Form and function : assessing the impact of mental representation on behaviour using computational models

This thesis presents three studies examining the methods used by human learners to construct mental representations to reflect external data patterns, and the impact the form of these representations have on subsequent behaviour. This involves three varied tasks in which representations are built and updated from experience: stereotype change, numerical estimation and learning consolidation. Each of these studies uses computational models of these processes to offer potential descriptions of the mechanisms used to construct our representations, and assesses the accuracy of these descriptions using both qualitative and quantitative comparisons with human behaviour. Such contrasts reveal the importance of the form of our mental representations on related actions: stereotypical beliefs are coloured by the organisation of group members, numerical expectations are dependent on the assumed format of numerical information, and stimulus choices are influenced by connections forged through experience. This then provides insight into the mechanisms used by human learners in these tasks, and the specific impacts of such mechanisms on related behaviour. We do however also note questions raised by the use of such methods on the accuracy of what may be highly-complex systems in describing human behaviour, and the algorithms that may be used to implement such systems in real life.