1 Introduction

Work is a part of life, but life is more than work. The introduction of computers in work has moved on into almost all other parts of life and are now completely weaved into our ways of living, in work as well as in all sorts of everyday activities. As the call for papers for this special issue says: “It is increasingly difficult to keep work and life separated … [W]ork tasks, personal tasks and leisure tasks blur into each other” so that “the neat distinction between which activities are work-related and which are not is becoming less and less meaningful”. However, many everyday activities are neither part of (paid) work nor entertainment and leisure; many everyday activities are in fact work tasks that used to be carried out by paid workers but are now performed by consumers or clients (like self-service shops, home banking and home care ). In this paper, we focus on the work that is delegated to citizens as public services become digitally automated: doing your civic duties is a kind of private activity, which is similar to both work and using consumer services.

The topic of the paper is the work involved in doing civic duties, and we use tax as an illustration. We discuss how this work can be supported by IT and what happens when it is automated. Automation of public services relieves the citizen from some work tasks and introduces some new. However, the automation simplifying the citizens’ tasks may also hamper their understanding of the public services and hence their possibilities to participate in democratic processes in society. Design of digitally automated public services therefore has to address a wide range of user needs and use contexts based in citizens’ life situations in private spaces as well as the work settings of the public service providers.

The paper is structured as follows: in the next section, we introduce our view on work and automation of work tasks as a basis for characterizing doing civic duties as work-like tasks. The next section presents a case study investigating the problems citizens experience when carrying out a civic duty. The following section discusses the challenges for design suggested by the findings from the case. We draw some lines to the larger picture in the following discussion and conclusion.

2 Work and the automation of work tasks

We draw on recent discussions in computer-supported cooperative work (CSCW) in order to get a better grip of the concept of “work”. Schmidt [1] discusses work as a socially organized activity characterized by its “purpose and circumstance” [1, p. 373]. “[I]n a system of social division of labour, constraints and requirements, etc. [work tasks] are typically externally defined, by other parties” [1, p. 373]. In trying to distinguish what is work and what is not, Schmidt refers to a distinction suggested by Urmson [2] between activities that “would be counted as ‘work’ in all standard contexts and those, which would be called ‘work” only for some purposes” [1, p. 374]: for a gardener gardening is work, but for an amateur garden owner, gardening is a hobby or just recreation. Building on Urmson [2], Schmidt suggests seeing work-like activities like gardening as work as they are “necessary or useful in a practical way” [1, p. 375] for having a well-kept garden. Schmidt refers to work-like activities as “secondary cases of work” also requiring “effort and concentration” in ways that are similar to “prototypical work activities”. What is work and what is not therefore depends on the purpose and circumstance, as well as on its similarities with the professional conduct of the activity in question.

Work practice is not just the “situated doing” of working; the concept refers to following “generally accepted principles of procedure” indicating that a practice is not a personal thing: it is shared and may involve collaboration with others who also expect regularity and share the criteria for “correct conduct” [1, 3]. The concept of practice is hence associated with concepts like skills, knowledge, experience and techniques. There is a normative dimension of the concept of practice suggesting that things should be done in the right way, also indicating that one can exercise the skills and the techniques involved and become a good practitioner.

Another take on work is represented in the work of Strauss and colleagues, who studied work in different contexts, including hospitals [4] and homes [5]. Their understanding of work includes many analytical levels where the larger “trajectory” and “arc of work” gives the frame for understanding how “primary” work tasks and “articulation” work tasks are weaved together [1, 4]. Taken together these theories help us understand work as tasks that are part of a larger, socially organized activity where the tasks can be carried out by different people and distributed over several contexts, all tasks contributing to the purpose of the overall activity. The work tasks can be part of different people’s practices and should adhere to procedures and socially agreed norms for these practices.

Within HCI, we see a broadening of use contexts and application areas in the notion of “the third wave of HCI” [6] contrasting earlier “waves” concerned with computers in work and moving from “human factors to human actors” [7]. Moving from workplaces into homes also implies shifting from located to mobile activities and from the desktop computer to networks of applications and devices [6]. In addition, the epistemologies change towards an emphasis on embodied interaction as a way to characterize HCI in the home [8]. However, the third wave co-exist with the second wave [6]: work is an important part of our lives—after all, most people in the world work for a living [1].

Exploring the characteristics of the third wave can be used to investigate what kinds of activities go on in the home. Bødker claims that “The second and the third wave seem to be stuck on either side of the divide between work on the one hand and leisure, arts, and home on the other, between rationality on the hand and emotion on the other” [6, p. 6]. Her view fits with other researchers seeing work as different from non-work. Sengers [9] describes leisure as time outside of work, where no “serious attention” is given to it; she describes leisure activities as consisting of a “mindless evening in front of the TV or Playstation” [9, p. 20]. Leisure is seen as having very different characteristics from work [10, 11]: while work life is focused on maximizing efficiency, leisure time is focused on the opposite. Sengers [9] argues that computer science has a Tayloristic basis that prevents fun and artful designs that trigger complex, emotional experiences in people’s leisure time.

The main dimension in the discussion about second and third wave is rationality—emotion: work as purposeful and efficient, leisure as “mindless” or consisting of rich and emotional experiences [9]. There seems to be nothing between fun, mindless leisure and effective, Tayloristic work [9]. And Bødker [6, 9, 12] appears to agree with this polarized way of characterizing work and leisure experiences.

The second dimension in which the second and third wave HCI differs is location; when people can work from anywhere, the workplace is no longer a characterizing difference between work and leisure. Moreover, work is often carried out on the move [13, 14] and not by sitting still at a desk [15, 16] In addition, many different kinds of work are carried out in the home: paid work, care work and housework [6, 17–19] to mention a few. Even when discussing working from home, Bødker [6, 9] seems to agree with the polarization of rationality and emotion emphasized in the first dimension.

