ChapterPDF Available

Chapter 3 Platforms at Work: Automated Hiring Platforms and Other New Intermediaries in the Organization of Work


Abstract and Figures

This chapter lays out a research agenda in the sociology of work for a type of data and organizational intermediary: work platforms. As an example, the authors employ a case study of the adoption of automated hiring platforms (AHPs) in which the authors distinguish between promises and existing practices. The authors draw on two main methods to do so: critical discourse analysis and affordance critique. The authors collected and examined a mix of trade, popular press, and corporate archives; 135 texts in total. The analysis reveals that work platforms offer five core affordances to management: (1) structured data fields optimized for capture and portability within organizations; (2) increased legibility of activity qua data captured inside and outside the workplace; (3) information asymmetry between labor and management; (4) an “ecosystem” design that supports the development of limited-use applications for specific domains; and (5) the standardization of managerial techniques between workplaces. These combine to create a managerial frame for workers as fungible human capital, available on demand and easily ported between job tasks and organizations. While outlining the origin of platform studies within media and communication studies, the authors demonstrate the specific tools the sociology of work brings to the study of platforms within the workplace. The authors conclude by suggesting avenues for future sociological research not only on hiring platforms, but also on other work platforms such as those supporting automated scheduling and customer relationship management.
Content may be subject to copyright.
Work and Labor in the Digital Age
Platforms at Work: Automated Hiring Platforms and Other New Intermediaries in the
Organization of Work
Ifeoma Ajunwa, Daniel Greene, *
Article information:
To cite this document: Ifeoma Ajunwa, Daniel Greene, * "Platforms at Work:
Automated Hiring Platforms and Other New Intermediaries in the Organization of Work"
In Work and Labor in the Digital Age. Published online: 14 Jun 2019; 61-91.
Permanent link to this document:
Downloaded on: 18 June 2019, At: 02:20 (PT)
References: this document contains references to 0 other documents.
To copy this document:
Access to this document was granted through an Emerald subscription provided by
emerald-srm:485088 []
For Authors
If you would like to write for this, or any other Emerald publication, then please
use our Emerald for Authors service information about how to choose which
publication to write for and submission guidelines are available for all. Please visit for more information.
About Emerald
Emerald is a global publisher linking research and practice to the benefit of society.
The company manages a portfolio of more than 290 journals and over 2,350 books
and book series volumes, as well as providing an extensive range of online products
and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner
of the Committee on Publication Ethics (COPE) and also works with Portico and the
LOCKSS initiative for digital archive preservation.
*Related content and download information correct at time of download.
Downloaded by University Library At 02:20 18 June 2019 (PT)
Ifeoma Ajunwa and Daniel Greene*
This chapter lays out a research agenda in the sociology of work for a type
of data and organizational intermediary: work platforms. As an example, the
authors employ a case study of the adoption of automated hiring platforms
(AHPs) in which the authors distinguish between promises and existing prac-
tices. The authors draw on two main methods to do so: critical discourse analy-
sis and affordance critique. The authors collected and examined a mix of trade,
popular press, and corporate archives; 135 texts in total. The analysis reveals
that work platforms offer ve core affordances to management: (1) struc-
tured data elds optimized for capture and portability within organizations;
(2) increased legibility of activity qua data captured inside and outside the
workplace; (3) information asymmetry between labor and management;
(4) an “ecosystem” design that supports the development of limited-use
applications for specic domains; and (5) the standardization of managerial
techniques between workplaces. These combine to create a managerial frame
for workers as fungible human capital, available on demand and easily ported
between job tasks and organizations. While outlining the origin of platform
studies within media and communication studies, the authors demonstrate the
Work and Labor in the Digital Age
Research in the Sociology of Work, Volume 33, 61–91
Copyright © 2019 by Emerald Publishing Limited
All rights of reproduction in any form reserved
ISSN: 0277-2833/doi:10.1108/S0277-283320190000033005
Downloaded by University Library At 02:20 18 June 2019 (PT)
specic tools the sociology of work brings to the study of platforms within the
workplace. The authors conclude by suggesting avenues for future sociological
research not only on hiring platforms, but also on other work platforms such as
those supporting automated scheduling and customer relationship management.
Keywords: Automation; hiring; algorithms; platform authoritarianism;
design; brokers
The sociology of work has long-linked analyses of macro-structural changes in
labor markets, institutional norms, and corporate organization with workers’
experience of the labor process, their investment in it, and their outcomes from it
(Kalleberg, 1989). This has only become more of a challenge in the information
economy, where work is increasingly organized by technological platforms whose
logic is opaque to employees (Aneesh, 2009; Schor & Attwood-Charles, 2017).
The reach of these work platforms exceeds the boundaries of individual organi-
zations, linking contractors across continents or tracking employee data across
other social spaces (Winter, Berente, Howison, & Butler, 2014). This restructuring
of management through data relations begins even before the employee’s rst day
of work.
Applying to work at a Target department store, for example, requires appli-
cants to spend hours on an Automated Hiring Platform (AHP) submitting their
work history, personally identiable information, and scheduling availability;
agreeing to background checks; and participating in lengthy personality and
skills assessments, all quickly analyzed by the platform and processed for the
hirer. The interface and analytics of the AHP is structured on Target’s terms.
Although the system is not designed by Target but by the data broker Equifax,
best known as a consumer credit reporting agency (Marron, 2007), the software
will process and sort applicants pursuant to the client’s criteria. There is a pre-
set menu of options for applicants to choose from at each point, adorned with
the company’s logo and colors. Even the rare open text eld found on the work
history page limits applicants to just 32 characters of description for each past
job. We term the sociotechnical phenomenon presented by this platform struc-
ture, platform authoritarianism (Ajunwa, 2018): the platform restricts the actions
available to workers on one side while offering new affordances to employers
on the other. Employers gain penetrating new insights into current or potential
employees, but the latter have no room to negotiate. Rather, job applicants must
engage with the platform as dictated or lose the opportunity to work. Other
major AHP vendors include Kronos, SnagAJob, and Recruit. Barber (2006) nds
that these recruitment and screening tools are now used by nearly all Global 500
companies, with paper applications increasingly unavailable except as accom-
modations for a disability.
In this chapter, we put forth a research agenda for the study of work plat-
forms such as these, using the conceptual and empirical tools of the sociology
Downloaded by University Library At 02:20 18 June 2019 (PT)
Platforms at Work 63
of work to open a new path in the platform studies conversation that is presently
dominated by media and communication studies. As an example of our proposed
approach, we explore the adoption of AHPs in the 1990s and early 2000s as a
case study.
What Are AHPs For?
By “platforms,” we mean digital intermediaries that invite submission of data
from one party through pre-set interfaces and structured protocols, process that
data via proprietary algorithms, and deliver the sorted data to a second party.
The raison d’être of platforms is data manipulability – data made more malle-
able through pre-set interfaces, which can be customized with applications built
for that specic ecosystem. Platforms are developed by a third party and rhe-
torically appeal to the data submitter as neutral intermediary, but almost always
have a nancial relationship with the analyzing party that purchased the plat-
form software or its analyses. “Platform studies” have largely developed within
the academic elds of media, information, and communication studies, where
rich analyses of social media platforms like Facebook, Twitter, and the like have
explored the construction of these digital spaces, user activities on them, and the
political economy of data within them. As the business model for these data inter-
mediaries emerged rst within the advertising-funded web services represented by
social media (Srnicek, 2016), it is not surprising that the analytic tools addressing
them developed within elds that have historically addressed the social impact of
advertising-supported media.
Yet, platforms are not isolated to social media. Rather, we are in the thrall
of a data-driven reorganization of the workplace that takes as its impetus, the
“workforce science” derived from nineteenth-century Taylorism and twentieth-
century Fordism (Ajunwa, Crawford, & Ford, 2016). Consider that automated
hiring requires a platform; an intermediary, in the form of an AHP drawing
on outside databases to screen applicants, deploying assessment tools honed
through engagement with thousands of applicants in disparate settings. But
there is a fundamental difference between Facebook’s social media platform and
Equifax’s AHP. While social pressure or market concentration within the sector
might make Facebook the most likely choice for certain online social activities,
its use is still voluntary and user submission of data are not coerced. Not so for
the jobseeker who must use an AHP to seek employment at Target or Walmart,
the barista who must get her schedule from Kronos, or the marketer who must
record her calls in Salesforce. At work, platforms represent a coercive force, even
when such coercion is “gamied” and invites participation with rewards rather
than overt punishment (Cohen, 2015). Recall that the true client for an AHP
vendor is always the hiring company who dictates the end goals and often the
process to accomplish them.
In this chapter, we describe how these social relations manifest in the design of
AHPs and the promises that accompany them, demonstrating the importance of the
sociology of work to investigations of what some have called “platform capitalism”
(Srnicek, 2016) and outlining a research agenda for the study of work platforms.
Downloaded by University Library At 02:20 18 June 2019 (PT)
We nd that platforms offer ve core affordances to management: (1) structured
data elds optimized for capture and portability within organizations; (2) increased
legibility of activity qua data captured inside and outside the workplace; (3) infor-
mation asymmetry between labor and management; (4) an “ecosystem” design that
supports the development of limited-use applications for specic domains; and (5)
the standardization of managerial techniques between workplaces. These combine
to create a managerial frame for workers as fungible human capital, available on
demand and easily ported between job tasks and organizations.
In what follows, we explore the insights that platform studies offer the sociol-
ogy of work regarding the circulation of data within platforms, as well as, the
interventions the sociology of work can offer regarding how organizations struc-
ture the labor process and the role of organizational intermediaries. These insights
are demonstrated through a case study: the history of automated hiring and its
evolution into a platform service through the 1990s and into the 2000s. We con-
clude by returning to our primary contribution – the managerial affordances of
work platforms – and draw on them to suggest avenues for future research.
Conceptual denitions of platforms and empirical methods to investigate them
are most highly developed in communication and media studies, which have
largely focused on social media. However, these tools are inadequate for the study
of work platforms. The design of work platforms is straightforwardly grounded in
hierarchical, prot-maximizing social relations that diverge from the assumptions
(if not the operation) of public, networked, conversational, and democratic social
relations embedded in the design of social media. The hierarchical social relations
within work platforms differ qualitatively from those in social media, with fun-
damentally different avenues of coercion and consent, many native to the work-
place, now digitally transformed, available within work platforms. Consequently,
work platforms must be studied as the successor to technologies of control long
analyzed in the sociology of work.
Platform Studies
Platform studies in other domains, such as communication and media studies,
prove useful in articulating how platforms differ from other types of technology
or software that may be used in the workplace. The “platform” label embraced by
advertising-funded hosts of user-generated content, such as YouTube, Facebook,
and Tumblr, develops from older denitions of the term – as computational infra-
structure, architecture, or political position – and has come to mean a neutral
intermediary advancing free expression, without the intermediary holding any
liability for the content posted by third parties (Gillespie, 2010).1 This is despite
the fact that the promotion of certain user-generated content, the censorship of
Downloaded by University Library At 02:20 18 June 2019 (PT)
Platforms at Work 65
others, and the ltering and ranking of content streams is the core commodity
these services offer.
Exploring this complex interplay of content, intermediary, user, client, and
software requires a complex set of conceptual and empirical tools. Van Dijck
(2013) denes platforms as
[…] the providers of software, (sometimes) hardware, and services that help code social activi-
ties into a computation architecture; they process (meta)data through algorithms and formatted
protocols before presenting their interpreted logic in the form of user-friendly interfaces with
default settings that reect the platform owner’s strategic choices. (p. 29)
Van Dijck’s denition is content-agnostic, though she largely limits her analy-
ses to social media platforms. Srnicek (2016) draws on a wider network of political
economy and labor studies scholarship to position platforms as economic inter-
mediaries in a new phase of capitalism in which data are the primary commod-
ity. He develops a exible typology of platforms based on their revenue models.
Advertising platforms such as Google provide free services to users, track them,
and sell their data to advertisers. Cloud platforms such as Salesforce lease digital
services (e.g., storage space and enterprise software) to businesses for use in their
own operations. Industrial platforms such as GE integrate data-mining features
into industrial production, turning xed capital into leased services. Product plat-
forms such as Rolls Royce use data-tracking to transform goods (e.g., aircraft
engines) into subscription-based services. Lean platforms such as Uber develop
software to outsource the operation of core business assets (e.g., cabs and their
drivers) to third-party contractors.
