Journal of Strategic Information Systems 33 (2024) 101838

Contents lists available at ScienceDirect

Journal of Strategic Information Systems

Fostering humanistic algorithmic management: A process of

A R T I C L E I N F O A B S T R A C T

Keywords: Unlike traditional employer-employee relationships, contemporary digital platforms use algo­
Humanistic algorithmic management rithms to control and regulate the crowd workforce. Although prior research has expressed
Algorithmic management concerns over dehumanization stemming from algorithmic management, limited scholarly
Complementarity
attention has been dedicated to exploring how human management can complement algorithmic
Crowd workers
Digital platforms
approaches to address these concerns. Leveraging a case study of an on-demand food delivery
COVID-19 pandemic platform during the COVID-19 pandemic, we developed a process theoretical model that uncovers
several drivers and mechanisms essential for the transition from mechanistic algorithmic man­
agement to humanistic algorithmic management. This model elucidates the dynamic substitution
and complementarity of human and algorithmic management through four key mechanisms:
replacing and dampening, compensating, enabling, and synergizing. It also delineates effective
humanistic management actions in scenarios in which algorithmic decisions are insufficient,
which contributes to both performance and humanistic outcomes. In the realm of contemporary
crowd workforce management, where digital platforms employ algorithms, this research sheds
light on the unique and timely insights into the role of human managers in enhancing the strategic
and humanistic values of algorithmic technology.

Introduction

Advancements in pervasive computing have facilitated ride-hailing and on-demand food delivery business models with thousands
of independent crowd workers on digital platforms (Deng and Joshi, 2016; Gol et al., 2019). Algorithmic management, enabled by
artificial intelligence and machine learning, optimizes operational efficiency and economic value by efficiently matching tasks with
workers (Kellogg et al., 2020; McAfee and Brynjolfsson, 2017; Möhlmann et al., 2021). However, the opacity of algorithmic man­
agement and data-driven approaches have unintended consequences (Marabelli et al., 2021; Marjanovic et al., 2022). Opacity,
symbolized by “black-boxed” performance evaluations, erodes worker trust and impedes performance improvements (Gal et al., 2020).
Datafication reduces real-world complexity to machine-friendly formats, neglecting contextual insights (Galliers et al., 2017). These
issues lead to dehumanization, in which workers experience information asymmetry and power marginalization (Möhlmann and

* Corresponding author.
E-mail addresses: [Link]@[Link] (T. Cui), [Link]@[Link] (B. Tan), [Link]@[Link] (Y. Shi).
1
Address: Room 4322, Level 4 Melbourne Connect, The University of Melbourne, Parkville, VIC 3052.
2
Address: Room 2096, Level 2 Quadrangle Building, UNSW Sydney 2052.
3
Address: Room 2121, Level 2 Quadrangle Building, UNSW Sydney 2052.

[Link]
Received 26 July 2022; Received in revised form 23 May 2024; Accepted 23 May 2024
Available online 4 June 2024
0963-8687/© 2024 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license
([Link]
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

Henfridsson, 2019; Nevo and Kotlarsky, 2020). The challenges of algorithmic dehumanization are accentuated during events such as
the COVID-19 pandemic, especially for crowd workers such as food delivery personnel facing safety concerns and limited personalized
support (Pryor et al., 2020). Firms should not blindly rely on algorithms but integrate human managerial expertise and contextual
information to complement algorithmic decisions (Grønsund and Aanestad, 2020).
Prior research has primarily concentrated on the integration of human intelligence and algorithmic intelligence, focusing on how
algorithms are used for worker evaluation and regulation (e.g., Aaltonen and Tempini, 2014; Cram et al., 2022; Gronsund and
Aanestad, 2020; Möhlmann et al., 2021). However, there is a notable scarcity of studies that explicate the roles of human managers in
complementing algorithmic management, particularly in addressing the challenges of dehumanization. Understanding the specific
processes and mechanisms involved in transitioning to more humanistic approaches is critical and under-examined. Addressing this
issue is crucial for multiple stakeholders. Organizations that adopt humanistic algorithmic management can improve worker satis­
faction, trust, and performance, enhancing strategic outcomes. Workers benefit from greater support, transparency, and autonomy,
fostering a more equitable work environment (Lehdonvirta, 2018; Wood et al., 2019). At the societal level, humanistic algorithmic
management aligns with ethical practices and sustainable crowd work ecosystems. Our concept of humanistic algorithmic manage­
ment offers humanistic insights for organizations to prioritize crowd worker well-being while optimizing operational performance in
fast-evolving industries like on-demand services (Kittur et al., 2013; Petriglieri et al., 2018).
This study aims to address two research questions (RQs): 1. How do drivers and mechanisms facilitate the transition from
mechanistic to humanistic algorithmic management?; 2. How can specific humanistic management actions be enacted in the context of
on-demand food delivery platforms? Employing dynamic complementarity notions from Huber et al. (2013) and Nevo and Wade
(2010), we examined algorithmic management and human management complements to achieve strategic and humanistic values. We
conducted an in-depth case study of an online food delivery platform during the COVID-19 pandemic, where workers faced heightened
safety risks and limited social protection (Perkiö et al., 2020). This necessitated human management interventions to safeguard the
effectiveness and well-being of platform workers.
We developed a process model that explains the dynamics of substitution and complementarity between human and algorithmic
management. In response to RQ1, our model posits four mechanisms and their associated contextual drivers: replacing and dampening,
compensating, enabling, and synergizing. In addressing RQ2, our model also characterizes humanistic management actions when
algorithmic decisions fall short, contributing to both performance and humanistic outcomes.
Specifically, flexible performance management softens automatic algorithmic evaluation and discipline, whereas algorithmic trans­
parency improves workers’ understanding and reduces resistance. Team and peer support enhances social interaction. Additionally,
algorithmic and human management enable each other, where adaptive behavioral guidance uses real-time algorithmic data surveil­
lance and analytics to provide personalized support. Our research also highlights a synergistic mechanism between humans (both
managers and workers) and algorithmic management, fostering co-determined work arrangements with more reciprocity and rewarding
relationships, which neither entity alone achieves.
Our research provides novel insights into algorithmic management, crowd work, and the IS phenomenon during COVID-19. First, it
enriches the algorithmic management literature (Cram et al., 2022; Grønsund and Aanestad, 2020; Kellogg et al., 2020) by developing
a process model of humanistic algorithmic management, and shedding light on the transition from mechanistic to humanistic algo­
rithmic management. As one of the initial empirical studies in this area, our research particularly emphasizes the role of human
managers in complementing algorithmic management when algorithmic decisions are insufficient or ineffective. Second, this study
contributes to managerial control literature (e.g., Harrower, 2019; Kellogg et al., 2020; Möhlmann et al., 2021) by advancing the
theoretical understanding of complementarity and substitution with specific effective management actions in algorithmic and human
management. Third, it contributes to the crowd work literature (Deng et al., 2016; Gol et al., 2019) by explaining how human
managers can improve worker performance and wellbeing. Fourth, we answer the call for an understanding of the IS phenomenon
during the COVID pandemic (Ågerfalk et al., 2020; Karanasios, 2022) by offering practical insights for crowd work businesses to cope
with operational challenges due to socially disruptive events.

Theoretical foundation

Algorithmic management

Advances in artificial intelligence are transforming how organizations manage their work and employees. Empowered by AI, or­
ganizations increasingly use algorithm-based decision-making to change traditional management practices. Algorithmic management
refers to the use of algorithmic technologies to automate decision-making, connect workers with tasks, and monitor and control
workers’ performance (Kellogg et al., 2020; Möhlmann et al., 2021). The phenomenon of algorithmic management involves human-
technology interactions in which humans design, implement, and use algorithmic decision-making systems (Marabelli et al., 2021).
Algorithmic decision-making benefits companies by deriving insights from big data and enhancing workforce efficiency. However,
studies reveal that different algorithmic controls (guiding and gatekeeping) negatively impact crowd workers’ usage and workarounds
(Cram et al., 2022; Wiener et al., 2021). One reason is dehumanization, raised by scholars (Gal et al., 2020; Marabelli et al., 2021;
Möhlmann and Henfridsson, 2019), driven by algorithmic management’s characteristics like opacity and datafication (Faraj et al.,
2018; Newell and Marabelli, 2015).
Opacity, refers to the lack of clarity about how decisions are made by algorithms in algorithmic decision-making, results in a lack of
transparency, eroding employee trust and hampering their ability to derive insights for better work performance (Gal et al., 2020).
Datafication involves aggregating and simplifying rich organizational data for algorithmic analysis (Aaltonen and Tempini, 2014).

2
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

This reductionist approach limits the adaptability of algorithmic decision-making to individual contexts and unexpected events (Faraj
et al., 2018). To address these issues, researchers advocate exploring the synergy of human and algorithmic intelligence in workforce
management (Günther et al., 2017; Rai et al., 2019).
Current research focuses on frontline workers complementing algorithms to align their performance with organizational goals. For
instance, crowd workers can develop market- and organization-like responses to address algorithmic control issues (Möhlmann et al.,
2021). Employees may use additional data sources for contextualized business reporting, countering algorithm-defined metrics
(Aaltonen and Tempini, 2014). Organizations must frame algorithmic analytics as a fallible technology and establish ethical guidelines
to address opacity, datafication, and nudging incentives from the managing application (Gal et al., 2020). However, these studies have
not explicitly explored the role of human management in mitigating algorithmic management side-effects.
Frontline employees may design algorithms or use the algorithmic output for daily work, while managers use it to evaluate
employee performance, allocate resources, and make strategic decisions. Delegating technology to manage the workforce and inform
data-driven decisions changes managerial roles (Sharma et al., 2014; Faraj et al., 2018). Some suggest managers should use algorithms
to “direct,” “evaluate,” and “discipline” employees (Kellogg et al., 2020). In situations where platform contractors submit low-quality
work or resist engagement, managers can practice “collaborative repair” by providing feedback, sharing responsibilities, and
acknowledging their contribution (Rahman and Valentine, 2021). Appendix A summarizes empirical research on algorithmic crowd
work management.
The aforementioned studies focused on how managers use technical and bureaucratic controls to direct the workforce and improve
their performance. The control perspective suggests that management should monitor and regulate employees’ work behaviors based
on institutional rules and algorithmic metrics (Kellogg et al., 2020). However, rule- and metric-driven management are insufficient for
addressing the dehumanization issue of algorithmic management. For example, crowd workers could adopt an adversarial attitude
toward using a platform app because of black-boxed performance evaluations and reviews they perceive to be unfair (Griesbach et al.,
2019). Furthermore, crowd workers experience information and power asymmetries because their communications with service
providers are mediated by the platform and they often receive template responses which do not address their contextualized re­
quirements (Rosenblat and Stark, 2016).
The issue of dehumanization becomes even more crucial when disruptive events such as crises or disasters occur. For example,
researchers have highlighted that “he implementation issue involves the wide-scale rollout of these (algorithmic decision-making)
systems during the COVID pandemic, in which human intervention is weakly included in the process (triaging) of assisting people
who cannot benefit from automatic diagnosis” (Marabelli et al., p. 2). To mitigate the unintended consequences of algorithmic
management, managers must address how to increase the humanity of automatic decision-making. According to Gartner’s updated
future work trends for 2020, the post-pandemic world will continue to bring workplace changes in terms of employee tracking and
governance (Barker, 2020). Therefore, exploring algorithmic management in the face of the unfolding pandemic and building a
knowledge base to advance theories of algorithmic technology management is opportune and critical at this juncture.

