Cruel optimization: People analytics and platforms of care

Abstract

Human resource platforms organize how workers participate in and experience work as informationalized relations and routines. The Workhuman Cloud, developed by a human resource tech firm, exemplifies how platforms and the organizations that use them work to engineer disciplined subjects. Platform studies scholars have shown processes of platformization reaching business management, labor and industrial relations, law and policy, and more. This article analyzes corporate plans and tech demonstrations for the Workhuman Cloud and industry expert reports promoting how to “make work more human.” What does it mean when platforms are enrolled to enact users as “human”? What kind of human do they constitute? We interrogate how actors deploy care and platforms in the differential management of workers. Incorporating people analytics, ideas about social recognition, gamification, and the automation of inclusion, the Workhuman Cloud is an expression of what we call “cruel optimization.” This concept names how platforms configure the persistent attachment to the belief that improving oneself through sociotechnical arrangements (i.e., algorithmic control, recognition, and rewards) instills a virtuous cycle of betterment even when such attachment works against one's well-being. Drawing from feminist and critical race STS, we scrutinize how the technopolitics of user configurations in contemporary work platforms ensconce surveillance and labor governance through the non-innocence of care.

When the COVID-19 pandemic began in early 2020, conditions accelerated the embrace of labor management practices like people analytics and artificial intelligence. Organizations shifted their operations from centralized spaces to accommodate public health guidance and regulations. Meanwhile, the exponential growth of flexible work and the expansion of remote workers created organizational challenges around worker discipline and retention (Dahik et al., 2020; Harter, 2023; Malhotra, 2021; McKinsey Global Institute, 2021; Rahman, 2023). Spatially distributed and absent from their administrative offices, many workers logged into their working environment through a range of video communications and work processing platforms. ¹ Google and Microsoft had long built a computational infrastructure to facilitate such remote work through platforms like Google Workspace and Microsoft Teams. This arrangement, while not new, treats workers as users and corporations as hosts or platforms. Companies see themselves as providing users with an infrastructure that facilitates interactions and relations (Fisher, 2020: 53–54). Organizational needs emerged for companies to coordinate, discipline, and retain workers with many industries looking to prior developments in data analytics to support their efforts. Rebranded as people analytics, the approach is a critical tool in the optimization of human resource management because actors believe it provides much-needed insight into employee behavior (Chamorro-Premuzic and Bailie, 2020; Giermindl et al., 2022). Others integrated or further expanded their use of automation and artificial intelligence to manage workflows, monitor worker behaviors, and more (McKinsey and Company, 2023).

People analytics translates workers into data, objects for intervention. “Better people analytics measure who [people] know, not just who they are” (Leonardi and Contractor, 2018: 70). The goal of people analytics, the article published by Harvard Business Review concludes, is for companies to use the “digital exhaust” generated by their operations, “data created by employees every day in their digital transactions, such as e-mails, chats, and file collaboration,” to gain “insights into their workforce.” People analytics employs operational data to inform human resource decisions in organizations, such as hiring, performance management, retention, and more. Paul Leonardi and Noshir Contractor (2018: 72), two university professors who work on management and behavioral science, want to align the performance of people analytics “with the hype.” In vogue across the business world and industries like pharmaceutics, technology, energy, and banking, the goal is to improve people analytics by integrating the data routines of an organization to govern its operations (Leonardi, 2011: 152; Leonardi, 2012). The arrangement conveys a need-to-know (people and their relations), although not necessarily out of some gracious exercise in social beneficence. Surveilled people are made into productive data bodies through statistical processes meant to shape human management decisions and future operations. Such governance enacts rationales of optimization—gradually improving the functioning of an organization through increasing efficiencies in its parts and their relations. Proponents of people analytics argue that labor requires constant surveillance and control in the maximization and anticipation of productivity. ²

In this article, we examine people analytics in relation to digital platforms and their data ecologies. Together, they constitute a labor regime that produces and configures users as responsive and self-improving workers. We study the kinds of user configurations drawn up in plans for white-collar workspaces. From video demonstrations and ads to corporate official remarks, plans and reports are structuring devices that mediate the contingencies of users, artifacts, and uses (Suchman, 2007). Our analysis of plans and user configurations builds on Devika Narayan’s (2024) call to platform studies to examine how digital platforms structure, coordinate, and manage organizational operations and labor. Feminist science and technology studies (STS) scholarship (Atanasoski and Vora, 2019; Murphy, 2017; Ong, 2006) shows how sociotechnical arrangements become the means through which neoliberal subjects are differently made. This article contributes to these fields by interrogating how digital platforms and neoliberal technologies of subjectivity, developed by designers and employed by companies, induce self-animation and self-government that, despite claims of enabling well-being, produce harm. This is what we call “cruel optimization,” an attachment to normative arrangements of productivity enacted through expressions of concern for oneself and the Other. Cruelty unfolds as users hold on to the promise that digital platforms care for them, their cultures, desires, and aspirations, just as user actions and interactions become sites for data colonialism (Couldry and Mejias, 2018). Cruel optimization names the obfuscation of a precarious order premised on the supposed liberation wrought by technological fiat.

