Creating Interpretative Spaces in and with Digital Infrastructures: How Editors Select Reviewers at a Biomedical Publisher

Introduction

Academic publishing today fundamentally relies on digital technologies. For editorial peer review, journals predominantly employ editorial management software (Taubert 2016, 2012) that helps them track submissions, communicate with reviewers, and eventually send manuscripts to production. Like other digital (Christin 2020) and evaluative infrastructures (Kornberger, Pflueger, and Mouritsen 2017; Krüger and Reinhart 2017), these editorial management systems (EMSs) do not just support existing peer-review processes but instead “reorganize processes of thinking, sensemaking and decision-making through categorization, classification, commensuration, calculation and other forms of ‘sorting things out’” (Bowker et al. 2019, 3). In addition, new digital tools are being developed and integrated into EMSs that aim to automate more and more tasks within editorial peer review, such as finding suitable reviewers (e.g., Elsevier’s Reviewer Recommender, Taylor & Francis’ Reviewer Locator, or Springer Nature’s Reviewer Finder), automating editorial decisions based on reviewers’ reports (Plotkin 2009) or even predicting reviewers’ recommendations (Checco et al. 2021). Peer review and editorial work, like other workplaces, have become a site of complex interactions between human experts and digital infrastructures.

EMSs fundamentally shape the peer-review process as a collective achievement. However, we still do not know a lot about how digital technology is actually used in editorial work. Despite their apparent proliferation, EMSs have only recently begun to receive more analytical attention (Taubert 2012, 2016). Thus, we are interested in understanding how digital infrastructures are employed in editorial offices.

While infrastructures have the potential to reorganize collective reasoning and decision-making (Bowker et al. 2019), their properties, functions, and effects are far from deterministic. The functions of digital infrastructures need to be actively accomplished by users in situated practices (Vertesi 2019), opening them up not only to the contingencies of the actual local and social contexts of use but also to practices of implicit and explicit negotiation, contestation, and resistance by users. This holds especially true in expert workplaces, where users engage in buffering strategies that limit and modulate the influence of infrastructures in order to protect their decision-making authority (Christin 2017). Recent anecdotal accounts by editors suggest that the interaction between human experts and digital infrastructures in editorial work can be highly contentious: EMSs and automated tools are criticized as taking unwanted and unhelpful “supportive” actions (Horbach, Ochsner, and Kaltenbrunner 2022) or as conducive to practices of bad governance and control within journals (Gershon 2018). Similarly, Taubert (2012) describes the EMS as exerting unwanted controlling influence over the editorial process.

As such, we need to examine how users strategically engage with and disengage from infrastructures as a way of protecting their control over processes and decisions. Acknowledging the complex relations between the roles of human expertise and digital infrastructures in editorial work, we ask: How do interactions between editors and digital infrastructures unfold? How are tasks and responsibilities distributed between infrastructure and editors? And, how do contextual factors and resources shape the ways in which editors engage with digital infrastructures?

Through a case study of a biomedical publisher of several journals, this article investigates practices connected to selecting suitable reviewers as a crucial step in the peer-review process. The article uses both qualitative (interviews and observations) and quantitative data (log data from an EMS). We trace the interactions of editors with different infrastructural functions, showing how editors strategically engage with and disengage from infrastructures, resulting in three different forms of interaction. The first is close engagement between editors and the infrastructure around issues of time-keeping. Here, acceleration emerges as a shared accomplishment (see Vertesi 2019) between users and the infrastructures, and the infrastructures consequently exert a great deal of influence over the peer-review process. The second is a form of mitigation in the context of establishing a collective editorial memory. Editors strategically draw on some infrastructural functions but avoid others, resulting in a significantly more limited infrastructural influence. The third is a form of disengagement when it comes to evaluating potential reviewers. Here, editors actively move away from digital infrastructures to achieve their goals in the peer-review process. Through these three forms of strategic interaction, editors create interpretative spaces for themselves that allow them to protect their considerable authority over the peer-review process. Furthermore, we show how editors’ ability to modulate their engagement with infrastructures in these ways crucially rests on the specific conditions of their work as full-time professional editors working in a shared office space. Building on these insights, we argue that spatial, social, and organizational factors all influence the degrees of freedom and discretion users retain when working with digital infrastructures.

