Binding the Future: Participatory Algorithmic Configuration and the Narrowing of Discretion

Introduction

Research on algorithmic management has expanded rapidly over the past decade, documenting how algorithmic systems are increasingly used to coordinate, evaluate, and direct work, and how they reshape managerial authority and worker autonomy (de Vaujany et al., 2021; Kellogg et al., 2020). This literature has shown how algorithmic management intensifies surveillance, reorganizes accountability, and contributes to new forms of precarity and dependency, particularly in platform-based and gig-economy settings (Curchod et al., 2020; Möhlmann et al., 2021; Newlands, 2021; Walker et al., 2021). These studies have been central in demonstrating how algorithmic systems transform the governance of work by embedding control in data-driven infrastructures that structure access to tasks, performance and evaluation of work.

Alongside this focus on consequences, a growing body of research has examined how algorithmic management is accomplished in practice. Rather than treating algorithmic control as technologically autonomous, this literature has shown how control is produced through ongoing human work, including interpretation of algorithmic outputs, local adjustment of system behavior, and coordination around digital interfaces (Bucher et al., 2021; Jarrahi et al., 2021; Kellogg et al., 2020). It has further demonstrated how routine and often backgrounded forms of digital work are necessary to sustain algorithmic systems and make them appear stable and objective in everyday organizational settings (Justesen and Plesner, 2024). Analytically, these studies have been useful for explaining how variability and contestation emerge in algorithmic control arrangements, and how such control requires active maintenance and organizational work.

While these two strands of research provide strong accounts of the consequences of algorithmic management and the socio-material practices through which it is enacted, they have offered less explicit theorization of how configuration work may shape the organization of work in advance. When workers collectively participate in configuring systems that allocate, prioritize, or evaluate work, such activity does not only sustain coordination in the present; it may also contribute to fixing how work is ordered, sequenced, and evaluated over time. Under such conditions, workers do not only maintain algorithmic systems; they also take part in shaping the conditions under which future work is organized and assessed. This forward-oriented dimension of configuration, and its implications for control and accountability in conventional organizational settings, remains insufficiently examined.

This problem is especially salient in conventional organizations – established bureaucratic settings characterized by relatively stable employment, formalized roles, and entrenched norms of accountability (Krzywdzinski et al., 2025; Monteiro and Adler, 2022). In such contexts, decisions about how work will be organized in advance are consequential because they are tightly coupled to questions of responsibility, justification, and evaluation within hierarchical structures. Digital interfaces play a central mediating role in this process by translating abstract rules and priorities into actionable sequences of work, thereby shaping how control is experienced and enacted in everyday practice. Algorithmic management in such settings must therefore operate through, and be compatible with, these accountability structures. To examine how these interface-mediated formalizations unfold in practice, this paper analyzes a case in which office workers in a conventional public-sector organization were tasked with configuring an algorithmic system to coordinate their work.

Empirically, the paper asks what happens when workers participate in configuring the algorithmic rules that organize their work. Analytically, I examine such configuring as the performance of technical self-control, a worker-led and technology enabled form of control. The argument is as follows: If control intensifies even when algorithms are simple, transparent, and configured by workers themselves, then algorithmic control cannot depend solely on opacity, the interpretative uncertainty associated with algorithmic systems (cf. Jarrahi et al., 2021), or lack of worker involvement (cf. Krzywdzinski et al., 2025). Indeed, as I show, it operates through the embedding of collectively defined priorities in the system’s configuration, shaping how work is ordered and how discretion is exercised over time.

In this paper, technical self-control refers to an emergent, practice-accomplished, and interface-mediated collective form of control. It combines elements of technical control (Edwards, 1979) with elements of concertive control (Barker, 1993; Sewell, 1998), but its distinctive feature lies in how collective deliberation becomes inscribed into durable technical configurations. Through this inscription, discretion is governed not primarily through retrospective evaluation but through the stabilization of an anticipated order of tasks. Drawing on a practice-theoretical reading of Heidegger (1962[1927]), the paper conceptualizes this process as the technical governance of projection, that is, the shaping of which futures of work become intelligible and actionable within a shared practice world through algorithmic means. As I show, increased worker involvement in configuration thus does not necessarily expand discretion; it can instead restrict it by shaping how future action is ordered and constrained.