A third dimension of the third wave HCI is the move from one desktop computer to a network of applications and devices, where one cannot see the design of a product in isolation. Every “thing” needs to be designed as a part of a larger network or system, where the user’s many meetings with products and services should be seen as part of a larger “journey” [20]. Bødker’s [6, p. 6, 8] concern is that the focus on art and consumer products in the third wave does not pay the same kind of attention to the users as the second wave did. “The emerging third wave seems strongly tied to a kind of consumerism that differs from the underlying co-determination framework” (of the Scandinavian societies) applied when designing for work contexts. Bødker asserts that one of the challenges from the third wave is to reach out to the home and private use of leisure-oriented consumer products. Our concern is that even very user-oriented approaches will be limited if the users are only seen as consumers.

In this paper, we focus on people as citizens rather than workers or consumers. As citizens, we have responsibilities and rights, and our focus here is on the “work” required to be a citizen. Some civic responsibilities are voluntary, like voting (which is a much appreciated right), and some are mandatory, like paying taxes. In addition to paying the taxes, providing personal data (in a tax return form) for the taxation calculations to be carried out by the tax authorities is also mandatory. We argue that civic duties are purposeful and work-like [1] and require a rule-based approach and a rationalistic cognitive effort to be carried out [21].

CSCW also discusses work compared with leisure and ludic activities. As CSCW has broadened its scope to also include studies of collaborative activities outside of work settings, there is a discussion whether the concept of “work” is a too narrow framing of CSCW today. Crabtree et al. [11] use ludic activities (leisure and gaming) to demonstrate how concepts from work (like routines, distributed coordination and awareness, constant interruptions, local knowledge and surreptitious monitoring) are useful for analysing and characterizing collaborate gaming. With the discussion about the second and third wave HCI in mind, one can debate whether gaming is purposeful practice and hence work-like. However, if we include the facilitators in our discussion: the people whose job it is to arrange and facilitate the game [11], we enter a more complex, socially organized activity where some people carry out paid work tasks and where other work tasks are distributed and delegated to people as well as to technology. If the game is part of a social happening at work where co-workers participate, the game is actually work also for them.

It is this more complex organization of work that we address in this paper. Public services often need information provided by the citizens for the public agency to deliver its services. As public services are increasingly digitized, more and more of the civic duties are delegated to computer systems doing tasks and supporting citizens in carrying out their part of the work [22]. New Public Management reforms require citizens to be “more self-directed and to take more responsibility for their welfare service, something which in turn has been central in reducing administrative bureaucracy” [22]. The relationship between the government and the citizens is therefore “increasingly … based on reciprocity” [22] indicating that citizens are enrolled in the government’s organization of work. We see the same enrolment of customers in commercial services, e.g. banking or health care, where the flexibility introduced by the possibility of self-service or doing it yourself anywhere (at home or on the move) and anytime according to your own schedule is seen as a benefit for customers even if they then have to carry out work tasks which was previously performed by professional employees [17].

This shift towards self-service also has the effect that “the responsibility—and risk—of obtaining adequate knowledge about one’s services is increasingly placed at the hands of citizens” [22] or customers. When the work task is a simple transaction [23], obtaining the necessary knowledge is easy; however, in complex services this is more difficult. Breit and Salomon [22] argue that the complexity of the governmental system surrounding a public service makes it very difficult for the citizens to obtain “adequate knowledge”, and this may increase the digital divide in society between those who obtain such knowledge and those who don’t. This makes the human call advisor an essential part of a digital public service providing the contextual “adequate knowledge” for interpreting and carrying out the delegated work tasks in the correct way. According to Cavanagh [24], “the self-service paradigm actually reduces task efficiency” in situations where the “self-service technology still requires the intervention of a trained expert”: such situations “are the result of poor design or inadequate performance support technology” [24, p. 469].

Digital public services—e-Government—assumes that computers can perform work tasks and support citizens in doing theirs by presenting enough information to the citizens for them to have “adequate knowledge” for doing their work in a correct way. This position is not new. Since the early days of industrialization, routine work tasks have been automated as a way to make production more efficient [1]. Mechanization of work tasks has taken over routinized tasks but also changed the arc of work to include both humans and machines [1, 6, 25]. Computers make possible automation of both physical and symbolic tasks basically by automating decisions [26], expanding the area for automation to include larger parts of the arc of work. Hence, human workers have to master a larger variety of manual, automated and semi-automated tasks in their everyday work practices [25]. But if both “psychomotor performance skills and cognitive recall” are delegated to computers, the human workers may be “losing the ability to function in key tasks without assistance” [24, p. 460].

Automation often translates complex activities into tasks that are easy to automate, leaving those that are difficult to automate to humans [17, 25]. Work tasks may be split and distributed; an example from health care is to split “watch over” a patient into automatic tracking of movements and human caring [17]. Delegating tasks to machines often requires people to do “machine work”, i.e. work to make the machinery work [5, 17]. This often creates new or additional forms of work (in particular articulation work), and people are required to do things in particular ways and be strictly correct [17, 24]. Automation hence may also increase the workload for some people.

Automation is a process of executing rule-based tasks that can be described in a pre-determined fashion. Automation of services often delegates some work tasks to a computer and the remaining work tasks to the customers or clients. If we take a closer look at automation, we see that automation often covers the routine tasks and leave the non-routine tasks to be executed by humans [25, 27]. When those tasks that can be automated are automated, only the non-automatable tasks are left for a human operator. These tasks are not easily described in a rule-based way and cannot be pre-determined. Also in situations where the automation does not fit, e.g. in a crisis or when deviation from routine occurs, the human operator will have to take over. Critical situations are by definition not pre-determined. Bainbridge [25] argues that the human operator will not be able to learn from running the system during its routine operations, but still has to step in when a more complicated situation arises. Hence, the human operator does not have intimate knowledge about the development of the process until the point where the automation resigns (Bainbridge’s analysis is based on industrial processes that previously were run by human operators). Bainbridge [25] discusses “the ironies of automation”; humans have to step in and do the complicated parts of the operations manually, but automation may reduce their abilities for carrying out these tasks because they have no practice from routine situations.