A dynamic complementarity view of algorithmic and human management

Our study employs the complementarity view as the theoretical foundation to examine the interplay between algorithmic and
human management approaches. This view suggests the logic of substitution and complementarity for the interplay of two entities.
Specifically, substitution involves replacing and dampening and complementarity covers compensating, enabling (Huber et al., 2013), and
synergizing (Nevo and Wade, 2010).
Previous research has employed various causal mechanisms to clarify complementarity and substitution (Bunge, 2004). These
mechanisms reveal temporal cause-and-effect relationships, providing insights into adapting algorithmic and human management.
Consistent with established causal mechanisms (Huber et al., 2013; Nevo and Wade, 2010), substitution encompasses replacing and
dampening. Replacing indicates the substitution of certain aspects of algorithmic and human management due to their functional
equivalence (Marabelli et al., 2021). In contrast, dampening suggests that algorithmic management might undermine the strengths of
human management (Gal et al., 2020; Rosenblat and Stark, 2016), as constant surveillance may signal distrust, negatively impacting
the supportive role of human management (Möhlmann and Zalmanson, 2017). In our study, we adapt a complementary perspective,
where algorithmic and human management exhibit unique yet mutually valuable strengths, involving compensating. Additionally, they
create favorable conditions for each other, characterized as enabling (Huber et al., 2013). Furthermore, we introduce synergizing as a
mechanism, emphasizing the harmonious coalescence of human and algorithmic management to yield emergent capabilities with
positive outcomes (Nevo and Wade, 2010). This mechanism depends on the compatibility between the capabilities of humans and
algorithms.
While algorithmic management emphasizes the importance of constant surveillance and automated decision-making (Kellogg
et al., 2020; Möhlmann et al., 2021), human management focuses on supporting employees and nurturing positive workplace re­
lationships (Becker and Huselid, 2006; Luthans and Youssef, 2004). Both types of management have their unique strengths. Algo­
rithmic management enhances managerial decision-making and delivers personalized feedback, making it crucial for effective remote
management (Rodgers et al., 2023). Human management has the potential to boost employee attitudes and behaviors, including work
motivation, organizational commitment, and skill development (Kehoe and Wright, 2013). Rather than viewing one as a weakness of
the other, it is important to recognize that the two management approaches are complementary, each enhancing the benefits of the
other. This perspective aligns with the concept of complementarity, where doing more of one approach (e.g., human management)
increases the advantages of doing more of the other (e.g., algorithmic management). This complements the traditional substitutional
view, which suggests that increasing one approach diminishes the benefits of the other (Tiwana, 2010). Using the existing logic of

3
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

substitution and complementarity, we aim to conceptualize specific humanistic algorithmic actions in a given context.
vation-driven collaboration is a particular form of inter-firm alliance that can range from informal arrangements initiated for
specific problem solving or incremental inno-vations to formal long-term strategic alliances for significant technological advances and
radical innovations. Motives for the establishment of innovation collaboration are several. Firms may decide to collaborate to spread
the costs and risks of innovation, especially in industries characterized by increasing development investments, such as the phar­
maceutical, telecommunication and aerospace industries. Collaboration between users and suppliers of new products and technologies
can be aimed at establishing technical standards and dominant designs. However, one of the most widely cited motives for collabo­
ration is the acquisition of knowledge and capabilities from partner firms (Hagedoorn, 1993; Hamel, 1991; Mowery et al., 1996). This
particular reason to engage in inter-firm collaboration is justified by the consideration that firms are char-acterized by heterogeneous
knowledge bases and that this heterogeneity comes from a process of cumulative learning (Dosi et al., 2000). Innovation-driven
collaboration is a particular form of inter-firm alliance that can range from informal arrangements initiated for specific problem
solving or incremental inno-vations to formal long-term strategic alliances for significant technological advances and radical in­
novations. Motives for the establishment of innovation collaboration are several. Firms may decide to collaborate to spread the costs
and risks of innovation, especially in industries characterized by increasing development investments, such as the pharmaceutical,
telecommunication and aerospace industries. Collaboration between users and suppliers of new products and technologies can be
aimed at establishing technical standards and dominant designs. However, one of the most widely cited motives for collaboration is the
acquisition of knowledge and capabilities from partner firms (Hagedoorn, 1993; Hamel, 1991; Mowery et al., 1996). This particular
reason to engage in inter-firm collaboration is justified by the consideration that firms are char-acterized by heterogeneous knowledge
bases and that this heterogeneity comes from a process of cumulative learning (Dosi et al., 2000).

Research method

The case research approach was selected because it is ideal for addressing “how” research questions (Corbin and Strauss, 2015;
Urquhart, 2022), and under-explored phenomenon (i.e., the nature and implications of the algorithmic management). Based on our
research objectives, we identified two criteria for the case selection. First, the selected case must be a digital platform that manages a
large and effective crowd workforce because we are looking to develop a theory based on proven, if not best, practices (Urquhart,
2022). Second, the selected platform should have a proven track record for deploying a wide variety of algorithmic management
practices, as this affords us a wider array of possibilities in our theory (Tan et al., 2020).
In line with our criteria, we selected one of the largest on-demand food delivery platforms in Australia, DeliVER (a pseudonym), for
our case study. This is because not only does DeliVER manage a large and vibrant crowd workforce consisting of over 7,000 registered
delivery riders, but it has established a range of algorithmic management practices that have proven to be highly effective in rein­
forcing its corporate policies as well, making it a revelatory or “extreme” (see Gerring, 2009, p. 653) case for the purpose of our study.
The algorithmic management practices instituted by DeliVER include a matching mechanism between workers and orders, a repu­
tation system in which customers rate workers’ performance, and a built-in navigation system that directs workers and reports their
whereabouts to the platform and customers.

Data collection

Data were collected in two phases. In the initial preparatory phase that spanned January 2020, we gathered data from a variety of
secondary sources, including internal corporate documents, industry reports, social media posts, and news articles. The information
from these “nontechnical” (Strauss and Corbin, 1998, p. 52) sources was used to increase our knowledge of the platform, the industry it
operated in, and its approach to algorithmic management, all of which formed the basis for formulating our subsequent interview
questions (Darke et al., 1998).
The preparatory phase was followed by a field visit phase that lasted from February to June 2020 and a follow-up phase from
February to March 2023. We interviewed representatives of DeliVER’s top management, including its Chief Information Officer, Head
of Operations, Digital Operations Manager, and several senior Data Analysts. We also interviewed several delivery riders to gain an
understanding of their perspective on the algorithmic management practices established by DeliVER. A total of 33 interviews were
conducted during this phase (refer to Appendix B). All informants interviewed during the field visit and follow-up phases were
identified via chain referral sampling (Biernacki and Waldorf, 1981). The interviews were based on a series of preliminary interview
questions that we sent in advance to our “gatekeeper” (Pan and Tan, 2011, p. 165) within the organization. This method of identifying
informants was adopted because, “gatekeeper” possesses more internal information to identify the most appropriate informants for the
questions (Pan and Tan, 2011).
We conducted interviews using semi-structured guides designed based on relevant themes in algorithmic management literature.
This approach is less rigid than a positivist case study aiming to validate pre-formulated hypotheses, striking a balance between
generative induction and pragmatic structure (Langley, 1999). Each guide included a standard set of questions about the nature and
implications of algorithmic management, tailored to the perspective of the platform or employees. We also had role-specific questions
for each informant (e.g., Senior Data Analysts discussed the data they collected and analyzed on delivery riders).
Each interview lasted between 30 and 60 min. We digitally recorded the conversations with permission from the informants and
transcribed them for analysis. We also used our observations of the on-demand delivery services as they were being performed and
other secondary sources (including internal corporate documents, industry reports, social media posts, and news articles collected in
the preparatory phase) as supplementary evidence for our interview data. After the data were collected, each member of our research

4
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

team examined the data carefully and independently to ensure consistent interpretation (Klein and Myers, 1999). If there were
conflicting interpretations, or if clarifications and follow-up questions were deemed necessary, the relevant informants were contacted
via email or instant messaging.

Data analysis

We performed data analysis concurrently with data collection (Corbin and Strauss, 2015; Urquhart et al., 2010). The voluminous
and poorly organized qualitative data were first condensed into manageable forms using narrative and visual mapping strategies
(Corbin and Strauss, 2015; Langley, 1999). The narrative strategy entailed the construction of a narrative that represented our account
of what had happened. On the other hand, the visual mapping strategy involved creating event timelines chronologically and doc­
umenting emergent theoretical ideas in a series of diagrammatic sketches. After the narrative and visual maps were constructed, they
were verified with relevant informants. This was done to ensure the validity of both our interpretation of the informants’ accounts and
the theory that we were developing (Urquhart et al., 2010).
Following the development of the narrative and visual maps, we organized and coded the data using the grounded theory tech­
niques of arriving at different levels of abstraction (Corbin and Strauss, 2015) supported by a three-level coding approach by Gioia
et al. (2012). Our data analysis involves both induction and deduction. Using inductive reasoning, we adhered to interviewees’
verbalization to code data. First, we assigned labels to the data (e.g., excerpts from interviews) to create a pool of first-order categories.
Second, we grouped the first-order categories into second-order themes based on common outcomes achieved by certain management
actions and induced a term, based on those common outcomes, to label that theme.
Using deductive reasoning logic (Corbin and Strauss, 2015), in the third step, we assigned theoretical interpretations to the themes
we developed through the application of existing substitution and complementarity logics: replacing and dampening, compensating,
enabling (Huber et al., 2013), and synergizing (Nevo and Wade, 2010). The data structure in Appendix C illustrates the coding scheme
that contains the definition or description of each construct. The use of both induction and deduction allows us to develop a theory that
preseves the narratives grounded in data and conceptualize the narratives into an adequate abstraction that provides theoretical
generalization (Gioia et al., 2012; Walsham, 1995).
When coding, new concepts, themes, or dimensions are integrated into the evolving schema. The schema adapts as needed, adding,
replacing, or removing elements, and coding proceeds accordingly (Urquhart, 2022). This iterative process involves continuous
comparison between theory and data, allowing the inductive model to gradually take shape. This process continued until theoretical
saturation was achieved, meaning the model could comprehensively explain the case data with minimal incremental learning
(Urquhart, 2022). Appendix C provides a sample data structure summarizing first-order concepts, second-order themes, and aggregate
dimensions related to replacing and dampening, compensating, enabling, and synergizing.

Findings

This findings section delves into the evolution of DeliVER’s management approach, exploring the transition from mechanistic
algorithmic management to humanistic algorithmic management. Overall, it investigates the mechanisms of substitution and
complementarity as DeliVER navigates the challenges posed by increasing demand and the ongoing COVID-19 pandemic. The dis­
cussion focuses on how the company dynamically balances human and algorithmic management through four key mechanisms:
replacing and dampening, compensating, enabling, and synergizing. Each mechanism showcases how DeliVER responds to the shifting
needs of its workforce, seeking operational efficiency while prioritizing worker well-being, flexibility, and satisfaction.