We observe cruel optimization in action through two products from Workhuman, a human capital management company founded in 1999 in Dublin, whose mission is, according to their website, “to help companies harness the power of their most valuable assets—their people.” The company seeks to fulfill such a mission by developing software applications meant to “build winning cultures” with gains extending “beyond the balance sheet”—making a difference in how workers feel about themselves and toward their workplace. Workhuman prides itself as “pioneering the human workplace” by developing a cloud platform that regulates interpersonal communications and governs labor through the instrumentalization of care and “recognition”—the practice of identifying and accepting someone while validating them and their actions. But the Workhuman platform, the core of the company's business, enrolls workers as users who generate data designed not just to recognize and reward workers. Such data are integral to the application of people analytics as a means of regulating and disciplining them. Meanwhile, concern toward making inclusive work environments, something we examine through Workhuman's Inclusion Advisor agent, is outsourced to a delegate tasked to shape worker interactions that may help minoritized workers feel seen and valued. Workhuman's platform and Inclusion Advisor negotiate a diverse set of industries and their particularities just as they enable user interactions with the goal of creating relational value—workers build social connections, workers feel seen, workers work harder. Yet recognition is not an end but a means. Rationales of care occlude how logics of optimization and practices of differential labor management place emphasis on the profit motive. Any possible user’s positive feelings grow to soften or numb the growing pressure to make oneself more productive. And if it feels bad to be lost within the crowd of unrecognized workers, then perhaps they should work harder. After all, this is how workers become “human.”

Digital platforms ought to be understood as labor technopolitical regimes that configure hegemony, and such hegemony makes cruel optimization possible. Platforms represent a style of technological development through which to manage and control workers as users/users as workers. Thinking with digital studies scholar Wendy Chun (2006: 19), human resource platforms “are ideology machines” that produce workers as users who experience their relationships with other workers through the representations afforded by interfaces. Hegemony unfolds through the successful enrollment of workers as users who obediently embrace and enact the logics of optimization and self-improvement configured by digital platforms. In the coming pages, we first situate the plans and actions of Workhuman within insights drawn from platform studies and feminist and critical race STS. This shows how digital platforms function as caring (labor) regimes—care operates as both an expression of concern and a coercive obfuscating practice. The second and third sections of the article address what life is possible when people’s aspirations for recognition and their practices of care are integrated into platforms as regimes of production, dispossession, exploitation, extraction, and circulation. Workhuman's incorporation of people analytics (discussed in Section 2) and the Inclusion Advisor (discussed in Section 3) to its cloud platform showcases how user configurations constitute and are made through the platformization of organizations in late capitalism.

Situating platforms as caring (labor) regimes

The future of work has many pasts just as its architects produce ever more futures—from the anxieties let loose by the automation of labor and the restructuring of the global division of labor into manufacturing and knowledge production enclaves in the early- and mid-20th century to the excitement for information networks and remote work at the turn of the 21st century (Baker, 2018; Handy, 1984; Hodgson, 2016; Noble, 1984). Even as the contours of the future of work in public debate are quite porous, three persistent threads can be found in how actors seek to make sense of, anticipate, and tackle a shifting sociotechnical landscape. Among the recurring areas of concern are what might be the jobs of the future and how technology transforms how people work; the dynamic relationship between jobs, demographic changes, technology, wealth, and social protection systems; and how actors in industry and the wider public interpret, prepare, and respond to these dynamic relationships. Within these historical areas of concern, machines, automation, and artificial intelligence endure as problem objects (Balliester and Elsheikhi, 2018; Michael, 1962; Moynihan, 1963). Rather than being abandoned, many participants today treat the integration and expansion of these problem objects as unquestionable. They are arrangements to be resolved. And digital platforms increasingly play a central role in these efforts, which are also about the management of labor and achieving its obedience without delay.

Techniques of labor management have undergone multiple cycles of transformation from a primary focus on the production process to an ever-expanding examination of both the production process and workers themselves. This includes the entangled discipline and control structures of the slave plantation with masters, overseers, and drivers enacting a “hieroglyphics of the flesh” (Spillers, 1987), all the way to the regulated rhythms of the factory assembly line and their production of efficiencies as conjured by Frederick Winslow Taylor’s (1919) “scientific management.” Writing about the intersecting histories of computing and industrial control, Meredith Whitaker (2023) shows, drawing on the work of digital studies scholar Simone Browne, that control of plantation labor through bureaucratic technologies and rationalist ideas helped ground distinctions of freedom and unfreedom, of Black people as commodities, “something-not-quite-human.” Management of labor has often operated as a means of governing degrees of freedom while guiding worker behaviors. Taylorist and Fordist theories of management and automation of the assembly line were, historian of technology David Noble (1984) argues, direct interventions in interrupting the building of worker solidarities and practices that could undo their exploitation. A manager's systematic surveillance of labor perhaps found no better sociotechnical ally than the clock, as it broke down the workday into equally discrete and measurable parts. This is not to say that the clock determined the habits and routines of industrial capitalism but rather, as historian E.P. Thompson (1967: 80) suggests, that industrial capitalism was concerned “with time-sense in its technological conditioning, and with time-measurement as a means of labor exploitation.” Time was embedded in the clock as sociotechnical measure and vector of differential and differentiating relations. Managing the relation between time, the worker, and work discipline was and continues to be integral to the making and reproduction of capital. Transformations to modes of production are not seamless processes ever improving over time, but emergences along the crooked line of the history of labor management (Burawoy, 1983) that include the plantation and its racializing regime (Johnson, 2001), the assembly line (Peña, 1997), the clock (Thompson, 1967), and the special economic zone (Ong, 2006) just as much as the algorithmic black box (Ajunwa, 2020). As the case of Workhuman shows, platforms today are embedded with techniques of labor management that represent both repetition and difference in the long history of racial capitalism.