The basic idea of self-service is “to take responsibility away from trained employees and place it directly into the hands of uncompensated customers” [24, p. 460]. In digital public services, the citizens themselves carry out parts of the governmental services in cooperation with or with the assistance of automated parts. “When consumers take responsibility for the delivery of the service, as well as their own satisfaction, they become genuine ‘co-producers’ of the service” [24, p. 464]. Designing practical solutions for the citizens, where benefits from the efficiency of automation can be combined with a space for action and possibilities for learning for the citizens, is a challenge. Automation makes it possible for the citizens to be passive and not pay attention; hence, a performative effect [28–31] may be that no learning or development of knowledge takes place. Automation of a public service like tax is aimed at efficient procedures for the public agency and at the same time simplifying the tasks for the citizens. A challenge is also to support “citizen overview, learning, and autonomy” [27, p. 99] through the design of the larger system of automated and manual tasks.

3 Doing taxes in Norway

Our research is about doing taxes as private, administrative work that many of us are doing from home [27]. Some of the tasks constituting “doing taxes” have always been carried out by citizens, but in Norway today many tasks are automated. The automation is based on a complex system involving employers, banks, social services and citizens all providing information for the automatic calculation of personal tax. However, even after automation is introduced, citizens are still required to do some work reporting personal information and, in some cases, providing documentation. All citizens also have to check the pre-completed tax return form that the Tax Administration provides. Doing taxes is intellectual work that requires some formal and mathematical thinking, skills that require extra cognitive effort in-between daily activities—skills that we are not equally set up with [21, 32]. Few citizens have any formal training or education in tax calculations. Citizens that have questions when they do their taxes can call the tax authorities’ call centre. The call advisors answer telephone calls that are organized in phone lines corresponding to different tax areas. The advisors provide general information about tax rules and regulations and give personal advice and guidance in how the callers can handle their tax issues. They often refer the caller to the online self-services and often walk them through them if they need assistance.

3.1 Co-listening to the citizens’ questions

There are approximately two million telephone calls to the tax authorities’ call centre per year in Norway, which has a little more than five million inhabitants. In order to find out why the almost fully automated tax service brought about such a high number of calls, we wanted to study the requests to the call centre and how the advisors handled the requests. Our main method for getting access to the citizen’s private work has been to co-listen to the telephone calls to the tax authorities’ call centre. Co-listening is a new method to gain insight into citizens’ private issues [27].

Our perspective is that of the citizen, but we have also looked at the work practices of the advisors. Our aim is to gain knowledge about how the automated tax supports the citizens with the purpose of designing better support for citizen’s autonomy in tax affairs. Our research challenges taken-for-granted assumptions about citizen autonomy and internet usage, and we position our research within a critical research paradigm [33]. Critical research aim to understand power relations and conflicts of interests, and takes an emancipatory stance towards less privileged groups in society [34].

We have listened in to 474 calls, sitting beside the advisor at the advisor’s desk in an open office landscape (Fig. 1). The advisor uses a wireless headset with a microphone for phone calls and a PC for looking up information in the many internal agency databases (Fig. 2). We used the same kind of headset but took care not to use the microphone. Many advisors make notes on paper during the calls and use small handheld calculators during the phone calls (Fig. 3). Co-listening gives insight into the citizens’ own descriptions of their tax issues, information in the databases, and the advisors’ responses and interpretations of the issues.

Fig. 1
figure 1

Call advisors sit at a desk in an open office landscape

Full size image
Fig. 2
figure 2

Advisor uses a telephone and a PC, wearing a headset with a microphone

Full size image
Fig. 3
figure 3

When needed, the advisor does calculations on a small handheld calculator

Full size image

Co-listening is a routine activity within the call centre for purposes such as quality assurance and training of advisors. All personnel doing co-listening—including researchers—sign the standard non-disclosure agreement of the Tax Administration. The call centre does not tape record calls. For privacy reasons, we did not record the calls for our research; hence, our data are paper notes and therefore includes few verbatim quotes from the calls. Most calls were described in a brief and simplified way during the co-listening situation. Many calls follow a quite similar pattern at least in parts. This made it easier to take note of (what we thought was) the most important information, bearing in mind that the filtering of information during long calls gives priority to what strikes our interest then and there. The necessity of simplifying and shortening the notes implies that some analysis was done while taking notes. Initial analysis started at the same time as the data collection. Analysis and data gathering went hand in hand and mutually influenced each other, as is common in interpretive research [35]. In-between the calls, the advisors often explained what they did and how they reasoned, which explained some of the issues that lead to the calls. These explanations helped us understand what the call was about, how the advisor interpreted the issue described by the caller, why the issue occurred and how the advisor responded to help the caller.

In addition, we carried out participant observation when present at the call centre. We also carried out 15 interviews with call centre advisors and managers, and with people in other positions in the Tax Administration, in particular employers working with online services. One employee working in a tax-related NGO was also interviewed to see the citizens’ challenges from outside the tax authorities. The interviews were recorded, transcribed and analysed in order to understand the advisors’ work practices and their opinion of the callers’ challenges. The interviews confirmed and enhanced the data from our co-listening. Documents like annual reports and strategic documents for the call centre and the Tax Administration were analysed in order to get a better grip of their responsibilities, challenges and dilemmas. We discussed the goals of the call centre (described in strategic documents) with the advisors and managers. The field notes and interviews were transcribed and coded to extract the challenges of the callers and how the advisors helped in different ways. Early analysis informed our further field work and guided later analysis [35]. Moreover, both authors have a long-standing professional relationship with the Tax Administration and personal experiences of doing taxes, and these experiences added to the formal data collection [36].

