Professors Bev Littlewood and Martin Newby to improve software for COVID-19 models
The work of the City, University of London academics will contribute to the increased confidence in the accuracy of decision making based going into COVID-19 software modelling.
City Emeritus Professors Bev Littlewood (Computer Science) and Martin Newby (Engineering) have been awarded funding from the Royal Academy of Engineering (RAEng) as part of a team of senior scientists who will be improving the software engineering quality of epidemiology models.
Led by Professor Harold Thimbleby of Swansea University, the team also includes Professor Neil Ferguson of Imperial College, whose report prompted the COVID-19 lockdown.
Epidemic modelling has informed global public health policy in managing the COVID-19 pandemic, by assessing strategies such as lockdown, mask-wearing, track-and-tracing, and the use of apps.
Throughout the crisis, however, there has been concern in the computer science community and elsewhere, about the quality of the software used in some models. A major concern, for example, has been the frequent lack of access to the code, and its documentation, which has hindered independent expert assessment.
This project proposes that these problems could be addressed via the institution of Software Engineering Boards (SEBs), which can be thought of as analogous to the Ethics Boards that are ubiquitous in most large research institutions. The project will consult with major players in the modelling and software engineering communities to win support for these ideas, and to scope the structure, membership and ways of working of SEBs.
Though the project has been motivated by COVID-19 modelling – and it is this area upon which the project will concentrate – the issues involved apply more generally to other scientific software. It is expected that a successful outcome of the eventual application of SEBs will include:
- Openness of software implementations of mathematical models to expert review;
Traceability and accountability for the model-based evidence and arguments supporting decisions;
Increased confidence in the accuracy of decision-making based on mathematical models;
Improved access to models by third parties via better documentation and validation of models and their software implementations;
Improved policy-making security by ensuring software is transferable and not dependent upon special knowledge of a single individual or group.