathematics, or medicine. Each concept has a sub-concepts list. For the “Computer Science” concept we propose “Algorithmics,” “Office Automation,” “Programming,” “Databases,” “Computer Architecture,” “Operating Systems,” and “Computer Networks.” The choice of objectives will be based on an interactive dialogue with the learner using a dynamic and user-friendly questionnaire (Figure 4).

Figure 4.

Objective identification process.

4.4 Assistance for curriculum selection

Once the choice is made (concept Ci, taxonomic level Nj), the next step will consist of finding adapted curriculums. Two possible situations: In the first, an exact match is found between the curriculum’s general objective and the learner learning objective, and in the second situation, we cannot find the right curriculum associated with the concept Ci.

4.4.1 First situation: exact match found

The selected curriculum will be used to identify the learning activities that must be accomplished by the learner. For example, we are looking for a course in “Computer Science” with a taxonomic level equals 2 (“Comprehension”); we found a bachelor’s degree curriculum in computer science that matches. But in such a situation, several equivalent curricula may be found. Faced with such a situation, we will use concepts from Artificial Intelligence (AI) to apply one of the classification algorithms either to aggregate pieces of curriculums found or to make a classification to recommend curricula to the learner [45]. Based on our contribution which improved the IEEE-PAPI learner model in a PLE context, we are detecting significant and useful variables (features) for unsupervised machine learning algorithms—ascending hierarchical classification also called clustering (Table 4).

Variable	Codification	Description
V1	LANG	Learner’s preferred Language: Fr, Ang, Ar, All, Esp.
V2	TYPF	Desired type of training: quick, medium, long.
V3	NBUC	Uses number of a given curriculum.
V4	NBAC	Number of completions on a given curriculum.
V5	RACC	Completion ratio on a given curriculum (RACC = NBAC/NBUC)
V6	NBOB	Number of objectives in a given curriculum.
V7	NATC	Average of marks awarded by learners on a given curriculum.

Table 4.

Sample of variables (features).

In Figure 5, we present a diagram describing the process of obtaining a personal curriculum.

Figure 5.

Curriculum selection process.

In Figure 5, we can see how the learner profile will be updated once the curriculum is selected. We know the curriculum is a couple formed by an objective (Ontology Concept C) and a taxonomic level of learning N. According to our learner model based on IEEE-Papi learner, we will access the “Learning Objectives” class to update it, either by adding this objective if it is not part of the learner’s profile or by updating this objective if the learner has evolved in learning by moving from one taxonomic level to another.

Previously, the classes representing the learner profile (Table 4) will also be updated according to the data provided by the learner, which is associated with the features involved in the ascending hierarchical classification algorithm (machine learning algorithm).

4.4.2 Second situation: right curriculum not found

For example, we are looking for training in BCNF (Boyce Codd Normal Form), but our system found nothing in the corpus. In this case, we need to go down the ontology and go to the “Normalization” node. It would then be necessary to work on the content of each curriculum concerning this node using its XML file and see if the associated block with the BCNF concept exists. The same thing here, we can find several equivalent blocks corresponding to our concept Ci and we must choose the most appropriate according to a classification strategy with always the same sample of variables.

To update the learner profile, we will use the same algorithm presented in the first situation.

5. Experimentation and results

In this section, we will experiment with a part of our approach (identification of personal learning objectives and curriculum recommendation). We have developed an assistant system, which allows any learner to use services offered without any technical or institutional constraints.

5.1 Web architecture

This is web architecture with a client using a browser (e.g., Chrome) containing our system, which will allow the learner to compose a personal curriculum and obtain the list of appropriate activities (Figure 6).

Figure 6.

Web architecture of our assistance system.

5.2 Web interface

In our system, which is based on our approach, we have a thin client which, through a Windows browser, offers the services planned and mentioned by our assistance system (Figure 7).

Figure 7.

Web interface of our assistance system.

By clicking on the “PLE” assistant, the learner could benefit from the offered services system: identifying learning objectives assistant, curriculums recommender, and activities recommender (Figure 8).

Figure 8.

Services offered by our assistance system.

