Abstract
This article investigates future visions of digital public administration as they appear within a particular regulatory process that aims to enable automated decision-making (ADM) in public administration in Finland. By drawing on science and technology studies, public administration studies, and socio-legal studies we analyze law in the making and identify four imaginaries of public digital administration: understandable administration, self-monitoring administration, adaptive administration, and responsible administration. We argue that digital administration is seen from the perspective of public authorities serving their current needs of legitimizing existing automation practices. While technology is pictured as unproblematic, the citizen perspective is missing. We conclude that the absence of an in-depth understanding of the diverse needs of citizens raises the question whether the relationship between public power and citizens is becoming a one-way street despite of the public administration ideals that express values of citizen engagement.
Keywords
Introduction
Public organizations are automating their decision-making processes in taxation, social security, immigration, and in many other areas. The discourses on digital public administration centre around technological innovation and the complexity of computer-based epistemic procedures (Katzenbach and Ulbricht, 2019; Koulu, 2021), raising doubts about the ability of regulatory action to keep up with the speed and transformative impact of digitalization (Brownsword, 2019; Floridi, 2018; Konrad and Böhle, 2019). The regulatory narratives about digitalization tend to form around two opposed ideas: either administration becomes more intrusive and pervasive, or more inclusive, responsive, and diverse (Katzenbach and Ulbricht, 2019). But which narrative is more persuasive? And what is the role of law-making in creating, enforcing, and realizing these narratives that shape social reality? Due to the social embeddedness and contextuality of digitalization, answers need to be sought through situated analyses and empirical knowledge that provide a better understanding of the ongoing regulatory priorities and institutional voids in democratic decision-making (Broomfield and Reutter, 2022; Lepinkäinen and Malik, 2022; Van Zoonen, 2020).
In this article, we examine future visions of digital public administration as they appear within a particular regulatory process that aims to enable automated decision-making (ADM) in public administration in Finland. The reform signifies an important step towards regulating not only automated but also more broadly digital systems in public administration. We ask: How is good digital administration envisioned? What are the perceived problems ADM presents for the relationship between public power and citizens? Our analysis maps out how law-making produces and mediates different visions about digitalization, with impacts on the relationship between citizens and public power.
As law is a potent form of social engineering (Cohen, 2019; Pound, 1942), regulatory processes on digitalization provide an interesting vantage point for examining the expectations, concerns, and hopes of our technological futures. Legal regulation, and the political process that produces it, participate in conceptualizing and steering digitalization by drawing the boundaries for acceptable and unacceptable use of automated systems. In other words, technological futures are discussed and negotiated, made, unmade, and decided upon, within, around, and in relation to legislative processes. Once a law is enacted, certain hopes for the future become fixed, although uncertainty about the law's fulfilment remains in the hands of those applying and enforcing the law. It is not only law that is affected, for framings adopted in legislation tend to become authoritative and decisive for broader meaning-giving in society. By looking at the ADM reform through the lens of sociotechnical imaginaries, the implicit assumptions behind these framings become visible.
Drawing from science and technology studies (STS), public administration studies, and socio-legal studies on law, technology, and society, we argue that technological futures discussed in the context of digital public administration, are not actually visions about technology itself but about articulating administrative ideals. These future imaginaries reshape the role of the citizen vis-à-vis administrative institutions, a relationship defined by power asymmetries and the potential for abuse that administrative law seeks to mitigate through accountability and oversight. However, these existing legal frameworks are often at odds with automation, contributing to the perceived urgency of regulatory action. For example, the legal rules on liability and accountability reflect an implicit assumption of human bureaucrats who are ultimately responsible for exercising public power and who are hence personally accountable for its misuse, an assumption that no longer holds true with automation (Koulu, 2020). The need to resolve these tensions between law and digitalization creates the venue for imaginaries to emerge but also forces the focus on technology.
Our study provides a qualitative content analysis of the Finnish ADM reform in the making. We present four distinct but overlapping imaginaries of digital public administration that come into play: understandable administration focusing on simplicity, explainability, and clarity; self-monitoring and autonomous administration that should not be burdened with external control; adaptive administration that goes beyond technology adoption; and, finally, a weaker imaginary of responsible administration that strives for equal and inclusive service to all citizens. All these imaginaries emphasize the perspectives of public institutions, whereas citizens’ perspectives are lacking from all but the last one. The average citizen becomes visible indirectly, either through his or her expected (lack of) capabilities and needs that remain subordinate to institutional perspectives, or through possible misuse of technological systems. Because imaginaries are persuasive, there is a danger of further distancing the citizen from digital public power, of less interactive and participatory communication, and an institutional predisposition towards ignoring the heterogenous individual and collective needs of citizens.
Our findings demonstrate that the imaginaries naturalize digital technologies, rendering their use unproblematic. This disguises the problems that originally led to the need to regulate ADM. We start by looking at the prior literature on socio-technical imaginaries, administrative ideals, and socio-legal studies. We then provide a contextualization of the Finnish ADM reform, followed by a description of our research design, data analysis, and the findings. In the discussion, we take up the materialization of imaginaries. We conclude by suggesting that a critical investigation of automation should be made in future research.
