PRE-PRINT prior to peer-review doi: 10.31235/osf.io/yr75f Accountability and neglect in UK social care innovation

Accountability structures in social care are critical. They facilitate democratic decision-making, responsibility and the equitable distribution of benefits. This study examines how innovation and technology is implicated in such structures. In the UK, innovation and technology researchers have predominantly imagined care as service provision, with accountability structured through paternalistic and technocratic configurations of people, materials and knowledge. Aligning with incumbent policy and interests, these structures neglect significant groups of actors and issues, with implications for ongoing vulnerabilities in the sector. This study empirically identifies diverse possibilities for how innovation could reconfigure accountability structures in inclusive, participative and less neglectful ways.


Introduction -accountability in social care
Technology's promises pervade policy responses to Covid-19. Technologists, funders and politicians continue to propose and finance innovation as solutions to vulnerabilities in the social care sector exposed by the pandemic (Dunn et al., 2020;Gigler, 2020;Knowledge Transfer Network, 2020). This is nothing new; technological innovation has long been championed as a palliative to vulnerabilities in care sectors, and more broadly across public policy domains (Pfotenhauer et al, 2019). Yet when it comes to initial responses to Covid-19, technology has been implicated in significant policy failures in the UK and elsewhere i . Technology's promoters are not holding up their side of the bargain. In so far as they can be called strategic, government and private sector responses can be characterised as attending to the speed and scale of innovation over its direction (Stirling, O'Donovan and Ayre, 2018). This is exemplified in the rhetorics of testing where success is measured narrowly in terms of scale-up of tests-per-day. Significantly less attention has been given by these actors to decisions of who benefits from such innovation, how benefits are distributed across society, and the underlying political choices of who decides and who is accountable for those decisions. Choices about who is prioritised for example, who is tested, and who is not tested. Structures of accountability within the social care sector are critical sites of concern. In this paper I address these concerns by asking: how has accountability in social care been structured and imagined in technology and innovation research? Accountability directs our attention to the fact or condition of being accountable. Analytically, it is usefully appraised in the adjectival form; to hold someone, or some group, accountable is to trace this relationship and make it operational. The job of the analyst then is to open up for analysis these relationships of responsibility and power. In this paper I do just that. I illustrate how science, technology and innovation are implicated in relationships of accountability, and how they are structured and configured in the social care sector through practices and outputs of innovation. In the field of innovation studies, accountability has traditionally been understood as an issue for those concerned with the consequences of science and technology (Genus and Stirling, 2018). In recent decades, discussions of accountability in research and innovation policy have tended to turn on anticipatory technological decision-making and inclusive deliberation. In response to emerging technologies such as synthetic biology, genetically modified organisms and more recently artificial intelligence and robotics, the focus of research policy tends to be the creation or strengthening of accountability structures that are capable of dealing with such technologies upstream within the research system or downstream in markets. The proliferation of responsible innovation programmes and ethics frameworks are examples of instruments designed to govern emerging technologies (Boden et al., 2017;Macnaghten, 2020). Technology and innovation also configure existing structures of accountability in more mundane ways. Technologies can open up bureaucratic administrations to broader forms of public oversight or even participation, strengthening relations of accountability between publics, public administrators and public representatives. Assessment methodologies that rely on audits and data production are designed to do just this, through star-ratings, league tables and alert systems for example (Cutler and Waine, 2003). But technology is also implicated in systems which allow powerful actors evade accountability. For example, by obfuscating responsibility through the imposition of data or algorithmic decision making, in turn hiding complexity and seemingly de-politicising decision making (Noble, 2018). Such socio-material accountability structures arise from often complex arrangements of people, technology, knowledge, data and infrastructures. Indeed, this latter case, historically excluded from studies of accountability in innovation, adds significant novelty to this study. Science and Technology Studies scholars, sociologists of health and infection and critical design thinkers have usefully conceptualised ways to imagine, build and perform socio-material arrangements in care sectors -what they call infrastructures of care (Langstrup, 2013;Weiner and Will, 2018). These concepts highlight the socio-material arrangements involved in establishing care as a set of relations, networks and obligations embedded in communities and constituent of their values. These concepts have diagnostic value -they can aid the assessment of how care is configured in places outside of acute settings such as hospitals. But they also have critical value. They draw attention to links between different objects and spaces that are integral to practices, beyond a narrow focus, say on emerging technology alone -often the flag pole element in research dedicated to innovation in social care (Prescott and Caleb-Solly, 2017). These concepts show us that care is not an automatic output of systems or infrastructures, but rather is situated within and amongst their constituent parts, and, crucially, is predicated on structures of accountability (Light and Seravalli, 2019). Accountability structures, like other sets of human and technical relations, are not given. They may be designed to narrowly emphasise centralised decision making and control, or conversely distributed widely, emphasising processes of democracy and collaboration in decision making. This is not to make a normative claim over any given structure, rather to note that the social (organisational structures, reporting rules, priorities) and material and technological (measuring tools, methods, statistics) aspects of accountability matter, and either explicitly or implicitly are often subject to reconfiguration following the imposition of new technologies and other kinds of innovation. Yet this renewed intellectual attention on infrastructures of care remains at odds with dominant policy framings in the UK. Here, innovation in social care continues to be framed under a service logic. Social care as a finite resource to be administered to recipients (Berridge et al., 2014;Mostaghel, 2016). In a budgetary era when Keynesian infrastructure spending is back, this framing has implications for the allocation of public spending to care sectors. Investment in national infrastructure for example carries the kind of rhetorical and economic heft that the shoring up of localised care services do not. And so, at stake in this study of how accountability is structured is first issues of resilience and vulnerability within care sectors -questions of who or what is neglected and who decides. And second, the salience of discursive framings of care that lend themselves to broader economic and political commitments to appropriate funding.