5.3 Assistance in identifying learning objectives

5.3.1 Input data set

To experiment with this component, we had the following data sets:

• Set of learner profiles with different scenarios: learner has never completed learning in a given concept, or has partially completed learning at given taxonomic levels, or has completely completed learning in a concept.

• Sample of learners requesting new learning curricula.

• A corpus of curriculums: each curriculum concerns a well-defined concept.

5.3.2 First situation: learner wants to learn “databases” level 1

Our system will offer him all the curriculums (“DB,” 1) from our corpus. Which one will we recommend to him? Firstly, our system will make a filter by taking into consideration the learner’s requirements and his/her profile. After that, our system will execute the unsupervised machine learning algorithm (ascending hierarchical classification also called clustering). After notification of the variables, we obtained a curriculum list from our curriculum corpus, including those dealing with “Databases” concept and the taxonomic levels. Given that the learner is interested in learning (“DB,” 1), the system extracts from our corpus all the “DB” curriculums with taxonomic level equals 1 (Table 5).

Concept	Level	LANG	TYPF	NBOB	NBUC	NBAC	RACC	NATC
DB	1	1	1	2	1855	1000	0.539	06.20
DB	1	1	1	2	2500	1500	0.600	08.00
DB	1	0.66	0.66	3	1985	600	0.302	05.40
DB	1	0.5	1	2	750	300	0.400	04.80
DB	1	1	0.66	3	2265	1265	0.558	06.10
DB	1	1	0.33	5	4578	4000	0.874	08.70

Table 5.

Extraction of curriculums related to (“DB,” 1).

Which curriculum our system will recommend to learner? The ML algorithm will calculate the similarity distances; before, it converts all the values in the interval [0…1], and sorts the curricula in ascending order according to d2 rubric (Table 6).

Concept	Level	LANG	TYPF	NBOB	NBUC	NBAC	RACC	NATC	Distance (d^2)
DB	N1	1	0.33	5	4578	4000	0.874	0.98	00.979
DB	N1	1	0.66	3	2265	1265	0.558	0.69	02.060
DB	N1	1	1	2	2500	1500	0.600	0.90	02.296
DB	N1	0.66	0.66	3	1985	600	0.302	0.61	02.720
DB	N1	0.5	1	2	750	300	0.400	0.54	03.507
DB	N1	1	1	2	1855	1000	0.539	0.70	02.632

Table 6.

(d²) sorted in ascending order by ML algorithm. Green color indicates the recommended curriculum(s).

As we noted, the curriculum with the lowest distance will be recommended, and in our case, it is the curriculum (“DB,” 1) with distance d² = 0.979.

5.3.3 Second situation: learner having “DB” levels 1 & 2 wants a “BCNF” level 1 curriculum

Our system searched in the corpus but found nothing. He turned back to his domain ontology to go back to one level. There, we found the concept “Normalization.” We know well that the BCNF concept is one of the normal forms encountered in database courses, containing the “Normalization” chapter. We will therefore search all the normalization curricula and detect the presence of the specific objective relating to the Boyce and Codd normal form. This means the system will work on the curriculum content (XML file) and its metadata is made up of the following sections: Concept, level, language, training type, objective number, description, and list of specific objectives.

To find the concept, the system will process the list of specific objectives contained in the XML files. As soon as we find concept_objs = “BCNF,” it will select the corresponding curriculum. After processing the already selected curriculums, we marked those which contain the concept “BCNF.”

An example of an XML block containing the concept “BCNF” is shown as follows:

<Curriculum>

<id>BD25879</id>

<Concept>NORM</Concept>

<Title>Normalization</Title>

<Level>1</ Level>

<Description>Basic Concepts of Normalization</Description>

<Duration>2h00</Duration>

<Nb_Objectives>3</Nb_Objectives>

<li>; <id_objs>1</id_objs>; <concept_objs>DF</concept_objs>; <Title>Functional Dependence</Title >; < Level_objs>1</ Level_objs>; <id_objs>2</id_objs>; <concept_objs>3NF</concept_objs>; < Title >Normal Forms</ Title >; < Level_objs>1</ Level_objs>; <id_objs>3</id_objs>; <concept_objs>BCNF</concept_objs>; < Title > Boyce Codd Normal Form</Title >; < Level_objs>1</ Level_objs>; </li>

</Curriculum>

To find the concept, the system will process the list of specific objectives contained in the XML files. As soon as we find concept_objs = “BCNF,” it will select the corresponding curriculum. After processing the already selected curriculums, we marked those which contain the concept “BCNF.” Following this processing, we obtain the following four curricula (Table 7).