Conceptualizing imaginaries and ideals in prior literature
To explain visions of better digital administration we draw on science and technology studies, public administration literature, and socio-legal studies (Jasanoff and Kim, 2009; Koulu, 2021; Rose et al., 2015). Despite their differing origins, these approaches operate with the concepts of ideals and imaginaries, that is, what is valued, desirable or risky, and address the role of technology in social change. Combining these insights enable the identification of sociotechnical imaginaries embedded in the ADM reform, as they relate to expectations of what constitutes good administration and how technology contributes to creating better public services. We start by looking at sociotechnical imaginaries (Jasanoff, 2015a), followed by the ideals behind e-government discourses to show how citizen perspective as part of the ideals of public administration has evolved over time (Rose et al., 2015). Finally, we discuss socio-legal studies with respect to values embedded in technological design, which contextualizes the debate of imaginaries in the realm of law and digital technologies (Brownsword, 2019; Koulu, 2021).
Sociotechnical imaginaries
Sociotechnical imaginaries are future-oriented visions about social and technological developments that relate to certain goals, often steering possible courses of action (Jasanoff, 2015a). Visions and imaginaries try to balance the hopes related to scientific and technological achievements while at the same time managing their risks. Sociotechnical imaginaries direct nation-specific technological projects and shape social responses to innovation as they are often accompanied by official policy instruments, which is why some imaginaries are collective and long-lasting (Jasanoff and Kim, 2009; Sismondo, 2020). At the same time, imaginaries are temporally situated, contested, and culturally specific, which makes them flexible and dynamic (Jasanoff, 2015b; Mager and Katzenbach, 2021).
There is ample literature on conceptualizing and constructing sociotechnical imaginaries in various geographical and political settings, e.g., health data and policy, energy, and digital transformation (Konrad and Böhle, 2019; Tidwell and Tidwell, 2018; Tupasela et al., 2020). Prior studies demonstrate that imaginaries centre around the unlimited benefits and potential of technological innovations, while risks are seen as manageable (Chan, 2021). Recent studies on the historical development and public imaginaries of AI in the media suggest that imaginaries rely on myths and vague terminology, which create ambiguity and uncertainty (Natale and Ballatore, 2020; Scott Hansen, 2022). Optimism over new technologies seems to prevail, emphasizing efficiency expectations (Lepinkäinen and Malik, 2022).
In the legal domain, sociotechnical imaginaries have not been utilized as extensively as they could have. We suggest they provide dynamic and flexible conceptualizations of future visions, in contrast to the normative rationalities of public administration. Their steering effect comes from outside the law, which is why they entail a future-oriented and situated concept for looking at legislative processes. The highly normative context of law-making restricts the creation and emergence of socio-technical imaginaries. To maintain coherence, new regulation must comply with existing legal rules and foundational legal concepts with their own historical background. Moreover, the law stipulates the procedures and practices of law-making, e.g., in consultation processes, which direct the modes of participation. These modes shape how imagining can happen. Although this domain imposes its own rationalities on those who wish to engage in the formation of ADM in public administration, space remains in which ideals and imaginaries emerge and become entwined with legal rationalities.
Ideals of digital public administration
In public administration, the interest in digital technologies has directed research towards changing values and ideals (Bannister and Connolly, 2014). Although administrative values are entwined with legal rules and structures as discussed, they cannot be reduced to one another. To demonstrate this, we draw attention to public administration studies that have traced the historical and ideological shifts and continuums of digital public administration. Scholars have suggested an ideological shift toward citizen engagement, responsive administration, and public value creation, leading to ‘platform paradigm’, or e-government (Janowski et al., 2018; Margetts and Dunleavy, 2013; Rose et al., 2015).
Rose et al. (2015) provide a categorization of four ideals of public administration that have dominated digitalization initiatives during the 2000s: professionalism, efficiency, service, and engagement. Each includes assumptions about the role of technology and its use in public administration. The perceived usefulness of technology reflects views of what constitutes good public administration, and these views are historically contingent on different value positions and ideologies.
The ideal of professionalism reflects the rule-oriented traditional bureaucracy articulated by Max Weber in the 1940s. This kind of bureaucracy aims for an independent and accountable administration and considers technology as infrastructural (Rose et al., 2015). Technology provides security and compliance with the law, enforcing bureaucratic rules. The efficiency ideal focuses on minimizing the waste of public resources and is associated with cost reduction and productivity, reflecting the value positions of the business-oriented New Public Management (NPM) of the 1980s. Technology serves productivity, enabling rationalization (Orlikowski and Iacono, 2001; Rose et al., 2015). The service ideal reflects the Public Value Management approach, which originated as a backlash against NPM in the 1990s. In e-government, the service ideal turns civil servants into responsive actors and service provision is guided by citizen orientation, aimed at producing social value (Grimsley and Meehan, 2007). Technology improves the availability of public services. The citizen becomes a customer and a subject. Finally, the engagement ideal highlights the interactional relationship between public administration, citizens, and civil society, thus aligning with the values of the New Public Service. Public administration supports the co-production of policy, which sees citizens as active participants. Technology becomes a networking facilitator and a social relations tool (Orlikowski and Iacono, 2001).
We perceive an important shift in the self-understanding of administration. The first two ideals are strongly internal to public administration, as they stress the functioning of the administrative system and the authority of a legitimate professional. The other two ideals, however, operate with an outward-directed and interactional administration vis-á-vis an active citizen. In comparison with sociotechnical imaginaries, these administrative ideals emphasize the social and administrative contingencies that also inform the digitalization of public administration, whereas imaginaries focus on technology and ensuing reimaginations. While the value ideals presented by Rose et al. (2015) provide a retrospective picture of the broader changes within public administration values, recent studies have pointed to a deepening interest in the interactional and personal dimensions of digital bureaucracy. Ranchordas (2021), for example, suggests that administrative empathy recognizes the diversity of citizens, placing the emphasis on the skills of citizens and the possibilities of alleviating the shortcomings of automation.