Materials and methods
In answering the research question, I analysed evidence from a series of research tasks conducted between 2018 and 2020 in the robotics and social care sector using situational analysis (Clarke, 2009). Evidence was produced through observing researchers and their innovation practices; interviews with researchers, technologists, care home operators and related experts; stakeholder workshops with roboticists; and reviews of academic and policy literature. In order to understand how accountability in social care is imagined at the level of research projects across the UK research councils, a study of research investments was carried out using observations from project abstracts. This built on recent methodological innovation in the study of situations using computational data (Marres, 2020). 127 projects relating to social care were identified in the UKRI's projects database, Gateway to Research ii . Project abstracts are well suited to the task. Typically, each abstract describes a prospective research situation. This description serves as a promise to carry out certain methods and answer certain research questions, and also to perform research in a certain way, matching a particular vision of how the social and technological world is and should be (Jasanoff and Kim, 2015). Project abstracts were systematically mapped to locate and analyse discursive arrangements of innovation, technology, and socio-material structures of accountability. These arrangements were categorised, and categories reduced until coherence could no longer be sustained. Three aspects of these situations are relevant to answering the research question. First, studying how accountability is imagined by researchers. This reveals something about the imprint of power on a situation and how subjects and objects are related within that situation. Second, how social care is described. For example, is care seen as a service, an arrangement of market players, an obligation, an ethical concern or something else entirely. Third, how technology and innovation is imagined within such a situation and what is seen to make a difference. Finally, I introduce two critical case studies cases (Flyvbjerg, 2006). The first, on data and artificial intelligence is strategically chosen to illustrate how technology reinforces incumbent accountability structures. The second, on assistive living robotics, shows participative research practices that attempt to broaden-out accountability.

