Behavioural cloning is a simple and effective technique for automatically and non-intrusively producing comprehensible and implementable models of human control skill. Behavioural cloning applies machine learning techniques to behavioural trace data, in a transparent manner, and has been very successful in a wide range of domains. The limitations of early behavioural cloning work are: that the clones lack goal-structure, are not robust to variation, are sensitive to the nature of the training data and often produce complicated models of the control skill. Recent behavioural cloning work has sought to address these limitations by adopting goal-structured task decompositions and combining control engineering representations with more sophisticated machine learning algorithms. These approaches have had some success but by compromising either transparency or robustness. This thesis addresses these limitations by investigating: new behavioural cloning representations, control structures, data processing techniques, machine learning algorithms, and performance estimation and testing techniques. First a novel hierarchical decomposition of control is developed, where goal settings and the control skill to achieve them are learnt. This decomposition allows feedback control mechanisms to be combined with modular goal-achievement. Data processing limitations are addressed by developing data-driven, correlative and sampling techniques, that also inform the development of the learning algorithm. The behavioural cloning process is developed by performing experiments on simulated aircraft piloting tasks, and then the generality of the process is tested by performing experiments on simulated gantry-crane control tasks. The performance of the behavioural cloning process was compared to existing techniques, and demonstrated a marked improvement over these methods. The system is capable of handling novel goal-settings and task structure, under high noise conditions. The ability to produce successful controllers was greatly improved by using the developed control representation, data processing and learning techniques. The models produced are compact but tend to abstract the originating control behaviour. In conclusion, the control representation and cloning process address current limitations of behavioural cloning, and produce reliable, reusable and readable clones.

Behavioural cloning is a simple and effective technique for automatically and non-intrusively producing comprehensible and implementable models of human control skill. Behavioural cloning applies machine learning techniques to behavioural trace data, in a transparent manner, and has been very successful in a wide range of domains. The limitations of early behavioural cloning work are: that the clones lack goal-structure, are not robust to variation, are sensitive to the nature of the training data and often produce complicated models of the control skill. Recent behavioural cloning work has sought to address these limitations by adopting goal-structured task decompositions and combining control engineering representations with more sophisticated machine learning algorithms. These approaches have had some success but by compromising either transparency or robustness. This thesis addresses these limitations by investigating: new behavioural cloning representations, control structures, data processing techniques, machine learning algorithms, and performance estimation and testing techniques. First a novel hierarchical decomposition of control is developed, where goal settings and the control skill to achieve them are learnt. This decomposition allows feedback control mechanisms to be combined with modular goal-achievement. Data processing limitations are addressed by developing data-driven, correlative and sampling techniques, that also inform the development of the learning algorithm. The behavioural cloning process is developed by performing experiments on simulated aircraft piloting tasks, and then the generality of the process is tested by performing experiments on simulated gantry-crane control tasks. The performance of the behavioural cloning process was compared to existing techniques, and demonstrated a marked improvement over these methods. The system is capable of handling novel goal-settings and task structure, under high noise conditions. The ability to produce successful controllers was greatly improved by using the developed control representation, data processing and learning techniques. The models produced are compact but tend to abstract the originating control behaviour. In conclusion, the control representation and cloning process address current limitations of behavioural cloning, and produce reliable, reusable and readable clones.

The educational benefits of involving learners as active players in the learner modelling process have been an important motivation for research on this form of learner modelling, henceforth referred to as participative learner modelling. Such benefits, conceived as the promotion of learners' reflection on and awareness of their own knowledge, have in most cases been asserted on the grounds of system design and supported only by anecdotal evidence. This dissertation explores the issue of whether participative learner modelling actually promotes learners' reflection and awareness. It does so by firstly interpreting 'reflection' and 'awareness' in light of "classical" theories of human cognitive architecture, skill acquisition and meta-cognition, in order to infer changes in learner abilities (and therefore behaviour) amenable to empirical corroboration. The occurrence of such changes is then tested for an implementation of a paradigmatic form of participative learner modelling: allowing learners to inspect and modify their learner models. The domain of application centres on the sensorimotor skill of controlling a pole on a cart and represents a novel type of domain for participative learner modelling. Special attention is paid to evaluating the method developed for constructing learner models and the form of presenting them to learners: the former is based on a method known as behavioural cloning for acquiring expert knowledge by means of machine learning; the latter deals with the modularity of the learner models and the modality and interactivity of their presentation. The outcome of this research suggests that participative learner modelling may increase the abilities of learners to report accurately their problem-solving knowledge and to carry out novel tasks in the same domain—the sort of behavioural changes expected from increased learners' awareness and reflection. More importantly perhaps, the research suggests a viable methodology for examining the educational benefits of participative learner modelling. It also exemplifies the difficulties that such endeavours will face.