A Framework for Model-Based Adaptive Training

Identify what is to be taught and how it is to be taught

A system specification is determined by mapping the problem specification (e.g., training need, task decomposition, expertise classification and trainee profile characteristics) to the appropriate training strategy, training plan and training units.

The outcome of training system specification is a separation of what needs to be taught and how to deliver the training. This specification feeds into the MOBAT design methods for the expert agent(s) and trainer agent.

Skeleton Training Plans

A skeleton training plan is essentially an instructional layout specification of sequential and parallel training unit arrangements. The layout of a set of training units, and the type of training units used, distinguishes each training plan from others. The trainer expertise is embedded in the selection criteria and characteristics of training plans and training units. This expertise must be captured via the trainer interface of the training system architecture.

A detailed discussion of pedagogical issues in training plan design is outside the scope of this report. The help of professional trainers and instructional technologists may be needed here. For example, see principles of instructional design [Gagne, Briggs & Wager 1992]. The trainer (author) defines the initial training plan comprising a set of the training units. The suitability of training plans for a given training objective and initial configuration of the training plan is gauged by a human trainer.

The trainer agent is expected to dynamically pursue training goals and re-configure the skeleton training plan depending on trainee needs, current profile (track record) and learning style. The specification of possible training plans provide a focused realisation effort for combining functionality and tools. The trainer agent takes these plans and determines the best ‘fit’ for a given goal.

Training Units

The elements in the training plan are calledtraining units. Thesetraining units specify the interaction between the trainee and the training system for a particular training goal. The training units can be pre-defined (off-line) in a training unit database or they can be generated (on-line) by the training system. Training units are domain specific instances of units of training.

For example; presentation, demonstration, practice lessons and assessment questions. The classes of training units needed are defined off-line by the trainer (author). Training unit classes identify the different flavours in units of training interaction.

General desirable characteristics for training units include the following: (1) training units need to be designed for specific generic tasks (which are relevant for a given training objective), (2) training units need to have measurable transfer of learning methods; and (3) training units need to be achievable for a given trainee within a reasonable time-frame.

Training units specify what (subject expertise) is to be taught. The training unit scripts, which specify how the training units should be delivered, are to be defined with the realisation of the trainer agent. Training units should be practical, written clearly and concisely. Training units are self contained fragments of training interaction. Training units should contain a clear statement of “What is success?” and “What is failure?”.Current attainment level and target attainment level for each task is maintained by the

trainer agent to measure trainee expertise proficiency. The expertise in each training unit gives an indication for the type and amount of transfer of learning possible. To measure trainee achievements, a trainer agent in a MOBAT application can use the general problem solving method presented in Section 7.8 (see also recognise-success and recognise-failure knowledge roles in Section 2.5.2).

Training units should be achievable within reasonable limits. If a trainee’s capability exceeds the training unit then the trainee will be bored. A trainee should be stretched to learn. A training unit should not be too much for a trainee to cope with at one time. There is a natural pull for development (motivation) when the training unit exceeds – but not overwhelms – a trainee’s capabilities.

Good trainers consider a trainee’s profile and reason to present training material in entirely different ways to different audiences. The trainee profile determines both what content is needed for a training unit and how a training unit can be achieved. Training units should be relevant to business needs. In manufacturing, many companies have long been using Just-In-Time logistics. This concept can be used for Just-In-Time training, as opposed to just-in-case training (this is education) where everything is taught on the off-chance it is needed.

If all subject specific tasks in a training domain are implemented with different types of training units (as well as alternatives) then there could be a very high number of training units. Training units should be defined only if there is a clear business need and training units should be defined for completion within realistic time-frames.

Important design work for training units in the MOBAT framework is in relating task features and training needs to the set of properties that make up each training unit. A summary of the specific training unit properties identified in the MOBAT framework is shown in Table 7-1. Training unit properties 1 to 10 are introduced in Section 4.8. Properties 11 and 12 are introduced in Section 2.6.1, Section 5.5 and Section 5.6.4.

Examples of training units properties have been presented in Proto-MOBAT, Scheduling-MOBAT and Workmanship-MOBAT. The set of training unit properties provide a principled basis for the creation of model-based adaptive training solutions. Placing these properties in the MOBAT framework provides great flexibility for a wide range of industrial training problems.

Table 7-1 MOBAT Training Unit Properties

Table 7-1 MOBAT Training Unit Properties

Flower Show
© | | Sitemap