Introduction to the digital welfare state: Contestations, considerations and entanglements

In 2019, a United Nations report issued a stark warning, urging the global community to pause and consider the genuine threat of societies ‘stumbling zombie-like into a digital welfare dystopia’ (Alston, 2019, p. 1). This cautionary message centred on the significant impacts of governments worldwide embracing the digitalisation, automation and datafication of their social protection systems. Human rights lawyer Philip Alston, the report's author, voiced concerns that had previously been expressed by scholars and activists but had not yet received widespread attention, especially regarding the possible weakening of privacy rights, increased surveillance and growing social disparities. Alston also highlighted concerns about governmental overreach and the further marginalisation of vulnerable populations due to barriers in accessing and navigating digital systems. The report was an important intervention, reframing the discourse from speculation about a distant future to recognising present circumstances. It underscored that ‘the digital welfare state is either already a reality or is emerging in many countries across the globe’ (Alston, 2019, p. 1), with concurrent risks and inequalities actively unfolding.

Understanding and exposing these current realities is an urgent task of critical scholarship. Yet the digital welfare state presents a challenging domain for research as it sits at the intersection of two sizeable bodies of social science investigation – digital technologies and social policy/welfare states – both of which are marginal to mainstream sociology, and involve a broad spectrum of rapidly changing technologies, actors and processes. Broadly speaking, the digital welfare state refers to a particular state formation in which wide-ranging digital technologies, including automated decision-making systems, algorithms, big data analytics and artificial intelligence (AI), are integrated into the administration and provision of welfare services by government agencies, which, in some cases, are then delivered through a complex set of contracted providers (see Dencik & Kaun, 2020; Larasati et al., 2023; van Toorn, 2024), which social policy scholars refer to as the mixed economy of welfare (Powell, 2019). These changes are occurring across a range of areas, from social security, public and social housing, healthcare, disability and child protection services, to food aid, border security, law enforcement and justice systems (Dencik et al., 2018; Dencik et al., 2019; Eubanks, 2019; Ferreri & Sanyal, 2022; Gillingham & Graham, 2017; Keddell, 2019; Mann, 2020, Moats & McFall, 2019; Rachovitsa & Johann, 2022; Redden et al., 2020; van Toorn & Cox, 2024; Završnik, 2021). At times, they have extensive overlap spanning various socio-technical systems, despite appearing to lack any clear connection. For example, biometric authentication systems, which verify individuals’ identities through distinct biological traits, such as fingerprints or facial features, are used in the distribution of welfare services and the securitisation of borders (Madianou, 2019; Magnet, 2011; Marciano, 2019; Singh & Jackson, 2021), while also enabling greater conditionality in social policy (Henman, 2011; Watts & Fitzpatrick, 2018). In various social service domains, citizens increasingly interact with governments through digital platforms and mobile applications (Henman, 2019; Schou & Hjelholt, 2018). Aiming to improve access and efficiency in service delivery, digital welfare states target substantial cost reductions by reducing dependence on manual human interventions and administrative labour, and integrating algorithms to automate decision-making processes related to eligibility, service levels and other aspects of social resource allocation for citizens and their households. Much of the adoption and incorporation of digitalisation has taken place in a context of global moves towards fiscal austerity and the criminalisation of poverty, homelessness, border crossing and asylum seeking, disability and mental distress (Dencik et al., 2019; McQuillan, 2022).

Viewed critically, the digital welfare state is seen to amplify these trends, allowing for increased scrutiny and surveillance of welfare recipients, new migrant arrivals and other ‘problem’ or ‘undeserving’ populations, alongside prioritising administrative labour cost-saving measures over meeting social needs (Bielefeld et al., 2021; Eubanks, 2019; Jørgensen, 2023; Mann, 2020; Schou & Hjelholt, 2018; van Toorn & Scully, 2023). This collection of papers aims to contribute to current debates in the area and simultaneously develop a deeper, more empirically nuanced understanding of positionality and relationality vis-à-vis the digital welfare state, particularly in relation to highly marginalised populations that are heavily reliant upon state provisioning to sustain basic levels of social, economic, physical and emotional wellbeing. This introduction to the collection seeks to highlight these important domains requiring closer ongoing sociological investigation and to provide readers with some introductory background understanding of the field.

