Future Trends

---

A great deal of the accomplishments in AI research, along with wide applications to MAS and e-learning, are fueling development of the intelligent support for MAS-based DLEs by applying AI techniques. However, to design more useful, more advanced, and more widely applicable MAS-based DLEs, we still have many challenges, such as the proper handling of the huge amount of data, the reduction of noise and error in data, the negotiation between agents in different educational institutes, the management and explanation of knowledge for DLEs, and so forth. Fortunately, these issues can be solved by using the latest advanced AI techniques, including data mining, fuzzy CBR, policy-based approach, semantic Web, etc. There are some positive trends in the development of intelligence for MAS-based DLEs. The main trends are as follows:

1. Applying data mining to student modeling to improve the model accuracy

2. Using integrated reasoning techniques to build more advanced intelligent support for effective e-education

3. Supplying policy-based negotiation ability for heterogeneous MAS in order for DLEs to provide effective and safe collaborative environments

4. Utilizing semantic-Web-based metadata for knowledge management in DLEs

Data Mining to Student Modeling

Data mining has been attracting a great deal of attention from researchers in the AI area. It has been widely applied to many areas, such as knowledge discovery, online analytical processing (OLAP), etc. But what is data mining? Generally speaking, data mining involves extracting or discovering useful knowledge or patterns from the large amounts of data, where the data can be stored in a database, data warehouse, or other information repositories (Han & Kamber, 2001). Data mining can deal well with the noise or errors in them by cleaning and integrating data and can effectively present the data through data transformation. With good data representation, data mining is able to build a more accurate model or knowledge presentation by applying different machine-learning techniques, as well as to select a suitable model or logical knowledge for the applications by evaluating the learned models with a well- designed evaluation approach. Student modeling is a vital issue for designing advanced DLEs. The processing of student modeling is extremely complicated. The data from the collected student behaviors might contain different formats, uncertainty, and noise, even errors, not to mention its tremendous size. Therefore, we need more advanced techniques for handling the complicated process of student modeling. Data mining comes naturally as a good solution for student modeling.

Integrated Reasoning to Intelligent Support for DLEs

Recently, many researchers have focused on integrating different reasoning techniques into a hybrid reasoning system that can solve more complicated real-world problems to satisfy the various requirements from the application domains. One such trend is to develop the hybrid CBR system by incorporating soft computing technologies, such as fuzzy set theory, neural network, genetic algorithms, rough set theory, and so on. For example, fuzzy set theory has been successfully applied to CBR (Pal, Dillon, & Yeung, 2002) to help match or describe the problems in linguistic terms. With the support of fuzzy logic, CBR can extend its ability in the applications where the transformation between numerical features and quantitative features is required for case indexing, case retrieving, and similarity measuring. Moreover, artificial neural networks (ANNs) are widely used to help CBR learn and generate knowledge and patterns. Today, such a hybrid CBR system is a common architecture for more complicated applications. To provide advanced intelligent support for MAS- based DLEs, it is necessary to build the hybrid reasoning systems that integrate several useful soft computing technologies into intelligent support systems for DLEs. In the learning systems, we have to deal with different knowledge representations for problem description, for student behavior, for domain knowledge, and for learning objects with both numerical and quantitative features. Therefore, it is obvious that the fuzzy-based CBR system is useful and helpful for us in solving the issues by providing more advanced intelligent support for MAS-based DLEs.

Policy-Based Negotiation between Agents in MAS

With the wide application of MAS-based DLEs in various educational institutions all over the world, the requirements for collaborating in education or learning will inevitably increase in the near future. To provide feasible, reliable, and safe collaboration environments for multiple institutions to share their education resources, for the learner to be able to select desired subjects or courses from the different universities, negotiation between the agents in heterogeneous MAS is a fundamental technique. Realizing such a negotiation function requires highly skillful AI techniques. One solution is to apply a policy- based approach to perform the negotiation in heterogeneous MAS. A policy- based approach, which was developed in the distributed network computing area, is an AI-based mechanism. It requires strong support from reasoning techniques such as rule-based reasoning, case-based reasoning, model-based reasoning, and so on. It has been applied to security management and privacy protection for collaborative e-learning systems (Yang et al., 2002). We believe the policy-based approach is useful, powerful, and feasible for realizing negotiations for agents in heterogeneous MAS.

Ontology-Based Metadata for Knowledge Representation in DLEs

As will be described in other chapters of this book, knowledge representation and management in DLEs is imperative. Metadata^[2] is an essential approach to representing knowledge or managing knowledge for building or reusing learning object or course material. It is also the fundamental building block of the semantic Web (Nilsson et al., 2003). An ontology is a controlled, hierarchical vocabulary for representing the knowledge system in the DLEs (Saini et al., 2003). Using ontology and semantic Web techniques, we can create an ontology for metadata in DLEs by using semantic Web schema. Furthermore, the designed ontology for metadata can be expressed with XML or DAML– OIL. Therefore, the usage of semantic Web, XML, DAML–OIL to express ontology for metadata in DLEs will be a fashionable topic in future research of MAS-based DLEs.

^[2]Metadata is defined as “data about data” (Berners-Lee, 1997).

---