Van Dijck’s image of platforms is that of a net: free-owing social activity is
caught, sorted, and the results delivered to the other – paying – side. This model,
as Srnicek notes, was rst developed for advertising platforms in the early days of
Web 2.0, and then generalized to other sectors. While Srnicek expands platform
studies’ scope to data commodication in a variety of domains, the eld still lacks
analyses of how platforms mediate employer–employee relations – necessitating a
closer look at platforms through the sociology of work.
A Sociology of Platforms at Work
Connecting platforms to the sociology of work advances the literature on work-
place technology, particularly in the context of control. Control, with or without
technology’s assistance, is established through different gradations and couplings
of coercion and consent. As workers, we are forced to consent to the information
asymmetries and command structures embedded in the workplace, because we
must work to afford housing, food, etc. (Anderson, 2017). At work, employees
face the possibility of managerial coercion through threats of ring or punitive
adjustment to duties or schedules, should they not adjust their level of produc-
tivity or comportment to the needs of management (e.g., O’Connor, Kmec, &
Harris, 2015; Sewell & Wilkinson, 1992). Complementing or substituting for
these threats are relations of consent – carrot over stick. Employers and their
representatives structure the labor process such that rewards, either material or
Downloaded by University Library At 02:20 18 June 2019 (PT)
norm-based, are offered to workers, who must then decide whether the bargain is
worth it (e.g., Hodson, 1999; Mears, 2015; Tuckman & Whittall, 2002).
New work technologies such as AHPs concretize social relations of coer-
cion and consent at work, a phenomenon we term platform authoritarianism
(Ajunwa, 2018). Edwards (1979) established the concept of technical control as
a vehicle in line with simple and bureaucratic control. Technical control mani-
fests most clearly in the assembly line, which binds workers to the rules embed-
ded in the technology’s function (Burawoy, 1983). Other technologies facilitate
control, even if the technology is not itself the controlling factor. Stopwatches
let managers time workers to set standards for control, scales weighed output
to control workers through piece-rate pay, and punch card systems provided for
greater of control of employees’ time (Ajunwa, Crawford, & Schultz, 2017; Ball,
2010). Current technologies appear to rely more on consent. Health monitoring
through wellness programs, for example, relies on the rewards promised to work-
ers in meeting personal goals and lowering insurance premiums. But coercion still
abounds (Ajunwa et al., 2017).
The sum of any combination of approaches to employee control through
technology is a greater power to employers. Direct control over workers can be
obtained through deskilling, while through consent, “freedom” is granted to
workers who are already bound to the interests of their superiors such that the
“freedom” may be used to advance those interests (Burawoy, 1983). Focusing on
direct control, Richardson (1996) explained that technology at work cannot be
thought of only in terms of its features or as some separate force, but that it must
instead be characterized socially to accurately reect the “transfer of power” it
In discussing the decline of labor unions, Richardson (1996) illustrated how
the power transfer becomes amplied. Technology eliminates or deskills the
trades that unionized and allows jobs to relocate or be conducted remotely to
avoid unions entirely. Workers lose their power individually and collectively
through technology’s control. The disruption to the mutual reliance that once
gave workers power with managers occurs at both levels, with once-skilled work-
ers able to be replaced by anyone (or no one at all), and once-unionized jobs able
to be performed anywhere.
Braverman’s (1998) seminal analysis of technology, Taylorist management,
and the labor process showed how managerial control is exerted through the
analysis and restructuring of the division of labor. This restructuring often car-
ries the threat of deskilling work tasks and removing labor’s monopoly of control
over the integrated process; “the brain moves up the chain” separating concep-
tion from execution. Key here are questions of skill and authority, and how the
workspace – its tools, its timing – is constructed to enhance or detract from either.
Platforms, as data intermediaries, can be used to automate these processes, creat-
ing new digital spaces to capture and analyze the data traces of employee actions,
making them legible, fungible, and replicable (Aneesh, 2009).
Monitoring technologies work differently but have a similar effect on the
power relationship between employers and workers. Despite challenging the cri-
tiques of workplace technology that present it as disempowering for workers,
Downloaded by University Library At 02:20 18 June 2019 (PT)
Platforms at Work 67
MacLarkey (1997) conceded that phone line tapping, video monitoring, keystroke
logging, document counting, and email reading had already become common-
place. Drug testing, GPS tracking, RFID tags, genetic testing, social media moni-
toring, random screenshots, “smart” badges, and various sensors in the workplace
advance monitoring, sometimes in ways unknown to the worker (Ajunwa et al.,
2017; Ball, 2010). Managerial control expands rst through work intensication,
as monitoring communicates that workers must be more productive, more ef-
cient, and more sensitive to employer goals (Chesley, 2014). Then, surveillance
creates a culture around what it communicates, giving norms, as a proxy for
management, power over workers (Ball, 2010). Such norm-backed control, cre-
ated through the use of surveillance systems, reveal the thin line between coer-
cion and consent (Ajunwa et al., 2016; Burawoy, 1983; Burawoy & Wright, 1990).
Ball (2010) reported reduced creativity under surveillance, and while not asserting
that as an intended effect as much as negative consequence, work could easily be
deskilled by purposefully limiting creativity through comprehensive monitoring.
While Burawoy (1983) suggested that bureaucratic rules replaced technical
control in popularity after Second World War, a more recent history of technol-
ogy through the sociology of work reveals that the bureaucratic rules are now
embodied in workplace technologies, furthering the power swing. For example,
the rise of electronic recruitment, as described by Ajunwa et al. (2017) and Ball
(2010), takes inexible rules once left to human application and brings them into
an even more rigid, digital system of control. Likewise, the analytics that convert
data into insights function on algorithms that are little more than a mystied
version of the rules once designed to remove judgment in the name of standard
procedure. The level of control, however, is increased, as the rules and their appli-
cation become more opaque and embedded through layers of machine learning
(Fourcade & Healy, 2017). However, humans still desire to match with an algo-
rithm because of both the excitement of feeling understood and the disappoint-
ment of being left out (Fourcade & Healy, 2017). By motivating workers to tailor
their behavior and attitude to the algorithm, rules control at an even deeper level
than before.
Meanwhile, surveillance no longer only records facts, but makes determina-
tions based on the rules, taking power from higher status workers (Fourcade &
Healy, 2017). These determinations ow from measurement to rankings, which
then turn into classes or categories, restarting the cycle. In this way, additional
control is gained via workplace technology: it re-categorizes worker status, on
behalf of management. New data, new forms of monitoring, new classications
and rules, and new ways of communicating them all combine to reduce worker
Within this history of technology tilting power out of workers’ favor, plat-
forms appear as the next logical step for research in the sociology of work.
Srnicek’s (2016) Platform Capitalism argues that the explosion of venture-
funded platforms in the early twenty-rst century would have been impossible
without the stagnant real wages, insecure employment relations, and asset-price
Keynesianism inherited from the late twentieth century (pp. 34–35). While he
does not enter into dialog with her ideas, Srnicek’s work can, from our position,
Downloaded by University Library At 02:20 18 June 2019 (PT)
be read as an elaboration of Beverly Silver’s (2003) immense body of research
on capital’s technological, organizational, and geographical responses to labor
militancy and falling prot rates. We cannot overlook that the rise of the work
platform is due, in part, to the technologies and political-economic conditions
preceding platforms.
Schor and Attwood-Charles (2017) lay out an empirical agenda for research
into the platform-based “sharing economy” that focuses on the extension and/
or marketization of social relations through platform sharing, labor conditions
within platforms like Uber, and the reproduction of existing inequalities through,
for example, racial discrimination against black Airbnb renters. Some of these
sites are work platforms as we understand them, others, like the home-rental ser-
vice Airbnb are better understood as “capital platforms” facilitating asset leasing
(Smith, 2016). Research into the “sharing” or “gig” economy like the above often
uncovers the unequal power relations within them, counter to the platforms’
promise of freedom and independence. What is important from our perspective
is to view management–by–platform not as an experiment on a new breed of gig
workers, but a general managerial strategy that manifests differently in restau-
rants, law rms, warehouses, etc.
Siciliano’s (2016) ethnography of audience analytics work usefully demon-
strates how platforms built as entertainment vehicles for consumers become,
through the ecosystem that emerges from the revenue model, elaborate systems
of control for workers. Design decisions in one part of the ecosystem cascade
out to organize labor in another area. He shows that work platforms, because
they are constantly updated through the cloud, are unstable means of both value
production and worker control. We share Siciliano’s belief that the sociology of
work must extend its analyses outward beyond the organization and build on it
in our analysis of automated hiring. However, we also suggest, empirically and
in our proposed research agenda, a countervailing tendency to his argument that
platforms’ continual updates and ecosystem effects may destabilize intra-organi-
zational managerial patterns. In some instances, we believe they will standardize
managerial approaches across organizations. AHP vendors offered to standard-
ize hiring practices across large chains’ retail locations, for example, and pitched
sector-specic solutions that standardized the data sources that pharmacies, for
example, drew on to vet applicants. It is an empirical question as to which ten-
dency holds in which setting.
Sharone’s (2017) in-depth interviews with users of the business networking
platform LinkedIn is, topically, quite close to our own interest in platforms
governing entrance into employment relations, rather than conduct within the
workplace. Particularly important are his insights into the social relations inher-
ent in LinkedIn’s interface, which place greater emphasis on physical appear-
ance earlier in the job search process, penalize non-standard work histories,
and force jobseekers to narrow their search to single sectors. That jobseekers
must comply to LinkedIn’s strictures is an example of platform authoritarian-
ism. This suggests a fruitful avenue for future research: exploring how platform
affordances shape the labor market signals of both hirers and jobseekers, creat-
ing new parameters for evaluation. There is a missing story, however, that may
Downloaded by University Library At 02:20 18 June 2019 (PT)
Platforms at Work 69
be elucidated by other methodologies: LinkedIn does not just design a box for
data freely submitted to them, they actively court the participation of large hir-
ers and headhunters in service of their mission to create an “economic graph”
of the global labor market. They are not just a piece of software, but an organi-
zation that shapes labor market signals to serve its ambitions. Platforms like
LinkedIn must be investigated as active participants in the labor market, rather
than neutral gateways. The organization, not just the people using its software,
must be an object of inquiry.
In the next section, we expand on these existing analyses and provide an
example for our research agenda for platforms at work with a history of AHPs,
focusing on their patterns of sectoral adoption and their role as labor market
To map the adoption and design of early AHPs, we parse what they were for
(i.e., what promises they held for enterprises in the future) and what they did
(i.e., how they re-structured existing hiring relations in the present). The two are
not necessarily equivalent. Businesses may embrace a technology for one rea-
son but use it for another. Distinguishing between future promises and existing
practices required triangulating between different archival sources (Toubiana &
Zietsma, 2017). We draw on two main methods to do so: critical discourse analy-
sis (CDA) and affordance critique.
Content analysis is a more widely used technique for textual analysis within
sociology and neighboring elds like political science, proceeding deductively
from the generation of a code-set based in categories of interests or themes
drawn from the literature, to the, often but not always, quantitative analysis of
texts of interest for the presence of the pre-established codes (e.g., Gilens, 2009;
Linneman, 2013; Saguy & Gruys, 2010). However, CDA is also used in sociologi-
cal studies interested less in frequency or distribution of textual themes and more
in the linguistic nuance of those themes and their interaction with other social
practices (e.g., Conrad, 2006; Rohlinger, 2002). For example, Barnard-Wills’
(2011) discourse analysis of UK newspapers mapped two competing frameworks
for discussing state surveillance: as appropriate counter-terror measure or inap-
propriate, Big Brother privacy invasion. Some cultural sociologists also combine
content analysis with CDA, using the former to examine the major grouping of
themes across historical periods and different types of texts, and then using the
latter to dig into a smaller subset of sources and describe how these themes play
out in the texts and how they relate to broader social context (e.g., Johnston &
Baumann, 2007; Smirnova, 2014).
