A Framework for Model-Based Adaptive Training

Training System Research - The Setting

Education and training are fundamental in order to maintain a modern industrial base. This is especially true in modern high volume and high value manufacturing areas where the continual introduction of new processes and products makes flexible training necessary for efficiency. Companies are now using Computer Based Training (CBT) technology in a number of training situations.

For example, an export control CBT can be used to deliver training in rules, regulations and restrictions for exporting goods to other countries. CBT advantages include self-paced learning, interactive student participation and lower training costs. CBT systems are currently being used for some training tasks but they have many disadvantages. CBT’s are very inflexible and cannot cope well with incremental changes in system operation, nor are they suitable for some complex tasks.

An alternative to these CBT programs is the use of relatively expensive full-replica simulation systems for training practice. A limitation in many CBT’s is the inflexible approach in training methods and the manner employed in bounding together the subject material taught. Conventional CBT places emphasis on rote learning (memorisation) of declarative knowledge (e.g., facts) whereas the use of simulators places emphasis on the learning of practical physical skills in realistic environments.

The Application of Artificial Intelligence Technology

The application of Artificial Intelligence technology to training has been investigated to enable the teaching of cognitive skills, such as trouble-shooting, and the development of mental models and concepts. The title of a book Intelligent Tutoring Systems [Sleeman & Brown 1982] has given the field the commonly used name of ITS. The typical architecture that has evolved suggests that an ITS needs:

  1. an explicit model of the domain and an expert program that can solve problems in the domain;
  2. a model of the student that identifies, at a fine-grained level of detail, what the student understands; and,
  3. a tutoring module that can provide instruction to remediate misconceptions and present new material [Clancey & Soloway 1990].

However, the practical difficulty of building reliable student models has questioned the central role of the student model [Self 1994]; and, alternative approaches have been developed with generic environments [Hill & Johnson 1993, Van Marcke & Vedelaar 1995].

The research in this work is based on the application of Artificial Intelligence technology, including the potential use of simulator tools in training, with a specific emphasis on industrial training problems. The fundamental approach presented here is to separate the knowledge that is to be taught, called subject expertise (domain expertise), from the way that it is to be taught, called training expertise (training strategies).

This separation allows different combinations of subject expertise and training expertise to be selected for a given training objective, thereby making the system more general and more effective for a given training need.

Flower Show
© | | Sitemap