Platforms are labor technopolitical regimes in the future of work as they coordinate, manage, and enact operations. ³ An extensive scholarship shows how platforms enable new work arrangements that draw from user free labor (Terranova, 2000) and commodify and promote the consumption of racial performance (Nakamura, 2008) just as platforms make possible microwork (Irani, 2015), crowdwork (Altenried, 2020), and gig work (Vallas and Schor, 2020). Irani, for example, shows how platform microwork is infrastructural by making piecework executed by thousands of workers into the computational power that others get to reimagine as the ghost in the machine. Microwork, she concludes (Irani, 2015: 729–730), constitutes boundaries of distinction between innovative and non-innovative labor, thereby raising the symbolic capital of the former. Infrastructural and dialectical approaches to platforms and labor are integral to how we grapple with the roles of distinction and differentiation at the core of racial capitalism (Robinson, 2000). Platforms understood as labor technopolitical regimes offers insight into the inner workings of what some call the ordinal society (Fourcade and Healy, 2024), platform capitalism (Srnicek, 2017), and surveillance capitalism (Zuboff, 2019). Platforms make possible a society dominated by an orientation towards, justifications for, and government through measurement. Tech companies such as Google and Facebook slowly understood that their data infrastructures—log files, transaction records, user actions, and user histories—were not merely the exhaust of operations but could be repurposed and value extracted from it (Fourcade and Healy, 2024: 11). Platforms turn user labor and user data into commodities, “something-not-quite-human.” We study digital platforms to understand the platformization of work, organizations, and business models.

Platform technopolitical regimes are diverse and diversifying. Digital platforms include people, institutions, practices, ideas, and material implements shaping the organization, enrollment, and activation of specific kinds of informationalized relations and routines. One of the challenges in platform studies can be found in the variation and indeterminacy of the “platform” concept (Cristofari, 2024). Digital studies scholar Tarleton Gillespie (2010: 3–4) points to four semantic territories: the structure that supports a generic or specific activity (architectural), the conceptual foundation for something (figurative), the beliefs of an individual or organization (political), and the infrastructure that supports the design and use of specific applications (computational). We find a combination of these senses in the discourse coming from tech companies like Workhuman, Microsoft, and Google. The Workhuman Cloud is described (Workhuman, 2023a), for example, as “more than a software platform” for it embodies “industry-leading expertise” derived from more than two decades of work in employee recognition. The platform is a figurative, conceptual foundation for it can “empower employees and companies” just as it is computational by supporting the use of specific applications that foment employee engagement and success—thriving through meaningful work.

The centrality of algorithms in platforms has led some to theorize the significance of these technologies in reconfiguring employer–worker and worker–worker relations of production within and across organizations. Algorithmic control in digital platforms is about directing, evaluating, and disciplining workers (Ajunwa, 2020, 2023; Kellogg et al., 2020; Levy, 2022; Zuboff, 1988). As a regime, it is a racial capitalist formation devoted to the management and administration of difference through recursively datafied relations of production, dispossession, exploitation, extraction, and circulation. Platforms trace boundaries around some bodies determined to be productive in specific ways and whose productivity enables the disciplining of other bodies (Gillespie, 2018; Irani, 2015). Worker experiences with platforms are not only varied due to the diversity of participants but also because these regimes build from and reproduce existing asymmetries. The Workhuman Cloud enables algorithmic control through a range of practices and arrangements like gamification, people analytics, and artificial intelligence which leads to the platform's association with discourses of innovation and technological progress. These discourses not only ground but internally animate the pursuit of optimal productivity.

Gamification extends beyond gig work into white-collar work. Workhuman's features of reward and recognition are part of a longer history of gamifying workplaces through the application of game design elements to non-game contexts using points, badges, and leaderboards to motivate and engage users (Deterding et al., 2011). Popularized by marketing and public relations departments in the early 2000s (Fuchs, 2014: 119–120), game elements are used to spur economic activity by influencing producer and consumer behaviors (Rey, 2015: 278). At best, gamification is said to foster empowerment among learners and increase ambition (Alrashedi et al., 2024), yet it may also take advantage of human emotions (Seaborn and Fels, 2015) and function as a tool for worker management and control (de Oliveira, 2021). Moments of accomplishment, we show, are visually represented by leaderboard statistics on Workhuman's platform. Gamification affectively transforms appreciation into a datafied process for defining, comparing, and optimizing employee success leading to more acute managerial control. These tools may simulate a sense of community and connection all the while enabling the feeling of being constantly watched, not only by managers but also by each other.

We contribute to platform studies by examining platforms as expressions of racial capitalist formations and sites for affective work and affective management. The first product we discuss is the Workhuman Cloud which uses gamification and people analytics to, in the language of the company, “make work more human”—optimizing organizational communications, interpersonal relationships, and a sense of belonging in the service of productivity. Workhuman offers its corporate users a platform with a suite of programs that mobilize insights from employee reviews to shape manager decisions and worker behaviors. People analytics and social recognition materialize cruel optimization. Informed by American studies scholar Lauren Berlant’s (2011) theory of cruel optimism, cruel optimization represents the persistent attachment to the belief that steady improvements of the self through sociotechnical arrangements (i.e., algorithmic control, recognition, and rewards) instill a virtuous cycle of betterment even when such attachment works against one's well-being. Optimization's cruel orientation occurs when users affectively embrace the measures and routines dictating what a good worker, a good workplace, and a good work culture look like even though such an embrace results in disavowing other future selves left unimaginable by optimization. Users are prompted to embrace what the managers in the platform reward and value as they incorporate this into their process of self- and sense-making. Cruel optimization represents the hegemonic triumph of platforms as ideological machines occluding practices of labor management and discipline like people analytics under expressions of care. Yet care is instrumentalized to build trust and lower user skepticism as all users should be concerned is that the platform loves them.