Although not all the papers directly address the nature of the (welfare) state with its digital transformation, all broadly work within a framework which moves away from seeing the state as a distinct, unified entity or actor, toward a conception of the state as a social relation (Jessop, 2002, 2007). The concept of the state as a social relation suggests that the state does not possess an inherent, static nature, such as being a neutral mediator between various social interests (for example, capital and labour), an independent entity with its own bureaucratic aims and interests, nor simply an instrument of the ruling class – as dominant accounts suggest. Instead, as captured across the papers in this special issue, the terrain of the state is shaped by the broader social relations in which it exists, particularly the dynamics of power among different social groups within a given historical moment, entailing processes of continuity yet simultaneously evolving and transforming over time (see Jessop, 2002; Peck, 2001). Given these dynamics, historically it remains clear that the state leans towards certain methods of action and certain types of actors, showing preferences – due to path dependencies long set in train – for particular policy and technological trajectories over others. Thus, analyses of state power should stem from an appreciation of the political struggles and contests unfolding in the state sphere, including practices of governing that aim to manage the external demands of social movements and political actors, as well as an understanding of the array of institutions existing within and forming the state. This is notably apparent in the body of literature on capitalist welfare state regimes initiated by Gøsta Esping-Andersen (1990). In his analysis of liberal, conservative and social democratic welfare state regimes, he took into account socio-cultural and political histories as well as social policy settings and trajectories. Although each of these distinct worlds of welfare capitalism embody different public values and approaches to citizenship, and evidence points to accompanying different modes of computerisation (Adler & Henman, 2005), there are arguments suggesting more recent convergence, particularly since the rise of neoliberalism and new public management (NPM) (Schmitt & Starke, 2011).

We suggest that this type of state theory is precisely what is necessary for examining the digital transformation of the welfare state and its administrative-institutional apparatus. Such an approach allows detailed empirical analysis of where different groups are positioned with respect to the changing balance of forces seeking to influence the forms, purposes and content of digital welfare. This deeper analysis is critical if we are to fully grasp ongoing shifts in the dynamics of political power and the interplay of current struggles, both internal and external to the state, that aim to shape the forms, purposes and content of welfare provision. For example, as Richard Titmuss (1955/1976) pointed out almost 70 years ago, welfare transfers and services are often provided as extensively to the professional and middle classes, and not just to the unemployed and very poor populations, as imagined in popular discourse. Take childcare subsidies and taxation relief by way of example. Monitoring and surveillance may be similar to other policy domains in relation to household poverty and parenting, yet due to individual and household capacity to afford, manage and integrate digital transformations, these transformations are experienced differently at the individual, household and political level, and for some groups (Henman & Marston, 2008) are considered much more convenient systems with AI and automated decision-making (ADM) implementation as they no longer have to ‘deal with the state’.

The types of digital transformations that occur within each area are thus imbued with long-standing political meanings and struggles coupled with historical dynamics of political, civil and institutional power. The point is not just that various groups occupy distinct relational positions concerning the state with its digital transformation over the last 20 years, nor that they are affected differently. Rather, it is the fact that their positionality of power in relation to the state impacts upon the practices of digitalisation that the state takes up and embeds across its welfare institutions and policy domains, with digitalisation often reinforcing and extending these differential relations of power vis-à-vis the state. For example, the growth of individualised or personalised service delivery advanced in the private sector (e.g. through customer relationship management systems) has been taken up in welfare states through profiling of citizen-customers leading to greater population segmentation and differential treatment (Henman & Dean, 2010; van Toorn and Carney, 2024) which have only accelerated with machine-learning AI.

