=Paper=
{{Paper
|id=Vol-3901/short_5
|storemode=property
|title=Addressing Deskilling as a Result of Human-AI Augmentation in the Workplace
|pdfUrl=https://ceur-ws.org/Vol-3901/short_5.pdf
|volume=Vol-3901
|authors=Firuza Huseynova
|dblpUrl=https://dblp.org/rec/conf/tethics/Huseynova24
}}
==Addressing Deskilling as a Result of Human-AI Augmentation in the Workplace==
https://ceur-ws.org/Vol-3901/short_5.pdf
Addressing Deskilling as a Result of Human-AI
Augmentation in the Workplace
Firuza Huseynova 1
1
University of Western Ontario, 1151 Richmond St., London, N6A 3K7, Canada
Abstract
The integration of artificial intelligence (Al) technology into the workplace has become a focal point of
discussion in recent years. While management scholars have typically advocated for an approach that
augments rather than substitutes human labour, little consideration has been given to the potential
drawbacks of human-AI augmentation strategies. Thus, this paper focuses on addressing employee
deskilling that can arise from human-AI augmentation in the workplace. Drawing on insights from eight
semi-structured interviews with experts across industry and academia, this paper details two key factors
influencing managerial decision-making in augmentation projects and touches on the blurred lines between
augmentation and substitution strategies—especially when augmentation erodes the satisfaction of human
work.
Keywords
Artificial intelligence, augmentation, automation, technological unemployment, deskilling, Al ethics, Al
governance
1. Introduction
Recent advances in artificial intelligence (Al) technology have led to a greater capacity for machines
to replicate nonroutine and nonrepetitive tasks which were once considered impossible to automate
[2, 9]. This has sparked significant debate on the role of automated systems in the workplace,
particularly around issues of technological unemployment, deskilling, and task encroachment [3, 4].
When it comes to the integration of Al into the workplace, management scholars typically advocate
for an approach that complements and augments human labour rather than substitutes it [1],
However, the increasing potential for Al to augment humans in more and more types of work has
made it important to study human-AI augmentation from a technology ethics perspective. Thus, my
research question is as follows: How does deskilling arise in the workplace when Al is deployed to
augment rather than substitute human labour?
This paper will begin with an overview of the broader debate between augmentation and
substitution, leading to a discussion on deskilling in the workplace. I will then outline the
methodology used to conduct eight semi-structured interviews with industry professionals and Al
researchers, setting the stage for a discussion of key findings. These key findings center around the
types of deskilling seen in the workplace, the key factors influencing managerial decision-making
around automation, and the blurred lines between augmentation and substitution. I will close the
paper with two recommendations for employers, synthesized from interview insights, on mitigating
potential negative consequences of deskilling in the workplace.
2. Background
2.1. The Augmentation vs. Substitution Debate
Among management scholars, the relationship between augmentation and substitution is usually
described as a trade-off [3]. Automating a task (used interchangeably with ‘substituting’ for the
purposes of this paper) involves handing it over to a machine with little to no human involvement,
usually for the sake of more efficient or productive operations [1]. On the other hand, augmenting a
7th Conference on Technology Ethics (TETHICS2024), November 6-7, 2024, Tampere, Finland
Q fhuscyno@uwo.ca
© 2024 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
task implies introducing heavier human-machine involvement, leading to an ostensibly
complementary balance between human and machine capabilities.
In a review of three New York Times best-selling books on the intersection of Al and business,
Raisch and Krakowski found that management scholars typically advocate for an augmentation
approach rather than resorting to substitution in the form of mass layoffs [1]. In public
communications, spokespeople for tech companies such as Microsoft and Google frequently state that
companies will use Al in the workforce to complement human abilities rather than replace or restrict
them [1]. IBM CEO Ginni Rometty suggested using the term “augmented intelligence” instead of
“artificial intelligence”, driving home the point that Al implementation should be aimed at
augmenting the existing capabilities of humans rather than replacing them [1].
