Intelligent Systems/Tools in Training and Lifelong Learning.

Learning is an active process clearly distinguished from simply being taught. Active involvement in learning helps learners build knowledge in their heads, which is one of the key issues advocated by constructivists. However, learners still need other kinds of help that instructivists might suggest. Not all learners can always be active. They sometimes meet an impasse they cannot resolve by themselves. What they need is not the extreme paradigms but a mild and friendly paradigm that enables them to be active when they want and guide them when necessary. Such a paradigm needs sophisticated intelligence.

MEDIA TECHNOLOGY FOR THE FURTHER EVOLUTION OF LEARNING PARADIGMS

Learning is a lifelong process where learners cannot always expect teachers around them. People need tools/systems to facilitate the learning.

One of the problems is adaptation to changes in the environment, which require evolutionary tools. In industrial settings, people are expected to catch up with rapid technological development--a reason why training is so important in industries. Billions of dollars are spent in employee training each year. One of the serious problems is lack of trainers. Advanced training systems of high performance also require intelligence to partially mimic such trainers' performance.

We have learned a lot about technology-based learning paradigms to date. They include:

* Computer Aided Instruction (CAI).

* Intelligent CAI (ICAI).

* Micro Worlds.

* Intelligent Tutoring System (ITS).

* Discovery Learning.

* Interactive Learning Environment (ILE).

* Intelligent Learning Environment.

* Simulation-Based Training.

* Computer-Supported Collaborative Learning (CSCL).

* Web-Based Learning.

* Life-Long Learning (LLL).

In the evolution of the paradigms, we could see the direction in which we were headed. It was a learner-centered, continued, network-supported, collaborative, and intelligent learning environment. This direction enjoys many of the characteristics of the paradigms we have seen thus far and compatible with most of them.

To learn more about such a new direction, we have this special issue on:

Intelligent Systems/Tools in Training and Lifelong Learning 

Fifteen articles were submitted and 10 articles were accepted with one invited article by Gerhard Fisher, Center for Lifelong Learning & Design ([L.sup.3]D), University of Colorado, USA. His article provides comprehensive views of lifelong learning. It is based on thorough investigation of the current situation of training and learning, and the conventional learning sup-authentic, situated, continued, domain-oriented, collaborative, and open environment for learning/training based on constructivism and technology support. It declares a departure from institutional education/training. He also presents his activities along with his theories.

The ten papers are classified into the following three categories:

An Architecture/System Based on Constructivism

1. Collaborative Innovation as a Process for Cognitive Development, by Madhumita Bhattacharya and Ranajit Chatterjee.

2. Learning at the Mental Gym--How to Get Mentally Fit for the Task You Have at Hand, by Eshaa M. Alkhalifa and Helen Pain.

3. Towards Adaptivity and Agility--A Fractal View on Learning Environment, by Sylvain Giroux, Richard Hotte, and Kim Dao.

4. Learning by Judging: A Network Learning Environment Based on Peer Evaluation, by Chuen-Tsai Sun.

Innovative Al Technology Applications to Educational Systems

5. Smex Web: An Adaptive Web-Based Hypermedia Teaching System, by Florian Albrecht, Nora Koch, and Thomas Tiller.

6. Pedagogical Ontology and Teaching Strategies: A New Formalization to Improve Lifelong Learning, by Sylvie Ranwez, Torsten Leidig, and Michel Crampes.

7. The GET-BITS Model of Intelligent Tutoring Systems, by Vladan Devedzic, Ljubomir Jerinic, and Danijela Radovic.

Report on Practical Experiences of Learning Environments

8. Micro-Robots Based Learning Environments for Continued Education in SMEs, by Pascal Leroux and Martial Vivet.

9. Arithmeticus: A DPS-based Model for Arithmetical Competence, by Joseph Klep. A (Retro)Prospective Study of the Elementary Conceptual Model for Learning

10.An Attempt to Define a Proper Relations Scheme Between Instruction and Learning, and to Establish the Dynamics of Learning in Relation to Modern Political Concepts as Study-Fairness, by Rik Min, Piet Kommers, Hans Vos, and Cor van Dijkum.

Those in the first category share a lot in common besides a constructivism-based approach. All of them deal with collaborative learning, and three of them assume Internet environments. Giving learners the initiative in the learning process, the systems/architectures exploit the capabilities of the Internet to enable collaboration beyond the limitation of distance. The common attitude towards learning support includes efforts in finding the best mixture of free exploration with the highest initiative of a learner and an appropriate guidance from the environment. The authors have come up with two types of solutions to this problem: one is to put the learner in collaborative settings and the other is the introduction of guidance capability to the system.