And while social relations are inescapable in such processes, cruel optimization is enacted under relative conditions of coercion in which a worker-made user and users-made workers are obliged to agree to the terms of use. This is what Anna Watkins Fisher (2020: 51) calls the coercive hospitality of platforms, or how access and sharing contain hidden costs borne by users. In addition to fulfilling their everyday work responsibilities, workers as users are expected to interact with coworkers on the Workhuman platform by showing how much they care for them. Care work toward coworkers is designed to make them feel recognized, valued, and seen. Data produced through these interactions becomes a valuable commodity for Workhuman to use in further calibrating and developing its platform, and for whatever organization employs it to manage and administer their working populations. Cycles of recognition are expressions of growing worker responsibilities and expansive data extractivism. Workers as users participate in cruel optimization to feel validated and seen but such fulfillment and sense of recognition are to be achieved by obediently maintaining productivity measures even if it negatively impacts workers.

The second product we analyze is an artificial intelligence Inclusion Advisor developed to coach users (manager and worker) through nudges that might lead toward a purportedly more inclusive and equitable workplace. Yet we see efforts in the automation of inclusion as expressions of what Chaar López and Sánchez (2024) call the “racial politics of care.” Care for the Other does not necessarily undo their subjection but rather is entangled with asymmetrical relations that make the Other vulnerable. Even as care holds together that which is of concern, feminist STS scholars have shown it is a non-innocent practice that can also sustain and reproduce harm (Murphy, 2015; Puig de la Bellacasa, 2011). In performing an attentive disposition toward their workers, corporate users of Workhuman's Inclusion Advisor delegate the transformation of their work culture to an AI agent. But as critical information studies scholars show (Grohmann and Araújo, 2021; Irani, 2015), AI agents are in fact never too far from the very human labor performed in data work (e.g., cleaning, tagging, and processing) including the trauma of moderating violent content (Gillespie, 2018; Gray and Suri, 2019) and the embeddedness of wages (Posada, 2024). The racial politics of care enacted by delegating to the AI inclusion advisor means that exceptional and unequal arrangements position some bodies, some communities, and some places as extractable so that others (white-collar workers) are equitably treated and included. Care for some is uncaring elsewhere, another expression in optimization's cruelty.

Care for workers shows how productivity was not the sole concern for companies prior to and after the COVID-19 pandemic. Scholars in industrial and organizational psychology show how worker well-being matters in the internal structures and operations of diverse organizations. One critical aspect of worker well-being is the idea of a work–life balance and how the redrawing of the boundary between these could have a range of impacts. Kossek and Lee (2017), for example, show how the energy, time, and behavioral demands of work can conflict with family and personal aspects such as marital satisfaction or physical health. Understanding the interdependencies between “work” and “home” has been of great concern to scholars (Greenhaus and Kossek, 2014) long before the explosion of pandemic remote work. Workhuman, we show, is part of this conversation as a key player not only in understanding worker well-being but in designing platforms that manage changing work arrangements. Their use of people analytics highlights how attention to work productivity and worker satisfaction and sense of fulfillment is part of a longer labor management regime. As the spirals of platform-based work grow outward, the COVID-19 pandemic represents an important conjuncture in the tightening screws of labor precarity.

Optimizing the human through social recognition and people analytics

To show how the Workhuman platform works, we examine how the company came to understand “social recognition” as an integral part of an organization and how people analytics offered a means to measure and instrumentalize such recognition. We argue that this conjuncture is an articulation of cruel optimization, as the desire for meaning and care is validated because it maximizes worker productivity. Platforms, like the Workhuman Cloud, become infrastructure by turning operational (user–worker) data productive to order, distribute, and coordinate labor in organizations. This often occurs by implementing gamifying logics that foment worker competition with themselves and each other through the accumulation of rewards and points. Such gamification underscores the cruelty of optimization by turning a desire for recognition into a hierarchy of value that could affect self-worth. Workhuman's diverse portfolio of clients who use its products, which include Merck, LinkedIn, Cisco, Citizens Bank, and British Petroleum, demonstrates its global reach in coordinating white-collar labor. Other tech companies like Microsoft and Google have sought to transform organizations and their operations through their own productivity applications (e.g., Word, Outlook, Docs, and Gmail) and platforms (e.g., Microsoft Teams and Google Workspace). We focus on Workhuman because its efforts offer insight into the epistemic infrastructures that make possible the platformization of organizations and work.

Workhuman has been concerned with worker well-being since its founding as Globoforce in 1999. The company offered other companies a service for managing gift-giving as rewards for worker performance. Workers would nominate a colleague for commendable behavior, and their company manager decided whether to approve this nomination for an award. Once approved, the worker received a notification with their award—both a letter with the words of recognition by colleagues and a gift, usually a gift card from a popular brand. Gift-giving, as sociologist Marcel Mauss (1990) argues, enables reciprocity and obligation, which in this case meant the awarded worker paid back its employer by continuing to meet productivity goals. Gifts as a practice of employee recognition revealed to Eric Mosley, Globoforce cofounder, that businesses and industries had a pressing need for feedback, and to treat “employees as people and ‘not human capital’” (Schawbel, 2016). This animated the motto to make work more human, which slowly morphed from company mission to company name, Workhuman.

