Dependence and non-intervention: Explaining drivers’ experience of taxi platforms’ algorithmic management in Oslo,Norway

Abstract

While attracting workers by promising autonomy, gig platforms exercise significant control through algorithmic management. This article explores the algorithmic management of taxi platforms in Oslo, Norway and how it is experienced by the drivers. Drawing on extensive fieldwork in the industry, it shows that although the algorithmic management is opaque, unpredictable, and non-negotiable, the drivers nonetheless highlight their flexibility and do not necessarily experience it as control. The analysis highlights two factors explaining these experiences. First, drivers in Oslo have access to multiple platforms and non-platform market segments and therefore remain economically independent of each individual platform. Second, the platforms’ algorithmic management is a non-interventionist form of control, operating primarily through market mechanisms and without intervening directly in the drivers’ labor process. This underscores the importance of market structures in conditioning algorithmic management and emphasizes how platforms are able to provide autonomy at work while maintaining authority over the market.

Introduction

Taxi driving is traditionally associated with significant autonomy (Hodges, 2020). While on the road, taxi drivers are alone in their cars, with no managers telling them what to do, where to drive, how to behave, or even monitoring how they do their job. Behind the wheel, drivers reign supreme. Many, furthermore, also own their own vehicle and can choose when and for how long they want to work. Hence, autonomy is a fundamental feature of taxi drivers’ workdays and labor processes.

In the last decade, however, the rise of taxi platforms, such as Uber, has challenged drivers’ freedom worldwide. This pertains, on the one hand, to how these have “disrupted” taxi markets across the globe, reshaping the taxi industry and transforming the conditions under which drivers make a living (Wells et al., 2021). On the other hand, drivers’ freedom behind the wheel is also restricted by the platforms’ “algorithmic management,” through which managerial control now enters the drivers’ cars. The rise of gig platforms has prompted significant scholarly attention to how these companies utilize the platform infrastructure to control the workers (Wood, 2024), and a dominant theme in the early research on platform work was how algorithmic management was an intensified form of digital control (Kellogg et al., 2020; Lee et al., 2015; Rosenblat and Stark, 2016). More recent contributions, however, have found that while platform workers generally work long hours under precarious conditions for low and unpredictable pay, many gig workers remain, perhaps surprisingly, content with their work (Vallas and Schor, 2024). Previous research explains this by emphasizing that platforms recruit workers from marginal segments of the labor market, for whom these jobs appear as one of few opportunities (Oppegaard, 2025; Wells et al., 2021). Some find that platforms function as a culturally legitimate venture for the downwardly mobile middle class to supplement their other income sources (Vallas and Schor, 2024), while others reference the platforms’ ideological subjectivations and interpellations of workers as “entrepreneurs” and “their own boss” (Barratt et al., 2020; Purcell and Brook, 2022).

Hence, with some notable exceptions (see Cameron, 2024; Maffie, 2022), the literature on platform work generally explains workers’ orientations to the “gig” with factors beyond the labor process. These perspectives are essential for understanding platform work and why the platform model has been so successful. This article, however, shows that there nonetheless are features of the platform workers’ labor processes that play a key role in explaining their sentiments toward the platform. Drawing on extensive fieldwork in the platformized taxi industry in Oslo, Norway, and concepts from the labor process theory (LPT) tradition, I explore how the platforms exercise control over drivers and how the drivers experience the algorithmic management. I show that the algorithmic management is an opaque, unpredictable, and non-negotiable managerial technique, and a significant source of frustration among the drivers. Yet, they nonetheless argue that they have flexibility in the labor process and generally do not perceive the algorithmic management as “control.” Exploring the organization of the drivers’ labor processes, this article offers two explanations for this tension. First, drivers’ experience of the platforms’ algorithmic management is shaped by their dependence on the platforms. Since taxi platform drivers in Oslo use multiple platforms simultaneously and in addition can find customers in other market segments, they do not depend economically on a single platform, enabling a slightly aloof attitude toward their managerial control. Second, the algorithmic management can be conceptualized as what I term a non-interventionist form of control. By operating as market organizers (Kirchner and Schüßler, 2019), taxi platforms use market mechanisms to determine and regulate the conditions under which the drivers work, without having to intervene directly in their labor process. By merely coordinating the market, this managerial control, furthermore, maintains an appearance of neutrality that provides legitimacy.

In the following, I first present the theoretical foundation of the analysis. I then describe the case of the platformized taxi industry in Oslo, Norway, my methodological approach, and empirical material. The analysis is structured as follows: I start by detailing the formal organization of the taxi platform drivers’ labor processes, before exploring three tenets of the taxi platforms’ algorithmic management—trip allocations, economic incentives, and evaluations and sanctions—and how the drivers experience each mechanism. The discussion section presents the two explanations for why the drivers do not necessarily experience algorithmic management as “control”: their independence and the algorithmic management’s non-interventionism.

The article offers three contributions to the literature on platform work and algorithmic management. First, it shows how features of the labor process are important for understanding platform workers’ orientation toward the “gig.” Second, it highlights market structures and economic dependencies as key factors shaping workers’ experiences. Local market arrangements and regulations thereby play a central role in conditioning the effects of algorithmic management. Finally, the article contributes by conceptualizing algorithmic management as a non-interventionist form of managerial control. It determines and regulates the conditions under which platform workers work, operating through market mechanisms rather than direct supervision, thereby appearing neutral and gaining legitimacy. Such a conceptualization of algorithmic management provides nuance to the narrative of intensified control, emphasizing how platforms are able to provide workers with a certain autonomy in the labor process while maintaining authority over the market.