Concept	Level	LANG	TYPF	NBOB	NBUC	NBAC	RACC	NATC	Distance; (d^2)
Normalization	N3	0,66	0,66	0,33	0,47	0,19	0,396	0,98	01,978
Normalization	N1	0,5	0,66	0,33	0,15	0,13	0,821	0,67	02,433
Normalization	N1	1	1	0,00	0,13	0,13	0,977	0,46	02,818
Normalization	N4	0,66	0,66	0,33	0,00	0,00	0,000	0,00	04,676

Table 7.

List of obtained curricula. Green color indicates the recommended curriculum(s).

The learner is looking for a BCNF curriculum with level = 1; our system recommends two, but the one with d² = 0.2433 will be best recommended.

5.4 Experimentation results

Nearly, a hundred learners enrolled in the first year of IT engineering took part in the experiment. At first, they passed a pretest to divide them into two similar groups according to their level. After that, the two groups were invited for a test (on the same day: 2 hours). We asked the learners to solve the same exercise (about relational databases normalization) by creating their own PLE. Each learner in the control group has to build his/her own PLE and therefore to solve the given exercise. However, learners in the experimental group have access to our assistance system installed in their web browser. To evaluate the effectiveness of our approach, we measured the time and scores obtained by the group that used an unassisted PLE and the one that used an assisted PLE. We observed firstly the time of realization for the same activity to the two groups (control and experimental) (Table 8).

	Group	Number of participants	Average (mn)	Standard deviation
Time	Control	50	108,4000	10,20022
	Experimental	50	20,8333	2,00144

Table 8.

Average of times activity.

The results confirm what we observed on the premises: the learner in the control group wasted a lot of time to find the appropriate resources to carry out the activity. We conclude that with an assisted PLE there is a gain in required learning time. On the other hand, we obtained the scores obtained by the two groups (Table 9).

	Group	Number of participants	Average (mn)	Standard deviation
Scores	Control	50	7,85	5,71086
	Experimental	50	14,64	1,64485

Table 9.

Average of scores activity.

We find that the mean of the control group is 7.9 with a standard deviation of 5.7. On the other hand, the results obtained in the experimental group are much better. Indeed, the average score is 14.6 (almost double) with a small standard deviation compared to that observed in the control group. This clearly shows that the use of an assisted PLE improves the learners’ performance.

6. Conclusion

Nowadays, with the emergence of information and communication technologies, learners have access to different learning systems: ranging from simple resources such as tutorials, e-Books, Moocs, and OER, to learning environments, which are generally led by an institutional framework whose teaching strategies are defined in advance: curricula, predefined teaching scenarios and methods, teaching tools and logistics, trainers, etc. In such a context, the learner only exploits these environments at the request of the teacher or the institution without being involved in the structuring of courses and pedagogical choices. The only possibility offered to them during their training is to personalize their environment during learning according to their profile and preferences. Such a strategy cannot surely satisfy all learners: Demotivation during learning, poor perception of certain modules included in the curriculum, and risk of repeating certain modules already validated elsewhere. All these are because of the design and implementation of training strategies that were developed based on teaching centered on the teacher. Such systems, which suffer from several failures: lack of harmonization between education systems and the professional environment and gaps in the design of training curricula, can no longer create value within training institutions.

In certain countries where learning strategies are frequently reviewed in the direction of continuous improvement and learner satisfaction, there is a trend toward a break with traditional systems centered on the teacher. We want learning centered on the learner where the latter would be involved at all levels of training: from pedagogical choices to the choice of resources in the learning environment. Such a system aims to be flexible at all times: learning without time constraints, learning anywhere, learning with any equipment, and total autonomy of learners in the choice of curriculum elements and even in the choice of their activities learning. This break is essentially justified by a better quality of learning with better motivation of the learner, who will move from the traditional passive mode to another comfort with the appropriation of his learning, from the choice of the objectives of his curriculum until the completion of the related activities. This would be ideal for working individuals who wish to acquire scientific and professional skills as part of lifelong learning. A question arises in this case: in practice, are there learning environments can offer this type of service to these learners?