We use CDA to map discussions of what AHPs are for. This is an inductive,
qualitative method that explores how different parties describe and explain their
social practices (Van Leeuwen, 2008). Thematic codes emerge from analysis,
rather than prior to it, and are progressively rened. The texts we analyze pro-
vide a story for technological adoption, tting new developments into existing
Downloaded by University Library At 02:20 18 June 2019 (PT)
ideas about how rms can and should function, adjusting those ideas as needed.
Fairclough (1992) locates the origin of CDA within a politically minded branch
of linguistics. His approach brings together sociolinguistics’ detailed textual
analysis, Foucauldian critiques of macro-sociological social practice, and micro-
sociological interpretive approaches to conversation analysis. Like Grounded
Theory (Strauss & Corbin, 1990), CDAs inductive, context-sensitive methodol-
ogy is also a critical-realist theory of social practice. CDA provides a set of tools
for approaching discourse as a constitutive element of social practice, based on
a set of assumptions of how meaning is made to matter in social life (Jørgensen
& Phillips, 2002). Chiapello and Fairclough (2002) present CDA as a critical tool
for the new sociology of capitalism; a method for an era where management
demands discursive work from employees, and where consent is elicited through
new ideologies of work as identity. In their mold, we nd CDA useful for iden-
tifying distinct managerial discourses and specifying both their inner logic and
their dialectical relationship to other organizational activities – here, the design
of AHPs by vendors and their use by hirers.
We also nd useful the related research in the sociology of work that conducts
CDAs of texts such as newspapers (Styhre, Backman, & Börjesson, 2005), indus-
try reports (Ness, 2012), and interview transcripts (Dick & Cassell, 2004) to exam-
ine the construction of workers’ identities. Critically, such studies do not assume
that such discourse creates the worker ex nihilo, but rather adopts a critical realist
approach wherein identity discourses work within the constraints and opportu-
nities afforded by organizational form. Closer to our object of study, Handley’s
CDA of UK employers’ graduate careers websites (Handley, 2017) show how
employers begin to manage workers’ expectations of the job before they are even
recruited. Like Handley, we explore management discourse that extends hierar-
chical labor relations beyond the rm and into the recruitment process, but we
move beyond her work by studying recruitment technology and its designers as
active agents in this process. We draw on prior work (e.g., Greene & Shilton, 2018)
critically analyzing mobile developer discussions about “privacy,” to understand
how discursive practices inuence technological design and vice versa.
We focus our discussion on technology and descriptions of it by examining
AHPs’ affordances. In its original usage in environmental psychology (Greeno,
1994), an affordance was a feature of the environment that offered an animal an
opportunity for or constraint on action (e.g., a cave affords shelter) that varied
depending on the characteristics of the species and their ability to perceive the
affordance. Media and communication studies have developed the idea to focus
on the relationship between design features, human perception of them, and the
relationship between the two that denes what different users can or cannot do
with a technology (Rice et al., 2017; Schrock, 2015). Sociologists and organiza-
tional scholars use affordances as a framework to study how technologies offer
specic uses to users while denying others (e.g., Hutchby, 2001; Leonardi, 2011).
MacKenzie, Marks, and Morgan (2017) analyzes interviews with older engineers
to show how the changing affordances of telecommunications infrastructure –
from analog to digital – made previous professional identities less accessible.
Siciliano (2016) presages our work on platform affordances at the organizational
Downloaded by University Library At 02:20 18 June 2019 (PT)
Platforms at Work 71
scale. His ethnography of analytics work demonstrates that “calculative cloud-
based information and communication technologies” afford different actions for
different parties: managers control the planning of analytics work, employees cre-
atively reorganize data within the cloud, and vendors adjust the form and func-
tion of the software on the y, frustrating both managers and workers. We build
from his approach, paying careful attention to the different affordances AHPs
offer vendors, hirers, and jobseekers.
Nagy and Neff’s (2015) concept of “imagined affordances” helps us focus our
CDA on the relationship between technical features and discourse about them.
They return to environmental psychology’s focus on the perception of environ-
mental features, suggesting that while affordances and constraints impose mate-
rial limits on human action, those limits are themselves the result of an ongoing,
back-and-forth interaction with the expectations of designers and users. CDA
helps us track how the expectations of designers and their clients are concretized
in design, and how users’ expectations for how the social world should work is
enabled or frustrated by design. This interactionist approach has been frequently
deployed in the study of technology, labor, and organizations, where research-
ers are interested in how discourse interacts with other organizational practices
and structures (e.g., Chouliaraki & Fairclough, 2010). For example, Greene and
Shilton (2018) analyze online discussion forums to show how mobile platforms
structure developers’ work practices and thus their operationalization of values
like “privacy.” Similarly, Liao (2015) demonstrates the dialectical relationship
between rms’ public-facing marketing discourse about augmented reality tech-
nologies and more-private design decisions.
Data Collection
Our inductive data collection and analysis began with an informal survey of cur-
rent AHPs. We conducted an informal survey of the hiring websites for top-20
largest private employers in the US as ranked in the 2017 edition of the Fortune
500, screenshotting each stage of the application process, and making note of
whether the rm required online applications (Appendix), and the features each
site had in common. These included listings of available jobs, work histories,
skills assessments, personality assessments, reference checks, background checks,
scheduling tools, and measures to verify eligibility for work and employer tax
credits. This, along with the names of current AHP vendors – available on the
hiring websites above, often in an “About” page – provided us with keywords with
which to query LexisNexis’ newspaper and trade press database, alongside more
general keywords such as “online job application,” and “internet job search.”
Historical studies of organizational, occupational, or technological change
in a single sector often draw on a single archive, particularly trade journals
(e.g., Arndt & Bigelow, 2005). AHPs, because of their nature as brokers sitting
between different constituencies, bring together multiple sets of actors with dis-
tinct interests: jobseekers, enterprises, and designers. Because of this dynamic,
and because of the need to distinguish between designers’ and users’ stories of
adoption, we identied three distinct archives of interest: popular press coverage
Downloaded by University Library At 02:20 18 June 2019 (PT)
of new technological trends in hiring and job-seeking, trade press coverage of
enterprise AHP use, and materials from AHP manufacturers themselves. The lat-
ter included marketing materials, instructional materials, patents, and technical
Our informal survey corroborated Barber’s (2006) nding that online submis-
sion of job search materials dominated the hiring practice of the largest global
rms by 2006. Working backwards from there, we were initially interested in
which types of organizations rst used online hiring, when, and how. We nar-
rowed our period of interest to 1990–2006, the period in which the Internet (rst
through bulletin board services and then the web) and thus online job search
popularized. LexisNexis delivered results related to our search times in the popu-
lar press (e.g., The New York Times and The Washington Post) as well as the
business and trade press (e.g., The Portland Business Journal and Chain Leader).
Results that mentioned AHP features offhand (e.g., a general job search story in
which online submission is incidental, or a stock listing for an AHP developer)
were discarded. Names of vendors and technologies that appeared in that initial
search were used to cast a wider net, frequently through the Internet Archive, to
nd online copies of marketing, instructional, and support materials that AHP
vendors used to solicit or train clients.
Of particular interest was Unicru – founded as Decision Point systems in 1987
and purchased by workforce analytics rm Kronos in 2006. Inductive analysis
revealed the difculty of identifying “rsts” in online hiring, because different
features were on offer at different times in different places. Instead what became
clear to us, and ultimately the focus of this chapter, was that Unicru was the
rst rm to bring these different features together into one product that could be
adjusted by different clients to t their needs. That is, they became market leaders
because they platformized hiring. Unicru’s dominance of the trade and popular
press coverage of automated hiring in our period of interest led us to focus fur-
ther on their design and marketing materials.
Data Analysis
We collected and examined 135 archival texts in total, combining popular, trade,
and corporate materials. We used them to build a timeline of AHP feature devel-
opment, and to conduct a CDA of AHP rollout and reception. The timeline
would go on to inform the high-level themes that emerged from CDA, helping
us understand the relationship between discourse about what AHPs were for and
their implementation on the ground.
Building a timeline of AHP feature development was relatively straightfor-
ward. AHPs did not emerge fully formed. Features (e.g., hyperlinked job openings
and automated background checks) emerged in ts and starts before we saw fully
formed online job applications in 2000 when Unicru ported their HirePro kiosk
software to standalone websites built along client specications. Features identi-
ed in one archival account would need to be corroborated by at least two other
accounts, preferably from different publications, before we could condentially
assert that a feature emerged in a specic year.
Downloaded by University Library At 02:20 18 June 2019 (PT)
Platforms at Work 73
The CDA of AHP rollout and reception required a closer, systemic reading
of the claims that jobseekers, technologists, and enterprise clients were making
about the design and use of hiring software. This method allowed us to explore
the spirit of AHPs: the emergent story about their role in the hiring process and
how they changed employers’ and applicants’ orientation to the labor market.
Jobseekers were generally interviewed in the popular press, where technologists
and enterprise clients (usually but not always executives who made procurement
decisions) were generally interviewed in the trade press.
Following data collection, each author did a rst pass on the entire collection
to classify each text as either popular press article, trade literature, or industry
materials. The three groups were then split in two for each author’s analysis, with
a code set developed collaboratively as texts were analyzed. Each author con-
tributed to a shared feature timeline, and conducted CDA on their own materi-
als. Importantly, coding the texts had to distinguish between who was making
claims about AHP designs and functions, in order to compare sales pitches with
more grounded journalism, and to compare the sorts of claims that could end
up in design (e.g., client requests for faster resume processing) versus those that
could not (e.g., applicant fears that no one would ever read their materials; an
important social fact about implementation but not a sentiment with the power
to inuence design). The code set for archival materials included: year; state;
archival category; economic sector; features described; AHP vendor mentioned,
if a specic piece of software was mentioned; whether the party making a claim
about AHP designs and functions was an applicant, AHP vendor, or client
(and among clients, an executive, hiring manager, or human resources staffer);
whether a text described current or potential features; and, eventually, a set of
inductively generated themes addressing AHP drawbacks and benets, for either
applicants or hirers. The latter forms the basis of the Triangulating Claimed
Benets section. Each author spot-checked pieces of the other’s collection to
corroborate their annotations and the emergent themes drawn from those anno-
In this section, we track AHP affordances as they rst emerged in the 1990s
and 2000s, the political-economic context for their adoption, the growth of the
businesses selling them, and reactions to them from hirers, applicants, and the
press. This story about the progressive platformization of hiring is built from our
empirical analysis of designers’ and adopters’ descriptions of these tools and
their integration into labor–management relations. The main protagonist here is
Unicru, the rst and largest AHP vendor for the hourly workforce, who beat com-
petitors to market, drove adoption by large employers, and developed the core
platform features that we see in wider adoption today. We highlight throughout
the affordances for automated hiring that emerged over time and, with an eye to
the next section, begin to abstract them from the specic domain of hiring to the
broader problem of management by platform.
Downloaded by University Library At 02:20 18 June 2019 (PT)
Job Boards and Experimental Software: 1990–1993
The rst major feature-set in the automated hiring space, appearing in the early
1990s, was job boards: simple postings of job openings, focused on salaried pro-
fessionals, the people who had the skills and means to get online at the time. There
were both local (e.g., LA Online) and national (e.g., ECHO) variants, as well
as boards hosted on larger, pre-Web Bulletin Board Services like CompuServe
and Prodigy (Bucy, 1991), all with the same fee-for-placement model as classi-
ed ads but with an added promise of disintermediation. Unlike a newspaper,
hyperlinked job boards enabled direct submission to employers. Elsewhere, The
Online Career Center charged applicants $6 to input a resume to their closed sys-
tem. They told employers this system would sift through the unemployed masses
generated by the early 1990s recession. Spinnaker Software Corp. and Data-Tech
Distributors, Inc. sold CD-ROMs to jobseekers, preloaded with databases of
employers and software to edit applicant materials and pre-ll forms for specic
sectors, like the federal government (Matas, 1993). Here we see the beginning of
structured data elds optimized for capture and portability within organizations.
Compared to print classied ads that prompted phone calls or letters, hyperlinked
job boards allowed for faster submission and capture of applicant materials. They
also begin, through companies like Spinnaker, to move beyond shared cultural
norms for resume structure and instead directly shape applicant data to the needs
of the hirer – incipient platform authoritarianism.