Managerial control and algorithmic management

In the capitalist labor process, labor process theory (LPT) argues, capitalists buy only workers’ capacities to work, their labor power, and therefore exert control to convert potential labor into concrete labor. Workers respond to managerial techniques, prompting dynamics of control and resistance in the workplace (Thompson, 2010). All workplaces therefore balance control and legitimacy (Wood, 2021), and managerial strategies vary according to multiple factors, such as organizational features, technologies, levels and types of worker resistance, and competitive dynamics (Thompson and Laaser, 2021). Managerial control might also aim to extract consent from workers (Burawoy, 1979) or provide autonomy at work while maintaining managerial authority over the process as a whole (Friedman, 1977). Still, no form of managerial control, LPT holds, can successfully stabilize the antagonistic capital–labor relations (Littler and Salaman, 1982: 264), and in practice, multiple and sometimes contradictory forms of control are often combined into hybrid control regimes (Joyce and Stuart, 2021: 162).

LPT has been an important conceptual starting point for research on platform work and algorithmic management (Tassinari and Maccarrone, 2020; Veen et al., 2019). In the early phase of research on gig platforms, algorithmic management was often described as a new and intensified mode of control (Lee et al., 2015; Rosenblat and Stark, 2016). Kellogg et al. (2020: 388), for example, write that algorithmic management “can be more encompassing, instantaneous, interactive, opaque, and disintermediating than the historical regimes of control that employers have used over the past two centuries,” while Veen et al. (2019: 401) describe how algorithmic management enables “platforms to produce tightly managed labour processes with marginally attached workers.” Others argue that algorithmic management can be seen as a “digital Taylorism” (Altenried, 2019; Brubaker, 2022: 119), standardizing, surveilling, and deskilling labor processes. Others highlight the opacity of algorithmic management (Jarrahi et al., 2021), which has been described as a trigger for worker mobilization (Tassinari and Maccarrone, 2020) and conceptualized as an “invisible case” experienced by workers as “as a system of control characterized by unpredictable updates, fluctuating criteria, and lack of opportunities to improve their scores” (Rahman, 2021: 947). Griesbach et al. (2019) use the notion “algorithmic despotism,” while Muldoon and Raekstad conceptualize platforms’ managerial control as “algorithmic domination” and argue that this control establishes relationships of domination between workers and the platforms, increasing “computational asymmetries between bosses and workers and allow the former to intervene at a more minute level in ways that are not feasible if required to be undertaken by a human supervisor” (Muldoon and Raekstad, 2023: 589).

Despite the intensive and extensive managerial control exercised through the algorithmic management, empirical examinations have increasingly found that gig workers nonetheless often express appreciation for the job (Vallas and Schor, 2024). Platform workers sometimes largely consent to the platforms’ practices (Purcell and Brook, 2022) and adopt an entrepreneurial orientation toward the job (Barratt et al., 2020; Wood and Lehdonvirta, 2021). Additionally, research emphasizes how platform workers can both successfully navigate these managerial techniques (Cameron, 2024) and actively resist the platforms’ control (Joyce and Stuart, 2021; Tassinari and Maccarrone, 2020).

This shows that workers’ experiences of algorithmic management are more varied than what Vallas and Schor (2024) call the “subjugation narrative,” which focuses on how algorithmic management deepens capitalist control over the labor process. The variety of effects of platforms’ algorithmic management is generally explained with reference to factors beyond the labor process. Algorithmic management has been described as contingent on workplace regimes, and thereby shaped by local regulations, market structures, and institutional factors (Wood, 2024). Within this strand of the literature, the platforms’ control is found to be irreducible to algorithmic management, relying on organizational techniques, such as classifying the workers as self-employed contractors and paying them on piece-rate models (Moore and Joyce, 2020). Gig platforms also draw on normative techniques, rhetorically promoting autonomy (Barratt et al., 2020), arguing that their arrangements enable workers to be “entrepreneurs” (Purcell and Brook, 2022). This reflects a broader tendency within LPT that analyzes managerial control in the context of political dynamics and factors beyond the labor process (Burawoy, 1979)

Some contributions indicate that nonetheless there also are features of how algorithmic management shapes how it is experienced by workers. In contrast to the “digital Taylorism” narrative, Wood et al. (2019: 64–65), for instance, find that algorithmic management among remote gig workers does intensify the labor process, but nonetheless simultaneously affords them significant autonomy and discretion. Cameron (2024), studying taxi platform drivers, argues that algorithmic management provides workers with structured yet frequent choices in the labor process, fostering a sense of agency and what is termed “choice-based consent.” Stark and Vanden Broeck (2024), furthermore, contrast algorithmic management to earlier forms of managerial control, as it is specifically designed to exert control in organizations without boundaries and coordinate decentralized actors, targeting flows of information to facilitate “self-management.” These contributions emphasize, as Kirchner and Schüßler (2019) argue, that platforms operate as market organizers, and that a key function of algorithmic management is to establish and regulate markets. According to Kirchner and Schüßler (2019), there are five elements in the platforms’ market organization. They first regulate the terms of membership, granting users access to the market without requiring the same contractual commitment as a traditional employment relation. Second, platforms determine the rules with which users must comply and enforce these through their algorithms. The third element is the platforms’ monitoring of conduct through data collection and, emblematically, evaluations from other users. Fourth, platforms sanction users by revoking membership (excluding them from the platform) or restricting access to certain features (such as bonuses). Finally, platforms establish a hierarchy wherein they, as market organizers and owners of the technical infrastructure, maintain a dominant position, giving rise to a fundamentally asymmetrical power relation between the platform and its users.

Although algorithmic management has received significant attention in the literature on platform work, the above discussion demonstrates that there still is need for empirical analyses of how this managerial control operates and is experienced in practice. In exploring how taxi platforms in Oslo exercise algorithmic management and how this control is experienced by the drivers, the analysis shows that while it causes significant frustration, drivers still hold that they have significant flexibility at work and do not necessarily experience it as control. This finding provides further nuance to the “subjugation narrative” dominant in the early phase of platform work research. In explaining this finding, the article builds primarily on the perspectives described above where the effects of algorithmic management are analyzed in light of how the workers’ labor process is organized, highlighting the drivers’ independence from the platforms and the algorithmic management as a non-interventionist form of control.

