Supporting policymaking with statistical emulation of complex models

Project ID: 2531ad1575

(You will need this ID for your application)

UCL Lead department: Science, Technology, Engineering and Public Policy (STEAPP)

Project Summary:

Many models of complex systems, like transport systems, are extremely large and complex themselves and take a long time to run. This makes it difficult for time-constrained policymakers to interact with the model and understand its insights. To reduce this problem, we can use an emulator: a surrogate model which can replicate the results of (emulate) the complex model on which it is trained, and can run orders of magnitude more quickly. However, emulators also introduce an additional layer of uncertainty.

The student will initially learn about emulation methods, then construct an emulator for a model of the UK transport system produced by the Department for Transport, developing skills in statistical analysis, statistical computing and visualisation. At the same time, they would take part in the STEaPP doctoral training programme in methods for public policy, to support their policy engagement. The major innovations of this project, developed by the student with primary supervisor Dr Erica Thompson, will be A) mathematical, looking at the scope and limitations of emulator methods for policy support, considering issues such as discontinuities or multi-valuedness in the output surface, and the uncertainty introduced to the emulator by collapsing a complex system into a restricted choice of input and output variables. The student will benefit from training and interactions in the UCL Department of Statistical Science, where a second supervisor will be located. B) interdisciplinary, drawing on insights from social science and public policy in the development of appropriate codesign and visualisation tools. The student will spend time with policy and technical analysis colleagues in DfT (subject to confirmation) and will benefit from STEaPP’s public policy training and research support including the UCL Policy Impact Unit

The student must have strong mathematical and programming skills, and an interest in science advice or public policy in the UK.