The New Classied Sections: 1993–1996
With the opening of the web in the mid-1990s, start-ups began to do more than
just digitize classied ads, two in particular: (founded in 1994 as
The Monster Board) and CareerBuilder (founded in 1995 as NetStart, renamed
in 1996). Monster did not just post listings, they also allowed jobseekers to post
résumés and charged employers for access to this database, which they could
then search at will (Ceron, 2000). CareerBuilder went even further. For $2,000
per month, companies could post listings to CareerBuilder’s database and review
jobseekers’ posted résumés. For a at fee of $5,000, clients could purchase
TeamBuilder software that allowed “non-technical” HR departments to design
job listings that integrated seamlessly into their own website, and then create
databases, scoring systems, and automatic forwarding trees (e.g., to specic hir-
ing managers for specic listings) that helped store and sort resumes received
(Chandrasekaran, 1996). CareerBuilder told potential investors this would cut
hiring costs in half, expanding clients’ applicant pool and reducing the time from
rst contact to offer (Selz, 1998). Humans still reviewed every step of this pro-
cess, but their work was sped up, reorganized, and networked. And because résu-
més would be surfaced within databases through keyword searches, jobseekers
had to ensure theirs was cleanly typed – and thus legible to optical character
recognition – and so began to draw from an ever-changing set of industry key-
words (Oram, 1997).
Structured data elds that shaped applicant data and directed it through the
organization proliferated at this stage, but other affordances emerge too. The path
Downloaded by University Library At 02:20 18 June 2019 (PT)
Platforms at Work 75
to platformization begins with projects like CareerBuilders’: off-the-shelf soft-
ware that can be customized by the client to t their organizational structure and
hiring needs. This is not the cloud-based platform of later years, where employers
select from a vendor’s menu and both can adjust the interface and analytics on the
y, but it begins the process. Further, the scoring systems on offer digitize paper-
based version of the same. They do not yet automate assessment of applicants’
qualities, but do sort and rank human hirers’ assessments. This is thus an increas-
ing formalization of the information asymmetry between hirers and jobseekers,
an affordance that the platformization trend will only further develop.
Decision Point Systems and the Kiosk: 1997–2000
Up to this point, online job applications and software for automatically process-
ing them had been focused on salaried, professional jobs – particularly in the
technology sector. But then, as now, hourly workers make up the plurality of the
US workforce. And turnover is higher in lower-wage retail or food service sectors
than it is in other, higher-wage sectors. This presented a large, untapped market
for interested rms, as well as distinct technological challenges. Automating hir-
ing in retail meant, compared to early ventures in professional services, process-
ing a much larger number of people for a given client, doing so quickly enough to
accommodate seasonal hiring spikes, tting them to more standardized positions,
and assessing them for sector-specic qualities (e.g., availability for night shifts
and propensity to shoplift).
Retail security rm Decision Point Systems recognized this opportunity in
1997. A major client had asked for assistance in evaluating applicants’ risk for
theft. Decision Point retooled their Multipoint document-scanning software as
HirePro, an AHP accessed through kiosks installed in stores. HirePro combined
a network of existing background check systems, into which applications would
be automatically fed, with standard work history questions and personality ques-
tionnaires meant to “pinpoint personality traits and characteristics desirable for
frontline retail jobs such as sales clerks and cashiers” (Rafter, 2005). Board mem-
ber Brian Ascher said, “No one is doing a full-service offering ….There are little
dot-coms that want to help you attract candidates, but they don’t help you evalu-
ate them” (Woodward, 2000). This the rst real example of platformized hiring:
not just automated but customized to the needs of individual enterprise clients,
the software acting as a broker between applicant and hirer, its standards and
lters adjusted remotely by the vendor.
Kiosks helped reach working-class jobseekers less likely to have home Internet
access (McConnaughey, Lader, Chin, & Everette, 1998). A large department store
would have two HirePro kiosks on the sales oor for applicant input and one in
the manager’s ofce where they would receive the system’s output. They looked
like large telephones, appearance customized to t client branding, with a small
screen and keyboard. Decision Point charged a small base fee for set up and a
larger subscription fee based on clients’ access to specic assessments, screen-
ing networks, and Decision Point’s own analyses of employee data (Brenneman,
2000a; Rafter, 2005). The latter, according to Decision Point’s home page in 2000,
Downloaded by University Library At 02:20 18 June 2019 (PT)
helped clients “easily track turnover, applicant volume, and increases in sales
…mak[ing] each location accountable for maintaining consistent hiring prac-
tices.” The dispersed hiring practices of large chains were thus centralized by the
platform: besides formalizing a set of procedures and goals across hiring units,
HirePro would also route hires to locations besides the one to which they applied
(Gilbertson, 1999).
Early AHPs requested work history, references, a personality question-
naire, and identifying information that would both verify eligibility for certain
employer hiring tax credits and be fed into a subset of the 40 independent ser-
vice providers with whom Decision Point contracted for sector-specic back-
ground checks. Everything was customizable. Good Guys electronics focused
on drug and alcohol use in their assessment, whereas Target asked for reactions
to employee theft or opinions on how many Americans cheated on their taxes
(Richtel, 2000). Between 10 and 20 minutes after completion, the hiring man-
ager at the location where an applicant applied would receive a three-page fax
or email summarizing work history and other data before giving a color-coded
rating of the applicant:
The applications are labeled red zone (don’t hire) and green zone (hire immediately). Yellow
zone issues warnings, such as an applicant does not follow rules, may not be honest and could
be argumentative with customers. (Brenneman, 2000a)
The HirePro report for hiring managers – called Proler – suggested follow-
up interview questions and areas of concern, based on jobseekers’ responses to
specic personality items or perhaps gaps in their work history.
The specic hiring needs of large retailers thus prompted the design of AHPs
able to quickly assess applicants, speed-up and deskill the interview process, and
centralize the assessment criteria and routing of applicants across retail locations.
Affordances that emerged at earlier stages proliferate, while new ones appear. The
sort of structured data elds that emerged at earlier stages of development were
deployed across the application. Data elds soliciting applicants’ work histories
or their attitudes toward coworkers’ behaviors could be adjusted based on the
increasingly legible data of both employees within the organization (e.g., sales,
schedules) and applicants outside it (e.g., a growing array of background checks).
This widens an information asymmetry between hirers and applicants – the latter
not only do not know the “right” answers in personality assessments but cannot
see the network of background checks HirePro queries. All this is made possi-
ble by software that links individual kiosks across the client’s locations, hosting
an ecosystem of single-purpose applications (e.g., personality assessments and
scheduling tools) that could be adjusted to client specications, and connecting
and standardizing hiring practices across locations.
Automated Hiring Becomes a Web Platform: 2000–2006
In 2000, Decision Point rebranded as Unicru “to forge its new identity as the
one-stop site for job recruiting in malls, stores and online” (Brenneman, 2000b).
Unicru already had kiosks in 4,000 stores in the US, around 12,000 individual
Downloaded by University Library At 02:20 18 June 2019 (PT)
Platforms at Work 77
units, but the name signaled a grander vision for the company: Moving the entry
point for automated hiring from kiosks to the web, making it a more iterative
process that constantly analyzed new data from external databases and inter-
nal workforce analytics. This shift prioritized interoperability between Unicru’s
AHP and other systems that could measure clients’ current employees to inform
future hiring. This is what made HirePro a platform. This software-as-a-service
model meant the AHP was more open to regular customization than a pre-
loaded kiosk. It hosted other applications and constantly adjusted its features to
manage the ow of data between clients, applicants, assessments, and screening
networks. In the summer of 2003, for example, Unicru announced partnerships
with TimeManagement Corporation to integrate Unicru’s hiring analytics with
their restaurant chain scheduling software, and ChoicePoint, to implement phar-
macy-specic background checks. Most of these screening networks and analytics
tools are of course beyond the view of the applicant, increasing the information
asymmetry between them and the hirer.
By integrating themselves further into clients’ existing operations and using
more and more measures on potential employees, Unicru began to accumulate
an enormous amount of data. These data were not sold to advertisers, as in the
social media platform model. Instead, Unicru recruited experts in industrial and
personality psychology and machine learning to develop applicant assessments
based on the performance of clients’ current employees. Unicru’s chief scientist
David Scarborugh promised that “Our system allows you to clone your best, most
reliable people” (Overholt, 2002). He led the rollout for the 50-item Frontline
Reliability Assessment, a revamped personality assessment drawn from “the
actual job results of 370,000 hourly workers in industries such as retail, grocery
stores, and food service” (Frauenheim, 2006). Here we see the affordances that
emerged previously expand, providing more possible actions for the hirer across
two dimensions. The platform’s reach is more extensive; with the move online net-
working various features together more quickly and allowing them to be adjusted
and updated on the y. And the platform’s reach is more intensive; acquiring
more and more data sources to examine the applicant and their potential t with
specic organizational roles.
Competitors such as Recruitmax and Kenexa attempted to beat Unicru
into new sectors, but largely failed because Unicru already owned the largest
slice of employers – hourly retail and food service – and had used that rev-
enue to build up the data, research staff, and technology to rene their product
for use elsewhere (Rafter, 2005). In 2006, they were acquired for $177.8 million
by “human capital management” rm Kronos, a leader in workforce analyt-
ics whose business began with automated timeclocks workers could not cheat.
Kronos’ nancial statements frame the intensive and extensive reach of Unicru’s
hiring platform as the perfect complement to Kronos’ existing workforce ana-
lytics. Unicru, renamed Kronos Workforce Acquisitions, would form the basis
of their new “talent management division” that “integrates with Kronos’ work-
force management products to link sourcing, selection and hiring strategy with
actual performance and labor planning” (Kronos Incorporated, 2006, F-24).
Downloaded by University Library At 02:20 18 June 2019 (PT)
The goal was to offer clients one platform that included automated hiring but
much more:
This integrated solution is designed to enable companies to cast a wider net for applicants, select
in people who t well and have higher potential, select out people who t poorly or present risk,
onboard and screen quickly, measure results and validate effectiveness, thereby driving top- and
bottom-line business results by continuously improving the quality and productivity of their
workforce. (p. 6)
This imagines a seamless ow of employee data from the time they become
interested in the rm to the time they leave it, with entrance and exit channeled
through a platform that disaggregates employees into the skills and dispositions
needed for a specic location of a specic clients’ business at a specic time.
Triangulating Claimed Benets
In the previous section, we reviewed the affordances that emerged over time
as the hiring process was platformized, primarily by Unicru. We believe these
affordances generalize to other work platforms in domains besides hiring, and
explore that further in the Conclusion. In this section, we draw further on our
CDA to examine how these technical affordances appealed to hirers in the 1990s
and 2000s – what benets AHPs brought to personnel selection. In the process,
we generalize from specic features and adoption claims to bigger trends in how
AHPs broker the relationship between management and applicant. We identi-
ed four primary benets, claimed by both AHP designers like Unicru and their
clients: loss prevention, reduced bias in hiring, reduced time spent hiring, and
increased retention rates. The master theme is of the reduction of friction in
the labor market, by fragmenting workers into discrete skills and dispositions.
By reading the claims of AHP designers against descriptions of design and
use, we pinpointed gaps between promises about designs and their implementa-
tion, showing how discourse about the potential power of AHPs relates to their
on-the-ground design and use.
Loss Prevention
Reducing theft of stock is a major concern for retailers. AHP vendors promised
that more precise employee selection would assist loss prevention in two ways.
First, vendors promised to reduce theft by eliminating from the applicant pool
anyone convicted of shoplifting or similar crimes, or who had previously been
red from a similar job for theft, even if they were not charged. The former
would draw from arrest and conviction records, while the latter would draw from
employee records shared among industry partners. For example, Unicru’s part-
nership with data broker ChoicePoint included access to four distinct databases
potentially available to pharmacy clients: Esteem®, the ChoicePoint National
Criminal File, the Health and Human Services List of Excluded Individuals, and
County Criminal Check (Unicru, 2003). Second, automated psychological assess-
ments promised to weed out applicants with thieving tendencies, even if they had
never stolen or been accused by an employer of doing so. For example, Unicru
Downloaded by University Library At 02:20 18 June 2019 (PT)
Platforms at Work 79
built a “dependability assessment” for Universal Studios Hollywood that would
“rule out individuals inclined to steal or skip work” (Rafter, 2005).
