The PATH-AI project1 was designed before the coronavirus pandemic arose. It aimed to understand seminal ethical dimensions (privacy, agency and trust: PAT) in Japanese and British developments in artificial intelligence (AI) in terms of cultural, social and governmental variables, and to explore the cross-cultural basis for international regulation. When COVID-19 and the societal and policy response to it swept the boards, PATH-AI decided to turn to them as a relevant context or ‘use case’ within which ethical values, governance and regulatory processes, and advanced technologies could be studied.
The response of policy-makers and society generally to COVID-19 is a phenomenon of immense complexity both within individual countries and globally. It brings together political, administrative, scientific, public health, health and social care, civil-society, economic, manufacturing, financial, logistic, communicative and other efforts. Beneath a common endeavour to defeat or contain the pandemic, tensions and conflicts between values, between different interests, and between policy strategies are exposed. Frustrations and recriminations have surfaced: for example, disagreements between central government and large metropolitan areas, between health professionals and government policy-makers, and even among Cabinet Ministers themselves. At various times, unsuccessful strategies have fuelled the ‘blame game’. Co-ordination of the large array of decisions and implementations is often precarious in this context of pressures for rapid decision and action amid great uncertainty, both of policy goals and the means for achieving them, as well as of the nature, likely path, severity, and consequences of the virus itself. The psychological and behavioural dimension of public life at individual and collective levels challenges policies and practices. This dimension reveals a host of anxieties, fears, and compliant or resistant attitudes that must be reckoned with if the overall response is to succeed. Defining what ‘success’ means is also an issue.
Photo taken by Andy Holmes.
These developments pose challenges for research into the response to COVID-19. What data to collect for understanding the welter of activities, participants and attitudes that constitute the response? How to organise, categorise, and sift the enormous outpouring of writing, speaking, and decisions when everything is potentially ‘data’? On which end of the global-to-individual spectrum to focus, knowing that all are important and that analytical resources – and the time to use them – are short? How to simplify and present the data to tell a coherent story of the response without leaving essential information out of account?
Constructing a timeline of significant events and developments in the British and Japanese responses was a logical first stage for PATH-AI in amassing data and comprehending the trajectory of activity. Recording who did what, when, then led us to adopt, as an early and provisional step, the analytical lens of multi-level governance (MLG)2 as a simplified framework within which to categorise different relevant actors and activities as the responses to COVID-19 developed over time. Frameworks for studying MLG have been used in diverse areas of subject-matter: e.g., understanding the European Union, environmental and climate policy, global finance, regionalism and federalism, industrial innovation, and the regulation of information technology and systems. Applying an MLG framework to any particular subject – in this case, the response to a pandemic – requires selectivity and tailoring to highlight the specific ways in which diverse phenomena or policy-making processes take place at and across different levels of activity and decision-making, and to assess their interrelationships and their contribution to the response.
Theoretical literature on MLG identifies two overarching types of structural configuration. ‘Type I’ features general-purpose jurisdictions with non-intersecting memberships, a small number of jurisdictional levels, and a system-wide, durable architecture; think Russian dolls. In our case, this would correspond to the formal, general governmental levels in Britain or Japan; for instance, the traditional, hierarchically arranged and ‘constitutional’ central and local government bodies, complicated by interstitial or overlapping layers for nations, regions, municipalities and the like.
In contrast, ‘Type II’ is less tidy and often improvised, It involves task-specific jurisdictions, intersecting memberships, many jurisdictional levels, and flexible design; think the folding and twisting of metamorphosed sedimentary layers of rock. The institutions of public health and of specific agencies for contact-tracing or immunisation within the overall pandemic response, and the actions that take place in and across them, may better correspond to Type II. This is because the response to COVID-19 has seen the marshalling and realignment of existing specialised roles and loci of decision and policy outside central Ministries and local councils: for example, structures within the National Health Service (NHS), in England its counterparts in Scotland, Wales and Northern Ireland, and the institutions and roles in the domain of public health. Add to this the creation of new ones specific to the pandemic – for example, the NHS Test and Trace Programme, custom-made partnerships among different units sharing the NHS ‘brand’, and the Scientific Advisory Group for Emergencies (SAGE). The binary framework of MLG types can only sum up a more complex depiction of intertwining and level-spanning that show that Types I and II are ‘good at different things and coexist because they are complementary. The result is a fluctuating number of relatively self-contained, functionally differentiated Type II jurisdictions alongside a more stable population of general-purpose, nested Type I jurisdictions’.3
Photo taken by John Cameron.
