Advancing the Value of Ethnography

Designed for Care: Systems of Care and Accountability in the Work of Mobility


Cite this article:

2018 Ethnographic Praxis in Industry Conference Proceedings, pp. 334–350, ISSN 1559-8918,

In this paper we explore the idea of a system of care through a city transit system. We argue that a systematic orientation to care is central to what makes a transit system work for people. Further, we suggest that this care orientation is recognized as such, even though it is not apparent in typical modes of systems management. Care is what knowing in this system is for. We examine how participants in the system navigate different epistemic bases of their work, focusing on how care work and information work intertwine. How is this work practiced and known? And how could we, as design researchers, use these practices to design systems of care? In service of these goals, we expand the notion of care work toward care of non-human actors as well as that of people. We focus particularly on the roles of automation and the risks automation presents for care. In a moment of increased automation in the workplace, what happens to the care of people and things? We argue that the systemic aspect of care, operating at multiple scales toward people and things alilke, is important for maintaining the goods the organization seeks to produce. And we propose a list of critical questions to ask when designing new systems to shape their orientation toward care.

In the logic of care exchanging stories is a moral activity in and of itself. But moral activities do not restrict themselves to talk, to verbal exchanges. They also come in physical forms. (Annemarie Mol, The Logic of Care, 77)


The radio crackles to life with the voice of a bus operator: “I slid tryin’ to make my turn onto Horsetooth and now I’ve got the median in front of me. I need assistance to back up.”

Inside a low set of office buildings and garages painted in tan and beige, several people sit in a control room in front of telephones, radios, and screens. These bus dispatchers are responsible for the operation of all the TransitService buses for this small city in the Western United States. It is clear to everyone present there that the call from this bus operator is potentially serious. But no one in the dispatch center has a clear idea of what has actually happened out at the intersection.

After a few beats of silence, the staff spring into action. One dispatcher summons a road supervisor to a special radio channel. Another asks the operator for more details, finding out that the bus is blocking two whole lanes. The dispatchers work to sort out the situation and find help. Twelve minutes later the situation is resolved. Once a spotter arrives on the scene it takes the bus about 10 seconds to back up and return to its normal route. But the process by which this result was achieved is deeply instructive for how the system operates.

During these 12 minutes, the road supervisor started a long trek across town in rush-hour traffic to get to the scene. Another bus operator who had just returned to base after finishing his shift was sent back out toward the incident in a company car with the thought that he might get there sooner than the supervisor. A dispatcher called the police to inform them of the incident. The dispatchers and the bus operator deliberated about whether it was safe to put out cones and to let a passenger off the bus while it was sitting in that location. The very reason the bus could not back up was itself for the safety of others: operators in this system are not allowed to back up a bus without a spotter, not because it is impossible, but because it is considered too risky for others in the vicinity. In the end it was yet another bus operator returning home from his shift who was first on the scene and provided the decisive help. And that was not by chance—a dispatcher knew this operator, and knew when, where, and how he would be returning home. That dispatcher called him directly to ask for his assistance.

As this event demonstrates, TransitService, like many service organizations, is distributed in space and deeply interdependent with infrastructures beyond its sole control. Corporations today are increasingly designing large-scale service systems—from ubiquitous electronic communications to expanding ride-sharing—that are interlocked with infrastructures managed by other parties. And service work within these systems is increasingly automated or eliminated, often displaced onto the users themselves. But we see in this particular mobility system a counter-argument to the prevailing logics of efficiency.

The overarching motivation for solving this problem was clearly to get the bus moving again. But throughout, there was a marked attention to the care of employees, equipment, and members of the public, all legitimate and important work that service organizations perform that is not directly business labor.

We selected the dispatch center as a site for our research with the objective to understand how distributed mobility systems operate, and how they might operate using increasingly automated vehicles. We were especially interested in the role and function of centralized control centers, here, Dispatch. Orchestrating the movement of people and vehicles around a city requires gathering information from a variety of perspectives: from the ground-level viewpoints of buses to the overview visions of dispatchers, to everything in between. The data that are gathered and tracked allow a mobility system to operate in a world that can be known only imperfectly and at a distance, mediated through recording devices, radios, and employees. But we found that the self-contained logic of information systems that reinforces practices of measurement and data collection—feeding the system with yet more information—is not enough to ensure that the goals behind the system can be met. In other words, information is not enough for mobility to be achieved. Another system is also necessary, and serves as the lifeblood of mobility services provided to people and publics. We describe this parallel and intertwined system as a system of care.

The system of care strongly contributes to the achievement of mobility: it organizes and shapes the purposes of knowing and measuring in this site. Actors seek to know the system in order to care for it and the people it serves. And this care overflows traditional regimes of measurement, prompting questions about how care itself can be known, assured, or documented (Strathern 2011, Adams 2015). Ethnographers expect to find ways, both helpful and harmful, in which work practice does not match up with regimes of measurement (Suchman 1995, Cefkin, Thomas and Blomberg 2007). But we extend this perspective to show that these practices may in fact become part of a systematic organizational approach to carrying out activity in the world, with implications for the design of increasingly automated service systems.