This benet emerges from multiple interacting affordances. Increased legibility
of employee behavior as data allows Unicru to model “dependability” and test for
it in applicants, and it allows data brokers such as ChoicePoint to sell to Unicru
lists of names that should be excluded. This massive information asymmetry
appeals to hirers: they have a network of resources to dene “t” in applicants
who is not reciprocated in the other direction. Finally, this system is customizable
for different clients or sectors.
Reduced Hiring Bias
AHP designers and clients also claim that automation reduces hirer bias, replac-
ing messy human decisions with a neutral technical process, for example, “From a
diversity perspective, articial intelligence can be very benecial because it’s blind
to things like color, age, sexual orientation” (Meredith, 2001).
But, algorithmic specication of “t” can itself become a vehicle for bias.
Unicru’s instruction manuals for clients makes clear that the offer to “clone
your best people” (Overholt, 2002) begins with the identication of existing
high-sales employees within client records. The process is certainly more data-
driven than most human hiring and provides a consistent, empirically driven
standard arising out of a rm’s operations – an attractive prospect for any hirer.
But both computer science and anti-discrimination law scholarship identies a
“garbage in, garbage out” problem (Barocas & Selbst, 2016) wherein systemi-
cally biased data sources produce systemically biased analyses, regardless of the
quality of those analyses. An extreme example: If stores were not hiring women
before they began using Unicru, or only giving female employees low-trafc
sales shifts, then that training data would lead Unicru’s model to identify more
men as ideal candidates. The prospect of “cloning your best people” comes from
both the increased legibility of work activity as data – used to inform applicant
assessments – and the centralization and standardization of hiring criteria –
local managers have their actions constrained by a platform implemented at the
corporate level.
Reduced Hiring Time
Those same affordances, plus the proliferation of structured, applicant-side data
elds that capture specic data of interest to hirers (e.g., skills assessments and
scheduling availability), promise clients a reduction in time spent hiring. What
Unicru’s training brochures called “faster associate capture” occurs through four
principle means: (1) applicants enter data that would have otherwise been entered
by hirers, (2) the rst round of vetting is fully automated, occurring within
minutes, (3) subsequent rounds of vetting are de-skilled, with interview guides
produced for hiring managers, and (4) applicants are routed to branches where
they’re most needed, instead of having paper applications led away where they
applied. This features strongly in client testimonials in high-turnover sectors like
retail and food service (e.g., Overholt, 2002).
Downloaded by University Library At 02:20 18 June 2019 (PT)
This reduction in time spent hiring points to a key contradiction: Unicru was
never fully automating hiring, despite their promises to the contrary. They were
automating rejection by culling the bottom 20% or so of applicants – the red
lights – from the pool before passing them onto hiring managers.2 Human manag-
ers remain in the loop but lose some of their local discretion. Variables of inter-
est are determined by corporate headquarters who purchase and implement the
AHP. The direction of interviews is shaped by how the platform solicits, records,
and presents applicant data, and then by its interview guides.
Increased Retention
Unicru commonly claimed – referring to their own private data – that they
reduced clients’ turnover rates 20–30% in the rst year of adoption by highlight-
ing for clients the best ts for open positions – those more likely to thrive in
their duties and less likely to quit. Rock Bottom Restaurants’ said their $400,000
investment in Unicru’s AHP delivered a 21% reduction in turnover from 2002
to 2003, with one general manager saying, “There is no unnecessary interview-
ing” (Crecca, 2004). The promise is one of transparency and predictability. The
AHP is supposed to help dene exactly what clients want from a hire, by analyz-
ing existing employee performance data garnered through point-of-sale systems,
performance reviews, etc. There are then, ideally, no surprises. The right types are
tted to the right tasks.
Relatedly, vendors promised to quickly process applicants in periods of high
need and, especially after the move to the web and the embrace of machine learn-
ing techniques, to analyze clients’ existing sales data to predict new needs of which
the client might not be aware. By linking Unicru’s hiring analytics with Kronos’
deep trove of workforce analytics, Kronos Incorporated (2006) promised to
“integrate employee selection strategy with actual labor performance, link labor
planning with sourcing and hiring” (p. 8). This is a vision for on-demand labor
that precedes the app-based gig economy by several years and vastly exceeds it in
scope, given the wide application of Kronos’ software across sectors. That vision
offered to corporate clients is grounded in concrete affordances: the increased
legibility of employee data that informs hiring analytics, the structured data elds
that measure applicant “t,” an ecosystem design that affords different measure-
ments of t for different sectors, and the increased information asymmetry that
intensively measures whether and where applicants might t – a viewpoint that is
of course not reciprocated.
Reducing Friction, Increasing Fragmentation
The four major claimed benets of AHPs – loss prevention, reduced bias in
hiring, reduced time spent hiring, and increased retention rates – can be
summarized as a reduction of friction in the labor market and the increased
fragmentation of jobseekers into discrete bundles of skills and dispositions, that
is, “human capital” (Adamson, 2009). Each of these benets depends on the
other, and they rely on different combinations of the affordances reviewed in
Downloaded by University Library At 02:20 18 June 2019 (PT)
Platforms at Work 81
the previous section. The platformization of hiring, embodied by Unicru and
its integration with Kronos, is the story of an emergent broker in the hiring
ecosystem, building an infrastructure for sorting out employee qualities and
directing them to hirers in the desired quantity. Increasing fragmentation to
reduce friction shifts the object of interest from the whole worker to specic
attitudes, behaviors, and capacities. Unicru’s Fast Company prole captures this
well: Blockbuster employee photos were overlaid with white text reading, for
example, “T or F: I pay close attention when people talk to me.” The jobseeker’s
self-presentation is restricted, dictated, and reprocessed in ways that are
unknown and non-negotiable to the jobseeker.
It is this breaking down of jobseekers into fungible parts that allows for the
smooth transition of those jobseekers from applying to interviewing to scheduling
to managing, because once the data prole has been built, it can be moved along
and scrutinized in different ways at different stages. What is reduced is not just the
time spent at any given stage but barriers to interoperability between stages. To
increase fragmentation and reduce friction, AHPs offer an enormous amount of
freedom to hirers. Affordances are customizable. The content of specic assess-
ments or background checks are adjusted either by request or in response to new
trends in employee and sales data. The extensive reach of the platform grows
and its insight into the applicant’s qualities intensies. But this same freedom
for the hirer is not offered to the applicant, who must play catch-up with each
new technical feature of the hiring process and who cannot even see many of the
measures being deployed against them. The jobseeker is coerced into this opaque
and non-negotiable process of AHPs, since workers must attain employment to
afford basic needs such as food and housing; thereby subjecting the jobseeker to
power disparities between themselves and the hirer that are created by informa-
tion asymmetries embedded in AHPs. This is platform authoritarianism (Ajunwa,
2018) and to conclude we suggest alternative domains in which to study it, and
possible methods to do so.
AHPs are intermediaries. They restructure the ow of information within the
labor market so that employers can select precisely the quantity and quality of
employees they need exactly when and where they are needed. Or, more accu-
rately, to reject applicants who do not t those needs. Jobseekers’ self-presentation
is limited, channeled, and re-processed by this new broker for labor market
signals. Platforms are distinct from other, previous workplace information tech-
nologies in their role as intermediaries implemented in similar fashion across
different organizations and workplaces and in their constant, cloud-based con-
nections to their designers – affording persistent adjustment and data analysis.
Workplace platforms are distinct from social media platforms, the predomi-
nant object of study in platform studies, because of the coercive nature of their
data ows and the high consequences that accompany their use in this domain.
Downloaded by University Library At 02:20 18 June 2019 (PT)
Our analysis of AHPs suggests ve core affordances which work platforms offer
(1) structured data elds optimized for capture and portability within the
(2) increased legibility of activity qua data collected inside and outside the
(3) information asymmetry between labor and management inside and outside
the organization;
(4) an “ecosystem” design supporting the development of limited-use, domain-
specic applications; and
(5) the standardization of managerial techniques across workplaces within large
organizations as well as between organizations.
Some work platforms will emphasize some of these over others. In the context
of hiring, these affordances bore four core benets to employers: loss prevention,
reduced hiring bias, reduced time spent hiring, and increased retention. These
come together in an overarching sociotechnical tendency to approach workers
as human capital, able to be ported smoothly between tasks and locations. These
affordances may bear different benets when applied to different workplace func-
tions (e.g., hiring and scheduling) or different parts of a rm (e.g., public relations
and human resources).
These ve core affordances that are offered to employers via work platforms
are important avenues for future research because they differ qualitatively from
the focus of the previous investigation that has dominated platform research and
methodology. Historically, platform studies and methods of investigation have
been largely conducted by communications and media studies focused on social
media. However, the design of social media is grounded in public, non-coerced
social relations, while work platforms are designed around hierarchy and prot
maximization. Thus, the methods of study that have been previously utilized
by communications and media studies to investigate social media platforms are
not adequate to study work platforms given the divergence in design and coer-
cion. The emergence of work platforms, these new technologies of control, are
grounds for future research in the sociology of work. To suggest future avenues
of research, we match different, contemporary work platform vendors to these
functions and rm locations (Table 1).
The primary benet of approaching these technologies from the sociology of
work is to examine how social conicts that dictate the terms of entry into, life
within, or exit from the workplace or are built into, diffused by, or decided by
platforms. Where Srnicek (2016) focuses on revenue models, our initial typology
of managerial affordances focuses on the technological content of labor-capital
conicts within and around the workplace. Power contests which might otherwise
be up for debate are rendered settled matters by technology and its constraining
functions, with the obscured reality that the power has been tilted in one direction –
in this case, in the direction of the employer.
Downloaded by University Library At 02:20 18 June 2019 (PT)
Platforms at Work 83
The stakes are even higher here than for other workplace technologies, whose
capabilities might not change much after purchase. Given platforms’ software
as a service business model, which implies cloud connections, constant updates,
and a broad swathe of enterprise customers, that power imbalance can grow over
time as more employee data are absorbed, more rms and sectors are brought
into the platforms’ reach, and new analyses of employees and rms emerge from
the vendor. As a broker between current or future employees and employers,
work platforms would seem, at rst glance, to be mere information intermediar-
ies facilitating freely adopted employment relations. But there is of course only
one paying customer for a Unicru or a Kronos and it isn’t the employee. The
information asymmetry work platforms grow between employers and employees
is part of the pitch. Platforms’ appeal to employers is their ability to disaggregate
workers into skills and dispositions and then sort and direct that human capital
on demand. They take workers apart and put them back together again, with the
precise recipe updated as needed.
In this chapter, we have formulated a research agenda that interrogates the role
of an emerging technology, platforms, in the workplace. How might the sociology
of work develop and pursue that research agenda?
Future Research Directions
Importantly, our CDA of automated hiring does not get at the use of these tech-
nologies on the ground and how designs may be frustrated or transformed by
different organizational contexts or worker adaptations. It is only a study of a
single platform type, its adoption, and early design. Other work platforms may
reveal different managerial affordances and different outcomes owed to different
sorts of workforces, regulatory environments, and labor processes. The study of
work platforms such as AHPs could offer greater insight into core sociological
concepts such as “social capital” (e.g., Fernandez, Castilla, & Moore, 2000). Two
future empirical directions are important to this topic.
First, interviewing human resource professionals about their use of AHPs
(e.g., McDonald, Damarin, Lawhorne, & Wilcox, 2019), or participation-observation
Table 1. Contemporary Work Platform Vendors, Their Functions
and Firm Locations.
Business Activities HR Functions Managerial Functions Publicity Functions
Recruitment LinkedIn LinkedIn
Hiring/onboarding Sapling, SnagAJob
Time-keeping Kronos
Surveillance Veriato 360, SpectorSoft,
Scheduling Kronos Sling
Internal communications Slack, Workplace, Facebook
External communications SalesForce Twitter, Facebook
Evaluation SalesForce, Sapience
Termination/off-boarding Integrify Kronos
Downloaded by University Library At 02:20 18 June 2019 (PT)
that monitored that use (e.g., Anteby & Beckhy, 2016), would lend greater clar-
ity to the distinction we draw between automated rejection and automated hir-
ing and the broader impact of computerized assistance on hiring decisions.