Driver: Increasing demand

With shifting consumer preferences, convenience, and the onset of the COVID-19 pandemic, which highlighted the importance of
contactless dining options, there was an increasing demand for food delivery services. DeliVER reportedly saw a dramatic surge in
customers signing up for their services since the outbreak, and it received an influx of new riders and restaurant partners. This abrupt
spike in demand posed challenges for DeliVER’s business planning and operations. DeliVER’s IT capabilities were key to managing
resources and processes to rapidly cope with the growing demand, and there was a need for a responsive and scalable online platform
as well as effective scheduling algorithms. DeliVER’s Head of Operations described the situation in which the platform found itself:
“This is a challenging time for us with the sudden rising demand, not only the increasing number of customers but also restaurants and
delivery riders. We need to be here to connect our restaurants, delivery riders, and customers to provide a reliable marketplace.”
- DeliVER Head of Operations
Delivery riders also experienced a demand spike, and one of them expressed the following:
“The workload has increased dramatically during the pandemic. I have to work longer with more orders to deliver, especially during peak
lunchhours. … This is tough. I felt exhausted after work and I have to start early again tomorrow because we do not have enough riders.”
- Delivery Rider L

5
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

Substitution logic of mechanistic algorithmic management: replacing and dampening

To meet the spike in demand, DeliVER adopted algorithmic management by Replacing and Dampening, which aligns with the
substitution logic of mechanistic algorithmic management. In particular, there was a transformative shift occurring at DeliVER as
algorithmic management took center stage, substituting for traditional human management functions such as human scheduling,
human monitoring, and performance evaluation. For instance, DeliVER used algorithmic resource optimization to replace human
scheduling, monitoring, and performance evaluation of delivery drivers. We observed this in our study in the way DeliVER optimized
how its riders are matched to orders by enhancing their allocation algorithms to fulfil a larger number of delivery orders as a result of
the spike in demand. In 2019, DeliVER started to develop real-time optimization algorithms that were capable of finding the most time-
and cost-effective routes for delivery scheduling. DeliVER’s platform also continuously evaluated riders’ performance by using
customer ratings and the riders’ acceptance rate of algorithmic order assignments. It also disbursed automated rewards and punish­
ments when riders were deemed to have performed well or fell short of their expectations.
A “technical infrastructure” for operational efficiency was established as a beneficial outcome of this rudimentary and mechanistic
form of algorithmic management. The algorithms enabled automatic and efficient resource coordination, allowing DeliVER to expedite
its delivery processes and respond swiftly to increasing demands without significantly increasing human managers or labor costs.
Improving resource optimization has resulted in unprecedented operational efficiency. DeliVER could now partner with more res­
taurants, coordinate a larger pool of delivery riders with minimal human intervention, and introduce new revenue streams from a
growing customer base. DeliVER’s Chief Information Officer described these algorithms as follows.
“We have been investing in building algorithms to help determine the best order scheduling and address the vehicle routing problem. …
This has been our long-term investment and has helped us automatically match customer orders with the closest delivery riders. … We
efficiently scaled our services to include more restaurants, delivery riders, and customers. We also scaled by introducing some new
delivery services.”
- DeliVER Chief Information Officer
In line with conventional remote management, algorithmic management dampened traditional work environments in which
workers engaged in social interactions with managers and peers. Platform-based delivery restricted access to interrelational and
instrumental support and could lead to social isolation. As delivery riders were geographically dispersed and might compete for de­
livery tasks, their work was often carried out in isolation or in contact only with customers because they were reluctant to share task
information with their peers. DeliVER’s Head of Operations elaborated on the unintended consequences of algorithmic management:
“As an app-based gig economy company, we do not have the traditional organizational structure or culture. Our delivery riders are not
like traditional organizational employees who work together in offices. So that, they do not have many opportunities to engage in social
interactions with each other or with our management team.”
- DeliVER Head of Operations

Driver: Resistance toward algorithms

The introduction of algorithmic management had far-reaching implications for DeliVER, not only within the platform but also for
its delivery riders. It fundamentally altered the work environment, shifting it towards a more solitary experience, where workers found
themselves operating in relative isolation. This shift was significant, as it departed from the traditional framework of human-to-human
interaction and oversight, affecting the social dynamics of the workforce. Moreover, this transition occurred against the backdrop of a
serious public health crisis that created further complications. In particular, the health crisis presented infection risks to employees of
the on-demand food delivery platform, adding an additional layer of concern. This has generated operational challenges for the focal
platform, intensifying the increasing demand for contactless services to mitigate health and safety concerns. The confluence of these
challenges created a perfect storm that crystallized an attitude of resistance toward algorithms among the platform’s delivery riders.
This resistance stemmed from a combination of factors, including sudden and profound changes in work processes, perceived loss of
control, and concerns over the potential implications for health and safety. Two delivery riders explained their frustration and negative
experiences:
“The recommended order scheduling is not intelligible sometimes. I do not have any information or explanations, and I have nowhere to
complain. This is even more frustrating during this unusual time.”
- Delivery Rider G
“At the beginning of the outbreak, everything seems uncertain and unsafe. But there were more orders than before. I felt I was asked to
take actions that might jeopardize my health and safety….”
- Delivery Rider J

Complementarity logic of humanistic algorithmic management: compensating

Collectively, these multifaceted challenges provided the impetus for a different management approach, one that is aligned with the
complementarity logic of humanistic algorithmic management. This approach aimed to reconcile the emerging algorithmic practices
with the realities of the organizational context (i.e., the unfolding health crisis) as well as the unique requirements of the business (i.e.,

6
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

on-demand food delivery). It sought to strike a balance between the automation and operational efficiency offered by algorithmic
management on the one hand, with the need to maintain worker well-being, safety, and job satisfaction on the other hand, thereby
setting the stage for the unfolding of interactions of algorithmic management and human management.
Driven by the resistance toward algorithms, compensating represented the initial step undertaken by DeliVER in line with this new
complementarity logic of humanistic algorithmic management. During this time, human management played the role of compensating
for the limitations and challenges posed by pure algorithmic management. This acknowledged that algorithmic decision-making, while
efficient, may not fully address the multifaceted needs and concerns of the workforce. The primary aim was to bridge the gap between
the operational efficiency offered by algorithms and the nuanced human and social aspects of crowd work.
First, we found that DeliVER’s algorithmic management for performance evaluation was compensated by a new system of flexible
performance management. More specifically, this humanistic management action entailed a softer and more humane approach to
algorithmic evaluation and discipline. Similar to other service sectors, food delivery platforms tend to use customer ratings as a quality
control instrument to manage worker performance. However, unlike the customer evaluation processes of previous eras, in which
managers were responsible for rewarding or punishing workers, algorithmic management tends to incorporate such feedback directly
into the technical control system. In the context of food delivery, delivery riders can instantly receive lower performance evaluations
without following detailed and rigid delivery guidelines provided by complex delivery instruction systems. Riders with low average
customer ratings may even be subject to immediate deactivation. DeliVER’s policy states that it may “deactivate” people whose last
100 ratings average less than 4.2 out of 5 stars. Algorithmic performance management tends to create greater frustration and stress for
workers for two main reasons: the immediacy of rewards or penalties; and the intentional opacity of the evaluation system. The
unfolding pandemic intensified the situation where riders struggled to balance their safety and performance, which gave rise to greater
resistance to constant evaluation.
To overcome this, DeliVER softened its algorithmic evaluation and disciplinary practices by not penalizing, but instead compen­
sating, delivery riders who need to take steps to protect their health. DeliVER also employed human “algorithmic auditors” responsible
for monitoring, auditing, and altering performance evaluation algorithms. For any of the deactivation decisions suggested by the
algorithm, the human “algorithmic auditors” would re-evaluate and re-assess each case before making a final decision. Delivery riders
with low customer ratings may be offered service improvement training instead of immediate deactivation by the algorithm. Through
this approach, DeliVER incorporated compassion and human managerial influence into the system and enhanced human responsibility
in the decision loop. Automated algorithmic decisions, such as performance evaluation, were perceived as less reductionistic (and
more meaningful or significant) because there was room for human input in these decisions. The ability of human “algorithmic au­
ditors” to contest prejudicial or erroneous algorithmic compensation or termination decisions also helped to protect delivery riders’
autonomy and lower the probability of stressful job demands such as job insecurity. These performance management initiatives
complemented traditional algorithmic management by shifting the emphasis from operational and systems efficiency to work-life
enrichment and sustaining multilateral value co-creation. Our informants at DeliVER described this softened approach to algo­
rithmic evaluation and discipline:
“To overcome the aversion reaction sometimes developed by algorithm users, our delivery riders … We, as the management team, need to
be empowered to change or intervene in an algorithmic decision. The rules should be set by us and not automatically by the algorithms.
We assign staff to reassess the decisions made by the algorithms. … Even a very small amount of power of intervention or modification
can significantly change how our riders are evaluated.”
- DeliVER Digital Operations Manager
“We were told we are not penalized by the rating systems for taking steps to protect our health. So I do not struggle with my need to
protect myself against the rating score I might get.”
- Delivery Rider I
Another specific type of algorithmic management tension originated from the fact that algorithms often provide limited trans­
parency and are perceived as highly opaque, which impedes workers’ understanding of their inner workings. In the case of DeliVER,
another humanistic management action we identified entails enhanced algorithmic transparency to compensate for algorithmic man­
agement and alleviate delivery riders’ resistance toward algorithmic management opacity. In this context, algorithmic transparency
refers to the disclosure of information about algorithmic control to workers. Such information could include, for example, the reasons
why specific algorithm-based work instructions are provided as they are, or how a worker is performing relative to others. Given the
complexity of algorithmic technologies, most DeliVER delivery riders did not fully grasp what types of data are being collected about
them, and how they are being used for performance evaluation. Whereas workers in a traditional human management organization
could typically use a standardized procedure to inquire about relevant criteria and metrics surrounding performance or pay, there were
limited opportunities to do so using an algorithmic management approach. In addition, the rapid development of algorithmic tech­
nologies led to the use of new data sources or methods that are not disclosed to delivery riders, which consequently reduced algo­
rithmic transparency. As a result, DeliVER riders felt obligated to follow the scheduling nudges displayed by the app and sometimes
worked in adverse conditions with high physical and mental demands against the backdrop of the unfolding pandemic, mainly because
they did not understand the extent to which compliance with these algorithm-based work instructions would be reflected in their
performance evaluation.
The DeliVER Head of Operations recognized the heightened importance of algorithmic management transparency during the
pandemic to foster positive perceptions among delivery riders. DeliVER used various channels to disclose information about input
variables influencing specific algorithmic outputs and instructions. Riders could access key information regarding input data that
directly impacted algorithmic instructions via the app or website. Extensive training was provided to help new riders understand how