EMSs as Digital Infrastructures for Evaluation

We approach EMSs as infrastructures for evaluation (Krüger and Reinhart 2017; Meier, Peetz, and Waibel 2017) whose primary function is to enable editorial peer review. To do so, they connect actors (editors, authors, reviewers, etc.) who have different goals and needs in different local contexts to coordinate their activities as well as provide functions such as data storage, data collation, and forms of monitoring. As such, they represent a crucial element in the collective socio-technical achievement that is editorial peer review. Infrastructure studies highlight the fundamentally generative role of such (digital) infrastructures in evaluations (Kornberger, Pflueger, and Mouritsen 2017): they establish values, objects, and qualities that were not previously defined (Bowker et al. 2019). As software for process management within editorial peer review, EMSs are particularly involved in establishing and shaping “collective reasoning, [the] structuring of attention and orchestration of decision-making” (Bowker et al. 2019, 3). As (mostly) off-the-shelf software systems that allow for relatively little modification by their end users, we could expect them to fundamentally transform editorial peer review, including both the organization of the evaluative process and salient evaluative criteria, and ultimately to redefine what it means to do editorial work (see also Bowker and Star 1999).

Yet, digital infrastructures should not be approached as deterministic. On the contrary, infrastructures “are constituted through and inseparable from the specifically situated practices of their use” (Suchman et al. 1999, 399; see also Star 1999), so digital technologies must be made actionable to become consequential (Büchner and Dosdall 2021). Functions and features of software are accomplished, that is, made salient and functional in socially and locally situated practices of use that connect infrastructures as well as various actors “in a moment of collective focus” (Vertesi 2019, 377). The specific functions of digital technologies are actively accomplished by groups of users, rather than the result of preexisting properties inherent to the software’s material design (Vertesi 2019, 371). Moreover, the accomplishment of complex workflows such as peer review is still contingent on human labor, even in highly automated settings: as Shestakofsky and Kelkar (2020) show, even highly sophisticated digital platforms require work that cannot easily be automated or delegated to an infrastructure, such as relationship labor, which represents a crucial element of editorial peer review. In practice, the effects as well as the concrete properties and constraints of digital infrastructures are shaped by the continuous discretionary work of a variety of actors, including users (Passi and Sengers 2020).

As studies of digital technologies in workplaces show, human users are not passive in the face of infrastructural properties and effects but rather are actively involved in modulating, shaping, and resisting infrastructures. In expert workplaces in particular, expert users strategically limit the influence of infrastructures on their work and discretion through “buffering strategies” (Christin 2017, 9) that typically take the form of resistance: users might refrain from using entire infrastructures or selected functionalities, engage in gaming, or openly voice their critique of infrastructures. These buffering strategies serve to protect experts’ decision-making autonomy from unwanted infrastructural interference and control (Christin 2017). As a result, human actors can create interpretative spaces for themselves based on whether and how infrastructures are used.

In the case of editorial peer review, human expert judgment is considered especially significant. As a cornerstone of academic self-governance (Reinhart 2012; Neidhardt 1988), peer review is fundamentally built around the figure of the “peer” that lends the process not only its name but its credibility (Shapin 1995). With such a fundamental reliance on human judgment (Schulz et al. 2022), we can expect that digital infrastructures in editorial offices will be contested, buffered, or circumvented to protect the authority of human experts. Accordingly, our analysis focuses on how editors interact with EMSs, tracing how editors strategically modulate their engagement with, and thus the influence of, infrastructures in editorial work. Through this strategic (dis)engagement, editors carve out distinctive interpretative spaces for themselves in the editorial process. Furthermore, to take the fundamentally situated nature of these use practices (Suchman et al. 1999; Vertesi 2019) seriously, we need to pay specific attention to contextual factors and resources (e.g., further infrastructures like office spaces or nondigital archives) that shape how editors might engage with digital infrastructures or refrain from doing so.

Data and Methods

How Digital Infrastructures Shape Peer Reviewer Selection: Acceleration as Common Goal between Editors and the EMS

The speed at which editors handle submitted manuscripts is a key factor in how researchers perceive the performance of biomedical journals, and as such is intensively scrutinized in the field (Andersen, Fonnes, and Rosenberg 2021). From researchers’ perspectives, the duration of peer review is seen as delaying the dissemination of research; while recognizing it upholds research quality, the review should be reduced to the minimum time necessary to render a decision.