Do hirers understand AHPs as empirical prompts to balance human biases? As
rhetorical cover for pre-rendered decisions? As trusted seers who override human
decisions? AHP designers clearly advertise their technology as the third, and occa-
sionally the rst, but it is ultimately an empirical question. Researchers could also
explore the operation of these technologies through a digital ethnography
(e.g., Ziewitz, 2016).
Second, studying applicants’ collective efforts to cheat these automated inter-
mediaries would help us understand how new information asymmetries affect
jobseekers’ self-presentation within the labor market; building from the sociol-
ogy of work’s long history of studying worker resistance (e.g., Anteby, 2016). We
are currently engaged in preliminary research in this vein, conducting discourse
analyses of forums wherein jobseekers conduct amateur audits of AHPs, sharing
responses to and results from personality questionnaires and background checks.
These collective efforts to uncover chain- or vendor-specic automated hiring log-
ics are, in many ways, ahead of academic researchers – although platform audit
studies are an emergent methodology (e.g., Hannák et al., 2017), inspired by past
sociological audit studies of, for example, housing discrimination.
Two other domains appear ripe for inquiry: automated scheduling and cus-
tomer relationship management (CRM). Automated scheduling platforms are
especially prevalent in retail and food service. They integrate with sales data to
predict stafng needs over the course of a year, a week, or a day. Individual shifts
are then broken up into increments as small as single hours. Employee schedules
become extremely unpredictable from week to week, with some rms expect-
ing employees to remain on call in periods of uncertain demand (Greenhouse,
2012; Scheiber, 2015). Kronos’ ability to integrate scheduling analytics with hir-
ing data, sales data, and a cross-location view of a rm’s operations makes it
a major player here. As with Unicru’s original hiring platform, this is not full
automation, but an expanded and systematized view of the process offered to
local management, who “are often compensated based on the efciency of their
stafng” (Kantor, 2014).
Worker surveys could classify scheduling patterns by sector, and their costs
(e.g., missed parenting time and longer commutes) by worker demographics.
A digital ethnography of the software itself, coupled either with participant-
observation of workplaces governed by these systems or in-depth interviews with
affected workers, could explore how these technological practices affect job per-
formance and work culture. Ideally, these are comparative projects. It may be
that software developed for certain large employers (e.g., Walmart) or certain sec-
tors (e.g., food service) then spreads to organizations where the scheduling needs
are fundamentally different, because they are all served by the same vendor or
because they admire the success of early adopters.
CRM platforms systematize sales teams’ relationship with current and poten-
tial clients, recording data to generate new insights but also restructuring the
contact process so that incidents (e.g., shortages and delinquent accounts) are
Downloaded by University Library At 02:20 18 June 2019 (PT)
Platforms at Work 85
either responded to automatically or routed to relevant staff as they happen.
Cloud-based SalesForce is the leader in this sector (Hardy, 2013; Rivlin, 2007).
Their software creates a unied view of individual client cases across a rm.
Customers gain an automated help desk to query the rm for specic orders or
questions. Managers can see every interaction recorded within the environment.
CRM platforms surface trends in employee and customer behavior for clients
and update the software on the y. Smaller app developers offer new features
that plug into the ecosystem. RingDNA, for example, integrates telephony into
SalesForce, allowing managers to record calls and oversee their staff’s customer
Single-site ethnographies like Siciliano’s (2016) are one route of investigation
here. Salesforce is certainly a cloud-based, calculative information technology
that facilitates a deep managerial view into sales staff’s practices and knowledge.
Whether staff fully consent to this deep managerial view is an empirical question.
Recording more of their customer interactions may give up more of their hard-
won knowledge, making them more expendable. A major challenge for research-
ing CRM software, however, is capturing its inter-organizational effects. A fuller
qualitative investigation would need to compare implementation and usage
between organizations, as in Barley’s (1986) study of hospital CT scanners, and
the effect on workplace hierarchy and intra-organizational cooperation. Barley’s
CT scanners, however, could not be remotely updated, patched, and augmented
by the platform provider and subsidiary services. Different organizations may
embrace this interactivity in different ways.
Each of these domains – hiring, scheduling, CRM – offers its own challenges
of study. Each domain is transformed by the platforms’ role as data interme-
diaries, making more worker activity legible as data and widening information
asymmetries between employers and employees. The sociology of work can, by
drawing on its traditional methodological strengths and embracing new concep-
tual avenues, map this technological terrain and the social relations within it. This
is a crucial step in creating more humane platforms and workplaces.
*. The authors contributed equally to this chapter.
1. See also, Section 230 of the Community Decency Act.
2. Many thanks to Miranda Bogen of Upturn for surfacing this insight during a work-
shop at Data & Society.
Adamson, M. (2009). The human capital strategy. Ephemera, 9(4), 271–284.
Ajunwa, I. (2018). Facebook users aren’t the reason Facebook is in trouble now. The Washington
Post, March 23. Retrieved from
Ajunwa, I., Crawford, K., & Ford, J. S. (2016). Health and big data: An ethical framework for health
information collection by corporate wellness programs. The Journal of Law, Medicine & Ethics,
44(3), 474–480.
Downloaded by University Library At 02:20 18 June 2019 (PT)
Ajunwa, I., Crawford, K., & Schultz, J. (2017). Limitless worker surveillance. California Law Review,
105, 735–776.
Anderson, E. (2017). Private government: How employers rule our lives (and why we don’t talk about it).
Princeton, NJ: Princeton University Press.
Aneesh, A. (2009). Global labor: Algocratic modes of organization. Sociological Theory, 27(4), 347–370.
Anteby, M. (2016). Denials, obstructions, and silences: Lessons from eld resistance (and eld embrace).
In K. D. Elsbach & M. R. Kramer (Eds.), The handbook of qualitative research methods:
Innovative pathways and methods (pp. 197–205). New York, NY: Routledge.
Anteby, M., & Beckhy, B. A. (2016). How workplace ethnographies can inform the study of work and
employment relations. Industrial and Labor Relations Review, 69(2), 501–505.
Arndt, M., & Bigelow, B. (2005). Professionalizing and masculinizing a female occupation: The recon-
ceptualization of hospital administration in the early 1900s. Administrative Science Quarterly,
50(2), 233–261.
Ball, K. (2010). Workplace surveillance: An overview. Labor History, 51(1), 87–106.
Barber, L. (2006). E-recruitment developments. Brighton: Institute for Employment Studies.
Barnard-Wills, D. (2011). UK news media discourses of surveillance. The Sociological Quarterly, 52(4),
Barocas, S., & Selbst, A. D. (2016). Big data’s disparate impact. California Law Review, 104, 671–732.
Barley, S. R. (1986). Technology as an occasion for structuring: Evidence from observations of CT
scanners and the social order of radiology departments. Administrative Science Quarterly, 31,
Braverman, H. (1998). Labor and monopoly capital: The degradation of work in the twentieth century.
New York, NY: NYU Press.
Brenneman, K. (2000a). Decision Point: DeKoning had to go. Portland Business Journal, November 12.
Retrieved from
Brenneman, K. (2000b). It’s adieu Decision Point and greetings to Unicru. Portland Business Journal,
November 19. Retrieved from
Bucy, E. P. (1991). High-tech way to job hunt: Computer bulletin boards are a valuable resource. Los
Angeles Times, March 4, p. A21.
Burawoy, M. (1983). Between the labor process and the state: The changing face of factory regimes
under advanced capitalism. American Sociological Review, 45(5), 587–605.
Burawoy, M., & Wright, E. O. (1990). Coercion and consent in contested exchange. Politics & Society,
18(2), 251–266.
Ceron, G. F. (2000). Where the jobs are: Recruiting websites have moved well beyond simply being
online versions of help-wanted sections. The Wall Street Journal, October 23, p. R34.
Chandrasekaran, R. (1996). Tapping into a web of aspirations: NetStart helps rms with online job
hunts. The Washington Post, December 30, pp. 13–14.
Chesley, N. (2014). Information and communication technology use, work intensication and employee
strain and distress. Work, Employment and Society, 28(4), 589–610.
Chiapello, E., & Fairclough, N. (2002). Understanding the new management ideology: A transdiscipli-
nary contribution from critical discourse analysis and new sociology of capitalism. Discourse
and Society, 13(2), 185–208.
Chouliaraki, L., & Fairclough, N. (2010). Critical discourse analysis in organizational studies: Towards
an integrationist methodology. Journal of Management Studies, 47(6), 1213–1218.
Cohen, N. S. (2015). From pink slips to pink slime: Transforming media labor in a digital age. The
Communication Review, 18(2), 98–122.
Conrad, B. K. (2006). Neo-institutionalism, social movements, and the cultural reproduction of a
mentalité: Promise Keepers Reconstruct the Madonna/Whore Complex. The Sociological
Quarterly, 47(2), 305–331.
Crecca, D. H. (2004). Keyed in: Online applications help Rock Bottom Restaurants hire the right peo-
ple, right away. Chain Leader, February.
Dick, P., & Cassell, C. (2004). The position of policewomen: A discourse analytic study. Work,
Employment and Society, 18(1), 51–72.
Edwards, R. (1979). Contested terrain: The transformation of the workplace in the twentieth century.
New York, NY: Basic Books.
Fairclough, N. (1992). Discourse and social change (Vol. 10). Cambridge: Polity Press.
Downloaded by University Library At 02:20 18 June 2019 (PT)
Platforms at Work 87
Fernandez, R., Castilla, E., & Moore, P. (2000). Social capital at work: Networks and employment at a
Phone Center. American Journal of Sociology, 105(5), 1288–1356.
Fourcade, M., & Healy, K. (2017). Seeing like a market. Socio-Economic Review, 15(1), 9–29.
Frauenheim, E. (2006). Unicru beefs up screening tool. Workforce Magazine, March 17. Retrieved from
Gilbertson, D. (1999). Computers replacing job-application paper: Target, Home Depot use high-tech
system. Arizona Republic, September 19: D1.
Gilens, M. (2009). Why Americans hate welfare: Race, media, and the politics of antipoverty policy.
Chicago, IL: University of Chicago Press.
Gillespie, T. (2010). The politics of “platforms.” New Media and Society, 12(3), 347–364.
Greene, D., & Shilton, K. (2018). Platform privacies: Governance, collaboration, and the different mean-
ings of “privacy” in iOS and Android development. New Media and Society, 20(4), 1640–1657.
Greenhouse, S. (2012). A part-time life, as hours shrink and shift. The New York Times, October 27,
p. A1. Retrieved from
Greeno, J. G. (1994). Gibson’s affordances. Psychological Review, 101(2), 336–342.
Handley, K. (2017). Anticipatory socialization and the construction of the employable graduate:
A critical analysis of employers’ graduate careers websites. Work, Employment and Society,
32(2), 239–256. doi:0950017016686031
Hannák, A., Wagner, C., Garcia, D., Mislove, A., Strohmaier, M., & Wilson, C. (2017, February). Bias
in online freelance marketplaces: Evidence from TaskRabbit and Fiverr. In CSCW Proceedings
(pp. 1914–1933).
Hardy, Q. (2013a). Salesforce wants to remake marketing—And itself. The New York Times, April 23.
Retrieved from
Hodson, R. (1999). Organizational anomie and worker consent. Work and Occupations, 26(3), 292–323.
Hutchby, I. (2001). Technologies, texts and affordances. Sociology, 35(2), 441–456.
Johnston, J., & Baumann, S. (2007). Democracy versus distinction: A study of omnivorousness in gourmet
food writing. American Journal of Sociology, 113(1), 165–204.
Jørgensen, M. W., & Phillips, L. J. (2002). Discourse analysis as theory and method. London: Sage
Kalleberg, A. L. (1989). Linking macro and micro levels: Bringing the workers back into the sociology
of work. Social Forces, 67(3), 582–592.
Kantor, J. (2014). Working anything but 9 to 5: Scheduling technology leaves low-income parents with
hours of chaos. The New York Times, August 13. Retrieved from
Kronos Incorporated. (2006). Annual Report, 2006. Retrieved from
Leonardi, P. M. (2011). When exible routines meet exible technologies: Affordance, constraint, and
the imbrication of human and material agencies. MIS Quarterly, 35(1), 147–167.
Liao, T. (2015). Augmented or admented reality? The inuence of marketing on augmented reality
technologies. Information, Communication & Society, 18(3), 310–326.