7
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

algorithms allocated ride requests and evaluated performance. Riders learned that their ratings and acceptance rates influenced
algorithmic outputs and future job assignments. For instance, DeliVER made introductory training sessions mandatory for all new
riders and offered service improvement training to those at risk of deactivation due to low customer ratings. Additionally, DeliVER
kept riders informed about any algorithmic management updates by sending emails about relevant modifications to their terms and
conditions. According to our interviewees, increased transparency in the algorithmic management process boosted rider engagement,
enthusiasm, compliance, and attitudes toward DeliVER. The Head of Operations and delivery rider at DeliVER explained as follows:
“Transparency is even more important during the pandemic. … Our new performance management initiatives aim to reduce information
asymmetries for riders and facilitate them to work better with the systems. … We regularly reach out to delivery riders by email to keep
them updated on new regulations and policies that may affect the algorithmic instructions they receive.”
- DeliVER Head of Operations
“At times, we had trouble observing what leads to certain instructions on the app… for example, I was not sure about what caused the
temporary blocking of my account, not obeying the traffic rules may be a reasonable explanation. … these trainings are helpful for us to
make sense of the app.”
- Delivery Rider G
In addition, we found at DeliVER that team and peer support compensated for algorithmic management by working in tandem to
cultivate a stronger sense of belonging and organizational identity among delivery riders. Self-learning algorithms were initially given
the responsibility of making decisions regarding the management of these riders, eliminating human input in areas that typically fell
under the purview of human managers. DeliVER’s riders also rarely socialized with other riders or the management team and may not
know anyone else affiliated with the platform. To address this issue, DeliVER tried to provide additional support and imbue its
approach to algorithmic management with personal touch. For example, DeliVER established a dedicated support team and service
hotline to provide information and answer questions to its delivery riders. On the delivery app, riders were also supported by a real-
time chat tool that allowed them to speak directly with DeliVER. In addition to employer support, a peer (or co-worker) support
approach was established. In the case of DeliVER, the platform tried to foster a common identity with an eye-catching square yellow
food delivery box using the DeliVER logo. Delivery riders were also encouraged to seek help from each other if required. At the time of
our study, the communication channels initiated and used by DeliVER were local WhatsApp and WeChat groups. Hundreds of local
riders sought information from others or had conversations about all kinds of issues via these platforms. Many riders also informally
created social contacts to connect and support each other. Our informants at DeliVER explained as follows:
“So, we started it by building our identity with a square yellow food delivery box. … It is good. It was like, you know whenever our
delivery riders pass and see others, they will immediately know they work for DeliVER. … We want them to know they are part of an
organizational community, there are support team, support staff with them…”
- DeliVER Head of Operations
“Delivery job is like an one-person alone job when we are riding. … Now it is good that we have a service hotline with a dedicated support
team for us. … I called once when I had a vehicle breakdown. I also enjoy mingling with other riders in online communities, airing our
frustration and gossiping.”
- Delivery Rider L
In compensating, it became evident that purely algorithmic solutions might not adequately address the emotional well-being, job
satisfaction, and social dynamics of delivery riders. Complementary human management actions were introduced to enhance algo­
rithmic decisions and actions. One key outcome of this was the establishment of “human-in-the-loop” processes supporting crowd
workers. These processes integrated the expertise, empathy, and adaptability of human managers with algorithmic automation and
data-driven capabilities. The aim was to strike a harmonious balance leveraging the strengths of both human and algorithmic man­
agement to better serve workforce needs. The “human-in-the-loop” approach fostered a more holistic management approach, focusing
not only on operational efficiency but also on worker well-being and satisfaction. It recognized that crowd workers are unique in­
dividuals with diverse needs, emotions, and challenges. DeliVER’s Chief Information Officer described this approach as follows:
“We believe in creating a harmonious blend of human expertise and algorithmic precision. It is not just about efficiency; it is about
recognizing our delivery riders as human. By integrating the empathy and adaptability of our human managers with the data-driven
capabilities of algorithms, we are better equipped to ensure the satisfaction of our workforce.”
- DeliVER Chief Information Officer

Driver: Perception of the lack of addressing personal needs

Compensating primarily focused on addressing the resistance to algorithms among delivery riders. While introducing compensatory
actions represents a significant step toward human-centric management, it became evident that these measures alone would not fully
address the diverse and evolving needs of delivery riders. The Enabling and Synergizing of humanistic algorithmic management
emerged as a multifaceted response to delivery drivers’ perception of their personal needs – such as well-being and work flexibility –
not being addressed. This approach integrated algorithmic capabilities with human empathy and adaptability to support workers’
emotional, psychological, and physical health. Acknowledging that workforce well-being is key to sustainable operations, especially
during a health crisis, the pandemic underscored the importance of flexible work arrangements due to fluctuating demand, safety
concerns, personal circumstances, and rapidly changing conditions. This required a dynamic management approach that combined

8
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

algorithmic and human elements to effectively address these evolving challenges. As the challenges and complexities of managing a
workforce became increasingly apparent during the pandemic, DeliVER’s Head of Operations shared this perspective:
“We realized that our riders’ needs extend far beyond just addressing algorithmic issues. Their well-being is essential, and that is not
limited to just their work tasks. It involves their emotional and psychological health, their physical condition, and their sense of belonging
to a supportive community. This shift towards maintaining well-being is about acknowledging the diverse needs of our workforce and
ensuring they remain not only productive but also content, especially when navigating a health crisis.”
- DeliVER Head of Operations

Complementarity logic of humanistic algorithmic management: enabling

To reinforce the strengths of both algorithmic and human management, DeliVER took significant steps forward during the Enabling
logic. Enabling refers to the association between human and algorithmic management, each enhancing the other’s effectiveness in
managing crowd workers. Algorithmic management, with its data-driven capabilities, could continuously monitor various aspects of
crowd worker performance and well-being in real time. It can identify patterns, trends, and potential stressors that may go unnoticed
by human managers. It formed the analytical backbone for decision-making, offering insights into individual worker needs and
preferences. On the other hand, human management brought empathy, adaptability, and the ability to understand the nuances of
human emotions and social dynamics. It could interpret the insights provided by algorithms and translate them into personalized
actions that resonate with individual crowd workers. Human managers were able to provide emotional support, encouragement, and
guidance that algorithms alone could not provide.
The enabling management action that DeliVER employed was adaptive behavioral guidance, which refers to DeliVER’s use of data
analytics and algorithmic systems to guide delivery riders based on real-time contextual information. Effective human management
was facilitated by algorithmic inputs in supporting delivery riders, addressing their empowerment and health and safety concerns
during the pandemic. The workforce required extra support to deal with challenges, including powerlessness stemming from algo­
rithmic management and health risks. Algorithmic management employed various practices and data sources for surveillance,
enabling real-time responses to influence riders’ behavior. Thus, algorithmic management could concurrently serve as a tool to provide
additional social support to benefit riders.
Specifically, DeliVER’s algorithmic technologies have enabled the constant surveillance of riders’ behavior and geolocation, which
has enabled new forms of personalized support. DeliVER’s data analytics team started to analyze individual patterns and detect ab­
normalities, such as irregular working hours or driving patterns, and matched known COVID-19 hotspots with riders’ locations. Based
on algorithmic surveillance and data analytic insights, they were able to send individualized advice and real-time nudges to riders.
Specifically, instead of using algorithmic management to nudge riders to work more efficiently and/or for longer hours, this initiative
was designed to send them over-hour work alerts to reduce work intensity (e.g., fatigue, work-life imbalance). To provide more support
to riders, another initiative incorporated algorithms that assigned riders to destinations close to their homes at the end of the shift. In
addition, DeliVER’s safety team compiled a comprehensive set of safety tips and recommendations that were specifically geared to­
wards contactless food delivery for their riders. These created value for riders while reducing potential infection risks. Some of our
informants elaborated on the adaptive behavioral guidance practiced by DeliVER:
“What we want to do is to match the COVID-19 hotspots released by the Department of Health with our real-time geolocation data of
individual riders. So that we can send an alert to them when they are around these areas.”
- DeliVER Senior Data Analyst A
“Every time when we check in in the morning, we were advised that our priority was our safety and well-being…. We got an app alert with
protection tips and certain hotspots. … I think these are good planning and good for us during the pandemic.”
- Delivery Rider B
At this stage of algorithmic management at DeliVER, the company provided ’personalized support’ tailored to the unique needs of
each crowd worker. This support went beyond generic, one-size-fits-all approaches and aimed to empower crowd workers to excel in
their roles, overcome challenges, and maintain a positive work experience even in the face of adversity. These new organizational
policies, practices, and treatments were the indicators of DeliVER’s support and commitment to delivery riders, which led to their
enhanced well-being and job satisfaction. This positive outcome was reflected in the interviewees’ comments:
“I understand that the app is operating as intended. The order sent to me is reliable and timely … The performance feedback I receive is
based on extensive analysis of past deliveries and can actually help us improve our skills.”
- Delivery Rider G
“I felt we had more interactions suddenly. It is not getting work through a faceless app anymore. I am happy to complete my delivery
orders and receive feedback from our customers.”
- Delivery Rider E

Complementarity logic of humanistic algorithmic management: synergizing

The Synergizing logic arrives at a harmonious collaboration between algorithmic and human management to realize humanistic
outcomes. It represented a shift from isolated efforts on the platform to collaborative synergy with its delivery riders. Algorithmic

9
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

management, with its data-driven insights and analytical capabilities, was seamlessly integrated with the adaptability and empathy of
human management. A dynamic partnership was formed, which leveraged the strengths of each form of management for mutual
enhancement. Specifically, algorithmic and human management were synergistically combined to create positive emergent capabil­
ities with beneficial outcomes. The synergizing mechanism was manifested in the co-determined work arrangements. Specifically,
algorithmic and human management were coordinated to provide delivery riders with work flexibility and more participation,
equality, and respect in the working relationship. Although algorithmic management processes were intended to be technically
rational and they were unable to force delivery riders to take an order, they nevertheless mimicked the authority of a work supervisor,
often hounding riders to complete their tasks. When a delivery rider rejects an order, the order can be repeatedly sent back to the same
delivery rider multiple times.
Algorithmic management at DeliVER also began to measure the rate at which delivery riders accepted orders they were offered
while they were logged onto the app. It then explicitly and implicitly rewarded those with the highest acceptance rates and penalized
those who repeatedly declined their orders. It also provided pay bonuses for riders who maintain an early access status during peak
demand periods, which is to start delivery work before 9:00 am every morning. Many algorithmic management platforms offer task
choices to a worker, but this tends to be implemented in a way where workers are disadvantaged by not agreeing: i.e., workers are
penalized for rejecting tasks or for not working during early peak periods.
To enhance delivery riders’ well-being during the pandemic, DeliVER decided not to penalize riders for rejecting tasks if they did
not feel well. Instead of providing incentives for longer working hours, DeliVER stopped the early access status scheme during the
pandemic to provide more flexibility and control to delivery riders over their working hours. DeliVER also offered compensation to
those quarantined. These approaches lessened the ‘always-on’ and unflexible form of work arrangements that tend to be associated
with algorithmic management. Having the power to decline tasks without implicit or explicit penalties or prejudice prevents delivery
riders from being frustrated by low-paid or alienating tasks.
In addition, it tended to be difficult for delivery riders, who are in a more vulnerable and power-imbalanced position, to say how
work is assigned and how performance is assessed. As a result, DeliVER arranged quality improvement sessions to allow riders to have
their voices heard and to have a greater influence on work arrangements. This provided delivery riders with more control to intervene
in algorithmic decisions, giving them a channel to question, discuss, or even contest automated algorithmic decisions. From the de­
livery riders’ perspective, this meant that the working relationship became more oriented towards mutually co-determined ar­
rangements, instead of being perceived as a subjugated working arrangement. In this case, DeliVER’s humanistic management mindset
was matched with the technological capabilities of the platform, which resulted in the co-determined work arrangement management
action. Our participants in DeliVER shed light on this management action:
“Empowerment is more helpful than control and penalty during the challenging time. We believe that emphasizing and engaging in
relational work with our riders can benefit us for the long term.”
- DeliVER Digital Operation Manager
“I know that DeliVER measures various aspects of our performance but it is good that they also consider our freedom to determine which
order I want to accept or reject. … That is why I signed up for this job, this platform, which is supposed to be flexible. We are not supported
to work 9 to 5 or feel forced to accept all orders sent to us.”
- Delivery Rider I

Fig. 1. A Process Model of Humanistic Algorithmic Management.