As the combined set of papers in this special issue suggest, understanding the digital transformation of the welfare state in relational terms creates opportunities to explore the multiple processes and relationships through which state power is constituted and expressed digitally. It also enables us to understand that, largely, digitalisation tends to further bed down – even literally encode – long-standing political, economic and social relations of power, enabling only incremental changes in welfare governance rather than bringing about deeper structural transformation. This perspective is embraced within many of the papers within this collection, which provide innovative explorations and methodologies of exploring, explaining and conferring these dynamics.

This raises a critical question as to whether digital, data-driven and AI-powered technologies signal a new governing paradigm or merely a superficial enhancement of past technologies. Contrary to the notion of a radical disruption implied by the recent ‘digitalisation turn’ or the more recent fetishisation of AI, we take the view that digitalisation involves change within continuity (Fussey & Roth, 2020).

Thus, digitalisation needs to be understood as part of a historical lineage of social policy in particular and society more broadly. How digital technology gets designed, developed and deployed is shaped by socio-political values and structural power. In doing so, it encodes particular ways of knowing and acting which in turn define relationships between citizens and the state. Thus, while digital technology is socially shaped, its operation also shapes society, in a co-evolving loop. Indeed, digitalisation is not new – it has not just occurred in the last 5 or even 15 years, but computerisation of the welfare state has a history, starting not long after the Second World War (Henman, 2022), and its operation can be seen within longer-term trajectories.

As technologies of both information and automation (Bellamy & Taylor, 1998), digital technologies contribute to the information and processual operations of the state. Higgs (2004), for example, sees digitalisation as part of a long trajectory of the evolution of the information state through centralised collation and calculation of information, that involves the reformulation of the world and how it can be perceived so it becomes legible and thus governable by the state (Scott, 1999). Scholars have long identified how digitalisation of information – stored in databases – prioritises well-defined and quantified knowledge over rich description, moulding people's lives into rigid categories and reducing administrative discretion (P. A. Busch & Henriksen, 2018; Henman, 1995; van Toorn and Scully, 2023). This theme of creating visibility and digital modes of truth is evident in this collection, as demonstrated by the works of Singh and Mateescu.

The welfare state has been intricately tied to the evolution of statistical forms of knowledge, methods and technologies aimed at delineating and managing populations, particularly the poor. Notably, the welfare state arose in the late 19th century and the first half of the 20th century by adopting insurance models of public policy to protect against age, disability and unemployment, based on actuarial knowledge. As statistical and computational techniques developed, so too did technologies of classification, social sorting and so on, thereby reconfiguring risk-based governmental rationalities. As Henman (2010) noted, ‘e-government is leading to a growth in the conceptualisation and conduct of targeted forms of risk government, whereby populations are differentiated and governed differentially’ (p. 11).

These observations beg the question as to what extent digitalisation is simply a tool to project broader socio-political values or whether it shapes those socio-political realities as well. In the public administration literature, it is widely recognised that the 1990s and the first decade of the 2000s involved an installation of NPM, a managerial approach inspired by neoliberal principles and adopted from the private sector, which was extensively implemented in government administration (Lane, 2002). Dunleavy et al. (2006) have argued that this mode of governance reached its limits when applied to digitalisation, resulting in a shift towards ‘Digital Era Governance’ that combines some elements of NPM with the technical capacities of digitalisation that enable service holism and cross-agency partnerships (Margetts & Dunleavy, 2013). While these observations are important in understanding institutional-level modes of administration and governance, they do not engage with the operations of power and the experiences of citizens vis-à-vis the digital welfare state that have been outlined earlier.

The operation of the state has thus changed through digitalisation. Bovens and Zouridis (2002) identify a shift from ‘street level bureaucracies’, whereby citizens transacted face-to-face with government officials in bricks-and-mortar buildings, to ‘screen level bureaucracies’, whereby citizens interact in person or on telephones or even via online chat with bureaucrats who are behind screens. The rise of the internet shifted citizen–state mediation from in-person to online. Bovens and Zouridis identify a rapidly emerging ‘system level bureaucracy’, whereby citizens interact with computers and government services and decisions are automated. Such system-level operation underpinned the design of Australia's infamous Robodebt Scheme, independent of its illegal algorithm (Whiteford, 2021).