Contextualizing imaginaries of law, technology, and society
Socio-legal studies elaborate the implications of digitalization on fundamental rights on practical and theoretical levels (Yeung, 2017). We draw attention to two tensions. First, there is an inherent tension in technology regulation between the need to regulate specific technologies while maintaining technological neutrality, i.e., the impartiality of legislation towards different technological methods and applications. A clear scope of application is necessary for legal clarity and predictability, but definitions should remain sufficiently general and flexible to accommodate technological change. Second, research highlights the ever-present challenges of technology regulation: the need and complexity of assessing the sufficiency of existing legislation, its effectiveness in advancing different regulatory goals, and the predicted impacts of new regulation (Bennett Moses, 2013; Brownsword, 2019; Hydén, 2022). Scholars also discuss the normativity of technological structures that guide and force human behaviour in different ways than legal structures. This social ordering and technological normativity emerges from the design choices of programmers and system architects and is distinct from the normativity of law, which obtains legitimacy through the democratic process (Hildebrandt, 2008; Lessig, 1999).
In sociotechnical imaginaries, law is often presented as reactive, as the imaginary of governable technological emergence is the naturalized narrative (Hurlbut, 2015). The idea of retrogressive law can be used to argue for industry self-regulation against government interference, where the latter is seen to hinder technological innovation. However, there is dynamic reciprocity between law and technology, as law structures technological development through various mechanisms, but is simultaneously and continuously shaped by these developments (Cohen, 2019). This entails that technology regulation and legislative processes become central for striking a balance between the contrasting interests of realizing the potential of technological development while mitigating the risks for social justice and fundamental rights (Cloatre and Pickersgill, 2020). Implicit assumptions about technology and human behaviour may become embedded and sedimented into regulation, both inadvertently and consciously (Koulu, 2020; Lindroos-Hovinheimo, 2021). Simply put, law drafting is the arena for imaginaries to emerge, to be negotiated and translated, and is the process in which they become sedimented into legislation as implicit assumptions.
Before proceeding to our research design, methodological choices, and data analysis, we will discuss the Finnish efforts within the public sector to contextualize our study.
Finland: a forerunner of digitalization encountering legislative problems
Described as a high-trust society, Finland has high levels of institutional trust accompanied by reliance on digitalization. In the EU Commission's Digital Economy and Society Index (DESI, 2022), Finland ranks overall 1st and in digital public services 2nd of all the member states. Hence, Finland forms a laboratory for digitalization and associated governance issues. Several governments have promoted digitalization and the currently aim to make digital services available to all citizens and businesses by 2023 (Marin’s Government programme, 2019).
Administrative automation has an established tradition in Finland dating back to the 1980s and 1990s. Several public authorities, such as the Tax Authority and the Social Insurance Institution have implemented their digital systems independently without an explicit legal mandate, based on the perception of digitalization as a technical issue of organizing work practices. The current ADM reform was initiated by the Ministry of Justice in early 2020 in reaction to increased pressure from Parliament's constitutional committee and public oversight authorities. These concerns highlighted problems related data protection in immigration (PeVL7/2019vp), social security and insurance legislation (PeVL78/2018vp), and ADM in tax administration (EOAK3379/2018). In response, the Ministry of Justice initiated its work to include ADM in the Administrative Procedure Act. Legal concerns relate to selection of cases for automation, the publicity of algorithms, good governance/legal protection, and liability.
In our understanding, any attempt to discuss ideals of digital public administration would be too limited without recognizing the role of legal rules and principles, particularly administrative law, and European technology regulation, which provide socio-legal context. Digitalization of public administration must comply with peremptory regulation that explicates the legal orientation of what is and has historically been considered good and desirable in public administration. These principles bring coherence, legal certainty, and predictability to administrative practice over long periods of time and societal change. However, good administration can be interpreted in many ways, affording different ways of exercising public power (Koivisto, 2014).
The principles of good administration are provided for by the Constitution of Finland (731/1999) and the Administrative Procedure Act (434/2003). Once enacted, the ADM reform will introduce new sections on general conditions for automation to the Administrative Procedure Act and new sections on liability and deployment to the Act on Information Management in Public Administration (906/2019). Although administrative law is typically bound to national jurisdictions, most share core values, which adds to the relevance of our analysis beyond national settings. These supranational administrative law principles include i.a. lawfulness, equality, impartiality, legal certainty, and participation, as exemplified by the Code of Good Administration of the Council of Europe (CM/Rec(2007)7; Stelkens and Andrijauskaite, 2020). Alongside transparency and publicity, the Nordic tradition of administrative law emphasizes the service principle, which requires public officials to account for their customers’ needs and inform the public about services and activities. Moreover, the service principle is continuously reinterpreted in legal praxis. For example, the highest instance of legality control in Finland, the Chancellor of Justice, interprets the service principle to include the obligation for appropriate user testing of digital public services (OKV/2019/1/2017).
Despite the different application scopes of national administrative law and European technology regulation, these legal regimes share values. Transparency and access to public documents and recordings are constitutional rights in Finland (Article 12 of the Constitution, the Act on the Openness of Government Activities (621/1999)) but transparency is also the cornerstone of European technology regulation, e.g., Article 12 of the General Data Protection Regulation (GDPR, 679/2016).