10
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

The outcome of Synergizing was the realization of “co-determined equitable work relationships”, signifying a profound trans­
formation in the platform’s workforce dynamics. These thriving relationships were marked by reciprocity, fairness, transparency, and
worker empowerment, which fostered trust, job satisfaction, and strengthened organizational identity, benefiting both the platform
and its delivery riders. This helped DeliVER navigate the complexities of a health crisis, ensuring the sustainability and success of the
on-demand food delivery platform. For delivery riders, having the power to decline tasks without implicit or explicit penalties or
prejudice enhances their work motivation. In addition, providing delivery riders with opportunities to question, discuss, or contest
automated algorithmic management decisions helped them maintain adequate levels of job autonomy. Our delivery rider informant
explained what they had been experiencing in new work relationships:
“My relationship with DeliVER has evolved into something that is more collaborative and equitable. We feel valued and respected
because we have a say on how things are done. It is not just about blindly following algorithms, it is about working together to make our
jobs better and fairer. When we have the freedom to decline tasks without fear and our concerns are heard and addressed, it is a win–win
situation for all of us. We are more motivated, and that makes our work better.”
- Delivery Rider G

Discussion

A process model of humanistic algorithmic management

By integrating our findings with the theoretical view of complementarity (Huber et al., 2013; Nevo and Wade, 2010), our process
model of humanistic algorithmic management in Fig. 1 depicts a transition from mechanistic algorithmic management (i.e., the most
rudimentary form of algorithmic management, where algorithms replace human management) to humanistic algorithmic manage­
ment, which involves replacing and dampening, compensating, enabling, and synergizing. The transition reveals how human management
can progressively complement algorithmic management to create better outcomes for crowd workers, in terms of support, empow­
erment, and equitable relationships. The transition is triggered by various drivers, leading to progressively better outcomes.
Mechanistic algorithmic management refers to a system in which managerial tasks and decision-making processes are predominantly
automated and driven by algorithms (Möhlmann and Henfridsson, 2019; Nevo and Kotlarsky, 2020). By contrast, human management
emphasizes the roles of human judgment, empathy, and adaptability in managing a workforce. This involves human managers who
actively engage with employees, understand their unique needs, and provide support, guidance, and oversight (Durward et al., 2020;
Taylor and Joshi, 2019). Humanistic algorithmic management hence refers to the interaction between human and mechanistic algo­
rithmic management, which leads to better work motivation and performance (Möhlmann and Henfridsson, 2019). The process model
represents a journey from mechanistic to humanistic algorithmic management. The process model starts with substitution, then
progresses to compensation, following by prioritization of worker well-being and co-determined equitable work relationships. This
transformative approach balances operational efficiency with worker empowerment, establishing a resilient ecosystem for both the
platform and its riders.
Although the transactional aspects of crowd work are evident, our study suggests that economic gains-centered work relationships
should be carefully reconsidered by linking humanistic management with algorithmic management (e.g., Wood et al., 2019). This adds
to the growing body of evidence suggesting that crowd workers are increasingly seeking a more co-determined and relational work
arrangement in addition to seeking opportunities for social interaction (Kittur et al., 2013; Petriglieri et al., 2018). The humanistic
form of algorithmic management uncovered in our study entails a greater extent of organizational support, reciprocity, and social
interactions than conventional approaches, and we suggest that it should be at the core of the working relationship between crowd
workers and the platforms they operate on.
As algorithms become increasingly important to organizations, humanistic algorithmic management represents a forward-thinking,
humanistic approach to managing crowd workers through a digital platform (Lehdonvirta, 2018). This provides a possible answer to
the critical question of how to make algorithms more human-centered (Faraj et al., 2018). Humanistic algorithmic management in­
troduces the interpersonal and empathetic aspects of people management into algorithmic control. Because the humanistic approach
enhances crowd workers’ work motivation, reduces labor turnover, and ultimately improves job performance, it sustains the in­
teractions between a platform, its management, and its workers from a long-term perspective (Fountaine et al., 2019).
The process model of humanistic algorithmic management paves the way for a more sustainable and worker-centric future. This
approach signifies a profound shift in the dynamics of work relationships, gradually moving from substitution to co-determined
equitable work relationships. As crowd workers’ expectations continue to evolve, especially in the face of unexpected crises such as
the COVID-19 pandemic, this model underscores the importance of incorporating humanistic elements into algorithmic management.
It anticipates that future workplaces will need to embrace a holistic approach that offers not only operational efficiency, but also
worker empowerment, support, and a sense of reciprocity (Brockman et al., 2019). This evolution towards humanistic algorithmic
management recognizes the collaborative potential of human-technology partnerships in elevating the overall well-being and job
satisfaction of crowd workers while fostering resilient ecosystems that benefit both the platform and its delivery riders (Kuhn and
Maleki, 2017). Essentially, it shapes the future workplace by integrating the best of human and algorithmic management, ultimately
redefining work relationships and nurturing a more sustainable and worker-friendly work environment.

11
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

Mechanisms of substitution and complementarity

In our study, a mechanism refers to a systematic and organized set of actions that describes how a particular transformation is
produced or brought about (Henfridsson and Bygstad, 2013; Wynn Jr and Williams, 2012). Mechanisms provide a detailed and
explanatory account of the causal pathways that lead to specific outcomes. As illustrated in Fig. 1, mechanisms of substitution
emphasize the shift towards automated processes where algorithmic management replaces and dampens traditional human man­
agement functions, such as scheduling, monitoring, and performance evaluation management actions. Conversely, complementarity
mechanisms – compensating, enabling, and synergizing – consist of five humanistic management actions (i.e., flexible performance
management, algorithmic transparency, team and peer support, adaptive behavioral guidance, and co-determined work arrange­
ments). These actions involve collaboration between human and algorithmic management.
Specifically, mechanisms of substitution in algorithmic management elucidate the process through which algorithmic elements
substitute human management components, resulting in a shift from human-led to algorithm-driven processes. Within mechanisms of
substitution, replacing involves identifying functional equivalents between algorithmic and human management functions. It sheds
light on how specific responsibilities that have traditionally fallen under human management have been gradually supplanted by
automated algorithms. A natural consequence of this replacement process is dampening, which explores the potential effects of
algorithmic management in diminishing the advantages inherent in human management (Huber et al., 2013). It delves into how the
increasing presence of algorithms – marked by continuous surveillance and automation – can impact human-centric facets of man­
agement, such as trust and support, often resulting in a shift towards work in isolation (Gerber and Krzywdzinski, 2019).
On the other hand, mechanisms of complementarity describe how algorithmic and human management can work together,
enhancing each other’s strengths and mitigating their weaknesses. They involve the integration of human judgment, adaptability, and
empathy with algorithmic decision-making. Hence, our model further suggests that humanistic algorithmic management complements
algorithmic management by providing additional facilitating conditions and incapacitating the weaknesses of purely operational
algorithms. Human actions aim to complement, circumvent, or support algorithmic actions, and adapt algorithms to address human
and social needs. Specifically, three types of human management actions compensate for algorithmic management, each precipitating
different implications for the focal platform. The management action of flexible performance management softens algorithmic evaluation
and discipline. Our model suggests that an on-demand food delivery platform must go beyond the mechanical accuracy driven by
algorithms and account for the emotions and attitudes of crowd workers. Similarly, enhanced algorithmic transparency helps delivery
riders develop a better understanding of and build more confidence in algorithmic technologies. The action of team and peer support
ameliorates the sense of isolation stemming from algorithmic management (Kurtessis et al., 2017).
Furthermore, algorithmic and human management create favorable conditions that enable the other. The management of adaptive
behavioral guidance empowers crowd workers with care and support through the capabilities of automatic resource coordination and
big data analytics enabled by algorithmic systems. Additionally, while enabling involves providing support and resources to enhance
the effectiveness of human work, synergizing goes beyond that to create a combined effect where the whole is greater than the sum of
its parts. This deep collaboration between algorithmic and human management generates synergistic outcomes that lead to equitable,
rewarding, and effective relationships. This is realized in the management action of co-determined work arrangements, which move from
managerial or app control to work-platform co-determined task arrangements, and more fully realize the advantages of human-
algorithm collaboration for enhanced job autonomy and motivation (Graham et al., 2017).
These five management actions in humanistic algorithmic management redefine evaluation, direction, and support (Kellogg et al.,
2020). Flexible performance management emphasizes adapting algorithmic evaluation and disciplinary practices with accountability.
Algorithmic transparency enhances efficient task direction, benefiting both riders and customers. These actions show that algorithmic
controls vary in nature and implications, influencing technostressors (Cram et al., 2022). Humanistic adaptation of algorithmic control
is crucial to address issues from datafication and reductionist analytics (Faraj et al., 2018; Galliers et al., 2017). Our study shifts from
efficiency-driven management to providing tailored support for crowd workers through holistic governance (Gerber and Krzywdzinski,
2019). Crowd workers require relational work arrangements, skill development opportunities, and social interaction (Duggan et al.,
2020).

Theoretical contributions

Our study offers several key theoretical contributions to the literature. First, we contribute to the algorithmic management liter­
ature by developing a process model of humanistic algorithmic management. This model explains the transition from mechanistic
algorithmic management, where algorithms replace human management, to humanistic algorithmic management. Prior research
highlights the negative effects of algorithmic management (e.g., Cram et al., 2022; Kuhn and Maleki, 2017), but there is limited
knowledge on how to reconfigure human management practices to establish supportive, reciprocal, and sustainable employer-worker
relations (Günther et al., 2017; Kellogg et al., 2020).
Our model moves beyond a managerial control perspective by explaining how human management complements the strategic
value of algorithmic management (Kellogg et al., 2020). It emphasizes the evolution of work relationships from initial mechanistic
control to increasing human management involvement, focusing on worker well-being and empowerment. This leads to the creation of

