Authors:
Antonia Krummheuer, Kristina Tornbjerg Eriksen
The integration of robots into workplaces is often touted for its potential to streamline processes and enhance efficiency. Their deployment, however, generates a substantial amount of invisible work for the individuals expected to benefit from such technological advancements. The term invisible work refers to the tasks and activities that are essential to the functioning of an organization or a system but are not formally recognized, documented, or compensated [1]. Often, individuals doing invisible work are not even aware they are doing it, because they have implicit knowledge, and it is deeply integrated in organizational routines and norms. This article delves into the invisible work that accompanies the use of robots in two different real-world healthcare settings.
Healthcare robots are seen as one of the key technological solutions for the demanding challenges of the system. In contrast to laboratory and testing environments, healthcare settings were not initially built with robots in mind. They are organized according to the social-material infrastructure and are inhabited by various actors performing a variety of tasks (e.g., visitors seeing patients, porters running errands, nurses taking care of patients). Robots that enter healthcare settings encounter the complexity of organizational work practices. Consequently, they will interfere with existing invisible work and will cause new invisible work. The following two examples—mobile logistics robots in a hospital and reminder robots in a life-support and rehabilitation context—illustrate that kind of invisible work, which became obvious in ethnographic explorations of human-robot collaboration. We will demonstrate and discuss how the observation of real-world human-robot collaboration enables us to highlight some of the invisible challenges when deploying robotic solutions in real-world settings.
Insights
→ Stakeholders' ideas about how a robot should work do not match the complexity of the invisible work needed to make a robot operate successfully.
→ Robot deployment needs to shift from focusing solely on technology to understanding the practices of various actors collaborating with, working alongside, or actively opposing the robots in real-world contexts.
Mobile Robots and the Invisible Work of Navigating Hospital Corridors
The first example reveals the invisible work involved in the deployment of two mobile robots in a hospital. Kristina Tornbjerg et al. [2] conducted an ethnographic field study, observing mobile robots moving carts with cutlery, glasses, and dishes in the basement of a Danish hospital. The study uncovered significant challenges on both material and social levels. Initially introduced to alleviate kitchen staff's workload, the mobile robots faced practical limitations, such as navigating narrow corridors, maneuvering around objects, and recalibrating paths when obstructed by humans who appeared suddenly. The hospital staff delayed their tasks, because they had to mind the robots, including removing obstacles to allow the robots to move freely. Some staff would also follow the robots, ensuring that they did not stop due to material or human hindrances, or to prevent sabotage caused by frustrated porters. Other new tasks introduced by the robots were positioning service carts for collection and continuous monitoring during task execution. These observations reveal a disconnect between expectations and reality: While management saw the robots as time-saving aids, the staff spent considerable time assisting the robots. Consequently, some staff members preferred not to collaborate or interact with the robots due to perceived inefficiencies.
The deployment of robots often brings to light overlooked challenges and complexities in real-life practices.
Despite the assumed simplicity of tasks, a web of invisible processes and interactions significantly influenced daily workflows and human-robot collaboration. Figure 1 illustrates the formal division of labor between robot and staff, as described by the kitchen's deputy manager. Following the mobile robots, however, uncovered a different reality. What was initially perceived to be a simple seven-step workflow for mobile robot task execution (Figure 1) turned out to be a complex series of 28 steps during actual operation (Figure 2).
 | Figure 1. The seven-step workflow for robot task execution as described by management (originally published in Tornbjerg et al.). |
The robot's navigation of hospital corridors interfered with other activities, demonstrating the invisible work of both humans and robot. While steps 1 and 2, and partially step 3, are identical in Figures 1 and 2, differences become apparent after that. For instance, in Figure 1 the robot approaches the service cart in step 4; in Figure 2 this does not occur until step 12. Meanwhile, the robot encounters a porter's service truck. Steps 4 to 11 (Figure 2) involve unmentioned activities (invisible work) in which the robot and porter manage navigation problems, including alert sounds, stopping, waiting, recalibrating, moving forward, and managing frustration (step 9). Similar navigation issues are observed in steps 13 to 19 and 23 to 25, delaying staff activities and causing frustration and anger (steps 16 and 24). At the end of the tour, the narrative matches the observation again: Steps 6 and 7 in Figure 1 correspond to steps 27 and 28 in Figure 2.
 | Figure 2. The 28-step workflow as observed when following the robot (originally published in Tornbjerg et al., highlighting added by the authors). |
Tornbjerg et al. [2] revealed numerous invisible events from unforeseen interactions between kitchen staff and robots that significantly affected collaboration. Seamless task execution in hospitals requires not only physical capabilities on the part of robots but also subtle adjustments by staff to accommodate them, such as clearing pathways, coordinating schedules, and providing manual assistance. Staff had to adapt their workflows and behaviors to address the challenges of robot integration, leading to misconceptions about the effectiveness of robotization. The kitchen staff invested significant time and effort in supporting robot operations, even though the robots were introduced to reduce the workload.