There is inherent overlap as well as tensions between EU-level and national regulations, which the Finnish ADM reform exemplifies. One of the drivers for the Finnish ADM reform was Article 22 of the GDPR, which bans individual automated decisions without consent or legal basis in national law. These overlaps and tensions are bound to multiply, as the EU's second digital agenda for 2020–2030 sets out to introduce various new legal instruments for artificial intelligence (AI) and data sharing, some of which are already passed or are in the legislative process. In February 2023, the proposal for the Artificial Intelligence Act (AIA, COM/2021/206) is still awaiting the vote at the European Parliament, but once enacted, its provisions will impose additional compliance requirements for AI systems used in high-risk sectors, including parts of public administration. In addition to directly binding EU regulations, Finnish public institutions must comply with EU directives, e.g., the Web Accessibility Directive and its national implementation (306/2019). These directives include obligations for public organizations to ensure the accessibility, understandability, and user-friendliness of digital public services.
It is relevant for our analysis that the Finnish ADM reform attempts to establish legal rules and principles for digital public administration. From the STS perspective, this means that sociotechnical imaginaries do not operate in a vacuum but are instead entwined with and limited by these legal framings. Ultimately, the law itself embodies ideals and imaginaries that might need reimagining – and which may be reimagined through lawmaking (Brownsword, 2019).
Research design
We focused on the ADM reform's initial stages because law in the making provides an opportunity to imagine futures for digital administration. Studying the pre-parliamentary phase reveals the existing imaginaries on technology and citizens’ roles that guide the renegotiations on the objectives of digital public administration. During consultation, actors can influence the law and the floor is open to interpretations of how the law should be reformed, what technologies should be included, and what is the position of citizens in this change. In fact, the pre-parliamentary phase can be considered more influential than the parliamentary phase when the content and scope of the law are determined (Lepinkäinen and Malik, 2022). The political is embedded in the public consultations, as societal stakeholders, advocacy groups, and civic movements are invited to participate. The outcomes can later be seen in the actual law, affecting what is changed and how the justifications are presented.
Methodology and data
Our primary data were the written statements during two consultation rounds in 2020 and 2021. In addition, the documents related to the law reform were used as background material. We regarded the written statements as critical points because of the openness of the consultation phase before the political parliamentary debate. The statements are a special form of text as they answer questions posed and structured by the legislators. However, statements serve as institutional responses through which the actors shape the public imagination and eventually have an impact on administrative practices. From our perspective, imaginaries also produce silences, as they select certain pathways and create path dependencies when suggesting possible futures.
Our data were collected through two official online services, which contained the documents and the statements. The first included documentation of the law reform and its timeline 1 the second the written statements. 2 All statements are accessible and can be retrieved from the online services free of charge. Statements were written in Finnish, although two stakeholders commented in Swedish.
The first round of statements was requested by the Ministry of Justice in 2020. The content of the request related to the problematizations of ADM in the assessment memorandum published earlier by the Ministry. The request included eight themes relating to the principles of good governance: the current state of ADM and regulatory needs, delimitation of ADM use, ensuring good governance and legal protection, information system regulation and the liability, publicity, and transparency of public decision-making, data protection, and other regulatory concerns. The Ministry reached out to 84 stakeholders, of which 58 commented on the request. An additional seven other actors provided statements online.
The second round of statements was requested in 2021, with narrow focus on draft provisions suggested by the working group of the Ministry, with specific yes-no questions. The citizen perspective was recognized mainly in relation to publicity and transparency, i.e., how administration must inform about ADM. The request itself directed statements towards legal technicalities and allowed less room for deliberation. The Ministry requested comments directly from 106 stakeholders, receiving 58 statements and an additional five statements from other stakeholders.
In 2022 the working group published a report in the form of a draft government proposal, which was merged with the Ministry of Finance's work on information systems and given to Parliament. The ADM reform is currently pending at the Parliament. Figure 1 shows the timeline of the law reform and the data.
Timeline of the law reform process and data.
Analysis
We started by looking at how the statements identified and discussed citizens, customers, and administrative subjects. These terminological passages functioned as entry points to the data, followed by inductive qualitative content analysis to identify, cluster, and categorize topics. This resulted in themes that ran through the data connecting framings of citizens, technology, and administration (Elo & Kyngäs, 2008; Graneheim et al., 2017). Our unit of analysis was a body of thought that included several sentences (a text excerpt). We followed three interconnected analytical phases: reduction, clustering, and abstraction from the data (Tuomi and Sarajärvi, 2018). First, we condensed the original expressions, after which we grouped them, forming sub-categories. These were then combined and collapsed into broader categories. Throughout the analysis we reflected on the similarities and differences between the categories to ensure they belonged to the same clusters. Finally, we formed main categories, resulting in four imaginaries. Table 1 presents a simplified example of the analysis after the reduction phase.
Forming the imaginaries.
The topics most often mentioned in relation to citizens were the publicity/transparency of decision-making and the program code (what the citizen needs to know about ADM and technology), the authority's responsibilities, and notifying citizens about automation (what the administration must tell), all of them being related to administrative law provisions. Additionally, non-legal topics, such as user interface solutions and technological characteristics of information systems, were discussed (what technology does and for whom). Regarding technology, the statements addressed the general aims and risks of ADM and AI use. In the 2021 statements, the focus shifted towards the description by the public authorities (how the administration must say they are using ADM).
As a result of cyclical reading of the data, we formed four imaginaries: 1) understandable, 2) self-monitoring/autonomous, 3) adaptive, and 4) responsible administration. As imaginaries are multiple, overlapping, and contested, we scrutinized how the imaginaries portrayed citizens, technology, and administration, what kind of argumentation was used, and how they naturalized ways of thinking about these interrelations (Jasanoff, 2015a; Mager and Katzenbach, 2021).