3.2 About citizens doing taxes

Until the simplified tax return was introduced in early 1990s, adult Norwegian citizens were responsible for gathering tax-relevant data about themselves and filling in their tax return form. On this basis, the Tax Administration executed the taxation calculations and generated a tax settlement notice that was issued to the citizens. There was a clear division of tasks and responsibilities between the citizens and the tax authorities. However, during the last 20 years, the Tax Agency has automated data gathering and tax calculations for employees and pensioners with the purpose of simplifying data reporting for the citizens and increasing the quality of the data while at the same time running a more efficient taxation process. The tax return process has become highly automated, to a degree where many citizens do not need to do anything to legally hand in the tax return form. However, the responsibility for correct reporting of data and figures is still with the citizens, but the tax authorities carry out the practical tasks of gathering data from banks, employers, finance institutions, public agencies and municipalities.

Interestingly, the automation of taxes took place while the Norwegian Government issued a white paper advocating “An Information Society for All” [37]. Taking this ambition seriously requires that all citizens can understand the basis for their taxes and be able to argue and complain if needed. Being able to handle one’s taxes is a matter of personal autonomy in economic affairs, but also has a place in the larger picture for the informed citizen in a democratic society.

Verne [27] has analysed the calls to the tax authorities’ call centre with the objective to understand the citizens’ challenges with doing taxes. She arrived at eight types of challenges that trigger a phone call with a request to the tax authorities, as illustrated in Fig. 4:

  1. 1.

    Circumstances in the lives of the citizen. A citizen can have a demanding life situation, or changes in the life situation trigger a need to check up changes in the taxation.

  2. 2.

    The “shape-sorting” box. Identifying and finding the right taxation category for incomes, estate, capital and deductions that correspond to items in forms, as well as finding the right name for a form or a document, similarly to category work described in Bowker and Star [38].

  3. 3.

    Using the online services. Callers do not find information, and the services are difficult to use.

  4. 4.

    Internal structures. Citizens’ requests cannot be carried out because of internal production deadlines, responsibilities etc. in the tax authorities’ organization.

  5. 5.

    Technical issues and anomalies. Built-in defaults and security measures can appear to the caller as technical errors. Both user errors and technical errors can occur and be difficult to differentiate.

  6. 6.

    Manual tasks and documentation. Some deductions need to be claimed and documented manually, and missing figures may need to be filled into the pre-completed form.

  7. 7.

    Laws and regulations. The rules and regulations need to be explained, interpreted and adapted to new situations for the callers.

  8. 8.

    Interactions with and between third parties such as municipalities (a), employers (b) and other public agencies (c) may complicate the situation for the citizen [27].

Fig. 4
figure 4

A simplified illustration of the eight types of challenges that can come between the citizen and a correct application of tax laws and a correct tax assessment. The numbers in the figure correspond to the numbers in the list above

Full size image

The main reason for citizens calls is to ask the tax advisors to help them find out and fit them into the tax rules and regulations. The second major reason for calling is to get a confirmation of their own understanding and interpretation of the rules and regulations [27]. Some citizens find the online self-service not possible to use and they need help from a human advisor in order to maintain their autonomy [22]. The advisors told us that many citizens struggle with the self-service solution, and surprisingly many young people call instead of using the digital service. One advisor commented that “[t]he winners are those who have used the old paper-based forms”: they have a better understanding of the tax system. We have seen that the tax advisor “disentangles” the caller’s question to create a space for action where the caller him/herself can take steps to solve the issue within the sociomaterial entanglement of doing taxes [39].

By seeing the callers as non-users of the Tax Administration’s online services, we argue that the advisors provide many different kinds of support for the autonomy of the citizens [40]. When citizens seem to be able to order a new tax card themselves the advisor help them to navigate the digital service and do it themselves. In contrast, when the advisor thinks that the callers are not able to handle the matter themselves, s/he provides more personalized help, e.g. orders a tax card directly from the internal databases based on the input from the caller.

The government needs to communicate with all citizens, also those who struggle with doing their civic duties. The call centre is enacting a relational approach with the aim that all citizens should receive the advice and help they need on an individual basis depending on their life situation and circumstances. The advisor and the citizen do the work of sorting out the taxes together.

4 When automation is not enough

The automation of taxes has changed the tasks constituting the work of doing taxes considerably. Our analysis emphasizes four kinds of tasks that the citizens still need to carry out manually in order to do their taxes. They are identified as old and new tasks, inside and outside of the automation.

In the following, we illustrate how automation influences and changes the citizens’ tasks of doing taxes. Figure 5 illustrates the task area of doing taxes, where the boundaries to other events and tasks in the citizen’s life are loosely defined. In Fig. 6, automation is introduced. The automation is rule-based, and these rules do not always cover the task area completely. There are some residual tasks left for the citizen to carry out.

Fig. 5
figure 5

Doing taxes is a loosely defined task area for the citizens. The boundaries are not clearly defined

Full size image
Fig. 6
figure 6

Automation does not cover the task area completely, and some residual tasks are left for the citizen

Full size image

In the next sections, we present our analysis of how automation changes the citizens’ tasks of doing taxes based on the four combinations of old and new tasks, inside and outside of the automation.

4.1 Old tasks covered by the automation: redundant tasks

The tasks that are covered by the automation are made redundant. A redundant task is a task that the automation carries out that used to be carried out by the citizen. For example, citizens often call to ask for confirmation that they do not need to submit the tax return form manually, as illustrated in the next example.

Example 1: The caller asks “Do we have to send in [the tax return form] or can we refrain?” Advisor Jan ToreFootnote 1 replies that if you agree with the pre-completed form you can refrain. This is ok for the caller. (20110427-10)

Redundant tasks can be carried out voluntarily by citizens, for instance by manually submitting the tax return form even when it is not strictly necessary. Another example of a redundant task is to provide documentation for an interest deduction, which is reported automatically by the bank.

