<p>The thesis examines how policy processes occur in practice over time, and advances a theory for understanding and explaining them using a complexity analytical lens. This lens provides a different way of 'seeing' policy processes holistically, in comparison with dominant alternative actor-focused, institutions-focused or idea-focused perspectives. The overall aims of the thesis are achieved through four phases. First, to construct the complexity lens, the thesis analyses the potential contribution of complexity theory, particularly as it relates to social systems and organisations. Second, existing theories of policy processes are scrutinised to identify areas where a complexity lens could provide new perspectives for their understanding, through a focus on: - 'wholes' of policy processes - policy problems and solutions - multiple participants - interactions within policy processes - dynamics within policy processes. Third, a study of New Zealand's tertiary education policy processes provides new empirical data. Data collected in unstructured interviews is represented in three differently-themed narratives, corresponding to three well-theorised analytical emphases: (1) participants; (2) institutions; and (3) ideas. The selection of three different perspectives takes into account the socially complex nature of policy processes. Fourth, the narrative data are examined through the complexity analytical lens and the results are compared with the views of policy processes obtained using the single lenses in the themed narratives. The four phases come together by demonstrating that viewing tertiary education policy processes through the complexity analytical lens provides a new perspective on policy processes which has implications for designing and intervening in policy processes. From this new perspective, policy processes are understood as complex social systems in interaction with other complex social systems. These systems consist of large numbers of interdependent and self-referencing participants, interacting with each other in ways that are nonlinear, influenced by prior experiences, and unpredictable in any precise sense.</p>
Dynamic calibration of agent-based models using data assimilation