Case, methods, and data

In November 2020, a liberalization of the Norwegian taxi market was enacted, repealing numerical restrictions on taxi licenses and the requirement for taxi owners to be affiliated with a dispatch center. This lowered the barriers of entry significantly, leading to a supply shock of steep increases in the number of licenses issued (Oppegaard et al., 2023). When Uber first launched in Oslo in 2014, it soon had to withdraw its most “disruptive” service, Uber Pop, which allowed drivers to use their private car to transport passengers via the platform, after over 130 drivers were sentenced for providing illegal taxi services. Uber continued to offer Uber Black, and relaunched multiple services after the liberalization of the taxi market, followed by Bolt in early 2021 and Yango a few months later. These platforms have since emerged in all major Norwegian cities.

In contrast to other segments of the Norwegian labor market, the Norwegian taxi industry is characterized by weak unions and has always been structured around self-employed taxi owners whose earnings depend on fares. This allowed the taxi platforms to develop within the established work arrangements of the industry (Oppegaard et al., 2025). The taxi owners own the cars and have to obtain a taxi license. They sometimes employ additional drivers to cover certain shifts, paying a commission on each fare, usually between 40 and 50%. When the taxi platforms entered the liberalized Norwegian taxi market, both new and already established taxi owners as well as their drivers could register on the platforms. Hence, the taxi platform drivers in Oslo are authorized as taxi drivers, with licensed cars equipped with traditional roof lights. This enables drivers to find customers through the taxi platforms as well as in non-platformized segments of the taxi market, such as street hailing and taxi stands, referred to as the “street segment.” It also makes the taxi platform drivers in Norway relatively professionalized compared to their colleagues in many other markets (Cameron, 2024; Maffie, 2022).

The emergence of taxi platforms extended many drivers’ customer bases by creating and giving them access to a new market segment but did not structurally alter the drivers’ working conditions. The work arrangements were already based on self-employed taxi owners and piece-rate models, and like the drivers in the pre-platformized industry, taxi platform drivers often work 60- or 70-hour weeks (Oppegaard, 2024). Furthermore, the taxi platform drivers, tend, like most traditional taxi drivers in Norway (Staalhane and Vassenden, 2022), to have migrant backgrounds. Many come to the taxi platforms from unemployment or physically demanding, low-paid, and precarious work in industries like transportation, logistics, and construction. Drivers therefore often argue that working for the platforms is not only one of few opportunities in the Norwegian labor market, but also a better job (Oppegaard, 2025).

My case study is based on extensive fieldwork in this industry. I conducted 69 interviews with drivers, primarily as “drive-along” interviews, where I ordered a ride as a passenger and interviewed the drivers from the backseat of their cars. ¹ Twenty-one interviews were conducted in 2018 and 48 between 2020 and 2022, after the platforms relaunched in Oslo. When entering the drivers’ cars, I presented myself and my research project, and asked if I could interview them during our ride. I emphasized that participation was voluntary and that everything they said would be anonymized. After each ride, I gave all drivers the largest tip the platforms allowed (NOK 50 [EUR 4.25]) and five-star ratings. The length of the interviews was restricted to the duration of the ride and usually lasted between 15 and 25 minutes, although some drivers wanted to continue the conversation after we had reached the destination. I asked all drivers why they started to work for the platforms and how they perceive the “gig,” but otherwise focused on specific themes—such as the platforms’ trip allocation algorithm, bonus campaigns, and rating system—in different phases of the fieldwork.

The drive-along interviews enabled an observational dimension to the data collection process. I could observe how drivers used the platforms’ mobile applications, how they juggled multiple devices, and use experiences from our ride together as interview prompts. Some of the drive-along interviews were recorded and transcribed but most were documented through fieldnotes written after each ride. I tried to remember or jot down specific phrases the drivers used while in the car, but what the drivers told me during the interviews has been filtered through my memory. In this article, I nonetheless present the interview data as direct quotes. The interviews were conducted in either English or Norwegian, depending on the drivers’ preferences.

These short and focused interviews provided, however, limited information about each driver, inhibiting further analysis of individual-level factors that shape perceptions of the algorithmic management. Furthermore, recruiting research participants through the platforms enabled access to a field that can be difficult to reach for researchers, but poses ethical and methodological challenges (Iphofen et al., 2022). Importantly, there is a risk that my dual role as a passenger-researcher made the drivers tailor their responses to paint a more positive picture of driving for the platforms. However, I found that they were as open to sharing their negative experiences and frustrations as the positive aspects of the “gig.” To calibrate my analysis, I also conducted traditional in-depth interviews with a total of eight drivers, three interviewed individually and five as a focus group, in November 2022. These drivers were recruited from a Facebook group for drivers in Oslo and received NOK 400 (EUR 35) as compensation for using their time to talk to me. These interviews were all recorded and transcribed. They covered the same themes as the drive-along interviews, but in more depth, and confirmed the overall findings from the drive-along interviews. My sample, furthermore, mirrors the general demographic characteristics of taxi platform drivers in Oslo (Staalhane and Vassenden, 2022): all but three of the drivers I interviewed had ethnic minority backgrounds, and all but one were male.

