How can data science contribute to developing evidence-based policy?

Reported by Dr Rebecca Van Hove CSaP Policy Fellowships Coordinator

CSaP’s Continuing Policy Fellows came together last month to discuss broad challenges in developing evidence informed policy, as well as the potential use of advances in data science and digital technologies for public policy.

CSaP Continuing Fellows are Policy Fellows who have completed their original two-year Fellowships and continue to engage in one-to-one meetings with researchers and group discussions. There are currently twenty Continuing Fellows.

The meeting started with an exploration of climate change science and the opportunities that big data can bring to climate policy. The discussion then moved to debate how government could engage more effectively with digital technologies and big data in general.

Mike Hulme, Professor of Human Geography at Cambridge, kicked off the debate about the nature of evidence and its uses, by illustrating how the relationship between science, evidence and policy-making has changed in the field of climate science throughout his career.

He demonstrated that climate data has been brought more firmly into policy decision-making in recent years, but argued that there now remains a need to foreground disagreement in the discussion of climate change. Mike discussed how differences in values and their worth, the understanding of risk as a socially constructed phenomenon, and the place of religion in public life, for example, help explain why people disagree about issues such as climate change.

Mike advocated for the importance of recognising such disagreement and taking it seriously, in order to better understand how one might harness the power of science and how it might affect public decision-making. He emphasised the need to recognise human values, convictions and cultures as deep-seated drivers of action, which must be brought into conversation with climate evidence to make effective policy-making possible.

Other participants added to this that the realities of politics similarly must not be underplayed. Building cross-party consensus on issues, as has happened in UK politics on the need for climate action – at least to a degree – allows for a stronger and more enduring reliance on scientific evidence.

Dr Emily Shuckburgh, Deputy Head of Polar Oceans Team at the British Antarctic Survey, gave a presentation on the ways in which attitudes to climate change had recently begun to transform in business and industry. Increasing requirements to report risks and liability in terms of climate change potentialities has led to a growing desire from businesses and investors to use scientific evidence to make decisions. This in turn means there is increasingly a need for evidence on a more granular level, which can provide information on the impact of climate change on either localised places or specific areas, such as for example human health.

This growing need is being matched by growing capabilities, as techniques for capturing data and building predictive scenarios develop. Emily also argued that attention should be paid to the ethical dimensions of these developments in data science, as more precise predictive modelling raises questions of who will share in this knowledge and the effects it will have on people’s lives.

How then can government best exploit data science: how can it be used to address existing problems, and which capacities can be used for specific problems?

The participants in the roundtable identified a variety of issues which are important to address in this regard, such as the need to find ways to share information about innovative uses of data science in policy, including at a local government level. Secondly, it was identified that there is a need to avoid the trap of focusing too much on tools and techniques, thereby losing sight of the actual problems themselves for which solutions are attempted. Thirdly, there are the challenges of data quality: how do we make sure we make full use of data, without giving rise to adverse effects in privacy and ethics?

Dr Eloise Taysom, who recently completed a PhD in Engineering at Cambridge and is now working for the Government Digital Service, presented recent findings from her work in the use of AI, machine learning and big data in public policy. She identified the difficulty of sharing data as an obstacle to the government’s utilisation of data science: while there is a large appetite for this, it is hard to accomplish from a security point of view, as the issues of privacy, cost and departmental boundaries currently limit what can be done.

Haydn Belfield, Academic Project Manager at the Centre for the Study of Existential Risk (CSER), brought attention to the potentialities of malicious use of AI. Recent research at CSER has identified three types of risks linked to AI systems which will have to be faced in the next five years: cybersecurity risks, risks to fiscal security, and political security risks.

Cybersecurity needs to prepare for the potentiality of malicious attacks becoming more effective and growing in scale, as techniques of attacks become increasingly automated. In fiscal security, the growing use of automated systems requires better protection against attacks on these automated systems.

Finally, in the political sphere, malicious use of AI – such as the development of fake content – could lead to information manipulation and put political stability in open democracies at risk. Haydn argued that awareness of the potential risks can help focus attention on finding solutions for prevention.

Will data science also be able to contribute to social science policy areas such as national well-being?

Discussion continued over dinner, where Dr Anna Alexandrova, Department of History and Philosophy of Science, also joined in the discussion. Anna, whose research centres on the philosophy of science and philosophy of social science, discussed recent trends in the use of data in the measurement of well-being. Anna questioned whether any one set of data – whether an economic one, such as GDP, or attempts at measuring happiness – can ever be usefully employed to measure well-being.

It was clear from the discussion that data science – and big data in particular – presents exciting and nerve-wracking opportunities for public policy. The complexities which advances in data technology bring with it – the ethical implications, potential effect on privacy, and risks of malicious misuse – make it clear that any implementations of data science need careful and coordinated consideration. Taking a step back to do so, and providing opportunities to share information, will be crucial to allow policy makers to explore how these new technologies can best be implemented.

Banner image from Sean David via creative commons 4.0