The importance of speed is also evidenced in our quantitative data: the EMS monitors the peer-review process by logging standardized events (e.g., “all reviewers assigned”) for each manuscript’s life cycle with the exact times at which they occurred, which allowed us to reconstruct the external review as a time line of collaboration between editors and reviewers (Figure 1). Editors initialize external peer review by contacting potential reviewers. In the average manuscript life cycle, the first potential reviewer accepts within a week after the first appearance of a manuscript in the database (seven days—all durations are expressed as medians), takes some time to complete the group of reviewers. At 8.9 days, all potential reviewers are assigned by the editor and those who accept the invitation will soon begin reviewing (9.6 days). Four to five days later, when enough reviewers have accepted the invitation to write a review, reviewer assignment is complete (14.3 days). Ten days later, the reviews are received by the editors (24.7 days), and the external consultation phase ends with all reviews received at 30.7 days. Clearly, the review process here is fast-paced.

OPEN IN VIEWER

Figure 1.

Time line of editors’ collaboration with reviewers. A time line of reviewer-related events relative to the first appearance of a manuscript in the database, visualized as box plots aggregated from all first-version manuscripts. As not all manuscripts go to reviewers (the desk rejection rate is ∼80 percent), not all manuscripts go through all of the events listed here. Occasionally events that typically happen only once per manuscript (e.g., “All Reviewers Assigned”) can occur several times in the records, because the log files are not retrospectively changed (e.g., when it is necessary to retroactively assign another reviewer for whatever reason). In the box plots, medians are marked as vertical lines inside the boxes. The boxes include all values between the 0.25 and 0.75 quantiles, the whiskers include all values between the 0.1 and 0.9 quantiles, and outliers were removed after computation but before plotting for better readability.

Acceleration is commonly hailed as one of the main benefits of automation and digitization. Our research visits and interviews show that timeliness is established as a common goal between all actors and mainly accomplished through two functions in the EMS: timekeeping and automation. For timekeeping, the EMS calculates and displays the duration between selected events to its users, which is particularly relevant regarding reviewer selection. Selecting reviewers who work fast is both enabled and encouraged by the EMS, which displays potential reviewers’ previous response times alongside their names and contacts. Editors consistently voice a strong preference for reviewers who work fast:

So, [I select] referees that I know and have worked with before that I know will usually deliver a referee report in a good time frame. (Mbio 3c, 11)

As such, the generative power of the EMS for timekeeping and automation can be seen in how the infrastructure records and displays time and sends chasers to delayed reviewers. Moreover, timeliness as a common goal between actors is furthered by infrastructure and editors alike: the editors complement the automated acceleration with a primer on timeliness and availability of reviewers during reviewer selection, and through relationship work that speeds up reviewers’ submissions. Thus, both the desire to speed up the review process and the practices by which this is achieved are observable in the task of reviewer selection and relationship work. In this form of close interaction, editors do not seem to need additional interpretative spaces for themselves.

Mitigating the Infrastructure: Collective Memory in the Reviewer Database

To select reviewers, editors enact certain expectations and criteria. In biomedicine, these expectations are both diverse (Glonti, Boutron, et al. 2019) and journal-specific but include expertise and proficiency, voluntarity or duty, ethical behavior, and an advisory rather than decision-making role (Glonti, Cauchi, et al. 2019). To assess these criteria, in our observed case, a bespoke database of past reviewers comes into play. When asked how they select reviewers, one editor immediately mentioned the database before any other criteria or considerations:

We have a data-, database. Yes, and the nice thing is that we evaluate all referees that we have ever used in there, with a simple ranking. Now we have one to five stars, so it really doesn’t go into detail, but so that we know, if we have ranked somebody badly, that we, or at least, maybe ranked them badly more than once, that we rather not use them anymore. (Mbio 2a, 22)