The emergence of Covid-19 -revealing accountability and neglect in social care
In the initial months of the Covid-19 crisis, Protect the NHS' was the political choice not only of an on-brand slogan, but of a national strategy (Chambers, 2020). The operating logic of this strategy was centralised command and control. Yet a combination of inadequate political planning and systemic infrastructure and service failings undermined efforts to adapt to the crisis in care sectors and especially in elderly care. Guidance from the national government was insufficient and late, arriving only on March 13th (Department of Health and Social Care, 2020d). Personal protective equipment was subject to supply and distribution problems (Hignett, 2020). The government made a series of incremental financial commitments for the sector. But the processes of distributing money through local authorities to where it was needed was slow (Department of Health and Social Care, 2020c). This prevented many homes increasing their most prevalent and adaptable resources, human carers. The UK government decision to give-up on Covid-19 testing outside of hospital settings meant staff and residents did not and could not know who was infected. And testing protocol for care homes was not reestablished until early May (Department of Health and Social Care, 2020e). These issues were exacerbated by the directive from the government to empty hospitals of old people for re-housing in care homes (Department of Health and Social Care, 2020b; National Audit Office, 2020, p. 47). As hospitals were cleared ahead of the expected Covid-19 wave, non-acute elderly patients were transferred 'into the community' whether they had Covid-19 or not. Jane Deith reports on the BBC's File on 4 (Deith, 2020) that despite government claims that "decisions to push out non-tested patients into care homes were made by middle ranking NHS managers, that was not the case. It was deliberate policy." Official guidelines even mention a "Capacity Tracker as a universal mechanism across the country to report bed vacancies and help manage demand during this incident" (Department of Health and Social Care, 2020a). This is remarkable. The primary unit the government was using to monitor and account for care in homes was not based on measures of care, health or even Covid-19 infection, but one of instrumental supply of beds. Tracing structures of accountability in care sectors and how they worked during these early months of the Covid-19 crisis is a difficult task. Rather than accountability, the dominant feature of the period seems to be neglect. Neglect of the interests of residents and users of care services, neglect of the staff and families who provide that care, crucially neglect of local and situated knowledge that might make a difference in individual situations. Moreover, this neglect was structured by inefficient and insufficient data, finance, communications and technological resource. And so, in the following sections I show how these conditions of neglect do not indicate merely a deficiency of innovation in a moment of crisis, but rather are contingent on policy and political decisions, market arrangements and technological interventions over many years.

The direction of innovation and technology development in social care
Remarkable historical improvements in health and well-being -at least for some -owe much to science, technology and innovation (Blume and Geesink, 2000). However, not all consequences of research and innovation are positive. Nor do any benefits unfold automatically -especially if they are to be fairly distributed to those who need them most (Stirling, O'Donovan and Ayre, 2018). Innovation is the practice of developing and implementing new ideas (Freeman and Soete, 1997). It is a form of conversation between needs and possibilities of designers, users and society, often contested and always political (Stilgoe, 2020). And so, understanding how innovation in its social, technological and institutional forms, has imagined and has been directed to address issues such as distribution of benefits, costs of implementation, or configuring of accountability within care systems is a fundamental challenge in answering the research question.

Policy and politics
Let's begin with the institutional context in which decisions about what to innovate are taken. In the UK, care that is administered outside of hospitals takes place 'in the community' -a strangely singular phrase for what is in fact a patchwork of communities, collectives and charities, around 19,000 'providers' in all (The King's Fund, 2019). In 2005 UK financing of long-term care as a percentage of GDP was broadly comparable with other advanced welfare states such as Germany, Australia and, Canada -although well-being outcomes lagged even then (OECD, 2005). In the 15 years since, spending in real-terms has fallen, while the population has risen (Age UK, 2019a). By the end of 2019, just before the pandemic hit, it was clear that resources such as finance, knowledge and people's time were insufficiently available across care sectors, and adequate finance was not getting through to where it is needed most (National Audit Office, 2018; Blakeley and Quilter-Pinner, 2019). One explanation is that the organisation of responsibilities within the care sector is highly fragmented (Care Quality Commission, 2018). Structures of accountability that would ground responsibilities locally are not coupled with adequate flows of finance. State funding is a mishmash of entitlements, needs based assessment, and local arrangements controlling access and levels of provision which is not supported by national commitments to adequate funding and good governance (Ranci and Pavolini, 2013). Untangling the historic threads that have led to this situation is not straightforward. There is no single policy, funding or service stream that is widely understood as 'long-term care'. In government, Labour, the Conservatives and Liberal Democrats have found substantive reform either impossible or undesirable. Instead, their policies have pursued the creation of quasimarkets and followed a health care logic of patient choice (Glendinning, 2017;Baxter, Heavey and Birks, 2020). These policies have been pursued consistently over the last three decades by all governments. This purposeful and systemic fragmentation of funding through local government, the NHS and individuals themselves has led to a significant accountability deficit (Shakespeare, Stöckl and Porter, 2018). In summary, innovation in the regulation of markets and choices imposed in pursuit of, for example, new public management, is a substantive factor in configuring the sector that has been left so vulnerable to crises.