Behind the legally flawed use of income averaging, Robodebt highlights other key impacts of automation for the operation of the (welfare) state and its relationship with citizen-customers. First, the speed of operations was greatly increased – notifications of possible debt increased 50-fold. Second, design shifted towards greater self-service, where citizens are tasked with supplying all necessary data, verifying calculations and presenting evidence to challenge algorithmic decisions. This involved the reversal of the onus of proof, which in the context of alleged debts involved what academic and welfare lawyer Terry Carney (2019) described as a form of extortion. Third, as governmental decision-making transitions from manual decisions made by humans to automated processes, it disrupts established government accountability mechanisms, which traditionally involved scrutinising the accuracy of human decisions, but now require scrutiny of algorithms. Four, Robodebt highlighted how automated decisions are typically more opaque, meaning the black box of government and wider social processes are not understood (O’Neil, 2017; Pasquale, 2015).

The integration of digital tools, AI and big data analytics into social policy and welfare administration does not merely enhance the capacity for surveillance and control; it also introduces new dimensions of governance that are predictive, pre-emptive and deeply personalised (Henman, 2020; Rouvroy & Stiegler, 2016). These modes of algorithmic judgement are particularly evident in risk profiling of the unemployed, which involves sorting people by likelihood of long-term unemployment (Desiere et al., 2019); in the profiling of children to assess risk of abuse/harm (Gillingham & Graham, 2017; Sacher, 2022); and in criminal justice systems to determine risk of criminal acts, threats to the public or recidivism (Brayne, 2020; McKay, 2020). Moreover, such profiling is highly prone to bias, reproducing and exacerbating bias in training data and designers, thereby reproducing social inequalities and discrimination with ‘objective’ data (Herzog, 2023).

These concerns form the backdrop and inspiration for this special issue. The collection evolved from an initial panel discussion held in 2022 at the Australian Sociological Association's (TASA) annual conference. The panel took an intersectional approach to analysing digital welfare. Panellists Lyndal Sleep and Shelley Bielefeld, along with panel organisers Georgia van Toorn and Karen Soldatić, explored the classed, racialised, gendered and disabling aspects of digitalisation, and the ways in which these intersecting power relations come to shape people's experiences of the welfare state – its institutions and provisioning – in the digital age. As empirical researchers with a commitment to community-based research, we were particularly interested in surfacing marginal perspectives. What emerged from the discussion was a particular sociological orientation towards the idea of digital welfare. This approach, while empirically grounded, is also politically committed and radical in its analysis of the structural forces that condition this particular historical moment of digital state (trans)formation. Key themes emerged around the ongoing neoliberalisation of social welfare, the role of digital technologies in the government of racialised others at and within settler state borders, and ideologies of eugenic nation-building that migrated – for example, to Australia – with the arrival of European settlers.

Following the panel, we wanted to flesh out in greater detail the positive contribution that a distinctly sociological approach can offer to understanding digitalisation and its transformations in/of the welfare state. We invited contributions from leading sociologists in the field. With the scope intentionally broad, we encouraged submissions addressing concerns at the local, national or transnational level of analysis in one or more policy areas internationally or within Australia. We encouraged a critical approach that scrutinises power dynamics, social inequalities and the implications of digitalisation on marginalised communities across diverse geopolitical contexts. The final collection reflects this diversity of standpoints. The papers explore a range of issues shaped by the concerns of local practitioner and service-user communities as well as by broader, sociological themes and questions. The collection as a whole highlights the role of technology in mediating state–citizen relations across different policy domains, including social care of the elderly and people with a disability, child protection and family services, education, migration and border control, and social security and welfare payments. Examples are drawn from many countries, including Australia, the USA, India, Denmark, Germany and Sweden. All the papers in their different ways critically engage with these broad themes while also highlighting the contextual factors shaping lived experiences of digitalisation in specific places, time periods, sectors and communities.