4.2 Old tasks outside the automation: residual tasks

The Tax Administration’s automatic gathering of tax-related information does not cover all kinds of information from all sources. There is still information that the citizen needs to provide to ensure a correct taxation. For instance, some deductions need to be claimed explicitly, and the citizen needs to provide documentation for these claims as illustrated below.

Example 2: The caller asks if diabetes type 1 will give the special tax allowance. Advisor Nils answers that this does not happen automatically anymore. He tells the woman calling to get a doctor’s certificate stating that her diet is regulated, and advise her to specify and show that the expenses are probable. He explains that undocumented special tax allowance for diabetes was terminated 4–5 years ago. He adds that it is ok that she forwards the doctor’s certificate if she already has submitted the tax return form. (20100428-19)

In this situation, the caller needs to fill in information manually in the tax return form. Values of registered cars, stocks, houses and other kinds of capital are usually reported and filled in automatically by the tax authorities. However, in some cases, for example when a car is new, this value is not pre-filled in the form.

Example 3: The caller says that she is filling in the tax return form online right now. She calls about the value of her car. Advisor Berit looks up the standard price table for cars etc. and replies that 30 % of the original price is used. The caller says, “Thank you!” and hangs up. (20100430-2)

A residual task is a task that is not covered by the automation, for example to provide information that is not reported from third parties, see Fig. 6. Failing to provide information that is not gathered automatically but should be provided manually by the citizen can in the worst case lead to penalty tax issued for an unaware citizen. A citizen needs to know which tasks are residual for them to carry out.

4.3 New tasks inside automation

The automation implements default rules that do not hold in all situations. Sometimes the citizens will need to override the default values. The following example concerns the tax card, which specifies the monthly deductions from the citizens’ wage payments.

Example 4: “Are there taxes on the minimum state pensions?” The caller cannot remember her National Identity Number, but says her name and birth date. Advisor Morten looks her up in the Population Register, retrieves the National Identity Number and looks up her tax card. He sees that her tax card specifies a 2 % advance tax deduction, but as a minimum state pensioner, she is eligible to an exemption card. Morten explains that only 85 % of the minimum standard deduction is included in the calculations determining the tax card. He makes some changes in the database and issues a new tax card. Morten calls this “stupid mathematics”. (20120111-18)

Advisor Morten overrides the internal default, which is a task the citizen does not have access to. If the citizens encounter technical errors or anomalies in the services, they will need to do extra tasks. Using online services may introduce additional complexity [41].

Example 5: The caller says “it concerns field 4.8.2 on the form, it needs to be added in the online tax return form”. Advisor Per tells him that he must create it himself in ALTINNFootnote 2 by choosing it from a drop-down list. The caller has tried but cannot find it. Per can find item 4.8.3, “is this it?” Per logs on to his own personal ALTINN tax return form, and even he cannot find item 4.8.2 in the drop-down list. Per suggests that the caller uses item 4.8.1 instead and explain why in the comment field. He asks for the telephone number of the caller so that he can look into it and call him back. After the call, he looks closer into the matter, but still he cannot find any explanation for why item 4.8.2 was not visible in the ALTINN form. (20110429-31)

Per suggests a workaround when an item was missing in the drop-down menu. The workaround is perfectly acceptable tax-wise, but the technical anomaly leads to extra work for the caller. Such new tasks can be carried out by the callers themselves by using the online self-services or with the help of the call advisors. Often the callers need the advisor to interpret and disentangle the situation for them to proceed with their doing of their taxes.

The new tasks are indicated in Fig. 6 by the automation extending outside of the original task area.Footnote 3 The new task is marked with the little circle inside the automation in Fig. 7.

Fig. 7
figure 7

Automation introduces some genuinely new tasks, indicated by the circle outside of the automation and the old task area

Full size image

4.4 Genuinely new tasks outside automation

While making some tasks redundant, automation may also introduce tasks of a different kind. The recent introduction of online distribution of the tax cards in Norway provides an example: the tax card is produced by the Tax Administration and since 2014 presented online instead of being distributed via mail like before. Employers request the tax card of their employees online and get access to their tax card information. This function is constructed in such a way that the citizens do not need to accept or give consent for the employers to retrieve their online tax card. However, the citizens can look up who has requested access. One of the authors found that two unexpected employers had requested access to her tax card. The online instructions state that if employers who do not need your tax card are listed, you will have to contact them yourself. No address or other contact information for these employers is given, and it is left for the citizen to decide if and how to contact them, find contact information and write a letter or make a phone call. This new task is illustrated by a circle outside of the original task area in Fig. 7.

A new task emerging after automation is the task of “finding out” what needs to be handled manually. Example 2 above illustrates a health condition which does not automatically lead to tax deductions; instead the deduction must be manually claimed by the citizen. A citizen needs to differentiate between those tasks that have become redundant and those that are residual—that is, between those that are automated and those that they need to handle manually. The task of finding out accompanies the residual and new tasks. The way that the automation is constructed influences how the residual and the redundant tasks will be differentiated.

4.5 After the automation

We have shown that after the automation, some tasks have become redundant but that there still are tasks for the citizens to carry out. Our analysis of the calls to the tax authorities shows that automation leaves some residual tasks for the citizens to carry out manually in addition to introducing new tasks. “Subtracting” automation from the task area in Fig. 7, we see the tasks left for the citizens to carry out manually in Fig. 8. The four combinations of new and old tasks are numbered:

Fig. 8
figure 8

Inverting figure and ground shows fragmented residual tasks together with new tasks inside and outside automation. The numbers are explained below

Full size image
  1. 1.

    An old task that is made redundant by automation is illustrated by a dotted circle. It may be carried out voluntarily or some citizens may ask for a confirmation that it is ok not to do it.

  2. 2.

    A residual task that must be done manually. It is accompanied by a task of finding out, which increases the workload of the original task. The increased workload is indicated by an enlarged circle that extends the original task area. The task can be to update personal data or manually provide documentation.

  3. 3.

    A new task inside automation, such as overriding defaults or handling errors.

  4. 4.