How innovation and technology researchers imagine accountability
So what is the role of research in innovation and technology in UK care sectors? In the UK, public investment in research is generally channelled through UK Research and Innovation (UKRI), and, prior to its establishment in 2018, its constituent research councils and funding agencies. In recent years, where social care technology and research has been prioritised in major funding initiatives, it has been as a recipient sector for the Strategic Priorities Fund, part of the UK's National Productivity Investment Fund, a programme designed to boost economic productivity (UKRI, 2020). In the Industrial Strategies Challenge Fund for example 'Healthy Ageing' is a flagpole issue area. At the level of research council strategy however considerations of accountability in the sector are secondary. Within these frameworks, the social care sector is understood narrowly as something of a container, to be filled with undifferentiated innovation, the purpose of which is to boost economic output. In order to understand the picture outside of these strategic funds, and at the level of research projects across the UK research councils, 127 projects relating to social care were identified in the UKRI's projects database, Gateway to Research, representing a total funding of £111,715,022 from 2006 to 2019. The search protocol was designed to capture projects that explicitly address innovation or technology in their project documentation. Table 1 shows the distribution of projects and funding across research agencies from 2006 to the end of 2019, and the relative distribution of funding on a project basis is illustrated in Figure 1. These projects range in funding and duration from one-off sandpits, conferences and seminar series to major 5-8 million-pound investments lasting up to five years with some centres for doctoral training funded for longer still. The 15 largest projects by funding account for 60% of the total funds. In some projects, social care is the central focus of the proposed research. Whilst in others, it is sub-domain of a larger problem set, or used in a justificatory sense, where beneficial future knowledge spillovers are promised.

How researchers imagine and understand social care innovation in their research
Using a situated analysis approach (Clarke, 2009), I mapped each research site, analytically creating a set of situational maps. Addressing the research question meant assessing what they revealed about the discursive arrangement of i) understandings of social care, and ii) how technology and innovation was implicated in accountability structures, or not. Put another way, what exactly do researchers understand as social care, and what kind of accountability structures do these projects imagine technology existing within. Coding project documentation revealed four distinct discursive arrangements of social care and accountability, visible in 111 of 127 project abstracts. These are outlined in Table 2. The dominant discursive arrangement was that of social care as service provision. Care was understood as a resource to be distributed by professionals, usually, but not exclusively, through market arrangements. A typical project here is Innovative Technology for Healthcare Delivery -The MIMIT: CIMIT Collaboration (project 4, supplemental data) which promises to develop technology for expert users engaged in the delivery of services in Manchester. This understanding of social care most closely matches the incumbent position found in contemporary government and regulator policy literature discussed above. Techno-deterministic understandings of care took care to be directly attributable, at least in part, to the introduction or maintenance of technology within a system of care. An example project is How Can Smart Home Data and Systems Improve Assisted Living Services (28). These understandings took technology and innovation to be an unquestionable good -challenges of innovation in these projects included increasing acceptability amongst users and ensuring rapid technology roll-out. Unsurprisingly, EPSRC and Innovate UK funded projects are overrepresented in this group compared with the corpus. Situated understandings of social care tended to recognise socio-material complexities of care. Social care was understood in terms of both market-based provision, and as emergent from community settings including diverse knowledge and built infrastructure. ESRC and AHRC projects are over-represented here. Participatory methods and interdisciplinary research design were common amongst these projects. As they were with relational understandings of social care, which stressed human-to-human networks and communities enrolled in care (as opposed to care literally situated within communities). The relatively small number of projects categorised in this way focussed on issues like informal carers, empowerment of users and developing care practices. These categories, emergent from the data, are not ideal types and are deliberately undertheorised at this point. Some overlap exists between them -in particular between technodeterministic understandings of social care and service provision. But despite this, it should be clear from this brief overview that considerable diversity exists in how social care is understood and ultimately researched, although the distribution of these understandings is far from even. Moreover, as illustrated in Figure 2, there is also diversity evident within research portfolios of individual research councils. And yet this diversity tends to be absent from policy level discussions.