As an infrastructure shared among all members of the editorial office, the database to a certain extent depersonalizes editors’ individual experiences. Through a five-star evaluation, knowing potential reviewers is transformed from a complex individual judgment into a “simple ranking, a process that again highlights how the infrastructure defines and opens up specific modes of evaluation (i.e. quantification)” (Bowker et al. 2019, 4). Interestingly, the five-star scale described by the editor above technically constitutes a rating (Heintz 2016) but is described as a “ranking.” The change in terminology signals that the evaluation moves from an individual noncompetitive evaluation (all reviewers in the database could have the same score) to a competitive comparison between potential reviewers (Brankovic, Ringel, and Werron 2018; Espeland and Sauder 2016), implying a hierarchy between the reviewers in the database. While the basic mode of quantification is enabled by the infrastructure, this competitive evaluative logic is an accomplishment of editors’ interpretation that goes beyond the properties of the infrastructure.

Although a lot of tacit knowledge gets lost in this process, the benefit of quantification is precisely this reduction in complexity through a combination of multiple ratings, that is, experiences, from different editors, different journals, and different times into an overall evaluation. When discussing how they work with the database, editors often used plural pronouns, highlighting the perceived benefits of such transferability of individual judgments. Here, the editorial team seems to blend into one collective actor through the database. Individual processes such as learning are collectivized and ascribed to editors as a group:

So, there’s a, as we document in our internal system, we…will…also do have a learning curve about how useful that input from that referee is, right. (Mbio 3a, 5)

OPEN IN VIEWER

Figure 2.

Editors sharing reviewers. Nodes represent editors and ties represent the presence of shared reviewers between editors in the four selected life sciences journals; the closeness of nodes indicates the number of shared reviewers (Kamada-Kawai network layout). In the database, user management is modeled so that a person is unique but can act in different roles at different points in time. A person can thus feature as an editor of one paper but as a reviewer or an author on another; roles may occasionally change during the manuscript life cycle. In our analysis, we ascribed fixed roles to persons for the entire manuscript life cycle ex post based on the manuscripts’ final list of editors and reviewers, occasionally leading to fuzziness in the data if a role assignment changed during review. To avoid accidentally counting infrastructural protocols as potential reviewers, we restricted this evaluation to persons who are in the final list of editors (as triggering persons) and reviewers (as affected persons) for the manuscript, which leaves us with 26 different editors and 3,987 reviewers.

OPEN IN VIEWER

Figure 3.

Reviewers contacted by journals. The diagram shows how reviewers are invited to review for the four different journals included in this project. Blue nodes represent reviewers, red nodes represent journals, and gray ties represent review invitations (without multiples). Of 3,987 referees, 502 (12.6 percent) were contacted for two or more journals. Network visualization is performed with the Kamada-Kawai layout, with unconnected nodes repelling from each other.

The collectivization through the reviewer database is noteworthy because the editorial work is otherwise organized with a high division of labor and relatively strict areas of individual responsibility: each manuscript is usually assigned to only one editor; the editorial office space is divided into distinct areas for each journal; separate journals also hold separate editorial conferences. Even though manuscripts are collectively discussed in the editorial meetings, and editors might offer informal advice on their colleagues’ decisions, editors still work alone on “their” assigned manuscripts most of the time. The reviewer database, however, represents the communal work of the editorial office, in which every editor contributes to building the database as a stock of knowledge, and benefits from it in turn.

In contrast to the EMS, the reviewer database contains few automated features or processes. In its basic function—storing information in an organized way—it is very similar to a physical archive, even though it might exceed such an archive both in storage space and ease of use. While this infrastructure might appear quite banal, and technologically, much more sophisticated infrastructures could be imagined, it is important to keep in mind that this is a bespoke infrastructure that, unlike the EMS, was designed for the needs and practices of the editorial staff and must thus hold a specific value for them. More precisely, as shall be discussed in the next section, editors by their own design still retain much of the often difficult and time-consuming work of selecting reviewers and utilize specific resources from their physical and organizational space to do so.

All in all, editors have established a collective memory to foster reviewer selection that works alongside the editors’ implicit and shared expectations and criteria. The solution of choice for keeping this memory is a digital infrastructure—a database—where past experiences with reviewers are recorded, rated, and stored for future reference. The database is consulted in the moment of reviewer selection, irrespective of boundaries between editors and journals. Unlike the EMS, the reviewer database is used as a mere tool, and the potential for automated reviewer selection is fenced off. So while editors choose to utilize certain modes of evaluation that are opened up by the infrastructure (see Bowker et al. 2019), for example, quantification, they often also mitigate them to reserve interpretative spaces for themselves.