This last decade has seen the emergence of a new concept, which is currently attracting the attention of several research teams. These are personal learning environments (PLE): do not confuse them with customizable learning environments. In our chapter, we have removed this ambiguity in order to clarify the PLE concept for researchers working on it. This concept, despite the efforts made by researchers, fails to reach a maturity threshold in terms of the stability of the concept, given the different educational and technological points of view presented in several works. Most mention the learning resource needs of learners but never the method of identifying objectives and selecting training programs. As if the learner is initiated into the creation of his curriculum, it is able to choose the appropriate objectives according to his profile and able to select the appropriate learning activities. In practice, only those who have teaching skills (trainers) are capable of doing all this. At this level, we asked ourselves several questions: Is the learner capable of identifying his or her own learning objectives? Do they need to introduce them to certain educational concepts to do so? Should it depend on an institutional structure to carry out this work? In other words, could the learner, without being an expert in pedagogy, identify his objectives and build his personal curriculum? All these reflections have been detailed in this chapter thanks to a new assistance for learners in a personal learning environment context and in a lifelong learning framework and which mainly based on artificial intelligence concepts for recommending appropriate curricula and activities.

In this chapter, we presented a brief overview of the state of the art on personal learning environments: we reviewed the basic concepts; the research carried out and the fit between PLEs and lifelong learning. Then, we detailed the different phases of our assistance approach in a personal learning environment: definition and identification of learning objectives, selection of a learning curriculum, and management of the learner profile. For the learner model, we adapted the IEEE-PAPI learner standard for PLEs in general and particularly for our approach (standards adopted have until now not dealt with the case of PLEs). In the different phases of our approach, we used artificial intelligence (AI) concepts and more precisely machine learning algorithms (ascending hierarchical classification) to offer learners appropriate curricula and activities. Our approach was implemented, and the system was developed and tested on a representative sample of learners. The experimentation results are analyzed and interpreted. It was carried out on two groups of learners, a first group working freely on their own PLE to solve a given exercise, with learning tools of their choice, without guidelines, method or approach, and another group, which had the privilege of using our support system while solving the same exercise.

Obviously, this contribution remains open to other actions, either in terms of potential for improvement or in terms of extension. The first action consists of activating the assistance phase which consists of offering appropriate educational activities to the learner after choosing their curriculum; knowing that on the same curriculum, activities can vary from one learner to another depending on their preferences, profile, etc.. The same machine learning process will be applied but this time on activities and no longer on curricula.

As for perspectives, the first will concern the learner profile manager, who is responsible for storing the learner’s information in a consistent manner so that his profile is up to date. A learner’s activities can come from several learning environments, and to update a learner’s profile, we must use a generic model to convert their data not necessarily from the same model. The goal is to prevent the learner from repeating activities already carried out in other environments.

The second perspective is an extension of our approach in terms of learning situations. We recall that we worked on two situations (finding exactly curricula aligned with learning objective and taxonomic level and finding objective and level including in others curricula).

A third situation not developed could be subject to an extension of our approach: it concerns the first approach phase of identifying objectives and the search for a curriculum: it is a situation where the learner introduces a concept Ci relating to a domain ontology and a taxonomic level N to our system, the latter will check in the corpus of curricula, but does not find a curriculum associated with the concept Ci and the taxonomic level N. However, in certain curricula, we find some pieces of curriculum (m_k), which, by aggregating them, can give rise to a curriculum approaching the desired one. Proposing a solution to this situation seems easy on the surface, but in reality, it is not going to be easy. How could our system assist a learner so that the latter can affirm whether such a piece m_k is part of his concept Ci? Generally, this is the mission of a teaching expert. In its absence, it would then be necessary to add an intelligent component (AI concepts) capable of detecting these pieces and recommending them to the learner.