How accountability is situated in research
Representations of accountability were determinable for 87 of 127 projects. Three distinct representations were identified and are listed in Table 3. Again, these are emergent from the data and some overlap exists.
Paternalistic representations are those projects in which agency to make substantive decisions about the provision of care is in some sense restricted. For example, limited to elite political

Understandings of care
Research funder actors or those with expert knowledge. Technocratic representations are somewhat similar, however in these cases accountability is understood to be embedded in technological and administrative apparatus and systems rather than residing with particular people. Technology provides care when it contributes efficiencies of action, data and knowledge to these systems. Projects that promised to enhance care services through advance data methods, by liberating and connecting data 'stuck' in case notes or legacy administrative systems or creating data efficiencies featured significantly here. 18 projects featured discursive arrangements of collective accountability, where accountability was emergent from participative innovation practices and collective action. Innovation in these projects sought to co-produce care services with people usually excluded from innovation processes (projects 33 and 25); establish convivial relations amongst communities and infrastructures (89,103,104,118); empower individuals through user-centred understandings of their environment (75); and build human capabilities (14,25,105). Remarkably, in the 127 projects identified, only in one did researchers explicitly state the goal of enhancing well-being and cultivating capabilities in carers rather than in those cared for (30). Again, these kinds of accountability structures are not ideal types, and some overlap was observed, particularly between paternalistic and technocratic categories. In both of these categories, accountability and decision-making agency tended to be constrained to select actors and networks. Conversely, although relatively marginal in the corpus, collective understandings privileged the broadening-out of participation in care practices and situated accountability and responsibility at multiple levels of governance, including communities. In 32 research projects, it was not possible to classify understandings of proposed or existing accountability structures. In some cases it may have been that this was not considered by researchers, in others, the space available to document the project may have been insufficient. Regardless, this absence is notable, after all, research councils in recent years have made some efforts to address issues such as responsibility in innovation, of which accountability is a major component.

The role of innovation and technology in social care research
Overall, what is apparent from this pilot study is that structures of accountability are more homogenous than understandings of care. Moreover, when mapped, co-occurrences of understandings of care and structures of accountability in individual projects take the form of a diverse set of relations. This is perhaps best understood with the aid of a visualisation of these co-occurrence relations in Figure 3. Incidences of Understandings of care are aggregated for each category on the left, weighted by project budget. These are mapped to structures of accountability on the right, again per project and aggregated by category. What's evident in Figure 3, for example, is that approximately half of the projects that understand social care as a form of service provision, also represent accountability in the sector as paternalistic. Remember, these are sets of co-occurrence relations, they do not indicate that forms of accountability have a dependant relationship with how care is imagined by researchers or funders or civil servants. Rather, what Figure 3 reveals is that for a given kind of understanding of care, the possibility of imagining accountability is not foreclosed, more than one possibility exists. Yet despite this possibility of diverse ways of imagining, and ultimately structuring accountability, we see in the corpus of projects dominant ways of understanding accountability that align with incumbent policy and actor orientations in the sector. At the level of research council portfolios, innovation and technology researchers are reproducing incumbent framings of accountability structures.

Innovation in practice
So how, if at all, is accountability structured in the practices and outputs of research? The following cases are illustrative of challenges and potentials for restructuring and redirecting the processes and outcomes of innovation for care.