For us as researchers, seeking to conceptualize how systems to perform mobility-work could be automated, the systemic nature of care confounds traditional ideas of how this labor could be formalized into computational procedures.

This paper is a call to ethnographic researchers, designers and strategists to take seriously the multifaceted system by which caring for people and things happens. In this paper we raise specific questions about what it means to design for care. Our particular interest lies in ensuring that increasing automation does not erase these desirable properties of caring systems.


The authors conducted one week of intensive fieldwork in the bus system of a medium-sized city in the Western United States. We collected fieldnotes in the dispatch center, during ride-alongs on buses, and in a number of staff meetings. We interviewed key staff members. And we collected photographs and hours of video and audio recordings per day in the dispatch center and on bus rides. Our overall engagement with the site was somewhat longer, including several calls in advance to scope the observation and identify useful sites to observe, and some follow-up to collect more documents to help us understand the service and its activities.

We analyzed the material with attention to how workers in this site come to know about the world and guide their actions as part of the bus system, with particular focus on the situated purposes and meanings of these actions. What types of knowledge are valued over others? Why are certain kinds of information collected? What is the affective engagement of employees in the work? Additionally, we analyzed the work-practice dimensions of the material, particularly how employees work together to solve problems. Methodologically, the authors transcribed and coded relevant selections of conversations for these themes, and performed video and interaction analysis using collected audiovisual data. To supplement our own data collection we were given the Operations Manual to study, and have returned to it to put what we saw in context.


In the ethnographic literature, care work appears as processual and relational work between people. Our notion of the system of care builds from Annemarie Mol’s concept of the logic of care (Mol 2008) and extends the way that care work is thought about in large, distributed organizations. In healthcare (Mol’s subject matter), choice appears under the guise of individual decisions and care is achieved over time by informing, providing support, cajoling, working with, or otherwise providing assistance to patients. Mol and others have found that care work is commonly not acknowledged by professional standards and measurement systems even though it is critical to the functioning of organizations (Nardi and Engeström 1999, Star and Strauss 1999, Bowker and Star 1999). Furthermore, care labor often occurs outside the market frame, and is uncompensated, situated in the home, and provided primarily by women (Cowan 1983), though some of this typically household labor is increasingly corporatized and built into business organizations seeking to provide for greater portions of their employees’ needs (English-Lueck and Avery 2014).

Service labor in the market sphere is often rendered invisible or otherwise “veiled” (Kreeger and Holloway 2008). Excavating the real content of the work day unearths all manner of labor that is illegible to official reports and not present in job requirements (Suchman 1995). But making it visible in order to gain official recognition is also a fraught enterprise,1 and measuring its value may be difficult or even impossible (Bryson et al. 2003). The services literature shows one dark side to making this work too visible: the routinization of service work. While standardized practices provide a sometimes-welcome buffer for employees from genuine interactions of care, they concomitantly reduce the space for individual problem solving and judgment (Leidner 1993).

We see the system of care as operating at multiple scales, for people and objects alike. Relational care work is closely tied to the work of customization, maintenance and repair (Star 1999, Darr 2008, Jackson 2014, Anand 2015). And all of these may be part of care work oriented toward larger agglomerations of people and things alike: from users to communities and publics that may be more diffuse and therefore less visible (Amirebrahimi 2006). Numerous studies have shown how transplanted technical systems may worsen rather than improve conditions if not maintained (for instance Mavhunga 2014), suggesting an emphasis on simplicity and reparability of technical interventions (De Laet and Mol 2000). Maintenance is therefore a key entry point for practices of care toward non-humans, which may yet expand beyond material repair into responsible use, preemptive action, and continued monitoring.2 Within infrastructure studies (Parks and Starosielski 2015, Larkin 2013) it is maintenance that allows technical systems to “just work” and be a largely stable and invisible background to a “modern” life (Edwards 2002, Anand 2015). Care and maintenance are particularly interesting here when they involve stewardship of a third space like the roadway, which is beyond the direct responsibility of either an organization or its customers.


Mol argues that the logic of choice is a neoliberal, modernist logic that puts the responsibility for the fate of the individual squarely on his or her own shoulders. So while it aligns with commonly held Western cultural values, it also represents a systematic abrogation of responsibility for individual people, communities, and networks of people and things. The logic of care, on the other hand, involves sustained relationships of responsibility, even if via some level of force. We see a similar dichotomy within the service organization, between care and efficiency. Efficiency, like choice, sounds appealing, but falls short by a similar abrogation of responsibility.

Many important facets of the service and infrastructure-related work described below are not easily amenable to measurement, and measurement tends to distort labor over time to focus more and more on what is measured rather than what is not (Adams 2015). Since so much care of people and things is hard to measure, it seems to face a great risk of being left by the wayside by a focus on automation and efficiency.

To refocus the discussion on the values of care within an organization requires us to look for instances of care in the process of work. What are the signs that this care is happening? How do people in this site know care? Care may be written into policies, procedures, and mission statements, but unless these are lived and practiced they might remain as just so many words. While regimes of accountability seek to provide quantitative evidence for certain kinds of care, from daily checks of bus tires to multi-year infrastructure projects, the caring properties of the system depend on the performance of values that are difficult or perhaps fundamentally impossible to account for by procedural means. Here we attempt to account for them ethnographically, by looking at the meaningful performances and narratives of employees in various roles within the service system: bus operators, field personnel, dispatchers, and support staff.