Findings
In the imaginaries of digital public administration, our focus is on the envisioned relationship between citizens and public power. In the first imaginary, technology is complex and risky if too much is revealed to the average citizen. The second imaginary discusses technology from the viewpoint of a well-functioning, customer-oriented administration. The third imaginary suggests that flexible technological solutions ensure interorganizational compatibility. The final imaginary frames technology as risky for vulnerable citizens, possibly leading to discrimination. Table 2 summarizes the imaginaries and how citizens, technology, and automation are perceived in them. In the following, we discuss the imaginaries in detail.
Imaginaries of public digital administration.
Understandable administration
The imaginary of understandable administration frames technology as too complex for citizens to understand. The relationship between digital public administration and citizens relies on the notion of an average citizen, emphasizing that the ‘average citizen’ should not be expected to understand technology or decision-making processes. Because of the complexity of automation, citizens should be provided with comprehensible descriptions of how ADM works, but no details on, e.g., source code. Public authorities are responsible for publishing understandable information on automation, but technological specificities are omitted as trade secrets or as incomprehensible to the ‘average citizen’.
The imaginary's proponents represent large institutional actors in Finnish society, such as ministries, industry, and social insurance institutions. By suggesting that automation is hard to understand, the citizen is constructed as ignorant and distanced from the administration by stressing that the ‘average citizen’ and ‘the public’ do not have to know too much. Although the ‘average citizen’ is a generalization, it serves to legitimize ADM, as the following quote suggests: The requirement for transparency in the use of automatic decision-making is particularly welcome. A clear description of the operation of the systems and notification of automatic decision-making is likely to increase trust in automatic decision-making. It is essential that the information is accessible and easy for the public to understand. (Technology Industries Association in Finland, 2020) Disclosure of a program code reveals the technologies, implementations, and potential misconduct controls used in broader cases. Disclosure of this technological background increases the realization of various security risks, because if the potential harm doer is familiar with the technologies used their vulnerabilities are easy to find. Instead of publishing a program code, it would be sensible to limit the published part to, for example, a logical model of the operation of the software or a description of the process or some other higher-level description. These are easier for citizens to understand and do not involve the disclosure of potentially confidential information. (The Social Insurance Institution of Finland, 2020)
As this excerpt shows, code functions as a line of distinction: it separates and defines what is understandable knowledge. However, the commentators did not mention the understandability of administrative processes, proposing that complexity is only related to technology. We suggest that not discussing the decision-making processes omits important questions on the organizational structures in which automation is used, as well as the complexities of the actual decision-making processes. Additionally, protecting the code and concealing it from citizens proves to be a paradox. The code is protected from citizens, who would not in fact understand it in the first place. Although understanding is not expected of the average citizen, he/she should be able to assess the validity of an ADM process and decisions based on general descriptions. Assessment requires skills to scrutinize complex processes with minimum knowledge. The following quote reveals that the issue is not physical access to technology but rather the skills needed to assess the decisions:
The description [of how automated decisions are made] should be accurate and easy to understand. The subject of automatic decision-making must be able to assess, through the description, whether the decision-making process and the final decision have been in accordance with the law. (Association of Unemployment Insurance Funds, 2020)
Drawing a line between understandable (public) information and concealed (private) code is not trivial. In the 2021 consultation, the issue of publicity was specified as a question of whether citizens should be given the name of an official for additional information. Most commentators took the stand that the authorities should provide information on the decision when needed. However, many did not consider it important to specify an actual person to provide that information, partly because the decision itself was automated. An actual person would be unnecessary if the authorities informed the citizen about the automation and provided a description of the process. Again, it was repeated that “the algorithms themselves are not very informative for the average administration client and therefore do not need to be included in the description. A process diagram or description of the automated decision-making process could be informative.” (Universities Finland, 2021)
We suggest that the imaginary of understandable administration operates with generalizations that disconnect the citizen from the administration and, more importantly, from technology. Public institutions are central for defining what is understandable and informative for citizens. The imaginary is strongly connected with transparency ideals and obligations, the core principles of good administration. Compared with Rose et al.'s (2015) ideals, this imaginary resembles the efficiency ideal, and we can speculate that transparency is strongly connected to efficiency in the Finnish context because transparency is thought to improve efficiency.
Self-monitoring and autonomous administration
The imaginary of self-monitoring and autonomous administration portrays a well-functioning administration that is process-oriented and efficient due to automation. To achieve efficiency, public administration should be free to utilize automation and monitor the systems internally as this ensures the materialization of the benefits of automation. A lean and error-free digital public administration ultimately benefits customers because ADM creates uniformity. From this perspective, regulation limits an otherwise well-functioning administration and the use of technology.