Case 1. Data and advanced analytic techniques in social care
The corpus of UKRI projects includes more than 20 projects which aimed to use data to address perceived knowledge deficits, economic and data efficiencies and a range of predictive and evaluative methods to enhance care. Typically, data is imagined as a solution to problems of efficiencies and monitoring. Indeed, one factor in explaining the vulnerability of the social care sector is the lack of data about the degree and distribution of system deficiencies at the sector level (National Audit Office, 2020, p. 42). According think-tank dotEveryone (Cory, 2019):

"Insufficient data is collected on the areas that matter most, including the outcomes for people who receive care. This fragmentation means it is not possible to understand what is happening across the social-care sector and inside the systems it's made up of.
Currently neglected is data that reveal the large disparity in per capita funding between areas.
The fragmentation that exists in the governance of social care is in part structured through data infrastructures it seems. In other words, sufficient data that would help locate and reveal vulnerabilities simply does not exist. Furthermore, the retrenchment of data collection has sustained inequalities and neglect within the social care sector. Given this retrenchment has Understandings of care Accountability in social care been predicated on budgetary choices and financial savings, this systematic data neglect is at least in part the result of political choice. At first glance the aims of these data focussed projects seem to address the structured neglect identified in Section 3. Yet in some cases, data that are being produced are also implicated in neglect. The Adult Social Care Outcomes Framework, for example, is used to "measure how well care and support services achieve the outcomes that matter most to people" (NHS Digital, 2018). The ASCOF was established to justify a market governance regime -the net result of which is to assure regulators that the provision of care is being measured, and in being measured can be controlled. This is to neglect important features of how care is practiced and experienced; how collaborative working improves the outcomes of older people in care for example (Cory, 2019). Moreover, because the framework reports on population level surveys and aggregated indicators, it reports the average experience of social care in the UK. Yet no individual experiences care at the mean. At stake here are vulnerabilities which occur at the margins, where care is insufficient or absent entirely. Arguably then, data frameworks like the ASCOF do more to obfuscate than make accountable. The point here is that big data and associated analysis techniques such as machine learning and other artificial intelligence technologies offer solutions to care as much in how they are configured (what real world practices and outcomes they measure, and what kind of statistical models they employ) as in the quantity of statistical inputs and outputs they process. In short, collecting and processing more data is not sufficient to address neglect, and can have negative impact on opening up structures of accountability to democratic scrutiny. Furthermore, these technologies require ongoing maintenance and indeed care of their own. In a recent House of Lords Technology Committee hearing, Stuart Butterfield of technology firm Canary Care, drew attention to operational and maintenance issues relating to data systems at user and community levels (Science and Technology Select Committee, 2020): Mr. Butterfield's point: these technologies require ongoing maintenance, skills and capabilities to manage and interpret data -aspects of social care infrastructure that are often overlooked in narrowly targeted research projects. And technologies that can, in their roll-out, actually increase operational and financial burdens on already stretched local government budgets. Finally, in recent years, civil society groups and critical scholars have raised serious concerns about societal impacts of these kinds of technologies. Bias in algorithms, increasing inequality, deficiencies in accountability, threats to democracy and the intrenchment of systemic racism feature (Benjamin, 2016;O'Neil, 2016;Noble, 2018;Whittaker, M., Crawford, K., Dobbe, 2018). Notably for this discussion, these issues are particularly prevalent in highly institutionalised settings (Benjamin, 2019;Eubanks, 2019) and disproportionately effect black and minority ethnic people, people who are more vulnerable to Covid-19 (HM Government, 2020). Despite attention to ethics and responsible innovation from the EPSRC and others (Boden et al., 2017;Macnaghten, 2020), these issues remain backgrounded in the discursive arrangements of accountability and care in individual projects as evidenced in Section 4.