Keeping the Digital Infrastructure Out: Individual Selection of Reviewers

Apart from shared accomplishments in timekeeping and support of the infrastructure as a collective memory, some aspects of editorial work remain largely nondigital. When it comes to the evaluation of a potential reviewer (their expected recommendations, their topical expertise), editors rely primarily on their own and their colleagues’ personal judgments even though, in principle, there would be technology available to help them handle that task.

This firstly becomes visible in the absence of a tool that would automatically match manuscripts with reviewers on topical and social characteristics. While reviewers’ areas of expertise are tagged in the EMS database, the topical assignment is still done by editors relying on personal judgment. This may be in part because tags are often self-defined by authors and reviewers and therefore not standardized. A human editor with knowledge of the field has no difficulty understanding if two different tags are closely related, whereas an algorithm can misinterpret them completely. Automating the process of topical matching would require developing and implementing a standardized classification, which, while not necessarily technically sophisticated, would still require quite a lot of conceptual and/or manual classificatory work. It appears that at least currently, this work is not seen as worth the effort.

Another instance where the potential for automation goes unused is in the exclusion of reviewers. Editors consistently report being mindful of authors’ requests to exclude certain reviewers and generally avoid inviting these reviewers (even though the reasons behind the exclusion might be personal or strategic). When showing the EMS to the investigator, an editor pointed to little text boxes pinned to some of the manuscripts’ entries, resembling sticky notes, containing the names of such excluded reviewers, sometimes with additional explanations about the rationale for exclusion. While the boxes seemed to be technically connected with the EMS, they did not show up in the data available to us. Judging by the ethnographic observation, the text boxes did not adhere to a standardized format that a machine could evaluate, but they were easy to understand for a human user. As such, they exploited the mnemonic function of the infrastructure but kept further interpretation for the editors. Even more than in the case of topical matching, the automation of reviewer exclusion (from a specific manuscript) is technologically basic, but it is also apparently viewed as undesirable, not worth the effort to develop, or simply not considered an option.

Rather than relying on infrastructure, editors instead engaged in highly complex and time-consuming nondigital evaluation strategies. These strategies represent forms of relationship labor that are difficult, if not impossible, to perform automatically (Shestakofsky and Kelkar 2020). In the interviews, editors mentioned seeking out personal acquaintance with researchers by visiting labs, going to meetings with research groups, and attending conferences. Editors still place a premium on face-to-face meetings to generate a type of personal knowledge that can be hard to acquire otherwise:

I guess what the experience gives is to know the less obvious biases, so by going to meetings we find out who has certain dislikes or preferences. (Mbio 3b, 23)

Yeah, so the referees can act very differently. So sometimes it does help to know them. [Interviewer and Editor laugh] (Mbio 4, 21)

Editors bring together the tacit knowledge created in face-to-face meetings and colleagues’ informal recommendations with knowledge derived from rereading reviewers’ previous reports to produce a complex and decidedly social appraisal of potential reviewers. These appraisals often revolve around highly implicit information about reviewers’ thematic and stylistic preferences or how their social relations can influence their evaluations:

Some referees will really like very, very biochemical papers or very, very mechanistic papers. (Mbio 3c, 13)

But yeah, I think in that sense…for example, that for a particular researcher you should not go to their colleague X, even if they think they are best friends. (Mbio 3b, 23)

How editors utilize the digital infrastructures leaves them with the lion’s share of the work, which then arguably counteracts other efforts to increase the speed and efficiency of the review process. However, this reliance on humans to select reviewers crucially protects editors’ evaluative authority and their understanding of what reviewer selection means—achieving this through algorithmic reasoning would be quite difficult. Rather than an inherent constraint of the infrastructure, this must be seen as a constraint that is produced through editors’ use practices and their interpretation of what it means to “know” and “select” reviewers. Here, editors are quite successful at establishing their criteria and evaluative processes as the correct approaches, which are processes that the infrastructure cannot (easily) achieve. In so doing, they effectively create and make salient a constraint of the infrastructure (see also Vertesi 2019), one that would perhaps not even be perceptible if editors did not place such an emphasis on knowing their reviewers in this way.