To realize full potential, the commentators suggested that the authorities should not be left with control mechanisms that are too burdensome. Instead, as the excerpt below shows, they should use self-monitoring and create uniform decision processes in which customers provide feedback that helps the functioning of the administration: Ex-post control emphasizes the organization's own monitoring at the system level, which must be reasonable in relation to productivity. Suggested means, such as spot checks, etc., are worth considering in further work. Regarding legal protection, the means of redress must be customer-oriented, and the realization of fundamental rights must be ensured at the system level. (Confederation of Finnish Industries, 2020) The benefits of automated decision-making are undeniable: it frees up human resources from routine tasks thus speeding up the processing of other applications. The automation of processing can also ensure the continuity of processing in various crises, such as the corona pandemic. (Finance Finland, 2020)
The reliance on automation was repeated in both rounds of consultation. Those favouring streamlined decision-making processes, considered automation to be a prerequisite for efficient administration: Legislation on automatic decision-making and an understanding of the use of artificial intelligence in public administration are essential preconditions for administrative reform and more efficient operations. For citizens, automation means better services. For Finland, a smooth public administration is a competitive advantage when competing for the locations and experts of operations. (Technology Industries Association in Finland, 2020)
In 2021, the public pensions institution KEVA argued against too limiting technology regulation, stating that “It is also worth remembering that the clients of the administration have the right to have their case processed without delay and the right to equal treatment. In this regard, automation and artificial intelligence can be very useful tools, meaning that the use of technology does not involve mere threats or risks (KEVA, 2021).
High-quality system design, regular monitoring, and correction of algorithms were argued to produce reliability and ensured the legal protection of administrative customers. The imaginary of autonomous administration pictures future administration in which error-free technology outperforms error-prone human decision-making. The control of technology is best organized through existing mechanisms to avoid additional administrative burdens. The customer benefits when public administration decides on ADM use. In this sense, the imaginary of autonomous administration reflects the notion that legislation lags behind technological progress, although this notion is based on an oversimplification of the reciprocal relationship between law and technology (Cohen, 2019; Hurlbut, 2015). Interestingly, this imaginary echoes two ideals of Rose et al. (2015), namely professionalism and efficiency, suggesting that a well-functioning public administration is a global competitive advantage.
Adaptive administration
In contrast to the previous imaginaries, adaptive administration shifts the gaze to the administrative practices that should be reconsidered and developed alongside of technology. Recognizing the diversity of current technological solutions in public organizations, the commentators paid attention to the design of information systems, their validation and testing. However, the commentators were concerned for interoperability between technological solutions of different authorities. The imaginary of adaptive administration connects the development of administrative practices alongside of technology with the equality of citizens and legal certainty.
Adaptive administration acknowledges the increasing use of technology and stresses consistency of decision-making guided by cost-related discussion. However, compared to the prior imaginaries, adaptive administration sees a need to develop interorganizational systems for data exchange. Instead of assuming an opposition between technology and human actors, the commentators suggested crossing the boundaries between organizations that need to communicate with each other. Thus, the statements emphazised that administrations need to develop technologies that suit the practices within public organizations: The situation shows the individual solutions of the various authorities. The need for regulation is justified from the point of view of legal certainty for citizens and for the development of information systems and the control of maintenance costs. As digitalization and automation increase, administrative functions should be re-examined rather than simply seeking to strengthen existing processes with new information systems. (Finnish National Agency for Education, 2020) The situation shows the individual solutions of the various authorities. The need for regulation is justified from the point of view of legal certainty for citizens and for the development of information systems and the control of maintenance costs. As digitalization and automation increase, administrative functions should be re-examined rather than simply seeking to strengthen existing processes with new information systems. (Digital and Population Data Services Agency, 2020)
Even though specific technologies were not categorically rejected, limitations were discussed: “It could be justified that, in well-considered situations, there would be no impediment to administrative decisions by AI in cases where the rights and freedoms of the individual would not be irreversible and easily remediable.” (Digital and Population Data Services Agency, 2020). The suggestion of AI use was repeated in 2021 although it was mostly connected to the needs of public organizations, not citizens. The demarcation between AI and rule-based technologies is important for the rule of law, as AI would only be suitable in assisting decision-making.
The imaginary suggests an adaptive administration that develops its internal practices in an environment of interconnected public organizations and information systems. The future focus is on the work practices of civil servants and not on technology or citizens’ needs. Compared to the ideals of Rose et al. (2015), the imaginary has similarities to the professionalism ideal. Also, the imaginary takes up the design processes but does not problematize them much further.
Responsible administration
The imaginary of responsible administration addresses the diversity of people's skills, competencies, and hindrances. This ‘weaker’ imaginary is based on a small number of statements by civil advocacy organizations. However, it is the only one to focus on citizens’ needs, suggesting a future where automation may introduce disadvantages. Automation is framed as potentially harmful and discriminating. Responsible public administration has a specific duty to protect people in vulnerable positions, drawing attention to the constitutional right of non-discrimination and equality. The imaginary problematizes the premises of technological design reflecting the discussions on administrative empathy, as the imaginary focuses on human as the contact point for administrative processes (Ranchordas, 2021).