The name of the company announces itself as a moral liberal platform. Workhuman organizes collective norms of obligation and the relationship between managers and workers. It is an imperative statement, a command—“work, human.” No laziness or stillness is accepted. The expenditure of bodily energy is a categorical imperative. Work is that which no one should, let alone can refuse. Yet the name of the company and its motto betray a contradiction. If the “human” must work, then work's inhumanity must be undone. The association of work with a kind of inhumanity speaks to what philosopher Hannah Arendt (1998) saw in Western antiquity's treatment of labor as that action taken in the service of maintaining life. This is labor's slavishness, a negative condition in any attempts at human affirmation. Such formulation endured in post-Enlightenment thinking as the rational subject was conceived, in contrast to the enslaved person and their inability to access certain faculties, as the person who had the freedom to reason, act, and judge (Kant, 1996; Wynter and McKittrick, 2015). How to bring about the human? In Workhuman's efforts, we find vestiges of the historical project of liberal humanism, which American studies scholar Lisa Lowe (2015: 39) shows emerged from “racial classifications and an internal division of labor.” The “human” materializes while an Other is unfree.

Workhuman vouches to create “human workplaces.” This is a “new work paradigm” where, according to Derek Irvine (2021), Senior Vice President of Strategy and Consulting at Workhuman, companies “leverage the power of human connection to build resilient, high-performing teams” by enabling and inspiring workers “to give their best.” Workers crave “recognition” which a Gallup and Workhuman report defined as, “praising, acknowledging or expressing gratitude to employees for who they are and what they do” (Gallup and Workhuman, 2022). The desire and need for caring and supportive relationships (“connection”), Irvine suggests, is translated into a work culture “where employees feel accepted, recognized, and rewarded for their unique qualities.” A core of making human workplaces is Workhuman's Social Recognition program, a continuation of gift-giving, whereby peer-to-peer interactions are allowed to formally acknowledge and reward good work. The game element of reward is also about fostering consumption (e.g., gift cards given to recognized workers) as it masks structural processes of exploitation (Rey, 2015). Workhuman's program facilitates employees as users to thank, acknowledge, and celebrate others within their organization at any given time—with managers and other supervisors able to see who is being recognized most often and for what. The publicity of recognition disciplines other worker-users into committing themselves to becoming self-improved and obedient humans.

In a promotional video from Workhuman (2019), we get a glimpse into the Workhuman Cloud platform in action as an arrangement that organizes employee feedback and well-being across diverse and multi-sited workplaces. A White male manager hops on a cab where he accesses the platform to record and deliver a video to a White female employee named Nicole, congratulating and thanking her for her presentation which he believed showcased how without “your ideas, your creativity, your dedication, we couldn’t have done it without you.” A Black female colleague at a construction site receives a platform notification on a tablet computer to approve or disapprove the nomination of Nicole for a “Delivering Excellence” award. She happily approves. Nicole then opens her virtual award and video message showing the number of points she received. She smiles while another male work colleague at an office setting opens the video post and comments approvingly of Nicole's recognition. The shot transitions to show Nicole sitting on a kitchen island in her home, an allusion to her flexible and remote work arrangement, when her two children come into the frame with their Black father. She proudly shows him her accomplishment at work. Racially and gender diverse workplaces are successfully organized by Workhuman's platform as users communicate and collaborate with one another in creating a “human workplace.” Computer-mediated collaboration facilitates not only encounter and coordination but practices of recognition that allow workers to feel valued and accomplished. The technoscientific scene represents a scenario whereby workers as users feel their work is meaningful rather than just a job.

Social recognition in the digital platform manages crowds through standards of performance all the while treating moral imperatives as if part of a voluntary moral economy. Eric Mosley (2012), one of the co-founders, believed “the wisdom of crowds” was strategically positioned to transform human resource management. Crowdsourced reviews allowed managers to draw on employee practices of recognition to collect, evaluate, and share information on employee performance. These reviews capture the “input from many, rather than a few or just one” thereby “extend[ing] performance evaluations beyond a single point of failure [in space and time] to reveal how employees are truly performing and influencing others in the organization.” Up-to-date evaluations draw not only from statements by one or two managers but turn a company's interactions between employees into extractable matter for analysis (Securities Exchange Commission and Globoforce Limited, 2023). Crowdsourced performance reviews rely on collected year-round instances of achievement, recognition, and worker mood and offer quantified measures that translate worker actions into parameters that managers handle and monitor on a dashboard (Figure 1). Combined programs (Workhuman, 2023a) constitute a composite image of a worker as a “whole human,” defined through: routines of and desires for recognition; the instrumentalization of conversations between workers and with managers as structured project development; the management of worker sentiments as barometers of corporate stability; prompts asking workers to disclose life events and work milestones; and the organization of sites/practices of social connection to fabricate a sense of corporate intimacy. These reviews, Mosley (2012) concludes, are meant to inspire workers rather than foster obligation to perform in a particular way all the while accountability expands through networked interactions. The cruel optimization of crowdsourced performance reviews has company managers in-source their evaluative work to those they oversee. This is among the things Kim and Werbach (2016) warn as an ethical issue of gamification: the manipulation of workers through the confusion of individual with corporate interests. Workers are coerced into a surveillant media network that records and processes organizational procedures.

Figure 1.

Workhuman manager dashboard.