The distinction between government and governance4 is reflected in these typological variations. If government refers to the conventional ‘textbook’ or constitutional institutions and their functions, processes of governance go beyond the formal structures of government to embrace a wider patterning of activities, networks and interactions that take place in and around the state’s bodies, the society, and even the international arena, that may be found in specific policy areas or domains, such as public health (and within that, pandemic response). This distinction emphasises that policy-making, or governing,5 is not fully understood by reference to the simplified picture of a constitutional or legislated arrangement of hierarchical or territorial governmental institutions alone. The timeline, actors and activities of the COVID-19 response show that action flows in myriad channels and involve novel combinations and interdependencies of roles and actors, a flow that goes more widely and deeply into other institutional and social constructs and informal arrangements besides formal governmental ones. Trust and trustworthiness are implicated in these innovated patterns. Crucially important in determining success or failure are individual and group behaviour and relationships at the level of the general public, whose agency, trust and attitudes towards privacy are an essential part of the response.
PATH-AI is interested in how the multi-level structures and behaviour affect, and are affected by, PAT in the COVID-19 response in both the UK and Japan. For example, how the quality and quantity of contact-tracing in the UK is shaped by the participation of many institutional and individual actors: central government and public-health participants, specially appointed agencies, local government health departments, private-sector contact-tracers, and members of the general public who are traced and who are asked to isolate themselves. Their relationships, including co-ordination, as well as their compliance (shaped, in part, by the public’s concerns over data privacy), arbitrate the success of the policy. The programme of vaccination depends upon scientists, pharmaceutical firms, transporters, varieties of health-service practitioner at central and local levels, and – again – individuals who present themselves (or not, partly owing to distrust in government or science) for vaccination. More generally, the part played by testing and contact-tracing in the overall response requires study at and across several of the Type I and II levels to construct a picture that moves through temporal phases, one that illustrates the way PAT shaped activities and perceptions.
In the landscape of governance, structures and relationships within the functionally differentiated public health sector, at several levels, do not align conveniently with those of government, and tap into a host of other sectors – public, private and voluntary – that follow different administrative logics. Academia as well as the institutional apparatus of biomedical, social and statistical sciences are also crucial players in the response to the pandemic, and also provide a major source of advice and guidance to serve the ‘official’ decision-makers. The relationships among all these participants, and many more, are only partly mandated and designed; many of them are un-coordinated or self-organised. Understanding the response to COVID-19 requires some purchase on how these configurations happen and change over time.
MLG’s typological categories therefore provide a convenient point of entry for comprehending the pandemic response, but further analysis will be necessary to open up the response to more finely grained insights into the processes involved. However, the perception that policy action and actors are found at several levels, or in a variety of arenas, does not mean that MLG models are the best way of approaching our research investigation. The notion of ‘levels’ as discrete and bounded layers may not correspond very well to many of the realities of the response; in any case, the meaning of ‘level’ is not always clear in the MLG literature. Groups, networks and roles might be more accurate concepts to employ for analysing the heterogeneous, temporally and spatially dispersed and open-ended activity and relationships that comprises the COVID-19 response.
In any case, MLG – although not uniquely – at least sensitises us to the need to concentrate on relationships of institutions, participants, roles, tasks and jurisdictions, within each of these types and across them, and turns us away from concentrating only on any one of these. As well as casting light on the pandemic response as a subject of PATH-AI research, further phases of this work may provide a case study for assessing the pros and cons of the MLG approach itself as an instrument for gaining insights into the response.
[1] https://www.turing.ac.uk/blog/introducing-path-ai-project-privacy-agency-and-trust-human-ai-ecosystems
[2] For a description, see https://en.wikipedia.org/wiki/Multi-level_governance; accessed 31/01/21.
[3] For a discussion of these types, see Hooghe, L. and Marks, G. (2003), ‘Unraveling the Central State, but How? Types of Multi-level Governance’, American Political Science Review, 97, 2: pp. 233-43; quotation is from p, 240.
[4] An exploration of the variety of ‘governance’ approaches, contexts and concepts can be found in https://en.wikipedia.org/wiki/Governance; accessed 07/02/21.
[5] ‘Governing’ can be seen as ‘the totality of interactions in which public and private actors participate, aimed at solving societal problems or creating societal opportunities; attending t the institutions as contexts for these governing interactions; and establishing a normative foundation for all those activities’ (Kooiman, J., 2003, Governing as Governance, London: SAGE Publications: p.4.)