Bus Operators

Hold on, hold on please! A loud beeping commences as the bus’s ramp lowers into place.

Hey Walter, how you doin’? You get to go first! says the driver, as he gets out of his seat to assist.

An elderly man with a walker gets on Bus 5 and moves toward priority seating.

Haven’t seen you in a couple weeks.

You good? The operator asks as the man gets settles in his seat. Does this [the walker] have brakes? The operator finds and sets the brakes.

Where you getting’ off? … Just ridin’ around?

In many ways, driving is not the core activity of a bus operator. As our informants suggested to us, “anyone can be taught to drive a bus.” This is not to say that safe operation does not require high levels of skill and attention. And indeed the operator’s manual tells that “The most important element of your job as a transit operator is Safety First” (Operations Manual, p. 159). While safe driving is of paramount importance when on the road, even at the cost of other metrics like on-time performance, it is not seen in practice as the distinguishing feature of the operator within this organization.

As routine interactions like the one above show—interactions which happen day-in and day-out for TransitService operators—the role of bus driver is as much about making people safe and comfortable as it is about driving. Operators are responsible for the critical work of aiding customers with mobility issues, such as helping passengers with walkers or wheelchairs board the bus and get secured into the appropriate restraints. This labor combines physical assistive activities that might be automated by a specially-designed vehicle, and communicative work that would be far more difficult, especially when customers themselves have issues communicating.

In the hiring practices of the organization itself, the position of bus operator is primarily a customer service role. The critical skilled task for the operator is interacting with people, a skill that at least some of our informants believed was innate and un-teachable. So they would pass up applicants with existing commercial driving licenses in favor of people with more service experience who would nevertheless require training to be certified as a vehicle operator. Operators must be able to attend to others, chat, cajole, assist, maintain order, and defuse situations. They, like the assistants in a store or the instructors in a classroom, are the stewards of the people who enter their space.

These interactional and service roles set up the operator as more than just a driver. She is a local information source, a help line, and a custodian of public trust, and in that way she joins the ranks of care workers. Operators are instructed by procedures to perform some tasks even when the bus is done for the day, such as to transport stranded passengers if their destination is on the “deadhead” route back to the garage.


Figure 1. Safety is so important that even the restrooms are not an escape from messages about safe driving. But much of what operators do on the road is properly customer service. (Here we have obscured this operator’s face and the service’s logo.)

Moreover, the orientation toward care configures other kinds of work that operators end up doing, much of which is not required by official procedures. For instance, operators may call in to dispatch to request welfare checks or emergency services for incapacitated people at stops or by the side of the road. They also report issues with roadway infrastructure such as street lights and traffic signals that are out, detour signs down in the roadway, and signage issues at bus stops. The operator-bus unit acts as a kind of roving information system, oriented toward service to the community. This kind of informal effort and attention contributes to making the world outside the bus a safer place.

Operator roles blend the care of people on and around their vehicles with the care of infrastructures and vehicles themselves. They are looking and listening for problems or potential problems, even ones that do not directly impact their own position or even the TransitService organization. In practice, the operators’ commitments to care appear to run deeper than what is regulated or required, and at times occur outside of existing regimes of measurement. As the sensory organs of the transit system, they sometimes do not know the significance of the things they report. They do not see the whole picture. But their attention and conscientious reporting are a key part of how the rest of that system, including road supervisors and dispatchers, knows what is going on.

Road Supervisors and Field/Training Coordinators

Dispatch calls out: 500 to an available Road Supervisor, Admin 1 please.

309 Admin 1, Lincoln and Linden, the RS responds, giving his location.

Shit, he’s on the wrong side of town, the dispatcher says as an aside, off the radio. Do we have a TSO out there? She turns to her computer to look.

Another dispatcher picks up the radio: 309, Route 12 says she slid trying to make her turn onto Horsetooth from Taft and needs assistance backing up. She’s currently blocking traffic at the moment. What’s your 10-20?

I’m at Lincoln and Linden, her 10-20? says the RS, requesting the operator’s location.

Horsetooth and Taft.

… There is a ten second pause.

OK I’ll head that way. I’m a bit out, but I’ll head that way, Taft and Horsetooth.

Bus operators are not the only workers who function as eyes and ears for the Dispatch Center. Road Supervisors and Field/Training Coordinators (FTCs) also roam the field to track of what is happening across the area serviced by TransitService. Piloting white vans stocked with essential tools that bus operators would not normally carry, they carry out simple field repairs: un-jamming fareboxes, opening and closing bus stops, and performing accident investigations. They check capacities at stations, deal with problem customers, perform welfare checks on bus passengers, and pull DVRs from buses in the event that accident footage is needed immediately.