12
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

co-determined, equitable work relationships. The model deepens our understanding of the interplay between algorithmic and human
management, reflecting changing worker expectations and a desire for balanced and relational work arrangements. It underscores the
importance of a human-centered approach, particularly during crises such as the pandemic, and offers organizations guidance on
building a resilient and worker-focused workplace.
Our study also addresses this issue by theorizing five types of humanistic management actions to explain how managers adjust
algorithmic decision-making for performance management, provide algorithm explanations, offer support and personalized guidance,
and create a flexible and rewarding relational work arrangement. It posits a dynamic process in which human management practices
tackle algorithmic dehumanization triggered by societal and operational challenges. It also suggests that the synergy created through
human and algorithm collaboration leads to work arrangements with greater efficiency and autonomy. Our findings demonstrate that
humanistic algorithmic management is effective in achieving both the performance outcome of operational efficiency. They also show
that the humanistic outcomes of workers’ increased trust in algorithms, enhanced organizational identity, heightened job motivation,
and enhanced well-being (Brockman et al., 2019).
Second, our findings enhance the theoretical understanding of complementarity and substitution by examining the interactions
between algorithmic and human management. The complementarity perspective encompasses established causal mechanisms: sub­
stitution involves replacing and dampening, while complementarity includes compensating, enabling, and synergizing (Huber et al., 2013;
Nevo and Wade, 2010). We adopt a dynamic view of complementarity and substitution, challenging and refining the causal re­
lationships between algorithmic and human management. Our findings reveal the dynamics between algorithmic and human man­
agement, following a trigger-mechanism/action-outcome logic. We identify four mechanisms − replacing and dampening, compensating,
enabling, and synergizing − that underlie the dynamic complementarity of human and algorithmic management. This study is one of few
initial studies that explore these dynamics and theorize their sub-mechanisms, responding to recent calls for algorithmic management
for crowd workers (Cram et al., 2022; Möhlmann et al., 2021; 2023; Tarafdar et al., 2022).
In addition, the synergizing mechanism we uncover enriches the theoretical perspective on complementarity by suggesting that the
‘mutual compatibilities’ of two entities involved in a complementary relationship lead to synergistic effects (Nevo and Wade, 2010, p.
170). In our research, mutual compatibilities manifest in the congruence between the humanistic management mindset of DeliVER
managers and workers and the technological capabilities of its platform. This differs from existing complementary mechanisms,
namely, compensating (one addresses the weakness of the other) and enabling (one creates favorable conditions for the other) (Huber
et al., 2013). Furthermore, we illustrate contextual events triggering changes in the dynamics between human and algorithmic
management.
Third, we contribute to the crowd work literature by elucidating the roles of human management in the context of algorithmic
management (e.g., Deng et al., 2016; Gol et al., 2019; Taylor and Joshi, 2019). While the crowd digital platform model offers effi­
ciency, previous studies have raised ethical and social concerns about the power imbalance between the platform and workers, which
could lead to exploitation and hinder the growth of the platform economy (e.g., Brabham, 2012; Kittur et al., 2013; Kuhn and Maleki,
2017). Our humanistic algorithmic management approach bridges this gap by explaining how managers can empower workers and
avoid marginalization, leading to increased acceptance of algorithmic management and improved job motivation and performance. By
investigating how to balance empowerment and marginalization in crowd work management, our study addresses a critical research
gap and provides early insights into mechanisms for crowd work governance (Gol et al., 2019). This includes mitigating negative
factors such as lack of social interactions, insufficient organizational support, and opaque performance evaluation to create a more
equitable and rewarding work environment – an imperative during public crises like COVID-19.
Fourth, our study contributes to the discourse on the IS phenomenon during COVID-19 by developing an understanding of how
human management, algorithmic systems, and crowd workers collaborate to effectively address societal and operational challenges
(Ågerfalk et al., 2020; Karanasios, 2022). Researchers argue that businesses must strive for a new and better normal post-COVID, as the
pre-COVID business-as-usual reproduces conditions that hinder effective responses to virus outbreaks and reinforce structural in­
equalities (Brammer et al., 2020). In other words, firms must prepare and adapt to cope with socially disruptive events. Our humanistic
algorithmic management model suggests that societal challenges become operational challenges for businesses during public health
crises, with human management mitigating the dehumanizing effects of algorithmic management. The model has broader implications
for crowd work businesses in proactively responding to socially disruptive events.

Implications for practice

From a practical standpoint, our research highlights the possibility of digital platform providers gaining strategic advantage by
engendering a more humanistic and effective form of algorithmic management. This may be achieved through purposeful reengin­
eering of their platforms to address the widely observed challenges of retention and supervision of crowd work (Duggan et al., 2020).
In doing so, this study corroborates the call for increased flexibility in algorithmic performance management, importance of the human
dimension, and provision of worker support. For instance, companies can consider naming an “algorithmic auditor”, who is respon­
sible for the monitoring, control, auditing, and alteration of prejudicial or erroneous algorithmic decisions.
Our identified action of enhancing algorithmic transparency could also clarify the ‘rationale for the selection, inclusion, exclusion,
or optimization of various inputs or outputs to the algorithm’ (Diakopoulos and Koliska 2016, p. 817). For instance, companies can

13
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

collect feedback from their workers’ experiences with the algorithm through surveys or focus groups, which they can analyze and
integrate into algorithms and work design. Furthermore, our study emphasizes the importance of preserving human involvement,
which can address the concerns and resistance workers may have when subject to algorithms or automated decision-making systems.
This study also offers practical implications for the pervasiveness of algorithmic management during the pandemic and potentially
the post-pandemic world. The concomitant rise of remote working, big data, and human resource analytics may further precipitate the
removal of the more interpersonal and empathetic aspects of people’s management. As such, our humanistic approach to algorithmic
management contributes to the current discourse in the industry by articulating how working relationships can be managed in ways
that address concerns surrounding power imbalances, fairness, and worker well-being, and by doing so in a manner that places value
on the development of a deep relationship of employees with its associated levels of commitment (Durward et al., 2020; Healy et al.,
2017). We also highlight the need for policy-informing bodies to develop policies that support the well-being of workers engaged in
algorithmic work as well as maintain reciprocity and rewards in such work relationships.

Limitations and future research

Although this study offers valuable insights into humanistic algorithmic management within the context of the gig economy,
several limitations should be acknowledged. First, it relies on a single case study of an online food delivery platform during the COVID-
19 pandemic. Although this in-depth exploration is informative, it may limit the generalizability of our findings. Future research could
use cross-case comparisons across multiple platforms or industries to validate the proposed framework. Second, future research could
include longitudinal studies to track the evolution of humanistic algorithmic management over extended periods. This approach
provides insights into sustainability and adaptability in dynamic and evolving operational environments. Additionally, future research
could investigate how humanistic algorithmic management practices vary based on the sophistication of algorithmic systems and
nature of worker tasks. A comparative analysis of companies that successfully implement humanistic algorithmic management with
those that do not reveal key success factors and challenges in promoting more humane gig platforms. Third, as our findings suggest
how algorithmic and human management practices can nicely strengthen and complement each other. Future research could explore
potential conflicts and tensions that arise when integrating algorithmic and human management practices. These investigations might
examine how differences in approach, values, or expectations impact organizational dynamics and performance. Understanding these
challenges could lead to effective strategies for managing and resolving conflicts in this evolving landscape.

Conclusion

The dehumanization of algorithmic management has contributed to the our human marginalization and inequality of crowd work
(Marjanovic et al., 2022; Marabelli et al., 2021). This issue has been heightened during the COVID-19 pandemic as lockdown strategies
across many countries have confined people to their homes and transformed on-demand delivery platforms and riders into essential
workers (Pryor, 2020). Through an investigation of an on-demand food delivery service provider and its delivery riders, our study
illuminates the challenges faced by the platform provider company. We develop a new approach to algorithmic management in
response to these challenges and fulfil the new societal mission, which we term humanistic algorithmic management.
Our study presents an in-depth understanding of what human managers should do when algorithmic management decisions are
insufficient or ineffective, as well as how to amplify the positive aspects of both human and algorithmic management, with a focus on
creating an equitable, rewarding, and effective work environment. This is especially crucial for coping with challenges associated with
the unfolding pandemic. We expect our theorization and empirical findings to set the stage for researchers and practitioners to
investigate and demonstrate how firms can create a more holistic form of value (e.g., social and ethical), as opposed to just economic
value, through humanistic algorithmic management.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper. This study was supported by National Natural Science Foundation of China (72172011).

14
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

Appendix

Appendix A
Summary of Empirical Research on Algorithmic Management of Crowd.

Source Research Context Main Findings

Rosenblat and Case study of Uber’s drivers • The information and power asymmetries produced by the Uber
Stark (2016) application are fundamental to its ability to structure control over its
workers.
• The rhetorical invocations of digital technology and algorithms are
used to structure asymmetric corporate relationships to labor, which
favor the former.
Deng and Joshi Case study of 55 crowd workers registered on Amazon • Crowdsourcing task characteristics (e.g., job autonomy, task variety,
(2016) Mechanical Turk (MTurk) task significance, task instruction, and task compensation) and a
digitally enabled environment (e.g., system affordance and MTurk
governance) to shape crowd workers’ continued participation in
crowdsourcing work environment.
Deng et al. (2016) Case study of 210 crowd workers participating in Amazon’s • Nine values shared by crowd workers share: access, autonomy, fairness,
Mechanical Turk (MTurk) transparency, communication, security, accountability, making an
impact, and dignity are uncovered.
• These values are implicated in four crowdsourcing structures:
compensation, governance, technology, and microtask. Two
contrasting perceptions—empowerment and marginalization—coexist,
forming a duality of microtask crowdsourcing.
Möhlmann and Case study of Uber’s drivers • The emergent tension between algorithmic management and workers’
Zalmanson autonomy is identified based on power asymmetry.
(2017) • Workers respond to and resolve these tensions by actively trying to
egain some of their lost control and sense of autonomy.
Griesbach et al. Case study of 55 food delivery crowd workers and survey of • Although all of the food delivery platforms use algorithmic
(2019) 955 food delivery crowd workers management to assign and evaluate work, there is significant cross-
platform variation.
• Instacart, the largest grocery delivery platform, exerts a type of control
called “algorithmic despotism,” regulating the time and activities of
workers more stringently than other platform delivery companies.
Gronsund and Case study of organizational members in an international • Humans and the algorithm are configured and reconfigured in multiple
Aanestad maritime trade organization that introduced automated ways over time as the organization dealt with the introduction of
(2020) algorithmic support for data analysis and prediction work algorithmic analysis.
• In contrast to replacing human work, the emergent configurations
require new roles and redistribution of extant expertise to augment and
improve the accuracy of the algorithm.
Möhlmann et al. Case study of Uber’s drivers, Uber executives and engineers • In the context of both algorithmic matching and algorithmic control,
(2021) crowd workers experience tensions relating to execution,
compensation, and belonging.
• These tensions trigger market-like and organization-like response be­
haviors by crowd workers.
Rahman and Case study of client managers and contractors in software • When managers use the tools to exert coercive control over contractors
Valentine development “gigs”on a platform labor market acting unexpectedly, it backfires and contributes to uncompleted
(2021) project outcomes.
• In contrast, when they refrain from using the tools for coercive control
in such circumstances and instead engage in collaborative repair, their
actions contribute to completed project outcomes.
Cram et al. (2022) Survey study of 621 US-based Uber drivers • Gatekeeping and guiding algorithmic control positively relate to both
challenge and threat technostressors.
• The technostressors have implications for continuance intentions and
workaround use.
• Gig economy organizations can use algorithmic control to enhance
challenge technostressors for their workers, thereby contributing to the
cultivation of a more committed workforce.
• Algorithmic control transparency may not always mitigate the negative
effects of threat technostressors.
Möhlmann, et al. Case study of Uber drivers and supplemented it by • Platform workers make sense of the algorithms managing them through
(2023) interviewing Uber executives and customers (n = 46). (1) focused enactment, (2) selection modes, and (3) retention sources.
• The multistep process of algorithm sensemaking allows platform
workers to keep up with algorithmic instructions systematically.
Tarafdar, et al. Case study of interviews with 15 Uber drivers, a netnographic • Algorithm is a multi-role sender with entangled roles, while the human
(2023) study of 1700 discussion threads among Uber drivers from two as role taker experiences algorithm-driven role conflict and role
popular online forums, andanalysis of Uber’s web pages ambiguity.
• Algorithm records all of the human’s task actions, it is ignorant of the
human’s cognitive reactions and undergoes ‘broken loop learning’.