One prominent theme was the connection of digitalisation with ‘the social’. Digitalisation is not conceptualised as a technical exercise that can be distinctly analysed separate to social processes. Perhaps this is unsurprising given that the authors are sociologists, but it is important and distinctive given common threads of implicit technological determinism or treating digitalisation as a technical activity dominant in much of the literature and research. In contrast with much public administration or management analytical approaches to digitalisation, whereby digital technologies are neutral tools (simply) implementing policy, political or management imperatives, the papers in this collection highlight the rich entangling of social values (e.g. control, surveillance, efficiency), forms of knowing and classifying, operations of power, and histories of relationships in which digital technologies operate and are enacted. This location of the digital welfare state within social and historical dynamics is an important counterweight to much recent scholarship that appears to treat AI and recent digital technology innovations as something completely new and without precedent. If there is a critique to be made, it is that the technical realities and affordances (Davis, 2020) of digital tools – and their diverse abilities – are perhaps given too little attention. One might be forgiven for thinking that adopted digital technologies are simply a reflection of society, a mirror that has no effect. Indeed, sociology is often only interested in humans, and non-humans become bit players or missing masses (Latour, 1992).

Van Toorn and Soldatić's paper situates modern digital and data practices within a longer historical arc, in this case relating to eugenic social policy. By examining contemporary applications of ADM systems in child protection services, immigration and border management, and social security policy, they reveal how these systems perpetuate discriminatory practices and exclusionary outcomes reminiscent of past projects of eugenic nation-building. The authors argue that although these systems may not explicitly endorse eugenics, they effectively replicate its principles by classifying individuals based on their perceived economic or biological value/risk to society, often through ableist and racialised lenses. Through their analysis, van Toorn and Soldatić shed light on the ways in which automated governance reinforces existing power dynamics and inequalities, with particular impacts on marginalised groups such as disabled people and Black, Indigenous and People of Colour (BIPOC) communities. Their paper underscores the importance of recognising the historical continuities in the use of technology for the governance of poverty, ‘problem’ families and national borders, and the need to challenge oppressive systems of governance.

Sleep's paper looks across ADM systems used in Australian social services, focusing on impacts of such systems on service users. Contrary to dominant motivators of greater efficiency, consistency and service delivery, her research reveals that these systems are often experienced as punitive, extractive, excluding and inhumane, reflecting anti-welfare ideologies in Australia where ‘social security recipients have been historically treated as economic scapegoats and punished for needing to access services’. For instance, the automation of processes previously handled by human administrators has resulted in the extraction of labour from service users, who are now tasked with navigating complex online systems and updating personal information. The paper also highlights concerns about digital exclusion, arguing that individuals lacking digital literacy or access to digital devices are effectively penalised, as they are unable to claim payments via online systems. Additionally, ADM systems are shown to struggle ‘to deal with the complexities of day to day life and the lived realities of being human’, particularly for people living with disability, whose unique needs and variations in daily functioning pose significant challenges for algorithmic decision-making (see also van Toorn & Scully, 2023). This difficulty in accommodating individual differences often results in exclusion from statistical calculations, further marginalising already vulnerable populations (Treviranus, 2020). Sleep argues that despite the involvement of humans in decision-making processes, the overall system is viewed as inhumane, with professionals often deferring to automated decisions rather than exercising their professional discretion.

