Book Review: Inside the Invisible Cage

Inside the Invisible Cage: How Algorithms Control Workers, by Hatim A. Rahman (2024). Oakland, CA: University of California Press.

Inside the Invisible Cage, Hatim A. Rahman’s new book, is a comprehensive ethnographic investigation of how algorithmically mediated labor-market platforms engender opacity through evaluation scores, rating systems, and search-result visibility. In doing so, Rahman’s book advances digital labor studies by providing evidence from an online platform for high-skilled labor, offering valuable insights into how algorithmic opacity plays out. As such, this shapes and constrains the lives of millions of workers in today’s platform economy, confining them in what the author calls an “invisible cage.” It is invisible because algorithms are dynamic, experimental, and opaque, dictating how workers are expected to behave; it is also a cage because these systems are irrefutable and they cultivate workers’ dependency, even though said workers can theoretically leave the platform at any time. Indeed, platform workers are stuck having to comply and contend with the unpredictability of the algorithm due to the “reputational interdependence” (163) created by the platform itself.

Rahman is trained in management studies and, as a result, his book is an ethnography of the management practices that organizations have increasingly begun to implement in order to manage and control workers. At the same time, he draws extensively on sociological and anthropological scholarship on work and labor, situating his study within broader conversations around technological control and contemporary labor regimes. Inside the Invisible Cage is based on six years of ethnographic research, conducted on a large online labor-market platform for high-skilled work – what Rahman refers to as TalentFinder (a pseudonym). A key argument is that algorithms do more behind the scenes than just matchmaking or acting as intermediaries; crucially, they control worker visibility, job prospects, and wages. They also compel workers to be more than simply part-time freelancers, as workers on the platform typically need to work full time in order to meet the algorithmic criteria.

To this end, Rahman efficiently conducts his ethnography in a digital setting, a challenging task given the participants’ geographic dispersion and the fact that much of their interaction occurred via computer. Rahman registered as both a client and a worker for firsthand insight into the platform’s communications and features. In addition to participant-observation and interviews, the author undertakes computational social-science techniques to keep track of the changes occurring in the platform, using natural language processing (NLP) models to examine large-scale textual data. His investigation traces the evolution of the platform itself, as well as the algorithm’s evolving nature, by analyzing its interface design, affordances, official announcements, and discussion threads posted by the platform’s managers and workers. Rahman’s ethnography, in a sense, goes against the grain of ethnographic tradition: rather than focusing on a people and a place, it examines a process.

Throughout the book, Rahman’s arguments are strengthened by the parallels that he draws between traditional organizations and algorithmically managed ones, thus clarifying what makes platform work unique. He provides substantial evidence of beta tests along with glitches and inefficiencies of the algorithm over time, which is useful for scholars and policymakers seeking to hold platforms accountable, especially at a time when platforms insist on exclusive data proprietorship.

The book’s theoretical framework of the invisible cage is an extension of the Weberian iron cage, which – when applied to the workplace – would point to the bureaucratic, rational, rule-based forms of compliance in traditional employer-employee organizations. As rating systems – pioneered by eBay – became commonplace in online product markets, they increasingly evolved into mechanisms of control within online labor markets. Similar to performance evaluations, ratings became the basis for rewards and penalties, thus shaping worker behavior and performativity. The key difference here is that, in online labor markets, ratings are given by clients and consumers, who correspondingly act as middle managers but in fact possess little expertise to evaluate workers. These ratings are then aggregated, sorted, and showcased on worker profiles in the platform by algorithms, which undergo constant experimentation and experience glitches that point to overall inefficiencies and inaccuracies. Moreover, workers face an information asymmetry when attempting to gauge the weights assigned to different metrics, when ratings increase or decrease, or in what form they are showcased in their profiles. Consequently, unlike performance evaluations, ratings in this case do not fulfill the task of helping workers improve on the job, as they are not made aware of the details and areas for improvement.

In the introductory chapter, Rahman defines algorithms and critically examines their application to social phenomena. While algorithms use fixed, deterministic formulae to predict outcomes, social phenomena such as a client’s satisfaction level with a service are highly subjective in nature and cannot be accurately calculated and indexed. Chapter 2 provides the trajectory of TalentFinder’s growth – its shift from a staffing firm to an online marketplace, to reap benefits of the global rise of the internet, the alleged “death of distance,” and new labor paradigms such as “Nikefication” – that is, the outsourcing of operations to lower-wage workers overseas (29). Conversely, workers signed on to TalentFinder for its promised freedom and flexibility, while clients joined to take advantage of easier payments and to save time and labor costs. Chapter 3 traces the evolution of TalentFinder’s evaluation system, the initial issues that it created, and how these issues led the platform to enact a series of reforms. Eventually, the platform changed its algorithm from a five-star rating system to a percentage score, which was even more non-transparent in its calculation.

One of the book’s key contributions is demonstrating how platforms use worker data, gathered through beta testing, to refine their algorithms, creating systems that become increasingly rigid and inscrutable. As TalentFinder began to reform its algorithm in a supposedly more democratic and collaborative manner, managers engaged with users in discussion threads; as Chapter 4 and 5 show, this interaction resulted in yet more changes. The platform shifted to a permanently opaque algorithm and further altered the interface design, calculation of evaluation scores, visibility of profiles, and dispute resolution affordances (among others) without giving prior notice to workers. The platform also made changes to its Terms of Service (ToS) without notifying workers or clients. The book provides detailed evidence as to how the ToS become an instrument of maintaining opacity. Chapter 6 focuses on the effects of such unpredictability on workers – such as their alienation from co-workers, incivility on the part of clients, few opportunities for them to grow, and constant experimentation with the algorithm.

Moving forward, Chapter 7 explains why, despite these cascading effects, workers cannot leave the platform due to the dependency that it creates. The algorithm publicly shows workers’ scores across their other social media and professional handles on the web – and the ToS stipulates that these can never be erased, even if one leaves the platform. Thus, the prospect of jeopardizing their careers or having to start from scratch keeps workers dependent on the platform, amplifying the paradox of freedom and flexibility that was initially promised. Chapter 8 outlines the theoretical and practical implications of the book, offering recommendations for workers, platforms, and policymakers. Chapter 9 concludes the text by imagining what the future may hold for workers in the age of algorithms.

Inside the Invisible Cage enriches our understanding of algorithmic management in platform work by providing ample ethnographic evidence. It also contributes to the scholarship on anthropology of work by examining algorithmically mediated work cultures. Given the fast-changing nature of work in the global market, Rahman’s book poses crucial questions on how the definition of work is evolving, how people find work and make a living, and how management strategies to control workers change over time. One promising area for future research is: what do spaces of resistance look like in online markets for high-skilled labor? Collectively, these insights make the book a valuable resource for undergraduate or graduate students with an interest in online labor markets. It will also be of interest to anthropologists of work, labor and digital studies scholars, organizers, and policymakers.