    A genuinely new task outside of the automation, e.g. contacting employers that have requested the tax card without any obvious reason (there is no support from the Tax Administration to do this). It also includes finding out.

These tasks form a fragmented and incoherent task area. The new tasks show little internal coherence and increase the fragmentation of the tasks that the citizens need to carry out. Submitting a tax return form when there are no changes has become a redundant task after the automation. Many callers ask if it is ok not to submit: they need a confirmation that the task is redundant. There is a risk of penalty tax involved in making a mistake. The task of finding out accompanies many redundant tasks and increases the work involved in carrying them out. Finding out also increases the work involved in carrying out a residual task, as the citizen needs to identify those tasks that are left for them to do and differentiate these from the redundant ones that are covered by the automation.

Finding out may require almost no effort, or it may take considerable work for the citizen and include one or more calls to the call centre. It may require little or no effort if the citizen has previous experience in handling the task, but the citizen may lack the knowledge on how to handle it if the issue is new to him or her.

The new tasks are a consequence of the way the automation is constructed, which influences the boundaries between the residual and redundant tasks. How the automation is constructed also decides whether a task is manual or automated, as illustrated in Fig. 9. A different design of the automation will make a difference. To find out which tasks are automated and have become redundant and which tasks are residual and need to be handled manually requires some understanding of the way that the automation is constructed. Making the tasks more coherent can increase the citizens’ understanding of the tasks left for them to do.

Fig. 9
figure 9

Different automations influence the residual tasks that are left for the citizen to carry out. The same task is indicated as redundant to the left and residual to the right

Full size image

In reality, tax in Norway is semi-automatic as there are still residual tasks [27]. Which tasks become redundant and which becomes residual are intimately related to how the automation is designed. In order to make beneficial choices for oneself, one will need to find out and execute the residual tasks and foresee effects of changes to the tax card or tax return form. Shipman and Marshal [41] make a similar argument when they write about the use of knowledge-based systems: “Users must learn the system’s knowledge representation, even if it is hidden by a good interface, or else they will not fully understand the effect of their changes” [41, p. 6]. Making changes without understanding their effects may lead to unfortunate choices with unintentional effects that are not in the citizen’s best interest [32].

5 Designing self-services for autonomous citizens

Many callers show very little competence and understanding of how the tax system works. This is to be expected as doing taxes requires the citizens’ attention only a few times a year, and many tax issues occur only once or a few times in life (like having a baby, moving). Bainbridge’s “ironies of automation” [25] describes how automation can prevent the human operators from acquiring experience and intimate knowledge about the normal routine operations of the system. When doing taxes, some of the required understanding concerns very mundane tasks such as keeping wage slips for the possible need for documenting a claim two years later when the tax is calculated; our empirical material includes a call from a young man who had to pay penalty tax because he could not document that he had not received wage payments from his employer who went bankrupt two years earlier. Citizens are used to the documentation of claims being done automatically. The automation is designed for easy—almost unnoticeable—submission of taxes rather than supporting learning about the system and the citizen’s role in it. Often hiding complexity is easier than explaining it.

Automation of taxes is designed by pushing the limits of automation by automating what can be automated. The tax authorities report that most citizens appreciate the fact that the difficult parts of tax calculations are hidden; they do not have to experience being bad at calculating percentages. Doing taxes was experienced as complicated also before the automated tax was introduced [27]. However, the division of work and responsibilities between the tax authorities and the citizens was very clear in the days of the old paper form: the citizen provided and documented personal tax-related information and the tax authorities executed the calculations and produced a tax assessment. The authorities would check and in certain situations change the information provided by the citizen.

There is still a division of work between the tax authorities and the citizens after automation. This division of work cannot be completely avoided. Many happenings in a citizens’ life have consequences for their tax and need to be reported by the citizen for a correct taxation as they may not be registered in the databases of the tax authorities (in time). In addition, there will always be a need for citizens to override defaults, make special claims or negotiate applications or interpretations of the rules. In principle, the legislation allows for discretion to accommodate for individual treatment [43].

We argue that supporting citizens’ capabilities to recognize and carry out their work tasks should include support for their learning and understanding of the larger system in which their work tasks are contextualized. A better understanding gives the citizens a larger space for action and choice when new situations or life events occur [27, 39].

In line with this, we suggest an alternative approach to design of digital public services aimed at supporting a citizen perspective by increasing the coherence of the tasks seen from a citizen’s position. Instead of pushing the limits of automation, the design approach aims for making the work tasks appear coherent to the citizens in a way that supports understanding and learning. Figure 10 illustrates this design approach for automation, where the residual tasks and new tasks are integrated into the larger coherent set of work tasks.

Fig. 10
figure 10

Same task area with a different automation design. The focus is on designing tasks that are coherent and easily understood by humans

Full size image

Our design approach, which is illustrated by the “wig” in Fig. 10, maintains the division of work between the automation and the citizen, but makes it easier to understand [27]. We argue for seeing the citizen as a member of a cooperative work ensemble, in need of support for (developing) a good practice. In order to support the citizens’ understanding of the division of work and hence the residual tasks they need to carry out, we suggest including support for the citizen to become aware of what the tax authorities and their automation do. Support for peripheral awareness means to provide mechanisms that make the citizens aware of actions carried out by the tax authorities without explicitly looking for it [44]. Here, we need to include being aware of the tasks that the “automatic collaborator” do as well.

Many of the tasks necessary for doing taxes follow a sequence in time. Drawing on service design “customer journeys” [20] and its focus on the whole: the journey, as well as the parts: the entry points where the customer meet the service provider, can address the trajectory but maybe not the longer life-cycle view. The notion of “trajectory” has also been used for orienting patients about the expected treatment sequence for their medical condition as a source for them to know what will happen, who does what, and their rights and duties. However, “customer journeys” and “patient trajectories” are not aimed at teaching the customer or patient about the underlying system. The notion of trajectory orients the designer to see the customer’s entry points to the service as parts of the larger service. The customer (or patient) is seen as exactly that: a customer, emphasizing the efficiency of the service for the customer as well as for the service provider.