Employee recognition efforts in Workhuman's Cloud produce data to be processed through people analytics. Beyond workplaces, people analytics emerged in the mid-twentieth century to leverage computational power to understand and anticipate human behavior—from international politics and warfare to public elections and consumer decision making (Lepore, 2020). But as a business and organization approach, it mobilizes quantitative analysis to raise the efficiency of operations. Diverse phenomena, from job traits and practices to employee histories, behaviors, and affects, are translated into metrics to be measured and recorded for future processing to improve operations, planning, employee recruitment and training, and overall performance (Bodie et al., 2017; Giermindl et al., 2022: 414). An underlying assumption in people analytics, as information studies scholars Giermindl et al. demonstrate, is that it offers the possibility to reduce human-error in decision making and allow organizations to anticipate human behavior. Work efficiencies, however, are not only justified from a productivity standpoint. People analytics’ goal of developing and shaping human behavior and character, organization studies scholars Gal et al. (2020) argue, also draws from the human relations movement. Workers have needs and desires, and improved management can secure their commitment (Grant et al., 2008). Human relations seek to reduce worker stress levels, increase their satisfaction, and expand opportunities for personal and professional growth. In its attention to self-fulfillment and self-growth, people analytics seeks to guide workers to voluntarily perform a virtuous cycle whereby they develop their practices and themselves.

Although well-intentioned, such an approach can simultaneously lead to negative, even cruel arrangements. Social recognition in the Workhuman platform instrumentalizes workers’ life stories, emotional interiority, and longing for acknowledgement, validation, and acceptance as vectors in the administration of organizational homeostasis. Labor is managed through the construction of a particular kind of humanity, one where recursive practices are performed in the service of self-improvement through the datafication of the flesh. In making awards publicly visible to other users of the Workhuman platform, like in the case of Nicole in the ad (Figure 2), awards function like badges that workers accumulate and show other workers what managers and other colleagues deem as proficient, efficient, and commendable work. People analytics grows entangled with gamification as badges, rewards, and points become symbolic expressions of social status. Badges can include a range of human dimensions like deliver excellence; respect for teamwork; leadership; be authentic now; ensure trust; make a meaningful difference; and pursue bold aspirations (Workhuman, 2019). Workers are led to internalize what is model work, the kind of effort they should strive to emulate. Scholars show how organizational and individual interests do not always align when fomenting job competition (Gilbert et al., 2009) and gamification (Kim and Werbach, 2016). Even as organizations may achieve notable work outputs that satisfy their productivity goals, deviation from model work means that worker efforts may go by unrecognized, potentially leading to anxiety about job security and doubts about self-worth.

Figure 2.

Social recognition awards.

People analytics and efforts at platforming work so that workers are singled out as users (both as individuals and populations) build on the hegemony of neoliberal logics and the longer history of labor management in capital. The human at this intersection is an entrepreneurial subject responsible for its well-being, where concern toward the moral improvements of the worker is not exclusively a humanist and virtuous orientation. Instead, cruel optimization redirects aspirations for moral growth toward production efficiencies. Neoliberal thought infiltrates everyday life, science studies scholar Aihwa Ong (2006: 4–6, 21–22) argues, by assembling arrays of knowledge and expert systems like people analytics which induce self-animation and self-government all the while regulating populations for optimal productivity. People analytics and social recognition are treated as techniques of use (Lin and Lindtner, 2021) for global capital in its management of work and workers as they instill a cruel optimism. Berlant (2011: 1) states that “[a] relation of cruel optimism exists when something you desire is actually an obstacle to your flourishing.” It can also be “a new habit that promises to induce in you an improved way of being.” By shaping workers to be an improved version of themselves, the Workhuman platform appeals to employees’ emotions—once they feel invested in, they in turn invest back into their company. This cruel optimization that Workhuman and the platform's configuration instill in users creates an attachment to their workplaces, to reach a desired goal or reward that exists if they do acceptable work. Optimization creates a “humanity” tied to how much value is placed on workplace achievements and in performing according to the standards established in high regard, which is the same as those in high reward.

Workhuman's categorization of commendable and exemplary performance creates standards of action that allow companies to differentiate between them. Practices of separation, differentiation, and distinction are embedded in the exaltation of those who are not alike, those whose work is not like that of others. The more recognized a worker becomes, the better strategically positioned they are in relation to others—creating inequality among coworkers, who now believe they must go above and beyond to receive the same rewards. But by the same token, recognized workers are now subject to the expectations they once met. The economization of life (Murphy, 2017) operates through a moral calculus determining the worth of subjects, practices, and visions of the good. The human subject in people analytics is a calculating and calculated entity shaped by as well as shaping “if, then” promises of perpetual improvement. Cycles of recognition in people analytics open workers to enacting attachments of cruel optimization where desires for feeling seen and accomplished govern the good and virtuous life as a productive life even if it may undermine their well-being.

Configuring the human and the racial politics of care

These cycles configure the human (user) to feel as if capital and its machines care for them. The integration of people into modes of production produces boundaries in terms of who and what values count as human, and under what conditions. The following technoscientific scene comprises artificial intelligence, techniques of nudging, and the disciplining of subjects under the cruel optimization of inclusion. Workhuman's product recognizes and maintains differences to facilitate the productivity of diverse bodies. Fair treatment of diverse workers within an organization occurs through the intensification of data extraction and the labor supply chains that support the development of machine learning models. AI's care for difference is also its cruel indifference. This is the racial politics of care, or how support and care for some is made contingent on the subjection of others.