Taking a similarly broad and critical view of the welfare state, Hjelholt's article investigates the implementation of digital welfare as a strategy to sustain the Nordic welfare model and silence resistance to it. The paper demonstrates the significant ease with which the digital welfare state is able to assimilate resistance to changes in the welfare system, even within traditionally more generous welfare regimes, notably the Nordic tradition. Drawing upon the work of Boltanski and Mathiesen, Hjelholt provides novel insights into the dynamics of power and resistance in contemporary welfare societies, particularly amidst their adoption of digital transformation strategies. Hjelholt suggests that digitalisation, within the context of late capitalism's regime of governance, is not merely a neutral tool for welfare ‘reform’ but rather a mechanism through which capitalist objectives are advanced and dissent is managed. In the case of Denmark, the government appears to respond constructively to societal critique of its digital initiatives, but, instead of making significant policy changes, it subtly adjusts its rhetoric to emphasise the importance of digital inclusivity. Denmark's digitalisation agenda is advanced, leaving behind persistent inequalities and contradictions, particularly affecting a segment of the population which remains marginalised and unable to fully benefit from digital welfare.

A second prominent theme evident in the papers in this collection relates to the role of digital technologies in truth-making. A key part of digitalisation is the collection and storage of digital data which are taken as representing the world in an objective manner. Within the digital (welfare) state, modes of identifying individuals and their status are critical in accurately delivering welfare. This necessarily involves the creation of digital versions of selves or so-called digital doppelgangers (Bode, 2005; Cheney-Lippold, 2017). These modes of knowledge are not, however, simply a reflection of the world. They are created as digital data ontologies that have social histories of standardisation and classification (Bowker & Star, 1999; L. Busch, 2011; Foucault, 1969/2013). The data structures embedded in algorithms then shape the world and come to reify ways of perceiving ourselves and others. Digital data are taken as truth and used for governing individuals and populations, reinforced by popular assumptions that such data are objective, and hence accurate. The problematic nature of this is evident in Eubanks’ (2019) account of people being labelled with ‘failure to cooperate’ and in risk classification based on biassed algorithmic models (Benjamin, 2019).

Sensor technologies – seek to read the world, enabling the monitoring of bodies, movement and mobility in real time. This makes sensory technologies particularly attractive in welfare systems that rely upon highly racialised, gendered labour regimes, such as the US Medicaid programme. Mateescu's paper examines the significant expansion of state surveillance into the everyday lives of disabled care recipients and the care workforce, largely poor women of colour, through the implementation of Electronic Visit Verification (EVV). EVV is a system mandated by federal legislation for Medicaid-funded in-home care services, aimed at reducing welfare fraud by tracking workers’ activities and locations through a mobile app. Mateescu maps the ways in which digitalisation not only enables new forms of surveillance of care workers against contracted labour time during home visits, but also the multi-functional surveillance it provides to assess Medicaid recipients’ need for services through ongoing care worker data collected on and about the recipient. Digital surveillance thus becomes a ‘tool for more finely tuning classifications of different types of paid and unpaid care’, delineating who is deserving of care and what level of care, with the aim of reducing so-called ‘fraud, waste, and abuse’ by both the recipient of care and care workers.

Bielefeld's paper is also illustrative of social security digitalisation that enables excessive state intrusion into the private lives of poor Indigenous social security recipients and their spending behaviours. Through systems of electronic surveillance, operationalised through the Cashless Debit Card the Australian state becomes capable of controlling and monitoring social security recipients’ purchasing habits, dictating what items can be purchased and where those items must be purchased. States are able to direct, even at vast distances, the everyday lives of the poorest poor, including controlling their spending on everyday household items such as toiletries and bathroom products, basic food items and so forth. Bielefeld's work juxtaposes the power of the card's designers and the role of private firms specialising in e-financial services through public-private contractual arrangements that have become so common with NPM practices under state neoliberalisation. The paper is demonstrative of the racialised practices of settler power that continue to stigmatise Indigenous peoples across settler colonial landscapes, positioning Indigenous peoples' dispossession within public portrayals of undeserving subjects requiring radical state intervention, even in their most private spheres of life.

