A Framework for Model-Based Adaptive Training

Trainee Characterisation

By focusing on either plant operation skills for operators or on circuit board fault finding for technicians, there is some indication of the target audience. However, a more detailed trainee characterisation is needed as individual trainee’s may have different preferred ways of learning or may be able to learn more effective at varying self-supporting levels (from trainer guidance) which requires different training strategies. The Proto-MOBAT mapping methods for learning modes and self-supporting levels are presented in this section.

As in the MOBIT and ITSIE projects [ITSIE 1992, D7], a learning mode is based upon the relation between the trainee’s preferred type of model (i.e., available knowledge) and the desired level of performance. With the types of models and three levels of goal expertise there can be 3 learning modes:

  1. rote learning mode through memorisation and recall of a given level of behaviour, where the knowledge representation model used for training is at the same level as the desired level of performance;
  2. inductive learning mode through a series of examples where the knowledge representation model used is ‘below’ the desired level of performance; and,
  3. deductive learning mode through the presentation of knowledge where the knowledge representation model used is ‘above’ the desired level of performance.

The actual learning mode adopted during a training session may be restricted by the available knowledge representations. If a learning mode based on what has been most effective in past training sessions can be determined, then realisation effort for different model types can be limited. However, if fewer model types are implemented then fewer learning modes can be supported and training choices based on trainee preferences are reduced.

Note: The term learning style was used in ITSIE and MOBIT projects for rote, deductive and inductive instruction modes (see Section 3.3.2). The term learning style is used in this research to indicate trainee preferences with readily understood descriptions such as learning by doing, learning from examples etc. (see Section 7.6.2). For rote, deductive and inductive types of learning, the term learning mode is used here to indicate the mode of instruction.

Extended Trainee Characterisation

Table 4-1 Extended Trainee Characterisation

Within the ITSIE project, the preferred trainee learning styles are mapped to three training strategies as follows:

  1. rote learning Þcoaching strategy;
  2. deductive learning Þtutoring strategy; and,
  3. inductive learning Þdiscovery strategy [ITSIE 1992, D7, MOBIT 1994, W1].

This is a direct relationship between a learning mode and training strategy. There may be many other trainee characteristics that could influence the most effective instruction strategy by a trainer. For Proto-MOBAT, consideration is given to further trainee characteristics based on trainee ability, experience, confidence and motivation. A level for each of these characteristics from high to low can determine a trainee’s “self-supporting level”. For example, 3 general experience levels for a particular training objective may be specified in the range from beginner (low) to competent (medium) and advanced (high). For trainee ability, experience, confidence and motivation any number of levels that can reasonably be defined within the specific training application can be used. It is beyond the scope of the Proto-MOBAT research to provide a detailed analysis into levels of trainee ability, experience, confidence and motivation. A trainee characterisation is presented in Section 7.6. For Proto-MOBAT, a trainee self-supporting level is formulated as a function of trainee characterisation levels and mapped to an instructional strategy (which is implemented by the training agent). A tutoring strategy is quite prescriptive and more commonly found in conventional CBT application. The discovery strategy does not typically provide instruction and can be more appropriate for on-line decision support and browsing application instead of a focused training application. A most effective strategy with high learning by trainees for the Proto-MOBAT application is the coaching strategy. This coaching strategy is refined in Proto-MOBAT with an additional facilitating strategy in order to adapt closely to trainee preferences with fine tuning of didactic and diagnostic methods (i.e., to deliver variations in structuring and explanation of training material).

There is no direct mapping between trainee characteristics and training strategies in Proto-MOBAT. Instead, the training strategies are implemented and adjusted as needed at run time by using trainer tasks and trainer methods. A principled approach for selecting training strategy using trainer task primitives is presented in Section 7.8.1. It should be noted that learning mode and strategy are attributes of training units, therefore at run time different modes and strategies within a training session can be used by the trainer agent. The differences discussed in this section for the ITSIE project and Proto-MOBAT experiments are shown in Table 4-1 Extended Trainee Characterisation.

© | | Sitemap