The paper makes three related contributions. First, it advances research on algorithmic management by theorizing configuration not only as present-oriented maintenance work but as a practice through which collective priorities become embedded in technical rule structures that shape the ordering of work over time. Second, it contributes to theories of organizational control by showing how participatory configuration may restrict rather than expand discretion when worker involvement contributes to the technical inscription of priorities. Third, it demonstrates how such participatory formalization may generate conditions for managerial recentralization by making work orderings visible, comparable, and thereby enabling standardization over time.

This paper is structured as follows. It begins by presenting a review of literature on technical control leading into contemporary algorithmic management and control. Next, it outlines the theoretical orientation used to conceptualize technical self-control. It then presents the research design and the empirical context. The paper concludes with the findings and their contributions.

Research approach and design

This study is grounded in a practice-theoretical approach (Reckwitz, 2002; Schatzki, 2001) that treats work as unfolding within shared practice-worlds rather than as the execution of individual intentions or the results of rigid structures. Selected concepts from Heidegger’s Being and Time (Heidegger, 1962[1927]) are used as sensitizing resources to study algorithmic control. In particular, Heidegger’s account of projection is drawn on to articulate how future-oriented possibilities of action are organized within practices (1962[1927]: 84), and his notion of das Man (the They) is mobilized to capture how shared expectations come to appear as impersonal and self-evident (1962[1927]: 219–224). I use Heidegger’s notion of projection to specify the temporality at stake in participatory configuration: control operates by stabilizing a horizon of actionable “next steps” in advance. In this case, the interface does not simply represent priorities; it materially organizes which futures of work become practically available and legitimate. Heidegger’s account of das Man further helps articulate how collectively authored priorities can become impersonal. Once translated into rule structures, “our” judgments are encountered as what “one does,” a taken-for-granted ordering that levels down differences into deviation from a shared normality rather than contestation of managerial authority.

These concepts are used here as sensitizing vocabulary for a specific empirical phenomenon – technical self-control and future-binding in a bureaucratic and accountability-oriented setting. From this perspective, adherence to or avoidance of algorithmic systems is not simply a matter of compliance or resistance but reflects different ways of positioning oneself in relation to socially produced expectations that govern work in such a setting in advance.

Data collection

The empirical material is drawn from a longitudinal qualitative case study conducted between the fall of 2014 and the beginning of 2017, spanning approximately 28 months. Data collection combined ethnographic interviews (Spradley, 1979) and observations, including periods of shadowing (Czarniawska, 2007). While the broader research project included multiple settings within the Agency, this paper focuses on one organizational unit in which workers were directly responsible for configuring the algorithmic rules that structured their work. Senior management at the Agency were open to me studying different units in so far as I shared my impressions of how events unfolded. This unit was selected because it offered particularly clear access to participatory configuration practices and their organizational consequences. See a list of key events below (Table 1)

OPEN IN VIEWER

Table 1.

Chronology of key events.

Date	Key events
October 2014	The initiative to develop the technology begins. Developers and top management decide on the overarching design of the algorithmic control interface.
March 2015	The studied office is chosen as a pilot setting. Small-scale experimentation with configuring and using algorithms in the work setting.
April 2015	Groups of workers are encouraged to configure own algorithms. Algorithmic allocation of work replaces manual allocation of work.
Second Half 2015	Disparate responses to technical self-control become evident. Some workers begin to voice discontent with the interface.
January 2016	Managers problematize technical self-control and the possibility of heterogeneity in algorithms.
April 2016	Technical self-control officially ends. Standardized set of algorithms are implemented.
Fall 2017	Data collection ends.

At the time of the study, the focal unit consisted of approximately 60 workers and two managers. The data set includes 29 interviews with workers and managers. Interviews were unstructured but practice-oriented (Nicolini, 2009) and typically lasted between 45 minutes and 2 hours. Interviews were done before the implementation in April 2015, during the implementation throughout 2015, and after technical self-control ended in 2016. See Table 2 below for an overview.

OPEN IN VIEWER

Table 2.

Interviews.

Role	Distinct participants	Pre	During	Post	Total
Managers	2	2	3	1	6
Caseworkers	9	5	6	4	15
Configurers a	4	2	3	3	8
Total Interviews	15	9	12	8	29

a

Note that the category “configurer” is a subset of caseworkers. These interviews include the chief configurer who was interviewed once pre-, twice during, and once after.