As we see in the radio chatter above, the Road Supervisor was the first line of contact when a bus got stuck and needed help to back up, the subject of our opening vignette. Road Supervisors and FTCs perform a range of such ad hoc tasks intended to keep the system on track, such as driving ahead of a bus that is running late (“running down” in TransitService parlance) to pick up passengers at select stops and thereby help the bus get back on time. They may also pick up passengers stranded by a bus that has broken down or missed a stop, or was unable to pick passengers up due to a lack of space on-board. Or, they may drive to bus stops and close them down during a detour, or re-open them after a detour ends, to make changes in service apparent to users trying to catch a bus.

Road Supervisors have a central role in making the physical world match up to the organizational, scheduled world of bus routings. And maintenance and monitoring of the built infrastructure that service depends on are a major part of their responsibilities. Field/Training Coordinators, who work with the Road Supervisors, are in part responsible for producing this organizational culture, which values the respect that operators show to the roadway and to others. This attention comes through in their educational tasks. From continued driver training to leading the bus operator staff in reviewing videos of “near misses” and “good moves” (pulled from bus DVR footage) that demonstrate and dramatize good driving, FTCs are a key part of the maintenance of a safety culture that protects the organization as well as the wider public. Road Supervisors and FTC’s are prototypical examples for how humans’ flexibility and adaptation helps to keep systems operating, and they are essential elements of the transit system of care.


Unit 46 to 700. A bus operator radios in on the Maintenance channel.

473, a dispatcher responds with his operator ID, identifying himself.

Uh, I just got into the South Transit Center off of Fairway, my low coolant light came on, and the bus has officially, uh, shut down.

Copy, so you are still on Fairway, or you actually made it into the transit center? The dispatcher pulls the bus up on the real-time service map.

Uh, I am still on, uh, Fairway just past the old pine going into the Transit Center.

Um, copy, um, probably what they’d have you do first is just shut down the bus, um, just turn it off completely. We’ll just see. I’ll get a hold of 700 and they’ll come on to the channel.

Copy. The operator responds, and works on restarting the bus.

The dispatcher picks up the telephone: Hey Mike this is Rob. We have unit 46 on Fairway, um, said that he had a low coolant light come on, buzzers and everything, and then the bus shut down. So it sounds like he’s possibly dead in the water.

Dispatchers are the control center for the transit organization. They track the buses, anticipate problems, collect information coming in from different sources, make sense of what is happening outside in the space of the street, and, as in the vignette above, connect people in the field who need help with resources that can assist them. Most critically, they are the source of ground truth for the organization. As one dispatcher described their role: “We are here to remember.” In the situation above, the dispatcher answered the operator’s call, got maintenance on the phone and then on the radio, coordinated in person with maintenance personnel as to where they would find the bus, and sent out an operator in a replacement bus. At the same time, she also filed incident reports, told the other operator where to put the bus once he got it started again, logged the operator onto the new bus when it arrived to speed the departure process, and used CCTV cameras to check the progress of the bus swap. So dispatchers are responsible for many kinds of communication and administrative work. They attend to and filter the details coming from different sources and media, and know when the answer to a question needs to change. The dispatch center, the dispatchers told us, is “a big giant filter.” It is responsible for knowing, and informing others: “at this moment, what is true?”


Figure 2. Dispatchers manage multiple streams of information. And they use tools like CCTV creatively to address and preempt problems, and to care for the service as a whole.

Dispatch is organized around a core of adaptable workers, who are oriented toward caring for the organizational system and those that it serves. The work of dispatch requires a kind of bricolage, using tools in nontraditional ways and making the best of what is available in order to provide service. We do not wish to suggest that dispatch has no set processes or standards—indeed they are the prime process-setting and standards-keeping body for the rest of the operational personnel, ensuring that bus drivers, road supervisors, and others are following the rules and meeting expectations. But, at the same time, the labor of dispatch is not at its heart about procedures or metrics.

Dispatchers’ work is therefore a kind of techno-service work, in which they may be managing information technologies one minute, doing administrative work another minute, and giving instructions to an operator at the check-in window the next minute. Dispatchers use a mix of computer systems, phones, radios, printed materials and both informal and formal information channels to sort through challenges. In several instances, public data sources were involved. We observed Dispatch reading Twitter and thus finding out about a potential disturbance at a busy bus stop, a prize drawing set up there as part of a local festival. In another instance that was reported to us, after a bus operator informed Dispatch of slow traffic, the dispatcher used one of the 4 computer screens in front of her to pull up and scan Twitter, finding reference to an accident at a major intersection near that location. In a different kind of situation, an operator reported an unexpectedly early street closing, due to a downtown special event, over the radio. This prompted a cascade of fact finding in Dispatch about what options were available, whether prior information was missed, and how to direct numerous routes around the streets.

This practice of flexible, bricolaged service work includes using tools in nontraditional ways. When we asked about their use of the security cameras to check on the status of the broken-down bus in the opening vignette for this section, one dispatcher noted that they aim to use the cameras carefully, looking only at TransitService property to protect the privacy of others in the community. But she also told us a story suggesting how important these cameras are as a general use tool. Once, when a bad snowstorm came through, most of the routes were out of service. Dispatch could page transit centers via a one-way audio link to make announcements but had no return channel for communications except for the security cameras. So they asked passengers at each station to raise their hands as to which direction they wanted to go, and then used the cameras to manually count customers. They then allocated and dispatched the appropriate vehicles to get the passengers to their destinations.