In contrast, design for coherent and meaningful tasks for the citizen sees the work of doing taxes from the citizen’s perspective, supporting the citizens’ activities and processes. Such support can involve tasks that are seen as “unnecessary” from the public authorities’ perspective, such as providing receipts or confirmations of tasks and of tasks done by others. An “unnecessary” task for the public agency can help the citizen understanding the whole underlying work process. When a citizen has changed her tax card online, she can immediately receive an electronic receipt confirming her action. Not receiving a receipt can indicate that the change request is not registered properly in the databases of the tax authorities; however, noticing the absence of a receipt requires a system that provides receipts regularly and reliably.

Within the division of work, the citizens need to find out which tasks are theirs to do. Finding out is a new task after the automation. Support for finding out will be important as it makes the division of work visible. Such support needs to be specific about which kinds of tax-related information the tax authorities collect, and what the citizens need to provide. Relatively simple designs such as lists of frequently asked questions (FAQ) where the issues are described from the citizen’s perspective will help with finding out. A FAQ that explains tax-related concepts and rules presupposes that the citizen can match their own life and circumstances with the tax rules and regulations. We have shown that this is difficult. A FAQ structured by occurrences and events that can happen in citizens’ lives will make it easier to recognize one’s own situation.

Design from a citizen perspective helps the citizen (aka user) to match their life situations to the tax rules. It provides helpful and informative confirmations that the citizen is doing it right. It supports learning by visualization tools that illustrate how taxation will change with varying income, fortune or debts and with life situations such as marital status. Visualization and simulation tools make consequences of such changes visible and available without skills in percentage calculations. Such tools can function as decision support in choice situations, for example when considering a job offer with a lower wage than the current one, and for learning about conditions that influence taxation—including advance tax which is most important for the running tax deductions.

6 Between work and life

In this section, we move from the case of citizens doing taxes to a more general discussion of our research findings and what they can teach us about civic duties and work, and about how to design for supporting citizens’ understanding and autonomy in doing civic duties.

6.1 Do-it-yourself

Automation for citizens’ self-service in public services has a parallel in preparations for do-it-yourself (DIY). In this paper, we use the notion of DIY to refer to homeowners improving and maintaining their homes themselves without the help of experts or professionals. DIY can be both leisure and work [45–47] and sits between “consumption (of materials and tools) and production (of changes to the home)” [46, p. 1]. Shove and her colleagues have studied the doing of DIY:

“the works, sweat, dust and frustration of mixing up bodies and their limitations with a diverse array of tools with which to transform a collection of materials to form the effect of a material change to the home, the product of labour. It is the work, of coordination of tools, materials, competence, confidence, body and the fabric of the home that places DIY at such a complex location in relation to the conventional boundaries of social scientific analysis, the boundaries between leisure and work, consumption and production, and ultimately between human and nonhuman.” [46, p. 4].

Watson and Shove [45] argue that the competence necessary for DIY “is perhaps better understood as something that is in effect distributed between practitioners and the tools and materials they use” and that product innovation enables “amateur DIY enthusiasts” to take on tasks that they before had to hire help to do (ibid., p. 77). Tools are enhanced in various ways: power tools make physically demanding tasks “lighter”, while other tools have been changed to modify the relation between the process and the result. One example of this is modern paint, which is easy to handle as it is fast-drying, non-drip and water-based, and looks good even when first-time painters paint. “[A]spects of the competence needed to paint … have been redistributed between person and technology, the paint having effectively absorbed capacities previously embodied in the individual wielding the brush” (ibid., p. 78). This fact that the tools that are used for DIY are designed to support the amateur in doing a competent job even when his/her competence alone is not sufficient: it is the amateur-with-tool that possesses the competence.

“[I]n structuring distributions of competence, objects indirectly structure possibilities of practice and consumption. … [T]he doing of DIY is itself of consequence for individual careers, emergent projects and future patterns of demand and product development, including demand for objects that help define the possibilities of future practice” (ibid., p. 85).

The “wig” (Fig. 10) represents a design approach that fits with this perspective as it takes the doings of the user (consumer, client, citizen) as the starting point and adds tools and materials that together increases the capacity for the citizen user to carry out a task as a non-professional—an amateur—with acceptable result. In amateur painting, the enhanced quality of the brushes and the paint is important for a good result: a parallel for civic services is that the tools support an understanding of the larger system so that the single tasks get their meaning and significance. Our analysis shows that some knowledge about the larger ecology of the public service is important for the citizens to understand their tasks—and to do them. In our data, we saw that the call advisors provided this understanding, not the automated system. We suggest applying a DIY perspective when designing automated public services in order to arrive at a set of automatic and manual tools and techniques that help the citizens to carry out their civic duties in competent ways.

The interactional nature of doing public self-services seriously suggests a relational understanding of the communication between the citizens and the public agencies. Relationally oriented public agencies accommodate and behave in forthcoming ways towards the citizens and their part of civic duties work. Design of digital self-services needs to take into account that the citizen will be using the services very rarely, perhaps only once in their lifetime. They will never be professional or educated in the topic area, and they will not be part of a community of practice sharing and building up knowledge over time. Moreover, citizens often happen to use public services when experiencing a difficult life situation, where more urgent matters of life and death are prominent, and civic duties seem less urgent.

6.2 Automated systems supporting collaboration

Large IT systems are often used by several user groups for different purposes. Digital public services are typically used by the public agency for providing a public service, like tax or social security. However, much of the data and information needed in order to give the service has to come from the citizens themselves. The automation of the Norwegian tax services has taken years before the data that used to be provided by the citizens (wage, bank deposits, etc.) are provided by employers, banks, insurance companies in a trustworthy manner. Only a few data items are left for the tax payers to register. These are irregular and non-standard because of the way the automation is designed, and their irregularity and uniqueness are not obvious to the citizen.