“Hi human!” the Workhuman Inclusion Advisor exclaimed from the pop-up screen (Workhuman, n.d.). The notification was triggered when the manager moved his cursor over the underlined text in his message to a coworker. The manager sought to celebrate the critical contributions made by his coworker, so he wrote, “You are a role model for other women on the team!” The Inclusion Advisor, a feature of the Workhuman Cloud platform powered by artificial intelligence, noted that such language suggested the coworker was “only able to impact those in their own gender sphere.” Warned that the language used could be perceived as “unconscious bias,” the manager was advised to deploy gender neutral language noting the worker's ability to impact “everyone on the team.” This improved and caring correction, although a small change, is part of how Workhuman seeks to improve worker relations in organizations. Interactions with the Inclusion Advisor as well as with other features of Workhuman's Cloud are designed to foster and “build cultures of belonging” that are attentive to its different members. Integrating inclusion into its user configurations, the cloud platform treats it as a programmable and reproducible routine designed to address issues of unfairness in the workplace and in a changing work landscape.

This scene, drawn from a video demonstration of the company's “workplace solutions,” sets the stage for the kind of future of work they anticipate and make possible: work that requires programmable routines of care enrolling artificial intelligence as a collaborator and delegate. The Inclusion Advisor enacts what Atanasoski and Vora (2019: 90–91) theorize as surrogate humanity, or the fantasy of “human-free social environments… concerned with replacing the racialized and gendered surrogates enabling freedom for the universalized liberal figure of the human with technological surrogates.” The future of work rehearsed in this scene enlists the AI advisor to conduct affective and political labor that corrects the manager's treatment of a woman coworker. Care work draws from and perpetuates a global division of affective labor (Duffy, 2011; Mezzadra and Neilson, 2013; Rodrigues, 2010; Vora, 2015) whereby mostly women of color become infrastructural for those within enrichment zones. Care work, labor that sustains, is often racialized and gendered invisible labor. Imagined as male, the manager outsources his communication with a coworker to the AI advisor while he is released from performing any semblance of attentive responsiveness and respect. This is fair treatment without the embodiment of fairness.

Making work more human promises to tend to the laborer as a complex subject through routines of care. It recognizes that there has been a failure of sorts in white-collar labor management by losing track of workers’ humanity. From the ravages of stress, anxiety, and mental fatigue triggered by the COVID-19 pandemic to the heightened attention given to gendered and racial injustices, Workhuman intervenes in this conjuncture by increasing the possibility that the rich and multilayered fabric of humanity can be more present in work environments. The future of work requires that workplaces recognize the diversity within their workforce while they maintain their inclusivity. Inclusivity is made into maintenance, reparative work to be performed through what Workhuman calls “micro-coaching.” The Inclusion Advisor, for example, is designed to foster a company's culture of recognition by advising platform users on the language they use while writing their awards. The Advisor, trained with a database comprising more than 70 million moments of recognition, identifies words or phrases that might “be perceived as bias” (Workhuman, 2022). The user is nudged by the Advisor that their word choice could be interpreted as troubling and lead to reproducing an uncaring work environment. Once the worker changes the word choice in their message, they can recheck it before sending it to their coworker. Micro-coached, users modify how they describe and name their diverse colleagues to satisfy the standards policed and maintained by the Inclusion Advisor. Because this interaction occurs multiple times a day and every week, micro-coaching becomes a technology of governing the everyday, a means of regulating how care is demonstrated and performed in recurring interactions between workers as users.

What Workhuman calls micro-coaching is described in the people analytics literature as nudging. Nudging in labor management platforms narrows the space of agency by pre-scribing pathways of action that ask users to entrust themselves to the system with minimal recourse for self-reflection. In human–computer interaction, a nudge comprises aspects of an interface architecture that alter how users behave and that seek to produce predictable behaviors in response to prompts. The practice draws from behavioral economics’ insights into influencing people's decisions without forbidding choice but rather narrowing it in service of maximizing health, wealth, and well-being outcomes (Thaler and Sunstein in Gal et al., 2020; Lin et al., 2017). Among myriad issues scholars have found with nudging are that: it prevents workers from pursuing their own outcomes and interests rather than those prioritized by the company; it does not afford workers the opportunity to meaningfully reflect and understand the situation where they were nudged toward certain behavior; and it is a manipulative practice that keeps workers from deliberating among themselves about meaningful choices and finding coherent interpretations that explain why they acted in such a manner (Gal et al., 2020). The integration of nudging to a chatbot (Inclusion Advisor) transforms dialogue into chat by breaking human conversations into discrete turns that chatbot design teams use as a metric of efficiency—a low number of turns in an exchange being the measure of a successful exchange (Slater, 2022: 187). The nudged subject is one that heavily associates its self-preservation with a division of labor premised on its well-being but directed toward the maximization of the production and capture of value.

The user configurations of the Inclusion Advisor enact a racial politics of care. Chaar López and Sánchez (2024) develop this concept by thinking with feminist and critical race STS scholarship that troubles the association between care and positive, good feelings. Care requires a disposition that tends to its messiness and contradictions, its non-innocence, and how phenomena are differentially (re)assembled (Martin et al., 2015; Murphy, 2015; Puig de la Bellacasa, 2011). Studies of racial politics of care tackle how practices of considerate and thoughtful attention equally maintain networked asymmetries whereby some bodies are made productive and life-sustaining for others. Its relational approach identifies discriminatory practices of thoughtful selectiveness, while allowing, as Katherine McKittrick (2021: 44–51) sharply argues about Black method-making, for agency and knowledge to unfold past understandings that posit them only as emerging from the violently subjected body. In this sense, care calibrates the study of matters of concern to unsettle its potential complicity in entrenching stratifications (Murphy, 2015). Workhuman's platform care for worker inclusion and recognition is meant to de-alienate workers, yet it intensifies exploitation (Fisher, 2012) across scales.