Unlike prior imaginaries, the imaginary of responsible administration attributes responsibility to public authorities at the early phases of planning automation. The commentators opposed technology optimism and efficiency, arguing that these should not be justifications for automation: The use of automated decision-making is generally justified by the need to save resources, speed up decision-making, and seize the opportunities offered by technological development. In this regard, we state that the primary criterion for the introduction of automatic decision-making should always be to increase the legal certainty of decisions and to maintain the legal protection of citizens. If these legal benefits might be jeopardized by automatic decision-making, it should not be introduced. (The Central Organization of Finnish Trade Unions, 2020) If the language and administrative culture are foreign, the customer may not receive the benefits or services to which he or she is entitled. This may be the case even though the matter is not particularly difficult to understand […] It may be that the client has been asked to make an application, but the client has not actually understood what is required of him, for example due to a lack of language skills. (The Central Union for Child Welfare, 2020)
Here, technological risks interconnect with the administrative practices unfamiliar to people who use public services. Technology use increases risks by removing the human element that is considered important for vulnerable groups. Responsible administration has a specific duty to advise their clients and consider their individual circumstances, skills, and resources: There are many clients of the Finnish Immigration Service for whom both the authority's duty to advise and its duty to investigate are emphasized, for example due to a lack of language skills and the unfamiliar nature of the Finnish system. There are also a particularly large number of vulnerable customers […] The risks of automatic decision-making in terms of legal certainty and equal treatment are highlighted for the most vulnerable. (The Finnish Refugee Advice Centre, 2020)
Although some acknowledged that ADM may bring about benefits through tailored solutions, the systems can profile people and user groups, making the relationship between customers and public power particularly fragile. Therefore, development was paramount: It is important that different user groups are involved in the design, development, and testing of systems so that the system to be deployed is not based on erroneous assumptions about the needs of end users (Finnish Disability Forum, 2020). Framing ADM as risky raises issues of inclusion because technological solutions may produce discriminatory systems: Especially decision-making regarding social and health services or benefits often involves a complex consideration of the individual circumstances of a person that cannot be outsourced to a machine. The state is a community of people, and the exercise of state power must also give priority to human activity in which a person in need of state help or support is encountered by another person if he or she wishes. (Finnish Federation for Social Affairs and Health, 2020)
Similar concerns were voiced in 2021 when the commentators observed that human interaction was especially important for people in a vulnerable position, for whom administrative processes in general and digital transactions in particular are challenging (Finnish Federation for Social Affairs and Health, 2021).
Whereas self-monitoring administration pictures humans as error-prone, the imaginary of responsible administration considers humans central to ensuring inclusion. The imaginary highlights the service principle of good administration in which public organizations are obligated to help citizens. Combined with vulnerability, the imaginary problematizes technological design, which together materialize possible discriminatory practices. The involvement of users in the system's design is suggested as a solution. Although the imaginary highlights the service ideal (Rose et al., 2015), it does not discuss engagement but rather the specific needs of vulnerable groups that require supportive human contact.
Discussion
We have asked how is good digital administration envisioned and what are the problems ADM presents for the relationship between public power and citizens? Based on our findings, we argue that good digital administration is defined from the perspective of public authorities and the Administrative Procedure Act reform serves primarily to legitimize existing automation practices instead of truly renegotiating digital futures. The absence of citizens’ perspectives raises the question whether the relationship between public power and citizens is becoming a one-way street despite citizen engagement ideals (Rose et al., 2015). Although all imaginaries address what citizens need to know and how public authorities must inform them about automation, the focus shifts to the formal descriptions of the public organizations.
The first three imaginaries assume an institutional approach, while the last imaginary considers digital public power from the perspective of vulnerable groups. The fourth imaginary provides a counter-imaginary to the others, which are dominated by technological optimism and operate on the ideals of traditional bureaucracy and new public management (Lepinkäinen and Malik, 2022; Rose et al., 2015). Instead of renegotiating the relationship between public administration and citizens, the imaginaries rely on traditional ideals of effectiveness based on automation.
As a result, ADM systems appear to be unproblematic to citizens, who are mostly missing from the picture. This is paradoxically associated with the notion of the average citizen. Prior studies have pointed towards similar reductions of citizens to ‘users’ (Broomfield and Reutter, 2022). By sketching the contours of the average citizen, imaginaries render technology complex but unproblematic, as its use is overseen by public authorities. When citizens are recognized, they are portrayed as disadvantaged groups without the skills or resources to defend themselves. Simply put, problems emerge only for those individuals who need special assistance. For the average citizen, technology does not impose difficulties, although, in the worst-case scenario, citizens might abuse the system. Therefore, imaginaries produce an ambivalent notion of the citizen. This notion is a dichotomous categorization of the average citizen and the vulnerable citizen, creating a distance between them, technology, and administration. Simultaneously, fear of citizens gaming the system forms a tension between technology and humans, as codes need to be protected from those aiming to misuse them.
Contextualizing imaginaries of law, technology, and society
After our analysis, the ADM reform proceeded from the pre-parliamentary phase to the government's proposal, which was given to the Parliament in September 2022 (145/2022 vp). The proposal initiates the parliamentary process and introduces new sections to existing administrative legislation. The proposal provides a vantage point for examining how the imaginaries have materialized during the law-making process. Although the wordings may change during the process, government proposals have descriptive as well as prescriptive authority as legal sources according to legal doctrine.
As stated, imaginaries differ in how they portray digitalization efforts that require regulatory action and concerning which legal principles they prioritize. The proposal includes a description of the policy problem and interpretation guidance through which we can examine which imaginaries have become decisive and how tensions between various ideals have been reconciled. The proposal describes the need to establish a legal foundation in accordance with Article 22 of the GDPR to allow automated decisions in public administration. Rather than initiating a political debate about the acceptable limits of digital public administration, the proposal aims to provide legitimacy for the current automation practices instead of questioning them. One could argue that this stance reflects, if not techno-optimism, then at least a neutral understanding that automation is a natural development in administration. At the very least, the naturalized acceptance of automation stands in contrast with the EU's ongoing regulatory initiatives.
The objectives are complemented by the more detailed justifications of new sections, as these seek to safeguard constitutional rights and the rule of law through provisions for appropriateness, transparency, accountability, and liability. The legal basis is established for rule-based ADM systems, but AI systems are excluded from the scope – and thus also from use. Transparency obligations include notifying citizens about automation. Accountability and control are provided by provisions on documentation and testing, quality control, and oversight, which are performed at the time of deployment.