There are all kinds of metrics implicated in this story, but the service criterion that was most clearly valued by dispatchers as they discussed their decision-making process was not to leave people behind. That desire not to strand passengers extends to directing Road Supervisors or Field/Training Coordinators out in vans to pick up passengers and take them to their destinations.

And all the reports that operators and Road Supervisors or FTCs make about infrastructural problems outside of the jurisdiction of TransitService do actually go somewhere. We witnessed dispatch reporting infrastructure issues to third parties, such as calling a local police department to let them know about problems with traffic lights. The desire and indeed sense of responsibility to report such things appears to us to be a recognition of the organization’s dependence on public infrastructure, and its shared use of the roadway. It is a way of giving back to a shared infrastructure the organization needs in order to survive, as well as a way of making things safer for others. This sort of care work is tied up with TransitService’s sense of good organizational citizenship.

Maintenance, Customer Service, and Management

Rounding out our description of the key organizational departments and employees of the bus system are the maintenance and shop personnel, management, and customer service departments.

Working behind the scenes to keep buses in good condition, the maintenance and shop personnel are a key part of TransitService operations. Of course, they are responsible for doing responsive repair work when bus operators identify issues in their pre- and post-trip checks. But preventative maintenance tries to anticipate issues that may crop up later, and requires flexibility and judgment. Conversations within Dispatch and between dispatchers and repair personnel made clear that buses have distinct personalities. They are not mass-produced, identical, and interchangeable. As dispatchers told us, “the worst thing in the world is a brand new bus,” a bus is “no good until it has been in service for a year.” These buses are hand built, and are all slightly different, so they will also fail differently. The teams of mechanics need to learn the quirks of each bus, and find ways to diagnose specific kinds of problems: one bus has a tendency for the front marquee to get stuck, and the best fix is to hit it from the outside with a broom handle. No manual could ever have provided that information!

There are also all sorts of business-oriented personnel working behind the scenes to keep TransitService running smoothly. Management is in charge of making the bus acquisition decisions, Analysts and IT track buses and passenger loads, Customer Service Representatives answer service questions, and Service Planners identify new routes to better serve the community. Many of these tasks are tied together by the data that make it possible to determine the value of current service. In a tech-focused culture in which machine learning and big data analytics consultants promise almost perfect knowledge of the world via the collection and use of ever increasing amounts of data, it is little surprise that managers at TransitService want detailed knowledge of what their vehicles are doing, how many people they serve, what times are busy and what times are not, etc. And yet, the organization does not collect everything it could collect given the tools it has. Though they collect lots of vehicle data already, they chose not to purchase an engine telematics tool offered by their service management software provider as they felt it would be to “surveillancey” for their drivers’ comfort. How data is used, or perceived to be used, is perhaps more consequential than what the data is (Nissenbaum, 2009). And these decisions about how to use the available resources rely on human judgments about role, responsibility, and relationships that are not amenable to programmatic treatment.

Beyond employee care, customer service is primary in the organization’s vision and mission statements. The first things in the operation’s manual after the table of contents are two pages of vision, mission, and goals. Among them is a page of verse, written about service to passengers, which is again repeated on page 182, in the middle of the bus operator’s section of the manual. This litany goes:

“Our Passengers

Are the most important people in our business

Our Passengers

Are not dependent on us, we are dependent on them

Our Passengers

Are not an interruption of our work, they are the purpose of it

Are not cold statistics; they are human beings with feelings and emotions just like you and me …”

While we are accustomed to thinking of vision and mission statements as aspirational texts that may have little impact on the day-to-day operations of the organization, it appeared to us that TransitService really did attempt to live by these goals. At one of the shift meetings we observed, staff announced that they had “lost one of [their] long-term passengers in a very unfortunate way,” and commemorated the deceased customer with photos on a section of the whiteboard. We observed these mission, vision, and customer statements performed in the everyday comportment of the organization, in ways not fully accounted for by the reductive sets of functional measures and efficiency metrics that an accounting of the organization is likely to stress. The sense that the organization needs to be a “responsible steward of transit resources” comes through in care of communal transportation infrastructures.


These individual situations of care work appear to form a system that goes beyond just the sum of the parts. Individually, a lone employee caring for passengers or equipment will make little difference to the operation as a whole. But care in this organization is systemic: it occurs at all levels within the organization, over varied timescales from days to years, and in ways that support each other across these categories through both formal procedures and informal, individual action. Buses will always have mechanical failures and break down. But multiple ways of attending to the problems of equipment—staggering bus acquisitions in time, replacing vehicles promptly at end-of-service-life, doing preventative maintenance regularly, performing pre- and post- checks daily, informal learning of how to attend to an individual bus’s unique problems, and dispatching appropriate buses to appropriate routes—sum together to balance the needs of buses against the needs of the organization and the publics that it serves. Nothing is run on the ragged edge. And service continues.

The idea that systems require a delicate balance of labor to maintain them is nothing new. But what we show is that much of this work, whatever else it may be, is also care work, and caring work. It involves taking care of people and things. And it is performed with a sense of meaning and responsibility, rather than just by a procedure. We have talked about the system of care in terms of the roles of its constituent people. But it is also possible to turn this view around, and look at what properties and things are cared for. This helps makes clear the systemic and mutually reinforcing nature of the care practices involved.