This indicates that the promise of simplifying tasks did not fully materialize. In addition, a dynamic hospital environment requires procedures to constantly evolve, adapting to changing circumstances and interactions between humans and robots. Finally, the intricate coordination of interaction between staff, robots, and the hospital environment is vital.
Reminder Robots and the Invisible Work of Reminding
The second example centers on the invisible work of reminding processes that is relevant for the development and deployment of reminder robots in healthcare settings. Matthias Rehm and Antonia Krummheuer [3] engaged in a participatory project to develop guiding and reminder robots for, and with, people with traumatic brain injuries and their care personnel. The project was based on an iterative series of participatory cocreation workshops exploring individual and organizational needs and practices, envisioning the robots' look and function, and, finally, building and deploying the robot prototype.
Rehm and Krummheuer point out that reminding is identified as one key function of robots in a healthcare context. While a clear definition is missing, most research seems to be based on a cognitive understanding that a reminder is a signal that triggers memory in a person. This cognitive approach neglects the invisible communicative work needed for executing reminding in healthcare, which the authors deem necessary for the successful development and deployment of reminder robots.
Their argument is developed by reflecting on the failure of one of their newly cocreated reminder robots when it was tested in real life. Several confirmation activities and positive tests occurred during the cocreation workshops. But when the reminder robot was deployed in real life, a person with a brain injury did not recognize the robot's reminder and became so agitated that she was unable to attend the appointment the robot had reminded her about. The authors revisited their video data of the cocreation workshops from an ethnographic and practice theoretical perspective, looking at the invisible but obvious work of how reminders are achieved by the participants in interaction with each other and the robot.
We should acknowledge the influence of invisible work in order to more successfully deploy robots.
They discovered that a central invisible assumption led the robot to fail, as they based the reminder robot on a cognitive idea of a reminder as a signal that triggers memory (Figure 3). This perspective has a strong focus on reminder technology and does not encompass the invisible communicative work of doing and preparing reminding in collaborative settings with care personnel and its organizational framing.
 | Figure 3. Cognitive perspective on reminder as a signal (originally published in Rehm and Krummheuer). |
The data review revealed a much more complex process in which reminders are prepared within organizational contexts and are individually adjusted by care personnel, depending on the individual preferences of the person they are reminding. Figure 4 visualizes the sequential unfolding of reminders over a period of time, across different stakeholders and activities, starting with the identification of, and commitment to, a reminder, followed by preparation work, such as scheduling of the appointment in a calendar and the organization of transport. Furthermore, they pointed out that reminders are not just signals but also individually constructed actions that have different communication forms. Depending on their distance from the appointment, care personnel use successive pre-reminders (e.g., talking about going to the hairdresser that day during morning routines, using an acute reminder when fetching the resident to go to the hairdresser [4]). As such, reminding is embedded in the joint activities of an individual and care personnel and is framed by the social-material ecosystem of a healthcare context, its daily routines, knowledge about individuals, and communicative practices.
 | Figure 4. Process model of reminding (originally published in Rehm and Krummheuer). |
This knowledge can be identified as invisible work: It is clearly visible in the data, and people recognize it when it is pointed out. Stakeholders were able to identify general routines (e.g., what appointments were necessary and how they should be scheduled) and the individual preferences. The invisible work needed to accomplish the activities, however, became visible only when the researchers treated the video-recorded interaction of the cocreation workshops as ethnographic data, taking a perspective of reminding as social practice.
Discussion
The two examples provide valuable insights into the complexities of human-robot collaboration within dynamic work environments, particularly in healthcare settings that are not specifically designed for robotic integration. Despite promises of optimized workflows and enhanced efficiency, the deployment of robots often brings to light overlooked challenges and complexities in real-life practices. These challenges are rooted in the social-material interplay of various actors and tasks, necessary objects and technologies, and the special arrangement of the setting and organizational routines and norms. The challenges encountered in both examples highlight the need to further investigate the invisible dynamics between humans and robots in complex workplace settings, starting at the development phase.