Routinized recognition highlights how Workhuman's care is a neoliberal sociotechnical arrangement. In a discussion of how Merck piloted the use of the Inclusion Advisor (Workhuman, 2023b), Christopher Cardarelli, Executive Director of Global Diversity and Inclusion, said the AI helped “remind people to be equitable and inclusive, both in terms of who they were extending recognition to and the words they were using to recognize them.” The Inclusion Advisor operated through pop-ups that people could “choose to click on” or they could determine that, “I don’t need that.” But Cardarelli states that most users did in fact click on the pop-up which nudged them toward using language the system deemed to be more inclusive. Inclusive behavior was not so much an expectation or an ethical imperative but an individual choice. Making inclusive behavior something that can be nudged away allows users to not click on the pop-up and to continue those practices the Advisor recommends against. Nudging treats inclusivity as a skill that workers can learn and develop in the persistent entrepreneurialization of life. And, in the case that the market makes inclusivity an unnecessary or discouraged commitment, which the second Donald Trump presidency has done, it can then be dropped in favor of other nudged behaviors. With thought and self-reflection expulsed from the process of recognition, nudging materializes cruel optimization by making well-being into a distributed assembly line of voluntary acts that individuals can follow or disregard in the perpetual self-alienation of labor.

Furthermore, the AI advisor is an obfuscating artifact drawing together heterogeneous elements in the racial administration of labor. It is an assemblage embedded with insourced and appropriated data from within the organization as well as outsourced labor for data management (Chaar López and Sánchez, 2024; Perrigo, 2023). Crowdsourced performance review data distribute the affective and creative labor to other workers who use the Workhuman platform. Their labor and work data are necessary for the functioning of the Inclusion Advisor. Such data is most likely “cleaned” through outsourced labor coming from “zones of depletion,” or areas where racial capitalism's human and natural resource extraction leaves behind toxic environments (Precarity Lab, 2020). This is what some scholars call ghost work (Gray and Suri, 2019). The labor that makes artificial intelligence possible is embedded in the programmable routines of care that managers perform for the Other. Not only does the AI advisor operate thanks to the invisible data-generating work performed by workers enacting the company's routines of care. It operates to reproduce the extraction and processing of operational data that allows the company to function as a platform. This is the feedback loop of cruel optimization. AI-sustaining labor is infrastructural for some to embody and perform a kind of (caring) humanity. The technosolutionist response is to improve the production and circulation of messages in collaboration with an AI delegate that requires the subjection of others in the liberation of some.

Conclusion

Innovation goes hand in hand with arrangements of cruel optimization. In public debate in the USA and other geographical areas, innovation is taken up uncritically as an inherent good. To innovate is thought of as necessarily improving upon what exists even when it often leads to the intensification and reproduction of an existing asymmetry of things (Suchman and Bishop, 2000: 331). In 2018, a Task Force (Autor et al., 2019, 2020) at the Massachusetts Institute of Technology was convened to grapple with how an emerging “age of intelligent machines” might transform how people work, collaborate with one another, and more. They found that technological development throughout the 20th century led to growth in productivity and wealth. The problem has been the increasing gap in wages and the unequal distribution of wealth created through rising productivity. They concluded the challenge was “building better jobs,” meaning equally distributing the fruits of increased productivity, increasing labor participation and upward mobility, and improving racial and ethnic disparities that lead to unequal opportunities (Autor et al., 2021). If “innovation fails to drive opportunity,” the leaders of the Task Force held, it can quickly “generate a fear of the future.” Innovative platforms, they reasoned, are bound to the creation of social, economic, and political improvements. Only when innovation fails are improvements not met or progress falls apart.

Digital platforms, this article shows, habituate users into scripted practices meant to enact efficiencies through discourses of belonging and care. Cruel optimization names the dynamic through which platforms build and maintain the hegemony of neoliberal logics. Understanding digital platforms as cruelly optimized labor regimes helps us grasp the embodiment of these logics and their material connection to the long history of labor management in capital. Our study of the Workhuman Cloud shows how its clients reintegrated remote workers into a single space where their actions and emotions were surveilled and disciplined through gamification, where workers internalized self-government in the benefit of optimal productivity. Workhuman's gamification is part of a post-COVID-19 future of work that promises to “make work more human” yet ensconces the “human” as an extractable and profitable object. Cruelly optimized workers represent the triumph of capital in this conjuncture, a nightmare scenario in “the age of intelligent machines.” But what the technoscience scene with the Inclusion Advisor demonstrates is that there are knotted dimensions to the growth of better jobs for people. Platforms of care for the Other so often come at the expense of many. That a priority is set to support some workers all the while others are subjected to routines of toxicity—of toxic content moderation, toxic dullness, and the toxic debilitation of enduring exhaustion and disappointment. One question we are left with here, however, echoes findings from the scholarship on Fordism (Peña, 1997; Watson, 2019). ⁴ How might cruel optimization enacted through digital platforms facilitate and dissuade worker solidarity? What “hidden transcripts” (Kelley, 1993; Scott, 1990) may we find in observing workers as users in action on digital platforms?

The value of life has no measure until it does. The ongoing push to leverage people analytics and platforms in the business world and in organizations needs to be understood in relation to the messy and violent history of measuring and valorizing life, although especially of drawing boundaries around who is made “human” and who is not. Company mottos (make work human) and research agendas (“Better people analytics. Measure who they know, not just who they are”) suggest that an analysis of people analytics and platforms of care requires a critical inquiry of differential technologies of use, one that is concerned with making heterogeneity productive to the benefit of the few. This is an innovation that enacts arrangements of cruel optimization justified through a racial politics of care.