As vehicles are cared for through fueling, daily checks, preventative maintenance and repair (and the vehicle fleet is cared for through the timing of acquisitions), so too are employees cared for through training, safe working conditions, responsible scheduling and information sharing. Customers are cared for via accurate transit information, safe and clean vehicles, on-time service, and welfare checks. Broader publics are cared for through avoidance of obstructions or delays caused by buses, by road safety, by reporting incidents with other infrastructures, and by responding to public desires.

And above all, these separate sub-systems of care work put together serve to maintain and allow for the movement of people and vehicles. Movement is maintained by coordination and timely communication within the vehicle system, and between the vehicle system and other municipal services. Movement is also maintained through the care of other constituents: the vehicles, employees, customers, infrastructures, and publics for and by which the vehicle system operates. Only because of attention to all the facets of the vehicle system can people continue to move through the system; without this attention to care: buses break down, employees get into accidents, customers are injured, the public is obstructed, and infrastructures degrade and become unusable.

Our research does not show that the system would cease to function as a system without any single one of these components. There is clearly flexibility in the practices of the organization. There are other kinds of work that are not done by this system, and one could have a successful mobility system with only some of these labors and not others. But take away enough of these activities, we think, and the system would collapse like Aramis, abandoned by Matra for a lack of love (Latour 1996). Either its ultimate effectiveness, or at least its qualitative character of competence and professionalism, would be compromised. By saying this is a system of care, we are saying that individual judgments are involved, but something is happening at a level above the individual, in a collective of multiple components that all pitch in to keep the system running.


“You just kind of take it all at the same time and just kinda sort it out, which one’s more pressing. Yeah I dunno, you just kinda do it.”

                    — A dispatcher describing how they manage incoming information.

Our purpose in examining TransitService was to learn about what it takes to move a fleet of vehicles around a city, picking up and dropping off passengers, in order to learn how a mobility service could be automated. Above, we have tried to describe the work of mobility, and the way that that work often overlaps and intersects with types of care. So our next question is: what implications does this have for the design of increasingly automated service systems? How can we use the system of care formulation as material for design?

TransitService stands as an example of the centrality of care to successful service work within larger worldly infrastructures. Someone needs to be watching out for slow-moving, dispersed, or emerging problems, even those that affect other systems outside the purview of the organization itself. What the care lens shows us in this case is that the official procedures of care are not the entirety of the organization’s care work. Within the grey areas beyond and around written procedures, the success of that work also depends on small decisions that are within the realms of individual’s decision authorities. This labor becomes visible to others via co-presence and collaboration—being part of a shared culture over the radio, in the Dispatch Center, and in staff meetings.

The situations in which care work is needed, and the flexible kinds of labor that are in evidence in these situations, make it likely impossible to strictly define and proceduralize all of this work. Real world decision situations present complex choices between often unknown alternatives, and actors are guided not just by procedures but by judgment and values. Increasing amounts of IT-based tools have helped Dispatch manage the chaos of the external world—it was much more disorganized, we were told, back when it was run using Excel spreadsheets rather than purpose-built fleet management software. But Dispatch was not fully automated: dispatchers who are trusted and expected to act with good judgment and autonomy are still a critical part of the organization.

And simply involving a few isolated humans is not enough. The uncaring strictures of human bureaucracies are unpleasant, and powerful, because of their indifference or antipathy to the people they affect (Herzfeld 1992; Graeber 2015). “Just following the rules” leads to helplessness in the face of a system that was not designed for, but against you. What TransitService shows is that care work is successful when humans are backed up by an organizational ethos, a culture, that gives them that autonomy and provides a field of cultural resources by which to orient themselves and from which to draw in order to solve the problems they encounter. As we saw at TransitService, an ethos of responsibility, customer service and risk-management is baked into the procedures, meetings, trainings, and everyday activities. Designing for care means creating a culture that fits the operations, as well as balancing human and machine roles.

Here, thinking about the system as a system of care overlaps with many of the social critiques of capitalism: the problem is that caring so often requires attending to the externalities of the system (in this case, the negative ways that buses wear down roads, or annoy neighborhoods, and the positive ways that they can be used as a platform for community health by reporting people who may be having medical issues near the roadway). And these externalities are not often captured in the value structures of the system because they are too fuzzy, obscure, or far-off to seem to matter.

So the risk associated with the design of increasingly automated service systems is that the positive, caring characteristics of the organization outlined here will be lost. Our research suggests five key questions that emerge as part of a careful design process for automated service systems.

First, is the design investing in skill and autonomy or trying to eliminate it? This intersects with the long debate between Intelligence Augmentation (IA) and Artificial Intelligence (AI). Are human skills being augmented and human autonomy increased, or is work being routinized and labor replaced wholesale? The care lens suggests investing in IA and aiding accountable autonomy. It suggests giving people better tools to handle the uncertainties in the world rather than hiding uncertainty behind a veneer of objectivity, or steamrolling it by inflexible process. As Mol suggests, care can involve restricting autonomy. But design for care needs to be very deliberate about the cases in which it does so, and should do so in the service of larger, processual accountabilities between the system and those around it.