Interviews with managers addressed the rationale behind the initiative, how its implementation was organized, and how its outcomes were perceived over time. These interviews were conducted both prior to implementation and during its rollout, with particular attention to anticipated challenges and emerging concerns. Similarly, I interviewed the same set of workers during the study as events unfolded. Most interviews were recorded and transcribed verbatim; in a small number of cases where recording was not possible, detailed notes were taken during or immediately after the interaction.

Interviews with workers were conducted in tandem with observations and focused on everyday work practices, collaboration within teams, and experiences of configuring and using the algorithmic interface. During fieldwork, I was present as a non-participating researcher with access to daily work activities, meetings, and informal interactions, but did not perform casework or participate in decision-making. Observations were conducted to gain insight into a range of activities, including casework, worker collaboration, configuration activities at the interface, and so on; and workers allowed me to ask questions and tag along during their workday.

In addition, I had access to internal documentation, user manuals, and email correspondence related to the implementation and use of the system. These materials were used to trace how expectations regarding the interface and its use were communicated, negotiated, and revised over time.

Setting the scene for technical self-control

In the beginning of 2015, the studied office presented its workers with the new interface for directing and evaluating work within teams. Prior to the introduction of the interface, teams of workers were expected to not only navigate systems to find distinct cases belonging to them but also know which tasks were required to be done within these. This manual coordination process proved intricate, as streamlined guidelines often lacked the specificity needed to effectively guide and assist novice workers. Consequently, teams frequently relied on the assistance of more experienced colleagues to effectively guide them to complete their work. The managerial objective behind implementing the interface was to encourage workers to utilize it for coordination purposes, thus bypassing the complexity of the existing systems.

The interface offered tools to display the overall workload assigned to a team and to organize the distribution of tasks among team members. In practice, the interface was to serve as the main way of collecting work instructions. A worker would open the interface and receive work through it automatically. For it to function, however, workers were expected to configure their own algorithms and insert them in the interface. The rationale behind the initiative was that workers would design algorithms together in their teams as they saw fit. For this to be possible, workers were expected to not only devise an own list of prioritized work tasks and distribution order but also learn how to translate that list into algorithmic form in the interface.

From the point of view of the project managers, the effort was framed to show how the Agency aspired to invest in technologies that increased workers’ sense of control. Local management drew on this rationale when tasked to implement the technology. While managers were keen on emphasizing the increased importance of worker involvement, there was an additional reason at play for letting workers engage in devising their own algorithms. To begin with, local management swiftly recognized the demanding nature of setting up the interface, given that it necessitated configuring algorithms. Even if the rationale was intended to reduce complexity, using the interface to control the work for the entire setting proved too complex for managers. Hence, managers figured it would be better to let this task fall to workers instead. In other words, the delegation of the task to curate algorithms was also a means for management to avoid a form of work they anticipated was to be too much of a hassle.

The initiative was met with mild enthusiasm among workers despite the emphasis on worker involvement. Prior to the implementation most workers were individually responsible for choosing their work tasks. While workers were used to work in teams, they mostly collaborated to assist each other as they saw fit. For instance, workers were used to cover for absent colleagues, to assist each other in when examining tricky cases, and in helping in the training of new colleagues. The new change implied the necessity to organize teamwork through the interface and its algorithms.

Casework in the Agency was subject to multiple and partly conflicting accountability pressures. Caseworkers were expected to process cases efficiently in line with organizational throughput targets, while also ensuring strict compliance with statutory rules and maintaining documentation suitable for audit and appeal. Although individual performance was not algorithmically scored, aggregate productivity and backlog levels were monitored at team and unit level, and deviations could prompt managerial intervention. Professional norms emphasized consistency, fairness, and defensibility of decisions, making deviations from established priorities difficult to justify ex post. These conditions rendered advance specification of work priorities both attractive and risky, as algorithmic ordering promised predictability while simultaneously constraining discretion.

Regardless, the initiative was not understood to be a major problem by workers. During early meetings, both managers and workers agreed on how the possibility to reduce the complexity of work task selection could be an important feature to train new workers and help the teams cover for absent colleagues.

