A Framework for Model-Based Adaptive Training

Task Analysis, Task Decomposition & Task Specification

Training objectives are the instructional goals for the trainer which for effective instruction need to be broken down into smaller units. These smaller units enable the trainer agent to find all unsatisfied sub-goals for a selected training goal and organise these units into training plans for training delivery. For Proto-MOBAT, three steps in structuring training objectives into smaller units are considered. First, a task analysis is performed for breaking a training objective into a set of smaller units which forms a profile of the tasks a trainee needs to master. This first step does not assume a computer-based training delivery. Second, the training tasks are further decomposed (if necessary) for generic task decomposition based on the approach by [Leitch & Gallanti 1992]. The trainer agent is domain independent, therefore this step is required to enable the trainer to recognise and act on the generic aspects of training tasks within the training goal hierarchy. And finally the tasks are specified (together with aspects of expertise classification and trainee characterisation) as units of training which can be used by the trainer agent to construct and deliver training plans. Figure 4-1 shows where task analysis, task decomposition and task specification fit in the MOBAT framework.

There is more than one way to decompose a training objective into smaller units. Two methods have been examined in Proto-MOBAT. One approach is the definition of a detailed task list for each training objective. Another approach is to identify the main topic area in the training objective and to create a list of sub-topics or list of concepts which make up the training objective topic. The latter approach is more common in typical training analysis when a training subject is broken down to show the contents of a training course. A domain specific task can be defined as an action attached to a concept. This means concept analysis is more general since it allows various actions to be related to the same concept. A concept defined as ( C ( trip logic input levels) can be used to define tasks such as ( name 8 ( trip logic input levels ) ) or ( verify ( trip logic input levels ) ). A partial analysis of a concept hierarchy for Proto-MOBAT is shown in Figure 4-3. Two analysis methods (i.e., task Þ subtask and topic Þ subtopic) have been examined in Proto-MOBAT. Once the tasks a trainee is expected to learn (to perform) are identified, both analysis methods ultimately lead to identification of the same generic tasks. The decomposition of training tasks into generic tasks is discussed below.

Partial Concept Hierarchy

Figure 4-3 Partial Concept Hierarchy

The Proto-MOBAT tasks have been linked to generic tasks which provides the training system with a limited set of types of training tasks. An example is shown in Figure 4-4. A domain specific training task does not always relate to a single generic task. Some less precise tasks are also defined. For example, training tasks involving “understanding” are often ambiguously stated, and can be more precisely decomposed into a combination of primitive tasks. Typical groupings of tasks are: (a) interpretation, identification and then decision; (b) interpretation, decision and then execution; and (c) interpretation, prediction and then decision. Each of these tasks can also be further decomposed into a combination of the same primitive tasks. This leads to a hierarchical decomposition until a base level is achieved.

Mapping Domain Specific Tasks to Generic Tasks

Figure 4-4 Mapping Domain Specific Tasks to Generic Tasks

The list of domain specific tasks shown in Figure 4-4 are not broken down from a single training objective. These tasks have been chosen as a representative selection for further specification and realisation within the Proto-MOBAT application. The base level task decomposition is used to identify the most appropriate software methods and should therefore be considered to be achievable by using one type of expertise and one training method. Each training unit is determined by one base level training task. Figure 4-2 Power Plant Presentation Training Unit shows an example screen for the training task hierarchy: ( understand ( plant process ) ) Þ ( describe ( plant process ) ) Þ ( interpret ( plant process ) ). The base level in this example is an interpretation task.

© | | Sitemap