Second, is the design attending to where work will shift? Automation does not eliminate work, it shifts it around. So where does work go, and what will happen if it falls off the map? If bus operators are computerized, who assists the passengers? If Dispatch becomes an automated load-balancing process, who is left to make judgment calls, find a lost item, or inquire for a customer about another service? It is tempting to simplify the system and cut costs by declaring some tasks out of scope, but that work does not disappear simply because the designers of the system wish it would. Design for care should account for where actual work-in-practice, not just theoretical work-on-paper, shifts.

Third, is the design creating a structure that will do the extra little bit to serve people well? Service, even of machines, has a human touch. So how will an automated system go the extra mile for the customer? Design for care needs to figure out a structure around automation that can make exceptions, address problems, and provide facilities for long-term restructuring. For instance, whether a route is technically possible does not mean that it is organizationally feasible, given TransitService’s values. We observed clear concern for not having buses turn around in neighborhoods, and not taking the buses onto streets where the heavy vehicles would cause undue wear. The system needs to be able to behave appropriately on the margins where numbers and procedures are not enough.

Fourth, is the design measuring the right things? And is it using them appropriately, and leaving room for uncertainty and judgment within reasonable limits? Care should be taken as to what is measured, and how it is used. To try to measure everything is tantamount to producing the Borghesian map, as complicated and unwieldy as the full world that it represents. And metrics are always open to being twisted or gamed. Vehicle location is a caring sort of automation when it is then used to track bus times, provide them to the public, and thereby better the overall service. But it has the potential to be a very uncaring one toward operators if it is used to surveil. Certain types of accountability may even be caring toward one group and uncaring toward another at the same time. These are difficult balances to be thought-through specifically, not left to chance. So design for care needs to pick its measures carefully, and combine measures with accountable judgment about their applicability and validity.

Finally, is the success of the design accountable to the unmeasurable goods that it is hoped to produce? If designing for care is ultimately about providing better service, we must recognize service has many meanings and definitions. It may mean different things to different people. And primarily it is going to be defined not by a number of people served, or a length of time in a trip, but by the values that structure the work. So design for care needs to ask: who keeps track of the success of things that seem important, but do not have a quantity or procedure attached to them? Caring means reserving a space for doing the things that are qualitatively “right,” for performing values of the organization, even when the results cannot be measured. And it means determining what those values are, what counts as “right” in a situated context. Here is a place for intervention via “reflective tools” to account for and shift organizational processes in a participatory way (Cefkin 2011). This is a method of keeping the organizational unit attending actively to qualities of its service that are experientially apparent but difficult to quantify.

These questions are not the end of the conversation. They are a starting point. But they should push us to think about care in organizations at an ecological level. As we have shown, care is about things as well as people, infrastructures as well as human relationships. Care is about flexibility, judgment, and values. Care is about balancing assistance and augmentation, autonomy and responsibility, in increasingly automated service work. In designing service organizations as systems of care, we need to take up long-term, responsible visions of the organization’s place in a wider world and ensure that automation and optimization do not wipe these out. Care is a process, not a product: it is in the relationships of living- and working-with that care happens.

Erik Stayton is a PhD candidate in the Program in History, Anthropology, and Science, Technology and Society at MIT. He investigates human interactions with AI systems, and currently studies the values implicated in the design, regulation, and use of automated vehicle systems. He also interns at the Nissan Research Center.

Melissa Cefkin is Principal Researcher and Senior Manager of the User Experience group at the Nissan Research Center in Silicon Valley. She has had a long career as an anthropologist in industry, including time at the Institute for Research on Learning, Sapient, and IBM Research.


Acknowledgments – We thank our colleagues and collaborators in our lab whose ongoing projects and open questions prompted and shaped this research. We also appreciate the detailed commentary and guidance from our EPIC reviewers. But we especially want to extend our heartfelt thanks to all the staff of TransitService, especially the operators we rode with, the dispatchers we observed in action, the road supervisors, managers, FTCs, and everyone else who spoke with us during this project, and was so willing to share about their work and endure the ethnographic gaze. We deeply appreciated the chance to see what you all do, and to render it into this text.

1. “Making work visible” is another way to talk about measuring or accounting. In the background of the literature and the rest of this paper, accounting and measurement are the joint fulcrum that both allows for the automation of work and threatens to overwhelm valuable and difficult to measure aspects of labor by prioritizing the parts that are easier-to-measure. An emphasis on quantification is especially distorting of the kinds of value that can be accounted for (Porter 1995), and it is in part for this reason that an ethnographic perspective is used to uncover these aspects of care work.

2. We think it is critical to expand the definition of care work, in this context, to include the care of nonhuman actors. Maintenance has procedural aspects, but competent maintainers use all their faculties and flexible judgment to anticipate, prevent, diagnose, and repair problems. They also need to maintain social networks at the same time as they maintain machines (Orr 1996). Furthermore, we find in our site a similar affect in the attention to objects and infrastructures as to people. In excess of regimes of formal measurement (such as maintenance schedules), maintenance proceeds by guesses, reasoned judgment, and intuition, motivated by values such as care.