Although emphasis was placed on highlighting how teams were to collectively engage with the interface, each team was expected to have one worker in charge of curating their algorithms. These chosen few were referred to as “configurers” and were expected to learn both how the interface worked and how to translate a team’s desires into algorithmic form. To find configurers, managers would either suggest possible candidates – often among those they perceived as technologically proficient or particularly experienced – or workers would suggest volunteers among themselves. As expressed early on among teams, it was not particularly hard to find volunteers who seemed interested in tinkering with the interface. Assigning this role was then not a contentious issue among workers, although selected “configurers” expressed varied degrees of enthusiasm about their new roles. While workers were expected to use the interface, configurers were expected to develop a deep familiarity with its technological infrastructure, surpassing the knowledge level of managers regarding the system. Although the initiative was introduced as a shared and relatively unproblematic change, workers’ subsequent engagements with configuring and using the interface diverged noticeably.

Configuring algorithms: Translating collective expectations into rules

To function as a coordination device, the interface required teams to actively configure the rules through which work would be ordered and distributed. While the system could recognize categories of tasks and cases through its integration with existing bureaucratic systems, it did not operate autonomously. For work to be directed through the interface, teams had to specify a set of rules, a manually arranged algorithm, that determined which tasks would be selected, when, and for whom.

Configuration did not presuppose formal programming expertise. The interface provided a simple but rigid structure for defining two elements: numerical values assigned to categories of tasks, and conditional if–then statements that modified these values based on specified criteria. At any given moment, the task with the highest numerical value was selected and automatically assigned to a worker. Once completed, the next highest task was selected in turn. In this way, configuration entailed deciding which tasks should count as most urgent under which conditions.

In practice, designated configurers translated collective expectations about priorities into this rule-based format. They assigned baseline values, introduced time-based escalation rules, differentiated outputs across groups, and occasionally added new categories. What teams discussed in meetings; how to balance complex cases, how to support new colleagues, how to avoid backlogs, had to be rendered into numerical hierarchies and conditional logic.

Despite the apparent simplicity of the system, this translation proved demanding. Collective understandings of appropriate sequencing did not map neatly onto numerical values. Multiple configurations could produce similar outcomes, making it difficult to determine whether an algorithm faithfully reflected team intentions (cf. Dourish, 2016; Lange et al., 2019). It was through this translation work that technical self-control took empirical form.

Through configuration, teams did more than adjust technical parameters. They specified in advance what would count as acting properly and efficiently. Algorithmic rules formalized the projected work order within their shared practice-world, orienting present action toward a collectively defined future rather than individually chosen trajectories (cf. Heidegger, 1962[1927]: 84). Once inscribed in the interface, priorities no longer existed as flexible understandings but as structured ordering devices that governed future action.

A key element easing the complexity of this process was the provision of a template algorithm. Rather than designing algorithms from scratch, most teams began by modifying this template incrementally. The template was widely perceived as sensible and familiar. Workers trusted it because it had been developed by a former colleague understood to “know the work,” and because it appeared to align with existing ways of prioritizing tasks.

As the chief configurer later explained, this familiarity was not accidental:

“My first step in working with this interface was to put in the priorities that many were already following. However, this set of priorities was actually derived from what had broadly been decided on by managers in the past.” – Chief Configurer (Pre-implementation)

The template thus carried earlier managerial priorities into the participatory configuration process. Although teams experienced configuration as an opportunity to define their own rules, the point of departure already formalized a particular ordering of work. Through incremental modification rather than wholesale redesign, collective expectations were translated into technical rules that committed teams to specific ways of organizing future work, often without explicit reflection on their origins.

Diverging engagements with technical self-control

Three recurrent ways of engaging with technical self-control were observed in the studied setting. The ways of engaging described below differ not primarily in attitudes toward technology, but in how workers positioned themselves in relation to the future that the algorithmic system articulated. These orientations can be understood as different modes of projection (cf. Heidegger, 1962[1927]: 184–188) within a shared practice-world: ways of taking up, reproducing, or distancing oneself from a technically formalized understanding of what future work was for and how present action should be directed toward it. The ways emerged over time as workers encountered the demands of configuring, using, and responding to algorithmically formalized expectations about future work.

The analysis below traces how these ways differed in their relation to configuration practices, the template algorithm, and expectations about future work. Each orientation reflects a distinct way of engaging with the same technical arrangement and its normative implications.