15
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

Appendix B
Interview Guide.

Stakeholder Group Interview Subjects Number of Interviewees Sample Interview Questions

Crowd task Delivery Riders with varying levels of 23 (17 in preparatory phase from February • What are the challenges you have faced in
Workers experience ranging from 1 months to 3 to June 2020 and 6 in follow-up phase terms of using the platform?
years from February to March 2023) • Have you obtained support from others to
cope with the challenges? If so, what type of
support?
• Have you experienced any changes in relation
to the company’s management practices? If
so, what are those changes?
• What are the roles or relationships between
algorithmic and human management?
• What are the outcomes you have experienced
or achieved from using the platform?
Management of Chief Information Officer, Head of 10 (6 in preparatory phase from February • What are the challenges you have faced in
Digital Operations, Digital Operations to June 2020 and 4 in follow-up phase terms of managing your workforce?
Platform Manager, and 3 Senior Data Analysts from February to March 2023) • What are the key practices in relation to
managing by algorithm that you have
established to cope with these challenges?
• What is the rationale behind those practices?
• What are the roles or relationships between
algorithmic and human management?
• What are the outcomes you have experienced
or achieved from those practices?

16
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

Appendix C. . Sample data structure and supporting evidence

Fig. C1. Data Structure

17
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

Table C1
Supporting Evidence.
First-Order Concept Sample Interview Quotes

Enabling Automatic Resource Coordination, Reducing • “We have been investing in building algorithms to help determine the best order scheduling and
Human Labor Involvement address the vehicle routing problem. … This has been our long-term investment and has helped us
automatically match customer orders with the closest delivery riders. … We efficiently scaled our
services to include more restaurants, delivery riders, and customers. We also scaled by introducing
some new delivery services.” - DeliVER Chief Information Officer
Monitoring Riders Remotely, Rewarding and • “These algorithms provide us data analytics and automatic decision-making capabilities to remotely
Punishing Riders Automatically monitor our delivery riders. … This replaced the staff required to continuously evaluate delivery
riders’ performance and give them feedback. …” - DeliVER Chief Information Officer
Restricting Social Interactions • “As an app-based gig economy company, we do not have the traditional organizational structure or
culture. Our delivery riders are not like traditional organizational employees who work together in
offices. So that, they do not have many opportunities to engage in social interactions with each other or
with our management team.” - DeliVER Head of Operations
Softening Evaluation and Discipline Practices • “They are generally employed on a casual basis, and would be unable to claim sick leave if they are
infected, or exposed to the virus. We offer compensation to delivery riders, whom diagnosed with the
coronavirus or placed in quarantine for up to 14 days.” - DeliVER Head of Operations
• “We are known as an app company, but we do not want to be directed by an app. The capacity to
influence or exert some form of human managerial control over the system evaluation and disciplinary
practices is what we want to achieve now.” - DeliVER Head of Operations
• “We were told we are not penalized by the rating systems for taking steps to protect our health. So I do
not struggle with my need to protect myself against the rating score I might get.” - Delivery Rider I
Allocating Human “Algorithmic Auditors” • “To overcome the aversion reaction sometimes developed by algorithm users, our delivery riders …
We, as the management team, need to be empowered to change or intervene in an algorithmic
decision. The rules should be set by us and not automatically by the algorithms. We assign staff to
reassess the decisions made by the algorithms. … Even a very small amount of power of intervention or
modification can significantly change how our riders are evaluated.” - DeliVER Digital Operations
Manager
• “If we got a low average customer rating, we would not be immediately deactivated on the platform. …
It was me, but I heard someone got contacted by DeliVER because of a low rating. … I think it is a kind
of protection to us as casual workers.” - Delivery Rider H
Disclosing Updated Information of Algorithms • “Transparency is even more important during the pandemic. … Our new performance management
initiatives aim to reduce information asymmetries for riders and facilitate them to work better with the
systems. … We regularly reach out to delivery riders by email to keep them updated on new
regulations and policies that may affect the algorithmic instructions they receive.” - DeliVER Head of
Operations
• “There was minimal recourse for delivery riders who seek to get clarity on what is going on “behind the
scenes”. This is common for the on-demand delivery sector. But eventually, we need to do the right
thing, which means acting with transparency and accountability.” - DeliVER Digital Operations
Manager
• “We got email updates of the terms and conditions. You know how it is like reading these long
statements. Usually, we do not read it. … But now with the pandemic, I started to read these emails and
attend information sessions because they keep us updated on what I should do and how to do it
appropriately.” - Delivery Rider L
Providing Information and Training Sessions • “At times, we had trouble observing what leads to certain instructions on the app… for example, I was
not sure about what caused the temporary blocking of my account, not obeying the traffic rules may be
a reasonable explanation. … these trainings are helpful for us to make sense of the app.” - Delivery
Rider G
• “It was compulsory for all delivery riders to attend an introductory training session when we sign up on
the platform. … This helped us understand the app instructions and how it works to evaluate us.” -
Delivery Rider E
Establishing Dedicated Support Team • “The technology is one side, the people on wheels is another. We set up a special team to support our
delivery riders during the pandemic, so they can contact us directly. As riders tend to work in isolation
without organizational identification; everything is mediated by our app platform. We aim to add
more interpersonal and emotional aspects of people management during this challenging time to
everyone.” - DeliVER Digital Operations Manager
• “So, we started it by building our identity with a square yellow food delivery box. … It is good. It was
like, you know whenever our delivery riders pass and see others, they will immediately know they
work for DeliVER. … We want them to know they are part of an organizational community, there are
support team, support staff with them…” - DeliVER Head of Operations
Building Online Communication Channels • “I used to feel my boss is the app, my colleague is the app…but now during the pandemic, I run a
WeChat group [a messaging and social media mobile app] with 285 DeliVER riders, and we are all
connected.” - Delivery Rider C
• “Delivery job is like an one-person alone job when we are riding. … Now it is good that we have a
service hotline with a dedicated support team for us. … I called once when I had a vehicle breakdown. I
also enjoy mingling with other riders in online communities, airing our frustration and gossiping.” -
Delivery Rider L
Detecting Behaviour Abnormalities • “We recently rolled out a contact-free delivery feature. Over the last two months, our tech and safety
teams have been working hard to provide additional Health and Safety guidance to its restaurant
(continued on next page)

18
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

Table C1 (continued )
First-Order Concept Sample Interview Quotes

partners and riders to ensure they have access to the most up-to-date information available.” - DeliVER
Digital Operations Manager
• “During the pandemic and all the city was in quarantine, our delivery riders were out for delivery with
increased demands. So, it is our responsibility to make sure they are safe. … These algorithmic nudges
were used to keep delivery riders working longer hours, but not now. … they can be used to advise
them not working overtime.” - DeliVER Digital Operations Manager
Sending Individualized Advice • “What we want to do is to match the COVID-19 hotspots released by the Department of Health with our
real-time geolocation data of individual riders. So that we can send an alert to them when they are
around these areas.” - DeliVER Senior Data Analyst A
• “DeliVER sent us in-app messages and advice from public health authorities reminding us of basic steps
we can take to help protect ourselves.” - Delivery Rider K
• “Every time when we check in in the morning, we were advised that our priority was our safety and
well-being. ... We got an app alert with protection tips and certain hotspots. … I think these are good
planning and good for us during the pandemic.” - Delivery Rider B
Lessening Controlled Work Arrangement • “Empowerment is more helpful than control and penalty during the challenging time. We believe that
emphasizing and engaging in relational work with our riders can benefit us for the long term.” -
DeliVER Digital Operation Manager
• “The same rejected order may come again and again. Essentially, we have to ignore it to decline. …
Now we were told we are not penalized by rejecting orders when we are unwell. Our health and safety
are the priority.” - Delivery Rider G
• “I know that DeliVER measures various aspects of our performance but it is good that they also
consider our freedom to determine which order I want to accept or reject. … That is why I signed up for
this job, this platform, which is supposed to be flexible. We are not supported to work 9 to 5 or feel
forced to accept all orders sent to us.” - Delivery Rider I
Involving Riders for Having a Voice • “We invite our delivery riders in with these quality improvement sessions. … They should have more
participation, equality, and respect in the working relationship.” - DeliVER Head of Operations
• “It is good to have people in the comapny who listen to us, … for our feedback and comments.” -
Delivery Rider E

References

Aaltonen, A., Tempini, N., 2014. Everything counts in large amounts: a critical realist case study on data-based production. J. Inf. Technol. 29 (1), 97–110.
Ågerfalk, P.J., Conboy, K., Myers, M.D., 2020. Information systems in the age of pandemics: COVID-19 and beyond. Eur. J. Inf. Syst. 1–5.
Barker, S. (2020). Gartner updates 2020 future of work trends for post-pandemic world. [Link]
post-pandemic-world.
Becker, B.E., Huselid, M.A., 2006. Strategic human resources management: where do we go from here? J. Manag. 32 (6), 898–925.
Biernacki, P., Waldorf, D., 1981. Snowball sampling: Problems and techniques of chain referral sampling. Sociol. Methods Res. 10 (2), 141–163.
Brabham, D.C., 2012. The myth of amateur crowds: A critical discourse analysis of crowdsourcing coverage. Inf. Commun. Soc. 15 (3), 394–410.
Brammer, S., Branicki, L., Linnenluecke, M., 2020. COVID-19, Societalization and the Future of Business in Society. Acad. Manag. Perspect. 34 (4), 493–507.
Brockman, P., Campbell, J.L., Lee, H.S., Salas, J.M., 2019. CEO internal experience and voluntary disclosure quality: Evidence from management forecasts. J. Bus.
Financ. Acc. 46 (3–4), 420–456.
Bunge, M., 2004. How does it work? The search for explanatory mechanisms. Philos. Soc. Sci. 34 (2), 182–210.
Corbin, J., Strauss, A., 2015. Basics of Qualitative Research: Techniques and Procedures for Developing Grounded Theory, fourth ed. Sage Publications, Thousand
Oaks, CA.
Cram, W.A., Wiener, M., Tarafdar, M., Benlian, A., 2022. Examining the impact of algorithmic control on Uber drivers’ technostress. J. Manag. Inf. Syst. 39 (2),
426–453.
Darke, P., Shanks, G., Broadbent, M., 1998. Successfully completing case study research: combining rigour, relevance and pragmatism. Inf. Syst. J. 8 (4), 273–289.
Deng, X., Joshi, K.D., 2016. Why Individuals Participate in Micro-Talk Crowdsourcing Environment: Revealing Crowdworkers’ Perceptions. J. Assoc. Inf. Syst. 17 (10),
648–673.
Deng, X., Joshi, K.D., Galliers, R.D., 2016. The duality of empowerment and marginalization in microtask crowdsourcing: Giving voice to the less powerful through
value sensitive design. MIS Q. 40 (2), 279–302.
Diakopoulos, N., Koliska, M., 2016. Algorithmic transparency in the news media. Digit. J. 5 (7), 809–828.
Duggan, J., Sherman, U., Carbery, R., McDonnell, A., 2020. Algorithmic management and app-work in the gig economy: A research agenda for employment relations
and HRM. Hum. Resour. Manag. J. 30 (1), 114–132.
Durward, D., Blohm, I., Leimeister, J.M., 2020. The nature of crowd work and its effects on individuals’ work perception. J. Manag. Inf. Syst. 37 (1), 66–95.
Faraj, S., Pachidi, S., Sayegh, K., 2018. Working and organizing in the age of the learning algorithm. Inf. Organ. 28 (1), 62–70.
Fountaine, T., McCarthy, B., Saleh, T., 2019. Building the AI-powered organization. Harv. Bus. Rev. 63–73.
Gal, U., Jensen, T.B., Stein, M.K., 2020. Breaking the vicious cycle of algorithmic management: A virtue ethics approach to people analytics. Inf. Organ. 30 (2),
100301.
Galliers, R.D., Newell, S., Shanks, G., Topi, H., 2017. Datification and its human, organizational and societal effects. J. Strateg. Inf. Syst. 26 (3), 185–190.
Gerber, C., Krzywdzinski, M., 2019. Brave new digital work? New forms of performance control in crowdwork. Work and Labor in the Digital Age. Emerald Publishing
Limited.
Gerring, J., 2009. Case selection for case-study analysis: Qualitative and quantitative techniques. The Oxford Handbook of Political Methodology. [Link]
[Link]/view/10.1093/oxfordhb/9780199286546.001.0001/oxfordhb-9780199286546-e-28.
Gol, E.S., Stein, M.K., Avital, M., 2019. Crowdwork platform governance toward organizational value creation. J. Strateg. Inf. Syst. 28 (2), 175–195.
Graham, M., Hjorth, I., Lehdonvirta, V., 2017. Digital labour and development: impacts of global digital labour platforms and the gig economy on worker livelihoods.
Transfer: Eur. Rev. Labour Res. 23 (2), 135–162.
Griesbach, K., Reich, A., Elliott-Negri, L., Milkman, R., 2019. Algorithmic control in platform food delivery work. Socius 5, 1–15.
Grønsund, T., Aanestad, M., 2020. Augmenting the algorithm: Emerging human-in-the-loop work configurations. J. Strateg. Inf. Syst. 29 (2), 101614.