In addition to these interviews, I worked part-time as a driver for the taxi platforms from May to late September 2021. I was employed by a taxi owner and drove 16 dayshifts—approximately 160 hours and 180 trips—in one of his cars, recording verbal fieldnotes while driving and transcribing these after each shift. This autoethnography provided me with a first-hand experience of the labor process and the platforms’ algorithmic management, as well as access to the platforms’ contracts, newsletters, and communication. My experience of driving cannot be generalized to the population of taxi platform drivers in Oslo. To make sure I did not become an undue competitor against other drivers working for the same owner and using the same car as me, I only worked during weekdays. These shifts are less hectic and lower-paid than weekend nightshifts, skewing my experiences. Nonetheless, the autoethnographic fieldwork was analytically valuable. Having experienced the labor process myself provided important insights into, among other themes, the platforms’ algorithmic management and the structure of the workday. Since I continued to interview drivers both during and after the autoethnographic fieldwork, my own experiences helped me to ask more pertinent questions and contextualize what the drivers told me during the interviews.

I analyzed the data by coding fieldnotes and transcripts, exploring both how the drivers described the platforms’ algorithmic management and how they experienced this form of control. In dialogue with previous research on algorithmic management, this analysis highlighted three techniques as particularly important: trip allocation, economic incentives, and evaluations and sanctions. The first round of coding revealed that the drivers argued that, despite frustrations associated with the algorithmic management, they experienced a certain flexibility in the labor process. I brought this finding back to the fieldnotes and interview transcripts, re-analyzing them in search of factors that could explain the drivers’ experiences.

The drivers’ work arrangement: Formal flexibility and piece-rate models

To explore how the taxi platforms’ algorithmic management operates and is experienced by the drivers, the analysis below situates this managerial control within the taxi market in Oslo. This case highlights the importance of formal flexibility to the platforms’ business models. Like other platform companies (Barratt et al., 2020), taxi platforms in Norway promulgate the freedom their work arrangements offer when recruiting drivers. On the websites where drivers can sign up, Bolt (n.d., b) writes: “Be your own boss. Start to drive and earn money.” Similarly, Uber (n.d.) writes: “You can drive with Uber whenever you want, day or night, 365 days in the year. It is always up to you when you want to drive, so it will never interfere with the important things in life.” This rhetorical emphasis on flexibility can be seen as aimed at regulating the drivers’ cognitive and emotional orientation toward the job, promoting opportunities and entrepreneurialism, and legitimizing their classification as self-employed contractors (Wood and Lehdonvirta, 2021). Yet, the taxi platforms go beyond rhetoric and do provide drivers a certain flexibility, allowing them to work whenever and as much or little as they want.

Still, the drivers remain dependent on working long hours. Earning a commission on each fare, they usually work 10 to 12 hours, six or seven days per week to make a living. “Compared to working in a grocery store, the hourly wage is very low, but you can work more. Or, you have to work a lot more hours,” one driver said (Driver 1, focus group interview, November 2022). The drivers also tend to register and log on multiple platforms simultaneously, a practice that has been described as “multi-apping” (Popan, 2024). Since they are authorized taxi drivers, they can additionally pick up customers in the street and at taxi stands, providing them with a formal flexibility many value highly, often citing it as the chief reason why they chose to drive for the platforms. “That is the beauty of this job, I decide when I work,” Driver 2 argued (in-depth interview, November 2022).

Since the drivers are paid on commission, their earnings are, furthermore, tied directly to their performance on the road. As Marx (1976: 697) wrote, piece-rates develops the “worker’s sense of liberty, independence and self-control,” a tendency illustrated by another quote from the driver cited above: “If you want money, you work hard, you work long, you sit in the car, you get annoyed by the people, you get annoyed by not getting trips, you get annoyed because the trip is shit, like cheap or whatever. But if you are positive, you can always earn money” (Driver 2, in-depth interview, November 2022). The piece-rate model thereby aligns the drivers’ economic interests with those of the platforms: completing as many rides as possible. This illustrates how gig platforms’ managerial control comprises normative and organizational techniques, in addition to the algorithmic management (see Moore and Joyce, 2020).

However, piece-rate models and inculcating in the drivers an entrepreneurial orientation toward the job is insufficient. In a working paper describing its business model, Uber wrote: “Driver-partners are free to work whenever they want and must be incentivized to provide services” (Hall et al., 2015: 1). Drivers’ flexibility hence posits a potential problem for the platforms: since workers can determine their own hours, the platforms’ business models need managerial strategies to ensure that they supply their labor when and where it is needed. The taxi platforms’ algorithmic management can be seen as a tool for solving this issue, exerting control over the drivers under conditions of formal flexibility.

Algorithmic management and drivers’ experiences

In the following, I explore the taxi platforms’ algorithmic management as based on three techniques: trip allocation, economic incentives, and evaluations and sanctions. The analysis highlights how the drivers experience these techniques and what effects they have in the platformized taxi industry in Oslo.

Trip allocation

The trip allocation algorithm coordinates supply and demand, sending customer requests to the drivers. While many of the drivers I interviewed believed that the requests are allocated to the closest driver, the exact mechanisms remain opaque. Requests appear in the drivers’ mobile application, giving them a few seconds to accept or decline. When a driver accepts a request, navigation to the pick-up spot starts automatically. The drivers “start” the ride in the application after picking up the passenger, and the navigation system directs the drivers to the destination. Upon arrival, they end the ride by pressing a button in the app. While en route with a passenger, the platforms might send drivers a new request—a “back-to-back trip”—directing them to the next customer immediately after dropping off the first.

From the drivers’ perspective, the trip allocation algorithm creates an unpredictable working day. Drivers do not know when the next request appears, and there can be significant waiting times between rides. As one driver said, “working, but not getting paid” (Driver 2, in-depth interview November 2022). Because they receive limited information about the ride, it can be difficult to determine whether the requested ride will be lucrative or not. Many of the drivers I interviewed were annoyed by long pick-up distances, work for which they are not compensated. Some drivers therefore prefer Bolt, which allows drivers to set a maximum distance to drive to the pick-up. “Uber,” however, “eats the car,” Driver 3 said (drive-along interview, January 2022), arguing that the platform made him to drive long distances, causing wear and tear on the car. Drivers also find it annoying when passengers cancel the request after they have accepted a ride or do not show up at the pick-up location. “We use fuel, pay toll charges, and so on, but get nothing. Like a slave,” one driver argued (Driver 4, drive-along interview, April 2022).

