Training for optimising internal task transfer in the acquisition of process control skills
The aim of this thesis is to investigate the acquisition of different elements of process control skills and how the transfer between task elements contributes to the acquisition of the overall task. Process control tasks are very complex cognitive tasks consisting of a number of subordinate task elements such as procedure-following, diagnosis, & monitoring, whose execution must be planned carefully in order to meet the system goal. In the past, research emphasis has focused upon training these subordinate tasks separately, ignoring the possibility that performance at one element would benefit from or interfere with mastery of another. Understanding these possible 'internal transfer' phenomena will influence training design. It would also influence issues of work design, including allocation of functions, since tasks designed to enable practice of the constituent elements to support each other would counterbalance problems associated with infrequent use of skills in automated plants. This thesis has focused upon the development of training methods to optimise transfer of knowledge and skills, assisting trainees to integrate different task elements within the overall process control task they need to master. The transfer of training literature was reviewed in order to identify major variables influencing transfer. To provide a framework for utilising previous empirical findings in examining transfer of complex process control skills, a model of Hierarchical Task Analysis was developed which described the task in terms of a limited number of operations and plans. A major hypothesis put forward in the thesis is that 'task elements with similar forms of plans and operations may prompt an individual to adopt similar cognitive processes and transfer will be observed'. The size of transfer, however, would be determined by the learning conditions under which the original task elements were acquired. To examine the influence of learning conditions upon transfer of task elements, four training methods were developed based upon a theoretical model of transfer which was integrated with the hierarchical task analysis. A large scale experiment was conducted in order to investigate the effects of the four training methods upon learning two similar tasks, in the context of starting-up a distillation column. This task was simulated in a microcomputer. The two tasks were designed to share common task elements but were different in terms of the required product specifications. Twenty-eight postgraduate students took part in a training course which lasted for eight hours approximately. The subjects were assigned to the following four experimental conditions: (i) the procedures-group which was provided with efficient procedures; (ii) the analysis-group which received additional explanations about the interaction of goals described in the procedures; (iii) the model-group which was provided with a structural model of the plant, but with no procedures; and (iv) the practice-group which learned the tasks by practice and which was used as a control condition. The model of learning was used to make transfer predictions and generate five experimental hypotheses which were tested in the main study; one hypothesis was concerned with the acquisition of the original task, two of them with the transfer of task elements and the other two with nonspecific transfer effects. For a number of performance measures such as speed, control performance and economy of operation, the procedures- and analysis- groups performed the original task better than the other groups. The model-group was faster than the practice-group; however, their control and economy aspects of performance were equivalent. An 'in-depth' analysis of the control actions and verbal protocols showed that the model-group continued to experiment with' the process, putting into practice the theory of plant which was taught to them. Performance at the transfer task indicated that all groups performed better than the practice-group, which provides support for the hypothesis that 'task elements similar in form may prompt an individual to adopt similar cognitive processes and transfer may occur'. However, the procedures-group degraded its performance and appeared to be inferior to both the analysis- and modelgroups, which may indicate that some kind of extrinsic information in the form of planning or conceptual knowledge would be required to optimise transfer. On the other hand, the analysis-group was faster than the model-group, but no significant differences were reported with respect to other measures of performance. On practice with a successive transfer task of the same kind, the observed patterns of performance changed. Only the analysis-group was significantly better than the practice-group, the other groups having scored in between these two groups. This finding has highlighted the role of practice in an interactive learning environment provided by the plant simulator. Although the experimental design attempted to control for nonspecific transfer effects by maximising the number of 'common' task elements between the original and transfer tasks, the individual differences observed and the changes in the kind of plans developed by trainees have indicated that such transfer effects have actually taken place. This was expected to occur to a certain extent, and two hypotheses had been formulated in order to examine this issue by looking into the types of planning and conceptual knowledge which supported transfer. As it was expected, the analysis- and model- groups achieved higher nonspecific transfer scores than the other groups which were measured in terms of the amount of disturbance caused to previously established parameters and the number of questions answered in a questionnaire administered in the end of the study. An interesting result, however, was that the practice-group appeared to be better than the procedures-group in this respect; this may be accounted for by the fact that the former group was more actively engaged in learning the original task. Finally, the thesis has investigated the transfer of the three subordinate goals common to both the original and transfer tasks. An important factor which has influenced the different degrees at which these goals were transferred was the degree of flexibility entailed in their performance. The thesis concludes with an evaluation of its own approach and suggestions for future research.