After comprehensive investigation of the role of collaborative innovation, Madhumita Bhattacharya and Ranajit Chatterjee introduce several useful tools for goal-oriented collaborative discussion for innovation and designed a unified system. The unique features of their research include basing the design rationale of the system on learning theories, and devising practical tools. Eshaa M. Alkhalifa and Helen Pain combine a virtual reality environment with simulated players and intelligent tutoring to give a learner appropriate guidance, keeping a situated learning environment. Chuen-Tsai Sun also employs group learning with peer evaluation through the Internet to realize the Delphi method. Collaboration with human and/or simulated agents give the learners motivation and active involvement in the learning process. These ideas also contribute to guiding learners and preventing them from being lost. To attain the same goal, Sylvain Giroux, Richard Hotte, and Kim Dao discuss an innovative idea of hypermedia-based navigation to realize high adaptivity and agility. A key idea here is to prepare a sophisticated multidimensional hyper-link mechanism that is adaptive to the context as well as to learners' and cognitive styles. Although the system assumes a single learner paradigm, it succeeds in realizing a learning environment of a good mixture of free exploration and guided learning.

Articles in the second category are different from those discussed previously because they are more enabling and technology-oriented. The common claim of the three articles is that artificial intelligence technology works well for adaptive tutoring/teaching to facilitate strong learning environments, compensating for negative aspects of free exploration of the information superhighway. Two of them deal with web-based learning environments, and two of them discuss ontology as an innovative enabling technology. Florian Albrecht, Nora Koch, and Thomas Tiller discuss an Internet-based approach and propose an adaptive, web-based tutoring system. They introduce Java applet to build learner model-based adaptive teaching in a web-based teaching environment. Sylvie Ranwez, Torsten Leidig, and Michel Crampes discuss a pedagogical ontology for dynamically composing web-based learning material. They put the adaptive navigation of hypermedia one step further to adaptive building of hypermedia as a learning material. Thus , these two articles try to find a good solution of active behavior of the learning environment. Vladan Devedzic, Ljubomir Jerinic, and Danijela Radovic discuss a sophisticated model for building an ITS. Their main interests include reusable component-based architecture that is deeply related to ontology. Their contribution is based on amalgamation of software engineering and AI.

In the third section, Pascal Leroux and Martial Vivet present a report on substantial activities in the real world, Small and Medium Enterprises: SMEs. It is about a year-long practice of well-situated and continued training using micro-robots, which have a lot to learn about how ideas have evolved.

Joost Klep describes the rationale and interactional consequences of his learning program, Arithmeticus. MathMirror is its front end that allows a student to express calculations by manipulating mathematical objects. The manipulations are interpreted as partial solutions and are recorded with time annotations. Comparing these students' expressions with solutions produced by Arithmeticus gives abundant information about the students' work. So, it is possible to qualify solutions in terms of effectiveness, speed, degree of automation, and rote knowledge used. Those qualifications are based on students' individual learning histories. Arithmeticus and MathMirror offer a learning environment in which children can develop their own strategies. There are no predefined expert solutions. A solution can be good or nicely related to previous learning history. The more strategies or facts learned, the smarter Arithmeticus will be and the more smart solutions Arithmeticus will expect from the student.

[INCOMPLETE]ditional reductionism to find the most elementary mechanism in learning, regardless of the control aspects around the learner. This article may be seen as a provocation to the various learning paradigms such as cognitivism and environmentalism, not to speak of constructivism. In the era of "zapping" and "staccato" browsing through the information ocean on the World Wide Web (Web) however, this notion about short exposure time and short-term memory (STM) may need this almost forgotten analogon in order to understand the mechanism of the feedback regulation system.

If completely free exploration is the primary key to implement constructivist ideas, then a set of books is the best solution. If the complete support of the learning process is necessary, on the other hand, ITS with stronger AI technology gives the best solution. No one believes, however, that either of both extremes is what we need. Appropriate stimuli should be given at appropriate times to the learners to maintain motivation, to help them when they reach a deadlock, and to let them reflect. All learners cannot behave actively to explore the huge space continuously. They will definitely come across difficulties that need to be overcome with help. These are areas where AI technology might contribute.

One of the main issues here, however, is that all AI-based methods need modeling, which is the key to enabling sophisticated and adaptive behaviors such as suggestion generation. Without modeling, a system can show hard-wired behavior (canned suggestion generation), which cannot be adaptive to the situation. The problem with modeling is that it does impose something negative on a constructivism approach, since it requires limitations of the target to model and such limitations might include a learner's cognitive state, learner's behaviors, interactions among co-learners, and so forth. This has been one of the main reasons why many of the AI-based systems have not been completely compliant to the constructivist approach. Articles we have in this special issue try to overcome these difficulties. Although the evaluation is up to the readers, the article should be informative and thought provoking.

As we have browsed all the articles, what you will find in this special issue are comprehensive and theoretical articles, sophisticated techniques to realize theories, and practical experiences in industrial settings. The co-editors would like for the readers to enjoy them.