Many automatic public service systems are examples of multi-user systems where one user group provides data for other user groups that deliver a service based on the data. The quality of the service normally depends on the quality of the data; hence, automated data gathering and quality checks will improve the quality of the service. Very often some data cannot be automatically gathered or their quality assured. This makes the citizen responsible for checking or providing data for the system (health care and tax systems are important cases [27, 48–51]).

Shifting the perspective from citizens as data providers to the public service system to a view where citizens are partners in the society’s organization of civic work paves the way for different designs. If the taxes are done in collaboration between the citizens and the tax authorities, citizens and tax case handlers are collaborators and the digitized system can be seen as a tool to support the collaboration.

Borchorst and Bødker [52, p. 173] argue from a similar perspective when they state that “civil servants and citizens have inherent different foci in the service provision process”, and that supporting a better common understanding of the process itself is an aim for designing technology support. Understanding the process for collaboration between a public agency and the citizens contribute to “effectively educating citizens in democratic thoughtfulness” [52, p. 189]. Our suggestion for a design approach: the “wig”, also aims to be a step towards addressing the challenge of “creating good design within the arena of citizen-government interaction” [52, p. 189].

Schultze and Orlikowski [53] warn that organizations that use self-serve technology risk creating an arms-length relationship with their customers. The relations between an organization and their customers are enacted through the work practices and interactions between customers and providers, but the relations deteriorate at the organizational level when the human relations are substituted by self-services [53, p. 105]. The work that the tax call advisors do should be appreciated from this perspective.

Doing civic duties is part of citizens’ responsibilities in a democratic society. Developing competence and skills to be able to act responsible and at the same time experience mastery—a space for action and choice—is important for the citizens in a democracy. Introduction of digital public services requires that citizens possess a new type of competence that includes skills in both the digital service and the topic area. Citizens who are unable to develop or acquire such competence may be disadvantaged by the digital services [22]. Moreover, a democratic society needs competent and responsible citizens. A design that hides the complexities of the inner workings of the rules and regulations of society makes it difficult to become competent. Becoming a competent and responsible citizen is about more than having a good user experience. A coherent design will support understanding, which is necessary for becoming a competent citizen.

6.3 Work and life

Doing civic duties disturbs the dualism between work and leisure. Is doing civic duties between work and life? We think work is a part of life. In the discussion about work in Sect. 2, many of the design challenges concerned moving from the workplace to the home and other less public places. Based on our analysis, we argue that the picture is more complicated than the suggested either–or: either work or non-work. Even the game example in Crabtree et al. [11] is not about non-work; it is about different kinds of work and multiple ways of being in-between and both work and leisure. It is obvious that the facilitators of the game work (for a wage), but also the game participants can be seen to do work if their participation is organized by their employer (even if the game happens in their spare time). Building on Urmson and Ryle [1, 2] Schmidt makes the distinction that gardening is work whether it is a gardener or a non-professional garden lover who carries out the gardening: it is work-like. Work and work-like activities are a part of life, and many of the work-like activities we do are duties rather than fun. We see civic duties as work that citizens have to do as members of a collective community/society. Unlike consumers and customers, citizens cannot choose if and which services they want to use: some of the services are simply civic duties. Using a public service therefore is work done in order to constitute the collective community or society.

Automation of public services delegates some of the civic duties to computer systems in ways that make the “waves of HCI” [6] less useful as a classification of the work involved since it maintains a simple division between work and non-work, where non-work is (more or less) mindless leisure activities. If we base our understanding of work on people’s lives instead of labour, we see that paid work is only one setting for work-like activities and that (paid) work can be mindless or fun just as work tasks done in one’s spare time can be really boring and difficult. All members of a society have to do work-like acts with or without payment as a way to support and be supported by the community in which they live. Such community-building social tasks are an important part of being a citizen and community member. Life contains important work beyond (paid) work.

7 Concluding remarks

Well-functioning public services are important for human well-being [54]. Personal taxation and the financial basis of a state are also important for democratic decisions in society. We have argued that support for citizens’ learning, understanding and space for action when doing their taxes is important when designing for automation of taxes. Citizens have duties, in contrast to customers or consumers; hence, public services need to be usable for everybody—even if the cognitive efforts necessary for handling the technological and intellectual challenges of doing taxes are not equally distributed among the citizens [21, 32]. Our data from co-listening to telephone calls to the call centre provided us with insights into the lives of non-users of digital services and to the life situations where tax issues emerge and become challenges [27, 40].

We argue that hiding the complexity of tax in ways that prevents learning and understanding may affect the democratic processes in society in negative ways. Automation introduces new tasks, and to navigate the new ecology, citizens need to understand at least parts of the ecology. Instead of learning the old tasks that were part of doing taxes, citizens need to learn tasks that are related to the design of the automation. We believe that our analysis of redundant, residual and new tasks will be relevant for HCI and design in other areas where the aim is to enhance citizens’ autonomy and mastery.

Doing one’s taxes is part of being a responsible citizen and is neither paid work nor engaged leisure. Being a citizen implies doing work-like tasks in some situations and in some phases of life. This work needs to be taken into account and supported when digital public services are designed. We have provided a framework and a general design approach for designing public services grounded in a citizen perspective: the “wig”. Establishing the public service as a coherent whole helps the citizens in finding out and doing their part of the service work.

Doing civic duties disturbs a duality between work and leisure: it is work that we have to do in our spare time as members of a society, making the society. Acknowledging the work of citizens (as well as customers and clients) for making a self-service work properly, adds complexity to design of automated services. The work done by the unpaid “workers” (i.e. the citizens, customers, clients) may presuppose competencies that cannot be required or even assumed. The concept of work, and the perspective of designing for and supporting the work, adds a dimension to design of the public service that protects the citizen from having to use a service where efficiency from the service provider’s perspective is dominating the design.