The platforms often conceal the passengers’ destinations. This frustrates the drivers, particularly if the ride takes them to low-demand areas, where they might have to drive back “with air,” i.e., without a passenger. Unpredictability, however, was not brought to the taxi industry by platforms. Taxi driving has always been characterized by volatile demand, information asymmetries, and waiting for fares. Like traditional cabdrivers (see Davis, 1959), taxi platform drivers employ different strategies to navigate this unpredictability. As mentioned, drivers in Oslo often log on to multiple apps at the same time to increase the likelihood and frequency of trip requests. They can decline and cancel rides—although this, as we will see, might cause low ratings and sanctions. Since drivers only earn money per completed ride, they accept most requests. Still, some experienced drivers try to maximize their earnings by declining or canceling rides they believe might be unprofitable. They might for instance decline a request if the pick-up distance is long or cancel the ride if they see that the destination is far away from higher-demand areas. Some drivers also use third-party software that routes requests from multiple platforms into one application. This software also enables the drivers to automatically decline requests for rides with pick-up distances longer than a customizable limit.

This shows that the taxi platforms’ trip allocation algorithm from the drivers’ perspective is both opaque and unpredictable. When I registered on the platforms, I had to watch a few introductory videos about proper driver conduct, but none of these explained the mechanisms of the trip allocation algorithm—nor any of the other techniques comprising the algorithmic management. However, the drivers’ maneuvering of the trip allocation mechanisms also illustrates, as is highlighted by previous research (Joyce and Stuart, 2021), that workers develop a working understanding of how the algorithms function, enabling them to strategize to potentially increase earnings. Nonetheless, the drivers’ incomes remain regulated by the platforms and the trip allocation algorithm, which determines when and how many requests drivers receive.

Economic incentives

Taxi platforms use different economic incentives to attract drivers. One technique is to adjust fares to fluctuations in supply and demand through dynamic pricing, referred to as “surge pricing.” The platforms calculate the price of ride by adding time and distance rates to the base fare but increase the prices in specific areas when demand is particularly high through a “multiplier” of for example x1.2 or x2.1 (Hall et al., 2015). The “surges” are visualized in the maps in drivers’ applications by showing certain areas with a distinct color (Lee et al., 2015). During the dayshifts I worked, “surges” were rare and I regularly experienced that they disappeared gradually as I approached them. Drivers called these “fake surges,” illustrating the opaqueness and unpredictability of the dynamic pricing algorithms. “I have no idea how it works,” one driver said (Driver 2, in-depth interview, November 2022), while another described it as “elusive” (Driver 5, in-depth interview, November 2022).

Some drivers do not adjust their driving to the price increases, arguing that the often small price increases matter little if they have to drive a longer route to the passengers. Other drivers, however, emphasized “surge hunting” as a viable strategy. One of them said:

I hunt the surge pricing, of course. [. . .] It works. For example, Saturday night, you see a surge all over the city, but you receive only requests for trips without surge. So, you sit there, “decline,” “decline,” “decline,” oh, two times the price, “accept.” We play the game. (Driver 6, in-depth interview, November 2022)

From this driver’s perspective, the dynamic pricing mechanism allows him to “play the game,” actively trying to take advantage of the opportunities for increased earnings the platforms offer. This illustrates that the platforms provide the workers with continuous choices, and although these are narrow and structured, they can provide workers a sense of autonomy (see Cameron, 2024).

Drivers’ sense of autonomy is also enabled by their opportunity to find passengers in the hailing and taxi stand segments of the market. These rides are more lucrative, since the fare is determined by the drivers’ taximeter rather than the platforms’ algorithms. One of my interviewees said: “You need a proper ‘surge’ and a really long ride to beat the taximeter” (Driver 7, focus group interview, November 2022). For this reason, some log off the platforms during high-demand periods—such as late weekend nights—to instead work the street segment of the market.

In addition to the dynamic pricing, the taxi platforms in Oslo also use bonus campaigns to recruit drivers and influence their behavior. The platforms tend to offer reduced “service fees” ² or a bonus payment if drivers complete a certain number of rides on the platform within a given period. These campaigns aim to increase the driver’s loyalty to the platform, and drivers often make sure they drive enough rides throughout the week to gain a lower “service fee” for the weekend shifts. Drivers can also get access to bonuses by foiling their car with the platforms’ logo. When launching in Oslo, Yango tried to gain market shares by combining the lowest prices with the most lucrative bonuses for drivers. Yango offered drivers a bonus that increased for every ride the driver completed, up to NOK 10,000 (EUR 850) for 15 rides. The bonuses, however, were scaled back after a month, requiring drivers to complete 35 rides to reach the cutoff at NOK 12,000 (EUR 1025). Yango later reorganized its bonus system completely, offering a bonus of NOK 1300 (EUR 110) to drivers if they completed 12 rides within given windows of time when demand tends to be high, illustrating how platforms use bonuses to attract drivers when entering new markets, only to reduce their offerings as they gain a foothold.