Case 2: Practices of innovation for social care in robotics research
Innovation in robotics for health and social care has been increasingly promoted in funding programmes in the UK and Europe (InnovateUK, 2013;Khaksar et al., 2016;Parliamentary Office of Science & Technology, 2018). Even before Covid-19, a discursive framing of crisis was driving this research. Ageing populations, insufficient finance in health and social care budgets and recently in the UK, a Brexit induced shortage of low-cost health care workers unwilling or unable to come and look after Britain's elderly are all presented as rationales for urgent innovation (Prescott and Caleb-Solly, 2017) despite concerns about societal risk analogous to those discussed about data (Sparrow, 2016(Sparrow, , 2020. Assistive living robotics (ALR) is one set of technologies that has received policy and funding support for use in averting people's early move into more institutionalized or acute care (Gallistl and Wanka, 2019). ALR research involves understanding how people and robots can interact "intuitively, safely and effectively" (Caleb-Solly, 2016) and combines mechanisms of behaviour modification, human robotics interfaces, surveillance technologies and data analysis such as machine learning techniques. Yet innovation in ALR faces a number of challenges in integrating complex technologies into even more complex social and physical settings -even ones as relatively stable as care homes. (Buhalis and Darcy, 2010). Some of these challenges are technical -such as the limited abilities of robots and algorithms to interpret unforeseen situations and complex environments, even relatively sedate places like care homes. Other challenges are social. Challenges relating to the complexity of care-related needs of people makes it difficult for assistive technology to stay useful over time, as people's needs change. Socio-economic challenges such as social isolation, ill health and poverty (Age UK, 2019b) which mean people who most need assistance are those least likely to gain access to research, design and testing environments. The implications of these challenges are two-fold: 1) Design processes which exclude the needs, wants and values of the most marginalised people further exacerbate neglect (Spanakis et al., 2016). 2) Because it's often easier to adapt people and environments to robots and not the other way around, robotics have a tendency to objectify the very people that are under care (Sharkey and Sharkey, 2012). The challenge for researchers and funders is to avoid the drive for independent living resulting merely in the structuring of new dependencies and risks of neglect mentioned in the previous section. Acknowledging these concerns, an interdisciplinary group of roboticists at Bristol Robotics Laboratory are producing technologies which are socially situated. That is, designed to work and respond to their institutional and material contexts. In a recent workshop, BRL's Praminda Caleb-Solly explained that the challenge of ALR is to enhance people's well-being across a range of dimensions, to "address unmet social needs, to add value, and to bring joy". iii Roboticists at BRL and elsewhere increasingly use open living labs in facilitating research methodologies for doing innovation together with different kinds of end-users (Ballon, Pierson and Delaere, 2005). This means designing and testing robots along with the infrastructures they rely on so that they will be acceptable and even enjoyable to use as-well as making efforts to ensure technologies meet ethical and cultural criteria as decided by users and care-givers themselves. In determining and addressing these criteria, the research progresses through technological advances as well as innovation in interdisciplinary methods. But even the most porous open living lab can't fully address acute social exclusions. The strategy of researchers at BRL to introduce robotics into social care sectors that might meet these needs requires that roboticists partner with actors in those sectors; people in need of care, their formal and informal carers, healthcare and service providers, clinicians and third sector organisations. At BRL, some researchers are trialling new ways to work with community organisations and city councils in a six-week reablement programme that combines technologies with social interventions. The important thing for them is recognising and responding to the complexity of care-related needs (Spanakis et al., 2016) rather than focussing on technology alone. Other roboticists at BRL are applying similar 'co-design' and 'co-creation' methods to the design of socially assistive robots for use in therapy (Winkle, Caleb-Solly, et al., 2019;Winkle, Lemaignan, et al., 2019). These methods draw on insights from human robotics interaction and science and technology studies to foreground participation and two-way learning between designers and end-users. These examples acknowledge and explicitly attempt to avoid tendencies towards technodeterminism common in robotics. Roboticists seek to allow, to some extent at least, participation in the design and steering of technologies by people usually excluded from practices of innovation. And so, even within highly technical settings such as robot labs, we see the possibilities of diverse ways of addressing both the ends and means of research for social care.