Additionally, according to a participatory study conducted with 54 knowledge workers across
seven fields, workers tend to hold favourable views on human-in-the-loop (HITL) approaches, which
aim to ensure an element of human oversight is integrated within workplace augmentation strategies
[10]. Proponents of the augmentation view generally argue that humans and Al should coexist in the
workplace; if humans do not learn to work with Al, they may risk falling behind the skill curve.
2.2. Deskilling
Although the impact of emerging technologies on human skillsets is very ambiguous and difficult to
quantify, potential deskilling as a result of human-AI augmentation should concern ethicists. The
concept of deskilling was formed in the 20th century to explain how automation leads to a loss of
practical knowledge and artisan skill sets such as looming or tilemaking [5], Outside of technical
skills, Shannon Vallor’s research in moral deskilling suggests that moral skills are just as vulnerable
to disruption by technological advancements as other types of skills [5], On an aggregate level,
deskilling is linked with technological unemployment, which, in its structural form, has the potential
to disrupt individual well-being, social cohesion, and entire economic systems [4], The negative
consequences of deskilling from both an individual and societal level make it crucial to consider
human-AI augmentation in the workplace a legitimate subject matter to study for technology
ethicists.
Machines are increasingly encroaching into tasks which require complex cognitive and emotional
capacities, previously only possible by humans. Task encroachment refers to the widening of tasks
that machines will be able to perform, from mere manual tasks to more cognitive and emotional ones.
[2]. Researchers typically distinguish between three types of Al systems: mechanically-intelligent
(manual), thinking-intelligent (cognitive) and feeling-intelligent (emotional) [6]. Manual tasks include
inspecting and operating equipment, controlling machines, scheduling work or activities, handling or
moving objects, recording information, and performing administrative activities [6], Cognitive tasks
include analyzing data or information, processing information, developing objectives or strategies,
thinking creatively, interpreting the meaning of information for others, solving problems, and
providing consultation or advice [6], Lastly, emotional work involves communicating with peers or
clients, assisting and caring for others, resolving conflicts, negotiating with others, developing or
building teams, training or teaching other people, and establishing interpersonal relationships [6],
Today, cognitive and emotional tasks which were once considered impossible to automate have been
automated, leading to deskilling in areas like customer service and nursing [5].
Due, in part, to task encroachment, workers in many industries are facing a redundancy of core
skills they trained for years to develop [5, 10], This is partly because businesses focus on automating
away tasks to achieve cost efficiencies in the short run, which signals industry competitors to pursue
substitution (ie. lay workers off) in order to remain cost-competitive and secure their position in the
market [10], However, in the long run, employees may lose some of the skills required to oversee and
alter the processes of Al-enabled tools— or, lose their jobs altogether.
The term ‘technological unemployment’, popularized by John Meynard Keynes in the early 20th
century, can be divided into its frictional and structural forms [2, 12], Frictional technological
unemployment occurs when workers shift between jobs in the pursuit of finding their ideal
occupation [2], Short-term frictional unemployment is considered natural and healthy in a market-
based economy [2], However, structural technological unemployment is a more long-term form of
unemployment resulting from a mismatch between the skills workers possess and the skills
employers are looking for [2], Unlike previous eras in history where new technologies eventually
created more jobs than they displaced, the current pace of technological advancement, particularly in
Al and machine learning, raises concerns about the ability of humans to upskill quickly enough [4],
The potential impact of structural technological unemployment extends beyond economic measures—
it can exacerbate inequality, strain social safety nets, and erode the dignity of work [4],
While deskilling is typically a term applied to workers who have lost their jobs due to automation,
it can also apply to workers who have experienced their work be augmented by AL In this paper, I
argue that it is misleading to present the difference between augmentation and substitution as a clean-
cut trade-off. The lines between augmentation and substitution are blurred, and deskilling can also
occur in workers augmented by Al, as will be discussed further on.
3. Methodology
3.1. Research Strategy
This paper relies on an inductive approach, wherein theories and interpretation are the outcome of
theory [7j. As to be expected when conducting inductive analysis, the central research question
changed multiple times throughout its journey. There was no central theory guiding the development
of research, but the research process was approached from a theory-building perspective [7j. Thus,
data was gathered and examined first and only then were theories crafted from observations.