To enhance the effectiveness and acceptance of robotic solutions, a shift in focus is required: from a technology-centric perspective to one that prioritizes the social and material contexts of work environments. The examples demonstrate that robots enter an already inhabited environment with existing routines that center on the organizational work that needs to be done and not on the robot's task (for public spaces, see Hannah Pelikan et al. [5]). This happens, for example, when robots navigate a hospital corridor in which other people run errands or enter a reminding process that is more complex than the pure production of a signal at a certain time.
Observing real-world practices enables us to unpack the invisible tasks that are glossed over in descriptions of a task. The robot is not just driving along the corridors: This navigation involves the mutual recognition of being aware of people encountering the robot and detailed coordination of activities to avoid collisions, such as waiting and giving way. Similarly, reminding is not just a signal. Rather, it is a situation-adjusted communicative practice that considers individual preferences and organizational contexts and can differentiate between pre-reminders and actual reminders.
How the robot invades people's workspaces and how people adapt to this invasion is of central relevance for the future deployment of robots, as it affects how people collaborate with, work alongside, or actively oppose the robot. Invisible work, however, poses two challenges to deployment processes. First, it is context sensitive. It is difficult to predict how a work environment will adapt to a robot before its deployment; therefore, it is difficult to predict how the robot should be tailored to the work environment. Second, invisible work is embedded in activities that we take for granted. Tacit knowledge is difficult to capture using classical talk-based methods, such as interviews and questionnaires. We should acknowledge the influence of invisible work from the beginning of the development process in order to more successfully deploy robots. Ethnography and practice theory are well equipped to uncover and map the taken-for-granted knowledge embedded in work practices. As such, we have the tools that will help us handle the challenges; we just need to apply them.
Conclusion
Our research has unveiled discrepancies between managers' and developers' visions of a seamless deployment of robots to make work more efficient and the complex realities encountered by hospital staff and healthcare personnel. We highlight the invisible work of humans (and robots) in accommodating each other's routines. While techno-enthusiasts and managers may have high hopes for robotic solutions, the findings underscore the necessity of understanding the complex interplay of human-robot collaboration and the nuanced experiences of individuals encountering robots in dynamic work environments. Ultimately, the research emphasizes the importance of mapping out the invisible work that accompanies work practices and how this work will be affected and what kind of invisible work will be added by the deployment of robots in the workplace. By gaining a deeper understanding of the challenges and complexities inherent in human-robot collaboration in real-world settings, organizations can make informed decisions about the integration of robotic technologies.
References
1. Star, S.L. and Strauss, A. Layers of silence, arenas of voice: The ecology of visible and invisible work. Computer Supported Cooperative Work 8, 1–2 (1999), 9–30.
2. Tornbjerg, K., Kanstrup, A.M., Skov, M.B., and Rehm, M. Investigating human-robot cooperation in a hospital environment: Scrutinising visions and actual realisation of mobile robots in service work. Proc. of the 2021 ACM Designing Interactive Systems Conference. ACM, 2021, 381–91.
3. Rehm, M. and Krummheuer, A.L. When a notification at the right time is not enough: The reminding process for socially assistive robots in institutional care. Frontiers in Robotics and AI 11 (2024), Article 1369438.
4. Amrhein, A., Cyra, K., and Pitsch, K. Processes of reminding and requesting in supporting people with special needs: Human practices as basis for modeling a virtual assistant? Proc. of the 1st Workshop on Ethics in the Design of Intelligent Agents. RWTH Aachen, 2016, 14–19; https://ceur-ws.org/Vol-1668/
5. Pelikan, H.R.M., Reeves, S., and Cantarutti, M.N. Encountering autonomous robots on public streets. Proc. of the 2024 ACM/IEEE International Conference on Human-Robot Interaction. ACM, 2024, 561–71.
Authors
Antonia Krummheuer is an associate professor of qualitative methods and technology studies in the Department of Communication and Psychology at Aalborg University in Denmark. She is a sociologist, and her research interests include exploring human-robot interaction and assistive technologies and combining video/ethnography and interaction analysis with human-centered and participatory design. [email protected]
Kristina Tornbjerg Eriksen is an assistant professor in techno-anthropology and human-robot interaction (HRI) in the Department of Sustainability and Planning at Aalborg University in Denmark. She employs a multidisciplinary approach, integrating perspectives from anthropology, HRI, and science-technology studies to enhance the understanding of robotic integration in real-world environments. [email protected]
This work is licensed under creative commons attribution international 4.0.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2025 ACM, Inc.
Post Comment
No Comments Found