Linneman, T. J. (2013). Gender in Jeopardy! Intonation variation on a television game show. Gender &
Society, 27(1), 82–105.
MacLarkey, R. L. (1997). Information technology and the workplace: An empirical investigation of
anti-technology theory. International Review of Modern Sociology, 27(2), 33–44.
MacKenzie, R., Marks, A., & Morgan, K. (2017). Technology, affordances and occupational identity
amongst older telecommunications engineers: From living machines to black-boxes. Sociology,
51(4), 732–748.
Marron, D. (2007). “Lending by numbers”: Credit scoring and the constitution of risk within American
consumer credit. Economy and Society, 36(1), 103–133.
Matas, A. (1993). Internet becomes an on-line opportunity for employers and job hunters. The
Washington Post, November 7, p. H2.
McConnaughey, J. W., Lader, W., Chin, R., & Everette, D. (1998). Falling through the Net II: New
data on the digital divide. Washington, DC: National Telecommunications and Information
Downloaded by University Library At 02:20 18 June 2019 (PT)
McDonald, S., Damarin, A.K., Lahore, J., & Gerber, C. (2019). Black Holes & Purple Squirrels: Human
Resource Recruitment and a Tale of Two Online Labor Markets. Research in the Sociology of
Work, 33, 93–120.
Mears, A. (2015). Working for Free in the VIP: Relational work and the production of consent.
American Sociological Review, 80(6), 1099–1122.
Meredith, J. (2001). AI identifying steady workers: E-recruiting rm offering tool to determine how
long job seeker will stay around. Chicago Tribune, July 16. Retrieved from http://articles.
Nagy, P., & Neff, G. (2015). Imagined affordance: Reconstructing a keyword for communication theory.
Social Media+ Society, 1(2),
Ness, K. (2012). Constructing masculinity in the building trades: ‘most jobs in the construction industry
can be done by women’. Gender, Work and Organization, 19(6), 654–676.
O’Connor, L. T., Kmec, J. A., & Harris, E. C. (2015). Giving care and perceiving discrimination:
The social and organizational context of family responsibilities discrimination. Research in the
Sociology of Work, 26, 249–276.
Oram, H. (1997). Technology rings the changes in recruitment and HR management. The Irish Times,
May 7, p. A6.
Overholt, A. (2002). True or false: You’re hiring the right people. Fast Company, February, pp. 110–114.
Rafter, M. (2005). Unicru breaks through in the science of smart hiring. Workforce Magazine, May 10.
Retrieved from
Rice, R. E., Evans, S. K., Pearce, K. E., Sivunen, A., Vitak, J., & Treem, J. W. (2017). Organizational
media affordances: Operationalization and associations with media use. Journal of
Communication, 67(1), 106–130.
Richardson, C. (1996). Computers don’t kill jobs, people do: Technology and power in the workplace.
The Annals of the American Academy of Political and Social Science, 544, 167–179.
Richtel, M. (2000). Online revolution’s latest twist: Job interviews with a computer. The New York Times,
February 6. Retrieved from
Rivlin, G. (2007). Software for rent. The New York Times, November 13. Retrieved from http://www.
Rohlinger, D. A. (2002). Framing the abortion debate: Organizational resources, media strategies, and
movement-countermovement dynamics. The Sociological Quarterly, 43(4), 479–507.
Saguy, A. C., & Gruys, K. (2010). Morality and health: News media constructions of overweight and
eating disorders. Social Problems, 57(2), 231–250.
Scheiber, N. (2015). Starbucks falls short after pledging better labor practices. The New York Times,
September 23, p. B1. Retrieved from
Schor, J. B., & Attwood-Charles, W. (2017). The “sharing” economy: Labor, inequality, and social
connection on for-prot platforms. Sociology Compass, 11(8), e12493.
Selz, M. (1998). Computerized employee-search rms attract investors. The Wall Street Journal.
January 13, p. B2.
Sewell, G., & Wilkinson, B. (1992). Someone to watch over me: Surveillance, discipline and the just-in-
time labour process. Sociology, 26(2), 271–289.
Sharone, O. (2017). LinkedIn or LinkedOut? How social networking sites are reshaping the labor
market. Research in the Sociology of Work, 30, 1–31.
Siciliano, M. L. (2016). Control from on high: Cloud-computing, skill, and acute frustration among
analytics workers in the digital publishing industry. Research in the Sociology of Work, 29, 125–153.
Silver, B. J. (2003). Forces of labor: Workers’ movements and globalization since 1870. Cambridge:
Cambridge University Press.
Smirnova, M. (2014). What is the shortest Russian joke? Communism. Russian cultural consciousness
expressed through Soviet humor. Qualitative Sociology, 37(3), 323–343.
Smith, A. (2016). Gig work, online selling, and home sharing. Pew Research Center, November 17. Retrieved
Srnicek, N. (2017). Platform capitalism. Cambridge: John Wiley & Sons.
Downloaded by University Library At 02:20 18 June 2019 (PT)
Platforms at Work 89
Strauss, A., & Corbin, J. M. (1990). Basics of qualitative research: Grounded theory procedures and
techniques. London: Sage Publications.
Styhre, A., Backman, M., & Börjesson, S. (2005). The gendered machine: Concept car development at
Volvo Car Corporation. Gender, Work and Organization, 12(6), 551–571.
Toubiana, M., & Zietsma, C. (2017). The message is on the wall? Emotions, social media and the
dynamics of institutional complexity. Academy of Management Journal, 60(3), 922–953.
Tuckman, A., & Whittall, M. (2002). Afrmation, games, and insecurity: Cultivating consent within a
new workplace regime. Capital and Class, 26(1), 65–93.
Unicru. (2003). Unicru and ChoicePoint expand partnership to provide customized screening
for chain drug store hiring workows. Business Wire, May 5. Retrieved from http://www.
Van Dijck, J. (2013). The culture of connectivity: A critical history of social media. Oxford: Oxford
University Press.
Van Leeuwen, T. (2008). Discourse and practice: New tools for critical discourse analysis. Oxford:
Oxford University Press.
Winter, S., Berente, N., Howison, J., & Butler, B. (2014). Beyond the organizational ‘container’:
Conceptualizing 21st century sociotechnical work. Information and Organization, 24(4), 250–269.
Woodward, S. (2000). Decision Point gets $16 million infusion. The Oregonian, May 4, p. B1.
Ziewitz, M. (2016). Governing algorithms: Myth, mess, and methods. Science, Technology and Human
Values, 41(1), 3–16.
Downloaded by University Library At 02:20 18 June 2019 (PT)
Employer Total
Application System?
Walmart 2.3 million Yes From their application FAQs: “We
do not accept paper applications
for hourly positions. We would
recommend checking for computer
access at your local library or
workforce solution center.”
Kroger 443,000 Yes
IBM 420,000 Yes
Home Depot 406,000 Yes Helpline representative: “We do not
offer paper job applications. All
applications are required to be
submitted online.”
McDonald’s 375,000 Yes, but individual
franchises may elect
for separate policies
Berkshire Hathaway 367,700 Yes, but subsidiaries
may elect for
separate policies
Amazon 341,400 Yes
FedEx 335,800 Yes
UPS 335,500 Ye s From the FAQs: “The process for
external candidates must be done in”
Target 323,000 Yes
Walgreens Boots
300,000 Yes
General Electric 295,000 Ye s
Albertsons Co. 274,000 Ye s
Wells Fargo 269,100 Ye s From their guide to the application
process: “The application process:
– Fill out and submit the application.
– You will receive an email conrmation
that we have received it.
– The recruiter or hiring manager will
review your prole.
– We will contact you directly if
your background matches our
hiring needs.”
AT&T 268,500 Yes
PepsiCo 264,000 Yes
Cognizant Technology
260,200 Yes
Starbucks 254,000 Yes
JP Morgan Chase 243,400 Ye s
Downloaded by University Library At 02:20 18 June 2019 (PT)
Platforms at Work 91
Employer Total Work-
Application System?
Lowe’s 240,000 Yes From their Careers FAQ: “To apply for
a job with Lowe’s, please follow these
– Search for jobs on
and click “Apply Now
– Create your prole / account (an email
address is required)
– Complete the application.”
Employer workforce data from 2017 Fortune 500 list of largest employers, exclud-
ing Yum China Holdings, who are based in TX but whose workforce is almost
entirely in China. This should be taken as a “best estimate available” since some
of the numbers include non-US employees (e.g., through other sources, we have
found that Walmart’s US workforce is closer to 1.4m).
Assessment of whether applications were online-only drawn from analysis
of employer websites and, if that was inconclusive, correspondence with their
human resource departments. To comply with the ADA, each company MAY
offer printed applications as a reasonable accommodation to disabled applicants,
depending on the nature of the disability and the request. This research was done
on the basis of able-bodied applicants, with conclusions drawn from posted poli-
cies and phone and online chats with company representatives.
Downloaded by University Library At 02:20 18 June 2019 (PT)
... The studies that do exist suggest that digital, internet-based technologies have enabled new labor market situations which complicate organizational boundaries and defy easy categorization as employment or independent contracting (Schor & Vallas, 2021;Vallas & Schor, 2020). Even for so-called "standard" employment, new digital tools have transformed the way workers and employers are matched (Ajunwa & Greene, 2019). ...
... Many algorithms employed by platform dLMIs can home in on personalized websites and social media profiles, allowing individuals and companies to connect more easily than ever before. Advances in AI-driven database management have spawned web-based interfaces that control the flow of information between job seekers and organizations (Ajunwa & Greene, 2019). Social media websites have expanded the set of online connections that facilitate the diffusion of information about job opportunities (Burke & Kraut, 2013). ...
... Job posting websites such as Monster and CareerBuilder began emerging in the mid-1990s (Ajunwa & Greene, 2019). ...
Full-text available
Dramatic changes in organizational forms and employee‐employer relationships have coincided with a proliferation of labor market intermediaries. Often digital and internet‐based, these new hiring technologies assist organizations in recruiting and screening potential job candidates. We identify three types of digital labor market intermediaries (dLMIs): connectors, curators, and comminglers. We examine the use of dLMIs through the lens of organizational theory, focusing on implications for organizational efficiency, power, and equity. dLMI use is patterned but variable across different organizations and has unintended outcomes that defy efficiency expectations. It poses new constraints for job seekers while allowing organizations and intra‐organizational groups to negotiate institutional pressures and power imbalances. Finally, dLMI use appears to reproduce pre‐existing inequalities among different types of employees.
... This stems largely from two concurrent shifts: the rise of internet-based communications and major transformations in the social contract that implicitly and explicitly governs employment relationships. The internet has spawned a variety of job posting boards, automated hiring applications, and social media platforms, which together have greatly increased the capacity of organizations to gather information about actual and potential job candidates (Ajunwa & Greene, 2019). At the same time, the new employment contract (Rubin, 2012) has transformed workers from stable employees into "free agents" who are responsible for continuously (re-)securing their own economic and social well-being (Gershon, 2017;Kalleberg, 2011;Smith, 2002). ...
... Perhaps for this reason, reports of screening social media profiles because "it takes little time and effort" decreased from 63% in 2011 to 34% in 2016 (SHRM, 2016). The early convenience of cybervetting appears to have decreased over time due to technology churn (Carr, 2016), notably the proliferation of online platforms, features, users, and content (Ajunwa & Greene, 2019). ...
... They are often based on stereotypical assumptions that are particularly damaging to members of disadvantaged communities. In hiring, algorithmic tools are often based on biased assumptions about "fit," which can reproduce rather than ameliorate discriminatory outcomes (Ajunwa and Greene 2019). ...
Full-text available
Cybervetting is the widespread practice of employers culling information from social media and/or other internet sources to screen and select job candidates. Research evaluating online screening is still in its infancy; that which exists often assumes that it offers value and utility to employers as long as they can avoid discrimination claims. Given the increasing prevalence of cybervetting, it is extremely important to probe its challenges and limitations. We seek to initiate a discussion about the negative consequences of online screening and how they can be overcome. We draw on previous literature and our own data to assess the implications of cybervetting for three key stakeholders: job candidates, hiring agents, and organizations. We also discuss future actions these stakeholders can take to manage and ameliorate harmful outcomes of cybervetting. We argue that it is the responsibility of the organizations engaged in cybervetting to identify specific goals, develop formal policies and practices, and continuously evaluate outcomes so that negative societal consequences are minimized. Should they fail to do so, professional and industry associations as well as government can and should hold them accountable.