According to the drivers, Yango’s initial bonus campaign was very profitable; one driver described it as the “best campaign ever” (Driver 7, focus group interview, November 2022). In general, the drivers in Oslo depend on the platforms’ bonus systems. One told me that that each ride pays very little and that “the money” comes from bonuses (Driver 8, drive-along interview, August 2021), while another argued that without the bonuses, “everyone actually lose[s] money” (Driver 1, focus group interview, November 2022). Drivers therefore structure their driving around campaigns: since they usually are registered on all three platforms, they choose which one to prioritize during a shift based on the campaigns. The drivers also switch between platforms to complete as many campaigns as possible. “If you are smart, you only drive campaigns. You take the campaign, log off, take the next campaign, log off. Next campaign. Then you have done the job. You have made money,” one driver said (Driver 7, focus group interview, November 2022). Still, the allure of the potentially significant bonuses does not always translate into actual earnings for the drivers. One of them told me how he often struggled to complete the necessary number of rides to obtain the maximum bonuses, particularly during Yango’s launch campaign. “They are tricking us,” he said, “it is baiting” (Driver 9, drive-along interview, February 2022). Another driver said: “I have been hunting these incentives. But some of them are horribly designed” (Driver 5, in-depth interview, November 2022), recalling how the bonus for completing a certain number of rides within a given time frame incentivizes dangerous driving: “If you have been driving for seven hours and you have ten minutes left, you drive like a crazy-person,” he argued.

This illustrates that while the campaigns might appear attractive, the drivers understand that the bonuses are deployed to influence their behavior. It also shows how the platforms’ economic incentives, like the trip allocation algorithm, are opaque and unpredictable. As a managerial technique, then, the economic incentives enable the platforms, acting as “market organizers” (Kirchner and Schüßler, 2019), to attract drivers and influence labor supply temporally and spatially through market mechanisms, e.g., prices and pay.

Evaluations and sanctions

The taxi platforms evaluate and sanction drivers based on multiple parameters. They measure how many of the requests drivers accept, decline, and cancel, but the effects vary between the platforms. Yango calculates an “Activity score,” withdrawing “points” if drivers decline or cancel requests. According to the drivers, they might lose access to Yango’s bonuses if their score falls too low. Bolt, the drivers said, might “deactivate” their account for a period if they do not accept more than half of the allocated requests. On Uber, however, the drivers argued that these scores do not have any significant consequences, enabling them to “play the game,” as the driver above described it, declining non-“surge” requests while waiting for better rides. In general, the drivers noted that “deactivations” are rare. One, however, told the story of how Bolt once “deactivated” his account: “I was like ‘what the fuck?’ It was because I had rejected too many rides, they said. What the fuck? I was angry. But now, when I know, I am better at going offline when I work other platforms” (Driver 2, in-depth interview, November 2022). This quote illustrates both the frustration the drivers can experience when their economic opportunities are limited by “deactivation” and the opacity of the platforms’ control mechanisms but simultaneously shows that drivers learn and adjust their behavior to work within the boundaries of the managerial techniques.

The platforms also assess the drivers based on the customers’ experiences. After each ride, the passenger give the driver a “rating” of between one and five stars, and the drivers’ average ratings are visible for passengers. Too low average ratings might result in the drivers’ accounts being “deactivated,” after having received a warning and period to improve their ratings (Bolt, n.d., a; Uber, 2016: 21). The threshold, however, is not revealed to the drivers. Yet, most of the drivers I interviewed did not worry about their ratings. My analysis suggests that there are two important reasons why.

First, five-star ratings seem to be the norm among customers in Oslo, and the drivers usually have high average ratings. One of the drivers showed me his average rating: “Ah, 97 [referring to his average rating of 4.97], fuck, it went down” (Driver 6, in-depth interview, November 2022), he said after checking his app. “Isn’t that good?” I asked. “No, it’s really bad. Two weeks ago, I had 99.” The drivers told me they rarely do anything in particular to receive good ratings, like offering candy, gum, or bottled water to their customers, as previous research has found drivers do (Rosenblat and Stark, 2016). “No, I am just polite and talk to the customers who what to talk,” one of the drivers said when I asked if he has a strategy (Driver 10, drive-along interview, December 2021). Another driver told me he never thinks about the “ratings”: “I just drive and act professionally” (Driver 11, drive-along interview, October 2021). Second, the drivers also often argue that their ratings do not have any significant consequences. As already noted, few of them had experienced “deactivations” due to too low average ratings. One driver told me he never has paid attention to his ratings: “I don’t care [about the ratings]. [. . .] They don’t matter, they affect nothing, so I don’t care” (Driver 12, drive-along interview, February 2022).

The rating systems are nonetheless opaque (see Rahman, 2021). Some drivers thought their ratings affect the number of requests they receive but were unable to confirm their suspicion. One described the platforms as operating like an “invisible hand” (Driver 13, drive-along interview, June 2021), as the drivers do not know which passengers give them each rating and struggle to determine why they receive poor ratings or complaints. Another driver said: “It’s fair enough when you know what you did wrong, but annoying when you get a bad rating, and you don’t know why” (Driver 14, drive-along interview, September 2021). A third driver described it as “humiliating” (Driver 15, drive-along interview, February 2022).

When asked if he had ever tried to contact the platforms’ support centers about these matters, one of drivers from the in-depth interviews said: “It’s no point. We sometimes complain, but it’s like throwing rocks against the wall. Nothing will happen” (Driver 6, in-depth interview, November 2022). This shows the non-negotiability of the platforms’ algorithmic management (see Issar and Aneesh, 2022). While the drivers usually are able to maintain an aloof attitude toward the platforms’ rating systems, they have essentially no avenues for voice and co-determination, and work at the mercy of a system that at any point and for whatever undisclosed reason might exclude them from their opportunity to make a living, without providing them with any means of negotiating or appealing the decision.