From technical self-control to managerial recentralization

Over time, the heterogeneity of team-level configurations and diverging ways of engaging with the interface became increasingly problematic from a managerial perspective. While worker-led configuration had initially been framed as a way to increase autonomy and adaptability, it produced a patchwork of algorithmic arrangements that were difficult to oversee and compare.

Management began to reframe this variability as a source of opacity, risk, and lack of control. Differences between team algorithms were described as undermining transparency and accountability, particularly in relation to organizational targets and performance monitoring. The very features that had justified delegation – local adaptation and participatory configuration – were now interpreted as obstacles to effective oversight.

In response, management moved to standardize algorithmic rules across teams. Central definitions replaced local configurations, and discretion over rule-setting was gradually withdrawn. Paradoxically, worker participation in technical self-control facilitated this recentralization. By demonstrating that algorithmic coordination could structure work effectively, even when configured by workers themselves, the initiative provided empirical justification for expanding and centralizing algorithmic control.

Technical self-control thus did not mark a transition away from managerial authority. Instead, it functioned as a transitional arrangement through which algorithmic governance was normalized and rendered necessary. Worker involvement stabilized algorithmic control in practice, making its later recentralization appear both reasonable and inevitable.

Discussion

This paper contributes to research on algorithmic management by showing that control may intensify through worker participation when collective priorities are translated into durable technical rules that pre-commit the ordering of future work. The central concept developed here is technical self-control: participatory algorithmic configuration through which workers help bind future action to collectively formulated or negotiated rule structures. This contribution suggests three broader implications: configuration may bind future work; opacity is not necessary for algorithmic control to narrow discretion; and worker participation may deepen rather than mitigate organizational control. I expand on these next.

As mentioned, enactment-oriented accounts have convincingly demonstrated that algorithmic control depends on ongoing work, including interpretation, adjustment, repair, and coordination around digital systems (Jarrahi et al., 2021; Kellogg et al., 2020; Neyland, 2015). This paper extends that line of inquiry by showing that participatory configuration is not simply maintenance work directed at keeping systems functioning, but – under certain conditions – a future-binding organizational practice. Once inscribed into the interface, these arrangements structured future action beyond the moments of deliberation that produced them.

Importantly, algorithmic control operates here without informational asymmetries or opaque computational systems (Curchod et al., 2020; Rosenblat and Stark, 2016; Walker et al., 2021). What is distinctive here is instead the temporality of control. Governance did not primarily operate through retrospective evaluation or episodic managerial intervention, but through the collective stabilization of projection: the shaping of which futures of work became practically available and legitimate within a shared practice world (Heidegger, 1962[1927]). Through configuration, teams did not simply “use” the system; they took part in articulating and fixing a horizon of legitimate next actions.

Recent discussions have increasingly examined participation, hybrid governance, and worker influence in algorithmic management (Krzywdzinski et al., 2025), often treating these as a counterweight to unilateral managerial imposition. The findings here complicate such accounts. Participation may mitigate certain forms of domination, but it can also stabilize control when it takes the form of collective formalization. Worker involvement does not automatically expand discretion; under some conditions, it may restrict it.

The findings also resonate with earlier concerns raised in studies of concertive control (Barker, 1993; Sewell, 1998), while indicating a distinct mechanism. Concertive control stabilizes norms primarily through ongoing social interaction, peer monitoring, and collective sanctioning. Technical self-control, by contrast, stabilizes collective priorities through technical inscription: shared understandings are rendered explicit, calculable, and materially embedded as rule structures that continue to order work even when interaction and deliberation recede. The crucial shift is therefore from norm internalization to rule inscription. In this respect, the interface did not merely display collectively defined priorities; it rendered them actionable and procedurally authoritative through the very process of configuration (cf. Lipp et al., 2025). What made this authority especially difficult to contest was precisely that the priorities had been collectively negotiated. The binding force of the ordering did not present itself as a managerial demand from outside, but as a stabilized outcome of “what we have decided.”

This impersonalization had two consequences worth raising. First, contesting the ordering of work became more difficult because it required reopening collectively formalized priorities rather than merely disagreeing with a manager. The target of critique was no longer an external authority, but a shared commitment, now encountered as an infrastructural fact. Resistance was thereby displaced from oppositional confrontation toward quieter modes of refusal, circumvention, or withdrawal. Second, active configuration frequently reproduced rather than subverted managerial priorities. Because teams configured their rules in relation to an inherited template, participation often deepened compliance with pre-existing bureaucratic logics. What appeared as worker authorship of priorities became a mechanism for stabilizing and extending organizational expectations in technical form.