Discussion and conclusions
In this paper I have addressed the question how has accountability in social care historically been structured and imagined in technology and innovation research? I have shown that accountability in social care is discursively framed by researchers in a diversity of ways. Yet despite this diversity, researchers in the majority of projects investigated reproduced framings of structuring accountability in the social care sector that align with dominant policy positions and incumbent actor interests. Specifically care imagined as a form of service provision, and accountability relationships structured through paternalistic and technocratic means. Beneficiaries of innovation research have tended to be national and local government through cost efficiencies; service providers through data and technological efficiencies; and individuals narrowly imagined as end-users. The implications of these findings is relevant for what and who is, and is not, attended to in infrastructures of care. For example, in only one of the 127 research projects assessed in Section 4 did researchers explicitly state that the focus of their project was improving conditions of carers (as opposed to the conditions of those being cared for or service providers). Carers are amongst the lowest paid workers in the UK, but during the Covid-19 crisis were amongst the highest valued. Yet they are almost entirely neglected in innovation research in the sector.
Research documentation I examined tended to neglect aspects of social care that are outside of market arrangements or service provision models. As discussed, typically neglected were professional and unwaged carers, but also families, wider communities and the social components of care infrastructures. Moreover, methods and opportunities to cultivate capabilities in the care sector that would mutually benefit neglected actors and build resilience and care capacity over time were almost entirely overlooked. I introduced two critical cases. The first illustrated how technology reinforces. incumbent accountability structures, showing how innovation in data can reduce accountability. The second offered examples of how participative research may broaden-out accountability by including in innovation people and communities often excluded. There are implications here for those who steers the direction of publicly funded innovation and technology research: the framings, objects, subjects and direction of innovation and technology research matter. Researchers have choices in how they frame and conduct their work. Reproducing framings and that align with accountability structures that even public agencies and regulators acknowledge as flawed (as evidenced in Section 3) is at best a waste of public funds. And policy decision makers and funders have choices in how they direct innovation in the sector, and these choices impact accountability.
What kind of research policy and evaluation criteria might usefully guide these choices? At the level of technology and data, perhaps what is required then is not simply big data, but thick data (Smith, 2018), that accounts for relations and obligations of care within communities of workers, end-users, and unpaid family carers and mitigates historic neglect. Specific data technologies are going to require specific governance models (Smallman, 2019). Recent work funded by the Nuffield Foundation makes a number of useful suggestions in this regard (Whittlestone et al., 2019). These might be augmented with re-invigorated evaluation criteria for research projects, trials, and procurement contracts with technology vendors. Criteria such as ensuring data infrastructure creates or add public value (Wilson et al., 2020). Criteria ensuring that the redistribution of this value is done in fair and equitable ways (Savona, 2019). Criteria that ensure the processes and outcomes of collecting, maintaining and repairing these data are fair and just to communities under-represented in social care governance. And evaluation methodologies themselves might pay greater attention to diversity and researcher capabilities required for participative and accountable research (Bone et al., 2019;O'Donovan, Michalec and Moon, 2020). And so, for those responsible for making decisions in care sectors adapt to Covid-19, precaution is warranted. For no matter the urgency, policy that promotes rapid, development, adoption and scaling of technologies without consideration of how innovation re-structures relationships of accountability runs the risk of advancing irresponsible innovation.

End notes
i These failures are evident in controversies of contact tracing technologies that did not and could not work (Boland and Wright, 2020; Lynskey and Veale, 2020) -not least amongst people living and working in care homes. In centralised testing infrastructures that didn't work as promised and give rise to mistrust instead of efficiency (Leach, 2020). And in technological solutions in acute care settings that have neglected the resource needs of long-term care in communities. Moreover, technology-led responses have failed to address new social situations as they have arisen (Chiusi, Fischer and Spielkamp, 2020). ii A description of the search protocol used, and an overview of the projects returned is provided in the Supplemental Data.
iii In a presentation to the European Robotics Forum, Malaga, Spain, March 2020.