In summary, editors, consciously or by habit, construct a boundary between the technical and the social that places certain tasks, responsibility, and information in the social rather than technical domain. They abstain from recommender systems for reviewer selection and from automated exclusion of reviewers. Likewise, they seek out nondigital sites to perform relationship labor (Shestakofsky and Kelkar 2020) and gain experiential knowledge about potential reviewers. These observations show how editors collectively and individually accomplish their editorial tasks while also keeping the infrastructure within limits, which leaves room for social appraisal of reviewers as a form of interpretative space for editors.

Contextual Factors of Editorial Work

The editors we observed established a division of labor between themselves and the infrastructures which, while it burdens the editors with most of the work and responsibility, also protects their evaluative authority as experts. Perhaps surprisingly, editors nonetheless describe the infrastructures as genuinely helpful. Such positive attitudes toward the digital infrastructures stand in marked contrast to Christin’s (2017) observation of buffering, in which users take a critical stance toward digital tools and (partly) reject them. It also differs from other accounts of EMS use, which all express a fundamentally critical view of these infrastructures (Taubert 2012; Gershon 2018; Horbach, Ochsner, and Kaltenbrunner 2022). With Vertesi (2019, 378), we can explain this contrast as contingent upon the context of the organizational practices in question. We have identified four sets of conditions that jointly shape how editors engage with the infrastructures.

First, the editorial work in our research site takes place under specific local and spatial conditions. With the shared and open office space, which contrasts with the remote work model described by both Gershon (2018) and Horbach, Ochsner, and Kaltenbrunner (2022), editors can draw on their shared physical space as an additional infrastructure to accomplish modes of collaboration that happen outside and independent of digital infrastructures. Their physical colocation allows for formats such as the daily editorial conferences or informal conversation, in which they can quickly and with little effort ask a colleague at the neighboring desk to recommend a reviewer, for example. By contrast, if the EMS represents the main or even only space for editorial work, as in the cases of Gershon (2018) and Horbach, Ochsner, and Kaltenbrunner (2022), editorial work will obviously be a lot more dependent on the functions and affordances of the digital infrastructures. The shared space thus provides editors with the opportunity to both configure the digital infrastructure and to circumvent it.

This shared physical space secondly also creates specific social conditions for the editorial work in our case. Editors achieve a high level of interaction density and collegiality with their coeditors, which enables them to collectively as well as individually reserve tasks for themselves and limit the use of the infrastructures. This is very different in Gershon’s (2018) account of her experiences in the highly problematic power dynamics that unfolded in an anthropology journal. She explicitly cites the lack of contact between editors outside of the EMS as one of the main reasons editors failed to challenge the strongly hierarchical organization of the journal that was encouraged by and inscribed in the EMS. ⁴ In our case, by contrast, editors have ample opportunity to build social relationships outside of and away from the digital infrastructure, thus enabling them to develop social relationships that were not planned, or not planned in this way, within the infrastructure. Together, spatial and social conditions enable editors to maintain their evaluative autonomy vis-à-vis the infrastructure.

Third, our case also exhibits specific organizational conditions. Compared to volunteer academic editors, editors we observed are employed to conduct their editorial work, allowing them a more stable position in relation to both journals and publisher. As regular employees, they are members of the formal organization of the publishing house, and their roles are defined and regulated in large part outside of the digital infrastructures. This also gives them a lot more power in negotiating with the publisher and, for example, enables them to shape the internally developed reviewer database according to their needs. In contrast, Horbach, Ochsner, and Kaltenbrunner (2022) discuss their experience as guest editors with an automated suggestion tool for reviewer selection. In their case as volunteer guest editors, the EMS featured as the main point of connection between the editors, the journal, and the publisher, effectively limiting the power guest editors had over the infrastructure and their engagement with it and making it difficult for them to advocate for infrastructures that would fit their needs better.

The fourth and last set of conditions concern the values of the wider academic field. In biomedicine, acceleration and efficiency are widely held values in academic publishing irrespective of infrastructures. These values then align with those inscribed in the infrastructures, resulting in editors perceiving the infrastructures as a welcome support. Meanwhile, in Taubert’s (2012) analysis in the social sciences, the emphasis on time and efficiency that is created by an EMS is perceived among editors as an unwanted external influence: an encroaching economic logic rather than support for a goal they already want to accomplish.