“If I want to be the guy who complains, I can complain”

The case of taxi platforms in Oslo thus demonstrates the tension between autonomy and control that generally characterizes platform work (Shapiro, 2018). The above analysis shows that the algorithmic management is an opaque, unpredictable, and non-negotiable form of managerial control, operating through allocating rides, economic incentives, and evaluations and sanctions. While drivers are frustrated by long pick-up distances, low fares, unpredictable bonus systems, opaque decision-making processes, and arbitrary and non-negotiable “deactivations,” the algorithmic management does not necessarily appear as control from their perspective. I asked one of the drivers whether the platforms are controlling him: “When I think about it, yes, that’s right. But I don’t have that taste or that feeling when I am working,” he said, and continued:

If I want to be the guy who complains, I can complain. Yes, if you think about it, it is really control. But on the other hand, I understand why [the platforms] do it. It is to improve the customers’ experience. We are motherfuckers, we are a little selfish sometimes, you know. And we try to be as efficient as possible and make as much money as possible, but that is at the expense of the customer and the availability of the cars. The apps try to make a compromise. I don’t want to complain. I can, but I won’t. (Driver 6, in-depth interview, November 2022)

His experience highlights a general perception among the taxi platform drivers in Oslo: while at work, driving for the platforms, the drivers experience a certain autonomy, and do not feel that they are being controlled. The drivers’ orientation is thus similar to what Vallas and Schor (2024) call a “stoic” view of platform work, wherein workers are critical of the platforms’ practices but see no benefits to complaining.

Discussion

The above analysis suggests that there are two key factors explaining the drivers’ orientation to the algorithmic management and why they do not necessarily experience this as “control”: first, that the drivers remain economically independent of each individual platform, and second, that the algorithmic management operates as a form of control that does not intervene directly in their labor process.

Drivers’ lack of dependency on platforms

The first explanation for why the drivers I interviewed argued they had significant flexibility at work and did not necessarily see the algorithmic management as control, is that they remain economically independent of each individual platform. Most drivers were registered on all three taxi platforms and could additionally, as authorized taxi drivers, find rides in non-platform market segments (the street segment). This reduces their dependence on each platform and enables them to make a living independently of their status on Bolt, Uber, or Yango. Drivers can switch between platforms to take advantage of market conditions and campaigns and are able to earn money through other platforms or the street segment if they were to be “deactivated” by one of the platforms. The way in which the platformized taxi market in Oslo is structured thereby enhances drivers’ independence and potential agency and has significant effects on drivers’ working conditions and their experience of the platforms’ algorithmic management. These dynamics are not unique to the Oslo market but are likely representative of other markets where workers can work for multiple platforms and/or have access to market segments outside the platforms.

Since I left the field, two important changes in the Oslo market have taken place, emphasizing the importance of dependence on the platforms for drivers’ experiences and working conditions further (see Klassekampen, 2025). First, Yango discontinued its Norway operations in 2024, leaving behind a duopsony of Bolt and Uber. Second, the government implemented price regulations for street hailing and taxi ranks, making this market segment significantly less lucrative for the drivers. Both developments have made the drivers more dependent on the remaining platforms, Bolt and Uber. With increased market power, Bolt increased the “service fee” they charge drivers to 25%, matching Uber. Both platforms have also reduced the prices—without consulting the drivers. There are also indications that they have scaled back their bonus system, illustrating that working conditions on platforms tend to deteriorate over time, as previous research has found (Maffie, 2022). These developments have caused massive frustration among the drivers and led to increased collective mobilization and even some protests.

While previous research has emphasized workers’ class position (Vallas and Schor, 2024) and ideological subjectivations inculcating an entrepreneurial orientation (Barratt et al., 2020; Purcell and Brook, 2022), the case of Oslo indicates that market structures also are important for explaining platform workers’ experiences and perception of the “gig.” This shows how local arrangements and regulations shape the effects of algorithmic management (see Wood, 2024). Specifically, my analysis suggests that how dependent workers are on the platforms significantly shapes their experiences and potential agency. This illustrates an important insight from LPT that market dynamics play an important role in structuring labor process and managerial control (Thompson and Laaser, 2021).

Algorithmic management as non-interventionist control

The second explanation I find is that the algorithmic management of the taxi platforms in Oslo in practice does not intervene directly in the drivers’ labor process. It does not endeavor to script the labor process, but functions as what I term a non-interventionist managerial control. This is a form of control that takes as its objective the conditions under which the drivers work, influencing behavior not by instructing workers on what to and not to do, but indirectly by incentivizing conduct. The trip allocation mechanism routes requests to drivers, allowing them to accept or decline the ride. The economic incentives of the dynamic pricing and bonus systems target this assessment, as well as making it appear more profitable to log on to the app and stay on the road. While not instructed to behave in particular ways, drivers’ performance is evaluated and sanctioned indirectly and post-factum, through passenger ratings and calculations of the acceptance and cancellation rates.

Taxi platforms thereby exercise control over drivers through market mechanisms, and, like other platform companies, take the form of what Kirchner and Schüßler (2019) call “market organizers.” As market organizers, platforms use their digital infrastructure to facilitate interactions and transactions by establishing and operating a digitally mediated market, and use the digital infrastructure to set criteria for access to the platform, determine the rules of interaction, monitor and sanction conduct, and maintain control over their users and market. This conceptualization highlights on the one hand how the platforms, in regulating their marketplace, essentially determine the conditions under which the drivers work. On the other hand, however, by operating through market mechanisms, the algorithmic management might nonetheless appear neutral. Rather than a detailed and direct control over drivers and their labor process, the taxi platforms’ algorithmic management operates by regulating the market within which drivers work. It structures the choices available to drivers but nonetheless organizes their labor process as a series of choices, which can generate what Cameron (2024) terms “choice-based consent.” Algorithmic management, in the case of taxi platforms in Oslo, can thereby be seen as a non-interventionist form of control: by targeting and operating through market mechanisms, the taxi platforms’ algorithmic management can govern conditions without coercion and intervening directly in drivers’ labor process, a modus operandi that provides the taxi platforms and their algorithmic management a normative legitimacy.