19
T. Cui et al. Journal of Strategic Information Systems 33 (2024) 101838

Günther, W.A., Mehrizi, M.H.R., Huysman, M., Feldberg, F., 2017. Debating big data: A literature review on realizing value from big data. J. Strateg. Inf. Syst. 26 (3),
191–209.
Harrower, K., 2019. Algorithmic decision-making in organizations: Network data mining, measuring and monitoring work performance, and managerial control.
Psychosociol. Issues Hum. Resour. Manag. 7 (2), 7–12.
Healy, J., Nicholson, D., Pekarek, A., 2017. Should we take the gig economy seriously? Labour Ind. 27 (3), 232–248.
Henfridsson, O., Bygstad, B., 2013. The generative mechanisms of digital infrastructure evolution. MIS Q. 37 (3), 907–931.
Huber, T.L., Fischer, T.A., Dibbern, J., Hirschheim, R., 2013. A process model of complementarity and substitution of contractual and relational governance in IS
outsourcing. J. Manag. Inf. Syst. 30 (3), 81–114.
Karanasios, S., 2022. The pursuit of relevance and impact: A review of the immediate response of the information systems field to COVID-19. Inf. Syst. J. 32 (4),
856–887.
Kehoe, R.R., Wright, P.M., 2013. The impact of high-performance human resource practices on employees’ attitudes and behaviors. J. Manag. 39 (2), 366–391.
Kellogg, K.C., Valentine, M.A., Christin, A., 2020. Algorithms at work: The new contested terrain of control. Acad. Manag. Ann. 14 (1), 366–410.
Kittur, A., Nickerson, J. V., Bernstein, M. S., Gerber, E. M., Shaw, A., Zimmerman, J., Lease, M., & Horton, J. J. (2013). The future of crowd work. In: Proceedings of
the 16th ACM Conference on Computer Supported Cooperative Work, January 2013, San Antonio, Texas.
Klein, H.K., Myers, M.D., 1999. Evaluating Interpretive Field Studies in Information Systems. MIS Q. 23 (1), 67–94.
Kuhn, K.M., Maleki, A., 2017. Micro-entrepreneurs, dependent contractors, and instaserfs: Understanding online labor platform workforces. Acad. Manag. Perspect.
31 (3), 183–200.
Kurtessis, J.N., Eisenberger, R., Ford, M.T., Buffardi, L.C., Stewart, K.A., Adis, C.S., 2017. Perceived organizational support: A meta-analytic evaluation of
organizational support theory. J. Manag. 43 (6), 1854–1884.
Langley, A., 1999. Strategies for theorizing from process data. Acad. Manag. Rev. 24 (4), 691–710.
Lehdonvirta, V., 2018. Flexibility in the gig economy: managing time on three online piecework platforms. N. Technol. Work. Employ. 33 (1), 13–29.
Luthans, F., Youssef, C.M., 2004. Human, social, and now positive psychological capital management: Investing in people for competitive advantage. Organ. Dyn. 33
(2), 143–160.
Marabelli, M., Newell, S., Handunge, V., 2021. The lifecycle of algorithmic decision-making systems: Organizational choices and ethical challenges. J. Strateg. Inf.
Syst. 30 (3), 101683.
Marjanovic, O., Cecez-Kecmanovic, D., Vidgen, R., 2022. Theorising algorithmic justice. Eur. J. Inf. Syst. 31 (3), 269–287.
McAfee, A., Brynjolfsson, E., 2017. Machine, Platform, Crowd: Harnessing Our Digital Future. W. W. Norton, New York.
Möhlmann, M., & Zalmanson, L. (2017). Hands on the wheel: Navigating algorithmic management and Uber drivers. In: Proceedings of the 38th International
Conference on Information Systems, South Korea.
Möhlmann, M., Henfridsson, O., 2019. What people hate about being managed by algorithms, according to a study of Uber drivers. Harv. Bus. Rev. 30 (August), 1–7.
Möhlmann, M., Zalmanson, L., Henfridsson, O., Gregory, R.W., 2021. Algorithmic management of work on online labor platforms: When matching meets control. MIS
Q. 45 (4), 1999–2022.
Möhlmann, M., de Lima, A., Salge, C., Marabelli, M., 2023. Algorithm sensemaking: how platform workers make sense of algorithmic management. J. Assoc. Inf. Syst.
24 (1), 35–64.
Nevo, D., Kotlarsky, J., 2020. Crowdsourcing as a strategic is sourcing phenomenon: Critical review and insights for future research. J. Strateg. Inf. Syst. 29 (4),
101593.
Nevo, S., Wade, M.R., 2010. The formation and value of IT-enabled resources: antecedents and consequences of synergistic relationships. MIS Q. 34 (1), 163–183.
Newell, S., Marabelli, M., 2015. Strategic opportunities (and challenges) of algorithmic decision-making: A call for action on the long-term societal effects of
‘datification’. J. Strateg. Inf. Syst. 24 (1), 3–14.
Pan, S.L., Tan, B., 2011. Demystifying case research: A structured–pragmatic–situational (SPS) approach to conducting case studies. Inf. Organ. 21 (3), 161–176.
Petriglieri, G., Ashford, S., Wrzesniewski, A., 2018. Thriving in the gig economy. Harv. Bus. Rev. 140–143.
Pryor, E., Riordan, T., Hoffstaedter, G., Robinson, R., 2020. Delivery workers are now essential. They Deserve the Rights of Other Employees. [Link]
com/delivery-workers-are-now-essential-they-deserve-the-rights-of-other-employees-134406.
Rahman, H.A., Valentine, M.A., 2021. How Managers Maintain Control Through Collaborative Repair: Evidence from Platform-Mediated “Gigs”. Organ. Sci.
forthcoming.
Rai, A., Constantinides, P., Sarker, S., 2019. Editor’s comments: next-generation digital platforms: Toward human-AI hybrids. MIS Q. 43 (1), iii–ix.
Rodgers, W., Murray, J.M., Stefanidis, A., Degbey, W.Y., Tarba, S.Y., 2023. An artificial intelligence algorithmic approach to ethical decision-making in human
resource management processes. Hum. Resour. Manag. Rev. 33 (1), 100925.
Rosenblat, A., Stark, L., 2016. Algorithmic labor and information asymmetries: A case study of Uber’s drivers. Int. J. Commun. 10, 3758–3784.
Sharma, R., Mithas, S., Kankanhalli, A., 2014. Transforming decision-making processes: A research agenda for understanding the impact of business analytics on
organisations. Eur. J. Inf. Syst. 23 (4), 433–441.
Strauss, A., Corbin, J., 1998. Basics of Qualitative Research: Techniques and Procedures for Developing Grounded Theory, second ed. Sage, Thousand Oaks, CA.
Tan, B., Pan, S.L., Chen, W., Lihua, H., 2020. Organizational sensemaking in ERP implementation: The influence of sensemaking structure. MIS Q. 44 (4), 1773–1809.
Tarafdar, M., Page, X., Marabelli, M., 2023. Algorithms as co-workers: Human algorithm role interactions in algorithmic work. Inf. Syst. J. 33 (2), 232–267.
Taylor, J., Joshi, K.D., 2019. Joining the crowd: The career anchors of information technology workers participating in crowdsourcing. Inf. Syst. J. 29 (3), 641–673.
Tiwana, A., 2010. Systems development ambidexterity: Explaining the complementary and substitutive roles of formal and informal controls. J. Manag. Inf. Syst. 27
(2), 87–126.
Urquhart, C., 2022. Grounded theory for qualitative research: A practical guide. Sage.
Urquhart, C., Lehmann, H., Myers, M.D., 2010. Putting the ‘theory’back into grounded theory: guidelines for grounded theory studies in information systems. Inf. Syst.
J. 20 (4), 357–381.
Walsham, G., 1995. Interpretive case studies in IS research: Nature and method. Eur. J. Inf. Syst. 4 (2), 74–81.
Wiener, M., Cram, W., Benlian, A., 2021. Algorithmic control and gig workers: a legitimacy perspective of Uber drivers. Eur. J. Inf. Syst. 1–23.
Wood, A.J., Graham, M., Lehdonvirta, V., Hjorth, I., 2019. Good gig, bad gig: autonomy and algorithmic control in the global gig economy. Work Employ Soc. 33 (1),
56–75.
Wynn Jr, D., Williams, C.K., 2012. Principles for conducting critical realist case study research in information systems. MIS Q. 36 (3), 787–810.

Dr Tingru Cui is a Senior Lecturer with the School of Computing and Information Systems at the University of Melbourne. Her research interests are interdisciplinary in
nature, exploring human-AI collaboration, responsible AI, business and learning analytics, and digital health. Her work has been published in journals such as Journal of
Management Information Systems, Information Systems Journal, Journal of Strategic Information Systems and Information & Management.

Dr Barney Tan is the Head of School and a Professor at the School of Information Systems and Technology Management (SISTM) of UNSW Business School. His research
interests include strategic information systems, enterprise systems implementation, electronic commerce, Chinese IT management and qualitative research methods. His
work has been published in journals such as Management Information Systems Quarterly, Information Systems Research, and Information Systems Journal.

Dr Yunfei Shi is a Lecturer at the School of Information Systems & Technology Management, Business School, UNSW Sydney. She obtained her PhD degree at the
University of Queensland in 2017. She undertakes research in areas including digital innovation and entrepreneurship as well as theories and methods in the IS
discipline, specifically with a focus on critical realist research. Her work has been published in journals such as Journal of Strategic Information Systems.

20