A third theme evident in this collection of papers is the importance of humans in making digital technology work. In contrast to the widespread rhetoric that a primary benefit of digital technology is that it automates processes and generates greater efficiency, several papers highlight how digital technology needs human involvement to enable its operation. This is also seen in much framing of AI and social media operations ignoring the human work involved in labelling content and content moderation (Le Ludec et al., 2023; Pinchevski, 2023). Such work also highlights how these human roles are often exploitative and powerless. Organisations, including governments, often automate processes to remove internal human processing, but this means that service users – in the name of self-service – will need to do more work in data input that may previously have been done by employees. In digital welfare states, automation has been observed to lead to greater administrative burdens as agencies become more data hungry to support complex eligibility testing, verification and compliance (Madsen et al., 2022; Peeters, 2023). This theme in turn also highlights the importance of maintaining human involvement in digital welfare, the ‘human-in-the-loop’ of decision-making – otherwise automated decisions may become inappropriate or unaccountable. These observations in turn highlight who gets to decide how digital welfare is designed and who has control in the model of digital welfare enacted.

Singh's paper explores how humans and technology are necessarily entwined within contexts marked by extensive digitalisation, growing welfare demand and limited state resources. His study, centred on Aadhaar, India's biometrics-based national identity verification system, illustrates the crucial role of intermediaries in helping citizens navigate complex bureaucratic processes. Using an ethnographic approach, Singh is able to show what macro-level analyses of digital welfare often overlook: that despite advancements in technology, human involvement at the micro-level remains crucial for any implementation – let alone effective implementation – of digital welfare services. This reliance on what Singh terms the ‘phatic labour’ of intermediaries reflects a deeper theme of human enmeshment in technological systems, emphasising that technologies are not standalone entities but are intricately intertwined with human agency and social contexts. His paper also speaks to the self-service logic of these systems, wherein citizens are expected to navigate bureaucratic procedures independently of formal state assistance – a form of outsourcing that relocates certain state functions and forms of ‘data work’ into areas of private, work and community life. Singh's paper in particular emphasises the nuanced dynamics of technological integration within welfare states, reminding us that human energies and interpersonal relationships bridge the gap between technological design and real-world application.

The paper by Zakharova, Jarke and Kaun critically analyses the dialectics of care and control within digital welfare systems through detailed study of digital technologies as they interact with human dispositions and systems oriented towards care. Digital technologies, they argue, more often than not adhere to a paradigm of control and surveillance, prioritising efficiency over nuanced engagement with care. And yet the human effort invested in developing, supporting and maintaining digital infrastructures is often rooted in caring relations – for example, between adult caregivers and their elderly parents or between school administrators and the student body. The paper presents three empirical vignettes, each one illuminating different aspects of this inherent tension. Vignette 1 illustrates how the integration of digital welfare geron-technologies into care arrangements for older adults involves invisible forms of labour to ensure functionality and acceptance, transforming care work to include ‘caring for digital technologies in order to maintain welfare services’. Vignette 2 explores the use of smart sensors in social housing facilities, highlighting the tension between providing safe housing and the focus on monitoring and controlling deviant behaviour. Vignette 3 discusses the digital transformation of German K-12 (kindergarten to grade 12) schools, where the maintenance of datasets becomes essential for school secretaries’ care for their school. This data work reveals disparities in recognition and reward between the predominantly female secretarial staff responsible for data input and the more highly valued data scientists or software designers and the ministries of education to whom secretarial staff report. In proposing an analytical framework for understanding the care–control dialectic in digital welfare, the authors emphasise three interrelated layers: the values guiding public care provision; the infrastructures organising care work; and the work involved in configuring and maintaining care infrastructures. Welfare technologies, while aiming to fulfil caring values such as convenience, security and safety, often overlook the diverse lived experiences of differently marginalised groups, and the complex, often invisible nature of care work in digital welfare undermines its quantification and integration into bureaucratic systems.

This special issue on the rapidly evolving digital welfare state highlights important themes requiring ongoing critical analysis and contestation. It also highlights how the now inexorable digitalisation of social life requires sociology to take seriously and as a default the inclusion of digital technologies in empirical, analytical and conceptual work, just as ideas of power have infused and animated sociological inquiry for more than a century.