2.9 Research Directions
3.1.2 Agent Technology for Knowledge Acquisition
The Delivery and Working Environments are grouping the functionalities of learning systems to enable the learning itself. Therefore they mainly fulfil requirements de-manded by learners. The DE presents the course, its structure, course metadata, enables course catalogue browsing, realises the registration and is responsible for other all func-tionalities that are directly connected with the presentation of and working with learning content during the learning process.
The WE is grouping functionalities for the support of the learning process. That refers to e.g. to classic requirements known from classroom learning. Components for web search as well as for the access to certain repositories are needed to get additional information about the topic of the course. It is important for the personal learning progress to be able to make private annotations to the course content and to manage
for collaborative work and time management and the access to office tools are needed under certain circumstances. Figure 3.3 visualises these chosen aspects for parts of the learning environments.
Delivery Environment Working Environment
Cata-logue Browsing
Course Presen-tation Course Structure
Registra-tion
Web Access
Reposi-tory Access
Annota-tion/Link
manage-ment
Private File System
Scheduler
Office
Figure 3.3:The Learning Environments (cp. [Mencke and Dumke, 2007a])
The learning environments need connections to the Administration and Interaction Environments and to the support layers. Administration for example is needed for the management of individual preferences; meanwhile interaction is fundamental for col-laborative learning tasks. As for the other environments the support layers are providing access to basic information, repositories and functionalities that are needed for the func-tionality of the actual environment itself.
In the following, chosen approaches for the usage of agent technology within the domain of knowledge acquisition are sketched.
3.1.2.1 Agent-Mediated Online Learning
The agent-mediated online learning (AMOL) architecture targets the automisation of a online learning process [Yi et al., 2001]. Therefore the authors assumed three parties of participants: the learners, the teachers and mediating education centers (cp. figure 3.4).
The difference to classic approaches is the existence of multiple education centers to provide the courses. A prototype was implemented with aglet technology (cp. section 1.2.6.4).
110 3 Agent-Supported e-Learning
Learner group
Education center group
Learner
Learner
Learner
Teacher group
Teacher
Teacher
Teacher Education center 1
Education center i
Education center n
...
Figure 3.4:AMOL architecture (cp. [Yi et al., 2001])
The implemented agents are mobile and their types are listed below:
◦ Pegagocial agent: tutoring based on task plan and feedback (answering the learner’s questions and judging his answers)
◦ Searching agent:searching for appropriate learning content based on learner request
◦ Querying agent:querying the various education centers for answers the pedagogical agent is not able to provide
3.1.2.2 Knowledge Assessment with JADE
A next architecture was described in [Anghel and Salomie, 2003]. It targets a special domain of e-Learning: the student assessment. The representativeness of this architec-ture is derived from its way of implementation. It is implemented by using JADE agent technology (cp. section 1.2.6.1) in an applet of a Web site. Parts of the architecture are visualised in figure 3.5.
Agent technology was chosen because of scalability issues for many users and bandwidth/latency related problems of the classic client-server model. The authors identified the following tasks for agents in their domain of interest:
◦ Personal assistant agent:for human-computer-interaction
◦ Server agent:coordination of evolving tasks (e.g. handling self-assessment requests, generating corresponding evaluation engines)
◦ Evaluation agent: evaluating the tests based on test information (questions, answer options, correct answer) and assessment process information
GUI
Intructor module
Student module
Admin module
Business logic Mobile agents
Data acess
Database management system
Utility server
Figure 3.5:Architecture for knowledge assessment with JADE (cp.
[Anghel and Salomie, 2003])
3.1.2.3 File-Store Manipulation Intelligent Learning Environment
The File-Store Manipulation Intelligent Learning Environment (F-SMILE) was pub-lished by [Virvou and Kabassi, 2002]. It is intended to teach novices the usage of a graphical user interface. Therefore it is protected and offers adaptive tutoring and help, based on the observed user actions. Used adaptation techniques are adaptive presenta-tion and adaptive navigapresenta-tion support [Kabassi and Virvou, 2003].
Learner modelling agent Learner
modeller Error diagnoser
Longterm user model
Tutoring agent Curriculum
generator Example generator
Advising agent Advice evaluator Advice formation Speech
driven agent Domain represen-tation
Figure 3.6:F-Smile’s architecture (cp. [Virvou and Kabassi, 2002])
112 3 Agent-Supported e-Learning
Four agent types are implemented (cp. figure 3.6):
◦ Learner modelling agent:observation of the learner’s characteristics and identifica-tion of possible misconcepidentifica-tions
◦ Advising agent: simulation of a tutor’s reasoning by the application of an defined formula that deals with the degree of similarity, typicality, degree of frequency, dom-inance to calculate the degree of certainty of the appropriatness of an given adice
◦ Tutoring agent: content, link and example adaptation based on learner information
◦ Speech driven agent: avatar for human-computer-interaction to provide entertain-ment and emotional function
3.1.2.4 Extended LMS “Samurai”
In [Ueno, 2005] the agent-based extension of the existing learning management system
“Samurai” and an analysis of its usefulness is described. Agents are used to provide optimized instructional messages to a learner. Therefore they identified nine primary variables of the user model as informational base for adapted message delivery.
A major part of their work was the comparison of courses held with and without the agent-based extension. The main results where:
◦ Reduced number of students, who gave up the course
◦ Improved test score
◦ Reduced variance of test score
◦ Increased learning time
3.1.2.5 Web-Based e-Learning Environment Integrating Agent and Computational Intelligence
A system for web-based elearning integrating agent and computational intelligence is described in [Giotopoulos et al., 2005]. The platform frontend, the student questioner reasoning and the student model agent, are connected with Web services (figure 3.7).
E-learning platform front-end
Student model agent Student questioner reasoner SOAP
SOAP SOAP
Figure 3.7:System architecture with Web-service-based interconnection (cp.
[Giotopoulos et al., 2005])
◦ Leading of the learner through the learning process
◦ Update of the learner model
◦ Access to possible interesting resources
3.1.2.6 Intelligent Learning Materials Delivery Agents
The Intelligent Learning Materials Delivery Agents (ILMDA) application was designed to deliver learning material to different students taking into account the content’s usage history and the student’s user profile.
The agents task is to learn from the available history data and to make assumptions about the appropriatness of learning material for certain students. The ILMDA archi-tecture is sketched in figure 3.8 [Soh et al., 2005a].
ILMDA Agent
Database
ILMDA reasoning Historical
profile, real-time behaviour
Parametric profile of student and environment
Retrieval instructions, profile updates, statistics updates
Examples, exercise problems, statistics Timely delivery
of examples &
exercise problems Lectures
Student
Figure 3.8:ILMDA architecture (cp. [Soh et al., 2005a])