... Recently, researchers have discussed fairness-related harms that can occur, such as people's individual experiences with AI systems [87] or how AI systems represent individuals in groups [4]. Of note, fairness in group AI (i.e., AI managing groups of people by distributing resources) is a particularly difficult challenge, as AI systems can unfairly allocate opportunities, resources, or information [17]. ...
Full-text available
Algorithmic fairness is an essential requirement as AI becomes integrated in society. In the case of social applications where AI distributes resources, algorithms often must make decisions that will benefit a subset of users, sometimes repeatedly or exclusively, while attempting to maximize specific outcomes. How should we design such systems to serve users more fairly? This paper explores this question in the case where a group of users works toward a shared goal in a social exergame called Step Heroes. We identify adverse outcomes in traditional multi-armed bandits (MABs) and formalize the Greedy Bandit Problem. We then propose a solution based on a new type of fairness-aware multi-armed bandit, Shapley Bandits. It uses the Shapley Value for increasing overall player participation and intervention adherence rather than the maximization of total group output, which is traditionally achieved by favoring only high-performing participants. We evaluate our approach via a user study (n=46). Our results indicate that our Shapley Bandits effectively mediates the Greedy Bandit Problem and achieves better user retention and motivation across the participants.
... Focusing on the use of cybervetting, this study adds to existing research on recruitment and management in a digital era (for example, Ajunwa and Greene, 2019;Halpin and Smith, 2019). The access to information posted on the internet alters the balance of power between employers and potential employees, and challenges the -supposedly objective -selection and decision-making procedures in the field. ...
The process of recruiting new employees involves the risk of hiring the ‘wrong’ person. Systematic and extensive information gathering is therefore used to support objective and rational decisions. Today, the use of cybervetting is part of the recruitment process, but prior research shows that emotions, contrary to the ideals of ‘objectivity’, are essential for sorting and selection decisions. Based on interviews with 37 Swedish recruiters, this study demonstrates how cybervetting is motivated, restrained and directed by recruiters’ feelings about the jobseeker and the practice of cybervetting. The study findings also emphasise that recruiters believe in a ‘professional’ means of managing emotions, and the notion that certain emotions represent a tacit knowledge with an emotional foundation that is difficult to articulate.
... In turn, organizations seeking to influence exchange rules may turn their focus increasingly to moves that are aimed at shaping the technologies of the marketplace, as much as shaping the market's cultural dimensions. The technologies used in automated hiring platforms, for example, allow managers to make new types of structured comparisons across workers (Ajunwa & Greene, 2019) enabling new exchange rules for labor that did not previously exist. ...
... Little is known about peer-to-peer interactions in these communities as a form of organizational or work-based knowledge management, and how the technical infrastructures of such spaces shape, and are shaped by, social structures, values, and interactions. The mediation of work by platforms is expanding, touching more aspects of professional and occupational life in processes like hiring (Ajunwa and Greene, 2019). Implicit, cultural conceptions of values are no exception, and are enacted and exercised through platform communications of workers. ...
Full-text available
Gig workers are typically thought of as individuals toiling in digitized isolation, not as communities of shared learning. While it’s accurate to say they don’t have the same information-sharing norms as people in traditional employment arrangements, some do gather, in part in digital communities. Online forums, in this space, have become popular sites for gathering, sharing information, and comparing practices. These behaviors provide an opportunity to examine gig workers as emergent communities of practice, and to analyze how work, identity, skills, and workspaces co-constitute each other as sociotechnical environments of work change. In this research, I examine workers’ interactions in an online forum, and focus on how they talk about scams. Analysis reveals that talking about scams is a way for workers to enact belonging in their community of practice. Victims are belittled by other workers, who frame vulnerability, and lack of foresight due to unfamiliarity with the forum itself, as a lack of authenticity. Repudiations are denunciations through which workers assert their belonging. These findings illuminate the practices of what I call “para-organizational” work, with implications for knowledge management in structures of algorithmic competition.
In the wake of media hype about artificial intelligence (AI)/human collaboration, organizations are investing considerable resources into developing and using AI. In this paper, we draw on theories of technology in organizations to frame new directions for the study of what it means to work “with” AI. Drawing on prior literature, we consider how interactions between users and AI might unfold through theoretical lenses which cast technology as a tool and as a medium. Reflecting on how AI technologies diverge from technologies studied in the past, we propose a new perspective, which considers technology as a counterpart in a system of work that includes its design, implementation, and use. This perspective encourages developing a grounded understanding of how AI intersects with work, and therefore ethnography, building on thick descriptions, is an apt approach. We argue that relational ethnographic approaches can assist organization theorists in navigating the methodological challenges of taking a counterpart perspective and propose several strategies for future research.
Full-text available
In China, truck drivers in the logistics industry have used social media to form various networked organizations for mutual assistance and protection of rights and interests. This study examines the organizational practices of Chinese truck drivers on social media (i.e., WeChat and Douyin). Using online communities, Chinese truck drivers have constructed a new type of solidarity that includes virtual and practical dimensions. Social media empowerment has expanded the social capital of truck drivers, promoting social integration and resource redistribution. This online self-organization provides a reference for collaborative governance among self-employed workers who want to promote professional solidarity. However, our findings also indicate that combating the exploitation of digital labor on capital platforms is fundamentally difficult for organized labor groups.
Full-text available
Algorithmic discrimination in employment. An overview Debates about the future of work in light of developments in artificial intelligence are held predominantly in the context of job losses and technological unemployment. Far less attention is paid to the challenges posed by the increasingly widespread phenomenon of algorithmisation of management functions in the modern world of work. Meanwhile, the shift to algorithmic management represents a significant qualitative change that, in addition to promising broadly understood modernisation and optimisation of decision-making processes, carries specific repercussions in the context of human rights protection, including in particular the prohibition of discrimination in employment. The article attempts to assess the adequacy of the EU antidiscrimination instrumentarium to the specifics of algorithmic discrimination mechanisms, and aims to encourage an in-depth scientific discussion on the need to develop effective regulatory responses at the national level.
Cybervetting refers to screening job candidates by evaluating information collected from internet searches and social media profiles. Relatively little is known about how organizational actors use this practice in hiring decisions. Interviews with 61 human resource (HR) professionals reveal that they cybervet in order to minimize hiring risks and maximize organizational fit. Their judgments are deeply rooted in assessments of job candidates’ moral character and how it might affect workplace interactions. Because it involves the construction of moral criteria that shape labor market actions and outcomes, we describe cybervetting as a morally performative practice. HR professionals express enthusiasm for cybervetting, but also concerns about privacy, bias and fairness. Importantly, cybervetting practices and policies vary substantially across different types of organizations. These findings deepen our understanding of how organizational actors define and regulate moral behavior and how their actions are moderated by market institutions.
Full-text available
This essay details the resurgence of wellness program as employed by large corporations with the aim of reducing healthcare costs. The essay narrows in on a discussion of how Big Data collection practices are being utilized in wellness programs and the potential negative impact on the worker in regards to privacy and employment discrimination. The essay offers an ethical framework to be adopted by wellness program vendors in order to conduct wellness programs that would achieve cost-saving goals without undue burdens on the worker. The essay also offers some innovative approaches to wellness that may well better serve the goals of healthcare cost reduction.
Full-text available
Mobile application design can have a tremendous impact on consumer privacy. But how do mobile developers learn what constitutes privacy? We analyze discussions about privacy on two major developer forums: one for iOS and one for Android. We find that the different platforms produce markedly different definitions of privacy. For iOS developers, Apple is a gatekeeper, controlling market access. The meaning of “privacy” shifts as developers try to interpret Apple’s policy guidance. For Android developers, Google is one data-collecting adversary among many. Privacy becomes a set of defensive features through which developers respond to a data-driven economy’s unequal distribution of power. By focusing on the development cultures arising from each platform, we highlight the power differentials inherent in “privacy by design” approaches, illustrating the role of platforms not only as intermediaries for privacy-sensitive content but also as regulators who help define what privacy is and how it works.
The Internet and social media have fundamentally transformed the ways in which individuals find jobs. Relatively little is known about how demand-side market actors use online information and the implications for social stratification and mobility. This study provides an in-depth exploration of the online recruitment strategies pursued by human resource (HR) professionals. Qualitative interviews with 61 HR recruiters in two southern US metro areas reveal two distinct patterns in how they use Internet resources to fill jobs. For low and general skill work, they post advertisements to online job boards (e.g., Monster and CareerBuilder) with massive audiences of job seekers. By contrast, for high-skill or supervisory positions, they use LinkedIn to target passive candidates – employed individuals who are not looking for work but might be willing to change jobs. Although there are some intermediate practices, the overall picture is one of an increasingly bifurcated “winner-take-all” labor market in which recruiters focus their efforts on poaching specialized superstar talent (“purple squirrels”) from the ranks of the currently employed, while active job seekers are relegated to the hyper-competitive and impersonal “black hole” of the online job boards.
This book discusses the transformation of firms into platforms-companies providing software and hardware products to others-that has occurred in many economic sectors. This massive transformation resulted from switching capitalism into data, considering them as a source for economic growth and resilience. Changes in digital technologies contributed much to the relationships between companies and their workers, clients, and other capitalists, who increasingly began to rely on data. Dr. Nick Srnicek critically reviews "platform capitalism", putting new forms of the business model into the context of economic history, tracing their evolution from the long downturn of the 1970s to the economic boom of the 1990s and to the consequences of the 2008 financial crisis. The author demonstrates that the global economy was re-divided among a few of the monopolistic platforms and shows how these platforms set up new internal trends for the development of capitalism. © 2019 National Research University Higher School of Economics. All Rights Reserved.
From the Pinkerton private detectives of the 1850s, to the closed-circuit cameras and email monitoring of the 1990s, to new apps that quantify the productivity of workers, and to the collection of health data as part of workplace wellness programs, American employers have increasingly sought to track the activities of their employees. Starting with Taylorism and Fordism, American workers have become accustomed to heightened levels of monitoring that have only been mitigated by the legal counterweight of organized unions and labor laws. Thus, along with economic and technological limits, the law has always been presumed as a constraint on these surveillance activities. Recently, technological advancements in several fields-big data analytics, communications capture, mobile device design, DNA testing, and biometrics-have dramatically expanded capacities for worker surveillance both on and off the job. While the cost of many forms of surveillance has dropped significantly, new technologies make the surveillance of workers even more convenient and accessible, and labor unions have become much less powerful in advocating for workers. The American worker must now contend with an all-seeing Argus Panoptes built from technology that allows for the trawling of employee data from the Internet and the employer collection of productivity data and health data, with the ostensible consent of the worker. This raises the question of whether the law still remains a meaningful avenue to delineate boundaries for worker surveillance.
For social analysts, what has come to be called the “sharing economy” raises important questions. After a discussion of history and definitions, we focus on 3 areas of research in the for-profit segment, also called the platform economy: social connection, conditions for laborers, and inequalities. Although we find that some parts of the platform economy, particularly Airbnb, do foster social connection, there are also ways in which even shared hospitality is becoming more like conventional exchange. With respect to labor conditions, we find they vary across platforms and the degree to which workers are dependent on the platform to meet their basic needs. On inequality, there is mounting evidence that platforms are facilitating person-to-person discrimination by race. In addition, platforms are advantaging those who already have human capital or physical assets, in contrast to claims that they provide widespread opportunity or even advantage less privileged individuals.
A discourse of employability saturates the higher education sector in the UK. Government and employers call on universities to produce employable graduates who are attractive to the labour market and can sustain their future marketability by taking responsibility for protean self-development. While the neoliberal assumptions behind this call have attracted robust critique, the extent to which employers shape graduating students’ subjectivities and sense of worth as (potentially employable) workers has escaped scrutiny. Inspired by Foucauldian analyses of human resource management (HRM) practices, this article examines employers’ graduate careers websites and explores the discursive construction of the ‘employable graduate’. The article contends that these websites function as a mechanism of anticipatory socialization through which HRM practices extend managerial control into the transitional space of pre-recruitment, with the aim of engaging students’ consent to particular norms of employability.