This shows how, as Vallas and Schor (2020: 281ff.) argue, platforms can act as “permissive potentates,” ceding control over certain aspects of the operation, such as how the workers carry out their tasks, while retaining the authority over others, for example pricing and task allocation. In their labor process, the drivers have, as shown above, the opportunity to switch between different platforms and between the platforms and other market segments. Furthermore, as this section emphasizes, the algorithmic management, by operating through market mechanisms and providing choices rather than intervening directly in their labor process, also inculcates in the drivers a perception of autonomy behind the wheel. While the drivers simultaneously experience the platforms’ algorithmic management as opaque, unpredictable, and non-negotiable, the flexibility and apparent neutrality of the algorithmic management obfuscate the constraints these digital control techniques represent. As market organizers, the taxi platforms use their technological infrastructure and algorithmic management to monopolize control over drivers’ economic opportunities and working conditions, implementing and maintaining a fundamentally hierarchical and asymmetrical relation of power between themselves and their workers (Issar and Aneesh, 2022). Through task allocation, economic incentives, and evaluations and sanctions, taxi platforms determine and continually adjust the market conditions drivers must navigate, without consulting them or providing any opportunities for voice and co-determination.

Conceptualizing algorithmic management as a non-interventionist mode of control contributes first to the scholarly and political understanding of platform work and digital managerial techniques. By highlighting how taxi platforms exert control without intervening directly in the labor processes, we gain a better understanding of both how these companies operate and why workers do not necessarily experience the control as such and develop an adversarial attitude toward the platforms. Second, it contributes to the theorization of the platforms’ business models and the tension between their employment models and the control imperative immanent in all capitalist labor processes. Gig platforms generally avoid employer responsibilities by classifying workers as self-employed contractors, a status legitimized by the significant formal flexibility workers are afforded. They nonetheless exert considerable managerial control, partly, as we saw Uber argue, precisely because the on-paper independent workers have to be “incentivized” to supply labor power (Hall et al., 2015). The platforms thus risk their workers being reclassified as employees if a case was to come to court. This would have major implications for their business model (Maffie, 2022: 350). From the platforms’ perspectives, the algorithmic management therefore must balance control with endowing workers with sufficient autonomy to legitimize classifying them as self-employed contractors. Rather than intervening directly in labor processes, operating at the level of market mechanisms can be seen as a strategy for walking this tightrope.

Finally, the notion of algorithmic management as non-interventionist control illustrates important but often neglected insights from labor process theory (Joyce and Stuart, 2021). It shows that, as Friedman (1977: 45) notes, workers’ autonomy in the labor process does not mean an absence of control: workers can have a relative control over their labor process while management retains absolute control and authority over the production process as a whole. In this sense, the non-interventionist managerial control exercised by the taxi platforms in Oslo enables achieving control “away from the point of production,” as Littler and Salaman (1982: 264) remark, making the control appear, from the perspective of the labor process, as a “non-issue.”

Conclusion

Taxi platforms like Bolt and Uber challenge the significant autonomy taxi drivers traditionally have enjoyed (Hodges, 2020). This article has used the case of Oslo to explore taxi platforms’ algorithmic management and drivers’ experiences. It shows that algorithmic management is an opaque, unpredictable, and non-negotiable form of control. The drivers nonetheless highlight their flexibility and do not necessarily experience it as control. The article proposes two main explanations for this tension. First, the structures of the platformized taxi market in Oslo provide the drivers an independence from each platform enhancing their potential agency. Second, the taxi platforms’ algorithmic management takes the form of a non-interventionist managerial control, operating through market mechanisms, endowing the drivers with a certain flexibility in the labor process and thereby taking on an appearance of neutrality and gaining a normative legitimacy.

The case of taxi platform workers in Oslo shows that, as elsewhere, platform workers are attracted by the autonomy the platforms promise—a promise that might not be broken by their algorithmic management (Wood et al., 2019). This provides important nuance to the “subjugation narrative” (Vallas and Schor, 2024) and notions of “digital Taylorism” that have characterized some strands of the research on platform work and algorithmic management (see Altenried, 2019; Brubaker, 2022: 119; Kellogg et al., 2020; Rosenblat and Stark, 2016). Importantly, this is not to say that the taxi platforms do not exert significant control over the drivers, merely that algorithmic management operates through specific mechanisms. In practice, the platforms still determine and regulate the conditions under which the workers work, illustrating the power asymmetries in these work arrangements. The platforms’ practices, furthermore, often frustrate workers and can lead to different forms of resistance, mobilization, and protest (Joyce and Stuart, 2021; Tassinari and Maccarrone, 2020). This is also the case in Oslo, illustrated in particular by the recent mobilizations following price reductions and commission hikes (Klassekampen, 2025). Since platforms tend to operate as market organizers (Kirchner and Schüßler, 2019), however, workers’ criticism of the platforms might, as this analysis and previous research suggests (Wood and Lehdonvirta, 2021), be based on an identification as market actors rather than “workers.” This is an interesting avenue for further research.

One might have expected the Norwegian labor market model—with its strong unions and working environment regulations—to curb the taxi platforms’ algorithmic management. The taxi industry, however, is an industry on the “fringes” of the Norwegian model, characterized by low unionization rates, and work arrangements that were “gigified” long before the rise of platforms (Oppegaard et al., 2025). As such, institutional mechanisms that could have shaped the algorithmic management were absent in the Norwegian taxi industry. The research literature has, furthermore, described factors beyond the labor process as essential in conditioning the effects of algorithmic management (Wood, 2024) and workers’ perception of the “gig” (Vallas and Schor, 2024). While these perspectives are essential for understanding the structures and dynamics of platform work, algorithmic management, and managerial control more broadly, this article has shown how the organization of the workers’ labor process remains a key explanatory factor. Specifically, this article suggests that the degree to which workers are dependent on the platforms and the extent of the algorithmic management’s intervention in the labor process play an important role in shaping working conditions and workers’ experiences.