All the imaginaries can be found in the proposal, although some of them more strongly than others. The understandable administration imaginary has become the most persuasive in the proposal, as the ideals of transparency and publicity are strongly present in the formulations. The imaginary is materialized in the obligations to make an official and public administrative decision about ADM deployment to an oversight body and to inform citizens about the automated handling of their case. If the proposal's formulation is accepted, the citizen's right to be informed requires that there should be a reference that the individual decision was produced automatically, as well as information about where the public institution's deployment decision can be found. The public institution is also required to publish the deployment decision on its website.
In contrast, a less explicit role is given to the self-monitoring and autonomous administration imaginary, which emphasizes that regulation hinders innovation. The narrative can be found in the overall logic, which lacks criticisms of increasing automation. Yet the autonomy of individual public institutions is limited through oversight mechanisms. The proposal includes provisions on the development and deployment process and grants an oversight board the authority to make inspections and require amendments when documentation is insufficient. The adaptive administration imaginary emphasizes equality and legal certainty but focuses on interoperability between public institutions. The interoperability perspective can be found in the provisions that harmonize development and deployment practices and new oversight functions.
Finally, the imaginary of responsible administration is the only one that is detached from the institutional perspective and recognizes the diversity of citizens, shifting the focus to service provision. Unfortunately, this diversity is mostly missing in the government proposal. The proposal recognizes certain vulnerable groups, such as the elderly and children, people who do not speak the official languages, and people with disabilities. Their rights and needs are argued to be improved by automation, as automation, according to the societal impact assessment, decreases the likelihood of human error and improves overall equality. Hence, the critical viewpoints of the imaginary have evaporated.
What is lacking in the proposal? What are the silences and problematizations that are left outside its scope? It seems that pressure was high to legitimize already existing automation. We can speculate that the need to legalize the status quo limited the potential for discussing alternatives. The ideals of the last imaginary could have provided an alternative, an emphasis on provision of public services, the diversity of citizens, and detachment from the technological focus. Instead, the proposal perceives automation to be relatively unambiguous and as something that improves legal protection by shortening the processing time. The social impacts of automation or the new ADM regulation are not assessed. A stronger focus on developing public administration and the welfare society could themselves have provided a different view of the advantages and disadvantages of automation.
Given the strong focus on the legitimation of existing automation, it remains to be seen whether the unproblematic nature of technology prevails in the upcoming parliamentary debate. If this is the case, we fear that the opportunity to discuss the relationship between public power and citizen has been missed. At the same time, it seems that the legal policy debate on digitalization threatens to lead to a one-way exercise of power in which citizens cannot participate. Despite the visions of participatory design and support for the digitalization of public services, citizens continue to be the object of digital power and are not an active party in the relationship. Finally, the ADM reform is a legal practice that operates in the language of law, which is tied to key legal markers, i.e., the limitations and fundamental rights found in legislation. Therefore, these legal markers determine the debate as legal principles are binding and officials need to fulfil them or face sanctions. The added value of imaginaries is that they make it possible to look at this discussion from the outside and consider the strong points without being bound to the law.
Conclusions
We shall now draw three conclusions and discuss the limitations of our study. First, the ADM reform is pending in Parliament, which is the venue for political debate on citizens’ as well as the administration's needs. However, the pre-parliamentary phase has eradicated problematic viewpoints, removing citizens from the picture. It remains to be seen if these perspectives will re-emerge in the parliamentary proceedings. Second, the imaginaries render technology unproblematic, which is in direct contrast to the reform's origins, initiated by the observed problems of automation and the lack of legal foundation for current automation practices. Arguably, such a lack of problematization is inherent in all legislative action and reflects law as a legitimizing social force. This non-problematization of ADM, however, is particularly worrying considering the specifics of the Finnish context, i.e., established tradition and the hegemonic discourse that emphasizes the inevitability of digitalization. These concerns are also relevant in the context of European technology regulation. Third, the lack of problematization makes it difficult if not impossible to raise sceptical viewpoints later.
Our contribution to sociotechnical imaginaries is analysing the context of law-drafting, which has not received much attention in prior literature focused on media texts (Natale and Ballatore, 2020; Scott Hansen, 2022) and policy analyses (Tidwell and Tidwell, 2018; Tupasela et al. 2020). Our approach complements public administration research by its orientation on imagining different potential futures. By identifying sociotechnical imaginaries of law in the making, our approach also contributes new empirical and theoretical knowledge to socio-legal studies, making it possible to uncover and question implicit assumptions about technology as they become sedimented to legal architecture.
Naturally, there are limitations to our study. We have focused on the official pre-parliamentary law-drafting process. The data can only provide a temporally limited snapshot of a long and complex process that takes place before and after the law is enacted. Imaginaries continue to emerge and materialize, although potentially in more limited ways as the debate proceeds to legal details. Imaginaries may also materialize differently depending on implementation strategies and actual ADM practices in different public authorities.
For future research, both regulatory studies and policy analysis, the Finnish example provides a landmark as the country is considered internationally a stable, trust-based welfare state that has invested in digitalization. Although the law reform works as legitimation of current practices, it draws attention to the fragmented development of automation ‘in the shadow of the law’. For future study, we suggest interdisciplinary work on law in the making that does not take for granted unproblematic and naturalized assumptions about digitalization and thus opens new spaces for critical perspectives, particularly those of citizens.
Footnotes
Declaration of conflicting interests
The second author gave statements in the consultation process. These were omitted from the data and analysis. The analysis was carried out by the first author only.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors have received financial support for the research from the Strategic Research Council at the Academy of Finland, grant numbers 335442 and 341434.