In summary, the observed social and local context complements other anecdotal accounts of editorial work with EMS. The colocation of editors in a shared office space fosters a high interaction density and collegiality among editors outside of the infrastructure in contrast to remote editors who have less freedom in how they use the digital infrastructure provided. Spatial, social, organizational, and cultural contexts allow the editors in our case to establish ways of engaging with or mitigating the infrastructures that further or hinder their goals and ultimately shape the peer-review process according to their preferences. As a result, editors also express a high rate of satisfaction with those infrastructures.

Conclusion

This paper set out to gain a better understanding of the uses and effects of digital infrastructures in editorial work. When looking at how interactions between editors and the digital infrastructures unfold, we observed three analytically different types of interaction: (1) editors and infrastructure jointly accomplish the acceleration of peer review, (2) editors mitigate the infrastructure when establishing a collective memory, and (3) editors disengage from the infrastructure when they evaluate potential reviewers. These differ depending on how the users delegate responsibilities to software and databases. We argue that editors strategically (dis)engage from editorial infrastructures. However, the said mitigating strategies are not to be misconceived as resistance toward the infrastructure but rather as a mode of interaction which can be more or less integrated with it.

When investigating how tasks and responsibilities are distributed between the infrastructure and the editors, we can conclude that the lion’s share of editorial peer-review labor is still accomplished by human work with only minimal infrastructural influence—especially when it comes to editorial decisions, which require highly complex and nuanced knowledge. Here, our results mirror Shestakofsky and Kelkar’s (2020, 884f) observation that human labor remains indispensable, especially when it comes to relationship labor. Much of what editors do in selecting reviewers and other editorial tasks can be characterized as “community management,” whereby they establish, appraise, and manage relationships between authors and reviewers, acting as ringmasters of their academic field. Here, relationship labor as social appraisal is reserved for human editors.

Editors, in fact, mitigate infrastructure to keep the majority of their work within their own expert domain rather than letting the infrastructure take care of it. They engage in ways that strategically ignore infrastructural functions or decouple them from editorial work and that enables them to configure the infrastructures in such a way that they retain evaluative autonomy, as well as most of the relationship labor involved in peer review. On the other hand, by choosing to have the EMS infrastructure support them, editors do not only reduce the weight on their shoulders, but also lose some degree of freedom.

As for the role of contextual factors in shaping these use strategies, our results illustrate the importance of physical, social, and organizational contexts for setting effective mitigating strategies, and even making them possible (see also Vertesi 2019). As full-time, professional staff in a shared office space, editors hold resources and tools that sit outside of the digital infrastructure. As a result, they can use infrastructures in ways that align with their goals and needs in the editorial process, and they express a high degree of satisfaction with the infrastructures available to them. This is different from Christin’s (2017) finding that experts critique infrastructures as a way to buffer their effects. Our findings also differ from existing critical accounts of EMSs, where editors describe their experiences with EMS as disenfranchising and controlling (Taubert 2012; Gershon 2018; Horbach, Ochsner, and Kaltenbrunner 2022). Our results strongly suggest that rather than an inherent feature or consequence of an overbearing infrastructure, such (dis)enfranchisement is produced through an interplay of digital, physical, organizational, and cultural factors. Arguably, the varieties of modes of interaction we have observed in our case are present because users have choices: the editors’ relative freedom in infrastructure use arise from the local and social arrangements in the editorial office.

It is safe to assume that in years to come, digital platforms and automated tools will only become more important in editorial offices as in other workplaces—a trend that accelerated with the sudden rise in remote work during the COVID-19 pandemic. Yet our case suggests that this development will take on heterogeneous forms and yield uneven effects across different areas and sites of academic publishing. In particular, our findings emphasize the importance of additional, often nondigital, infrastructures, and organizational contexts in shaping how digital infrastructures are used and what effects they generate in editorial and other types of work. If we want to ensure the ongoing digital transformation of academic publishing does not result in disenfranchising human users (editors, reviewers, and authors alike), we should pay specific attention to these factors and contexts that condition the power human users hold in their interactions with digital infrastructures.