Adams, Vincanne, ed.
2015     Metrics: What Counts in Global Health. Durham, NC: Duke University Press.

Amirebrahimi, Shaheen
2016     The Rise of the User and the Fall of People: Ethnographic Cooptation and a New Language of Globalization. Ethnographic Praxis in Industry Conference Proceedings 2016.

Anand, Nikhil
2015     Leaky States: Water Audits, Ignorance, and the Politics of Infrastructure. Public Culture, 27, 2:305-330.

Bowker, Geoffrey and Susan Leigh Star
1999     Sorting Things Out: Classification and Its Consequences. Cambridge, MA: MIT Press.

Bryson, John, Peter Daniels and Barney Warf
2013     Service Worlds: People, Organizations, and Technologies. London: Routledge.

Cefkin, Melissa
2011     Comment on Jane I. Guyer: Blueprints, Judgment, and Perseverance in a Corporate Context. Current Anthropology 52, S3 (Supplement to April 2011):S17-S27.

Cefkin, Melissa, J. Thomas and J. Blomberg
2007     The Implications of Enterprise-wide Pipeline Management Tools for Organizational Relations and Exchanges, GROUP ‘07: 61-68.

Cowan, Ruth Schwartz
1983     More Work for Mother: The Ironies of Household Technology from the Open Hearth to the Microwave. New York: Basic Books.

Darr, Asaf
2008     The Rise of the Techno-Service Sector: The Growing Inter-Dependency of Social and Technical Skills in the Work of ERP Implementers. Ethnographic Praxis in Industry Conference Proceedings 2008.

Edwards, Paul N.
2002     Infrastructure and Modernity: Scales of Force, Time, and Social Organization in the History of Sociotechnical Systems. In Modernity and Technology, Thomas J. Misa, Philip Brey, and Andrew Feenberg eds. Cambridge, MA: MIT Press.

English-Lueck, J.A. and Miriam Lueck Avery
2014     Corporate Care Reimagined: Farms to Firms to FamiliesEthnographic Praxis in Industry Conference Proceedings 2014.

Graeber, David
2015     The Utopia of Rules: On Technology, Stupidity, and the Secret Joys of Bureaucracy. Brooklyn, NY: Melville House.

Herzfeld, Michael
1992     The Social Production of Indifference: Exploring the Symbolic Roots of Western Bureaucracy. Chicago, IL: University of Chicago Press.

Irani, Lilly
2015     Justice for ‘Data-Janitors. Public Books, January 1, 2015.

Jackson, Steven
2014     Rethinking Repair. In Media Technologies: Essays on Communication, Materiality and Society, T. Gillespie, P. Boczkowski, and K. Foot, eds. Cambridge, MA: MIT Press.

Kreeger, Lisa and Elizabeth Holloway
2008     Now You See It and Now You Don’t: Consequences of Veiling Relational Work. Ethnographic Praxis in Industry Conference Proceedings 2008.

De Laet, Marianne and Annemarie Mol
2010     The Zimbabwe Bush Pump: Mechanics of a Fluid Technology, Social Studies of Science, 30,2: 225-263.

Larkin, Brian
2013     The Politics and Poetics of Infrastructure, Annual Review of Anthropology 42, 1: 327-43.

Latour, Bruno
1996     Aramis, or, the Love of Technology. Cambridge, MA: Harvard University Press.

Leidner, Robin
1993     Fast Food, Fast Talk: Service Work and the Routinization of Everyday Life. Berkeley, CA: University of California Press.

Mavhunga, Clapperton
2014     Transient Workspaces: Technologies of Everyday Innovation in Zimbabwe. Cambridge, MA: MIT Press.

Mol, Annemarie
2008     The Logic of Care: Health and the Problem of Patient Choice. London: Routledge.

Nardi, Bonnie and Yrjö Engeström
1999     A Web on the Wind: The Structure of Invisible Work, Computer Supported Cooperative Work 8: 1-8.

Nissenbaum, Helen
2009     Privacy in Context: Technology, Policy, and the Integrity of Social Life. Stanford, CA: Stanford University Press.

Orr, Julian
1996     Talking about Machines: An Ethnography of a Modern Job. Ithaca, NY: Cornell University Press.

Parks, Lisa and Nicole Starosielski
2015     Signal Traffic: Critical Studies of Media Infrastructures. Urbana, IL: University of Illinois Press.

Porter, Ted
1995     Trust in Numbers: The Pursuit of Objectivity in Science and Public Life. Princeton, NJ: Princeton University Press.

Star, Susan Leigh and Anselm Strauss
1999     Layers of Silence, Arenas of Voice: The Ecology of Visible and Invisible Work, Computer Supported Cooperative Work 8:9-30.

Star, Susan Leigh
1999     The Ethnography of Infrastructure, American Behavioral Scientist, 43: 377-391.

Strathern, Marilyn, ed.
2011     Audit Cultures: Anthropological Studies in Accountability, Ethics, and the Academy London and New York: Routledge, 2011.

Suchman, Lucy
1995     Making Work Visible, Communications of the ACM, 38:56-68.