These dynamics speak directly to debates in algorithmic management that frame worker responses primarily in terms of compliance, gaming, and overt resistance (Bucher et al., 2021; Kellogg et al., 2020). Echoing longstanding critical concerns regarding empowerment, responsibilization, and self-governance (Fleming and Sturdy, 2009), the case shows how worker involvement may redistribute rather than diminish control by embedding collective commitments in technical infrastructures. This redistributes the moral and social costs of dissent. Resistance becomes more individualized and avoidant than collective and confrontational. In the present case, this was visible through avoidant distance, where discretion was preserved by bypassing the interface rather than by reopening deliberation about the rules themselves.

The study also sheds light on how participatory algorithmic configuration can generate conditions for managerial recentralization. Algorithmic management scholarship has often emphasized the decentralizing and fragmenting effects of digital systems (de Vaujany et al., 2021). The present case points to an additional dynamic: participatory formalization produced visibility; visibility produced comparability; and comparability rendered standardization both feasible and legitimate. Variation in locally configured priorities became visible across teams, making differences in work ordering increasingly comparable and open to managerial evaluation. The very process that initially distributed control also created the informational and organizational basis for its later consolidation. Technical self-control thus contained conditions of its own reversal, leading to a general tightening of managerial control of work in the long run.

The dynamics identified here are most likely to emerge in conventional, accountability-intensive organizations where algorithms are transparent and rule-based, where workers might be invited to configure task allocation within pre-given interface architectures, and where work priorities are tightly coupled to justification, audit, and comparability. The claim advanced here is therefore conditional rather than universal, but analytically relevant beyond the immediate case wherever participation becomes intertwined with technical inscription in organizing work.

Although the empirical case examined here concerns a transparent and relatively simple rule-based system, the mechanism identified is relevant for contemporary forms of AI-enabled algorithmic management. Contemporary AI-enabled systems do not displace this dynamic but may deepen it by making configuration more granular, adaptive, and continuously recalibrated. Rather than binding future work through relatively stable rule structures alone, these systems can dynamically refine priorities, recommendations, and classifications in ways that remain shaped by worker involvement while still limiting discretion in consequential ways. Under such conditions, collectively authored and negotiated priorities may become even more difficult to contest, not because they are necessarily opaque, but because they are increasingly embedded in evolving infrastructures that quietly reorganize what counts as appropriate action. The critical question is therefore not simply whether workers participate in configuring such systems, but what forms of ordering participation help to stabilize over time.

Conclusions

This study has examined what follows when workers participate in configuring the algorithmic rules that organize their work. It introduced the concept of technical self-control to capture a distinct form of governance that emerges under such conditions. Technical self-control designates a situation in which collectively articulated expectations are translated into technical rules that continue to order work after deliberation has receded.

The argument of the paper is conditional. The case concerns a conventional, accountability-intensive public organization using transparent and rule-based systems. Under such institutional and technical conditions, participatory configuration can generate technical self-control because expectations are tightly coupled to audit, justification, and comparability. Other settings characterized by opacity, precarity, or limited worker involvement may produce different dynamics.

More broadly, these findings suggest that algorithmic governance in conventional organizations cannot be evaluated simply by asking whether workers are involved. Participation does not automatically counteract control. Under certain institutional conditions, it may deepen it as discretion is narrowed not through exclusion but through involvement in shaping the arrangements that organize future work.

This raises a more uncomfortable issue for organizations pursuing digital transformation. It is not sufficient to claim that systems are legitimate because “humans are in the loop” or workers helped design or configure them. Closer scrutiny of what kinds of technologies are introduced, what forms of rule inscription they require, and how they reorganize the horizon of possible action over time is required. Involvement can redistribute responsibility and render governance more acceptable, yet it can also make control more difficult to contest precisely because it appears as something “we” have decided. If that is the case, then the critical task is not merely to insert humans into algorithmic systems, but to reflect on which forms of technological ordering become formalized and stabilized, and with what consequences for discretion and accountability.