3.2. Literature Search
The literature search was primarily collected through two databases: Google Scholar and Western
University’s OMNI database. Thanks to the advanced search engine capabilities of both databases, a
thorough search was done on combinations of keywords including: “Al”, “deskilling”, “automation”,
“augmentation”, “substitution”, and “task encroachment”.
Since the field of information systems is highly interdisciplinary, it was essential to search for
studies outside the field to uncover as much relevant research as possible, including fields like
business ethics, sociology, and human-computer interaction. To obtain high-quality literature, results
were filtered on peer-reviewed academic materials and journals. As well, given the high-paced growth
of new technologies, priority was given to articles published in the last five years. Lastly, publications
with a higher number of citations were prioritized, as this metric was taken as a proxy for how
credible and influential an article was in its respective field.
3.3. Interviews
To add depth to research findings, semi-structured interviews were conducted with two target
groups: managers with experience in AI-implementation projects, and researchers focused on Al
ethics. A semi-structured interview is one in which the interviewer is able and willing to switch the
order of questions asked or ask unique questions to each interviewee depending on their experiences
[7j. As this project’s inductive investigation was based on asking open questions, semi-structured
interviews were chosen to allow for flexibility and adaptability during the interview process.
A total of eight interviews were conducted. Interviewees were primarily selected based on their
experience in the field. See Table 1 for a complete list of interviewees along with their credentials
(names eliminated for anonymity).
Table 1
Interviewee Overview
Number Experiences
II Chief Al Officer at major Canadian university
12 CEO of soft skills training company, doctoral candidate researching Al and society
13 Independent journalist, host of award-winning tech podcast
14 Research and policy analyst at Responsible Al Institute (RAI)
15 Al strategist at Big 5 Canadian bank, alumnus of Oxford Internet Institute
16 ________________Doctoral candidate at McGill University researching Al governance
17 ____________Doctoral candidate at University of Toronto researching Al ethics and policy
18 Senior digital workforce transformation consultant at Big 4 audit firm
3.4. Interview Protocol
Before commencing each interview, consent was established to record audio and take notes during
the interview. Interviewees were informed that the collected material would be processed
anonymously and confidentially to safeguard identities. The interview protocol contained six base
questions, although the questions were adjusted liberally to align with the interviewee’s relevant
experiences and the overall direction of the discussion. The questions are listed below:
1. What is your name, what does your organization do, and what is your role in the organization?
2. Are you involved in any Al implementation projects? If so, which projects are you involved in?
3. What do you think are the most important factors a business leader should consider when making
decisions about automating tasks currently performed by humans?
4. An alternative to substituting labour is using technology to augment people's work. But, even
when work is augmented, skills can be eroded. How do you decide which skills people need to
keep and which ones are okay to lose?
5. Which type of workers are more likely to lose skills when Al is introduced to augment their
workforce?
6. What are new skills that workers or employees need to develop to adapt to the adoption of Al in
the organization?
3.5. Limitations
When it comes to selecting a research topic, one’s decision about what to investigate always precedes
the methodological issues of how the research might best continue [7j. It is undoubtable that many
researchers hold values driving them to choose their research topic, especially in the case of a topic
related to ethics, as mine is. However, I do not subscribe to the view that research should be values-
free; on the other hand, I believe value-commitment is a good thing for researchers to have as they
can use their passions to direct and interpret their investigations while remaining open about the
types of biases they may have. In my free time, I mainly consume content (ie. books, podcasts, and
videos) that is critical of Big Tech and technological determinism. Thus, my paper reflects this
perspective. However, I believe I have developed a fair and balanced paper that considers the
perspectives of techno-optimists and tech skeptics alike.
Still, one limitation of the paper is the semi-structured nature of the interviews led to a lack of
standardization among many interview questions. Although valuable insights were gained from each
individual interview, it was difficult to compare and contrast answers given to any of the questions
in a statistically significant way. However, the qualitative nature of the data allowed for a deeper
exploration of diverse perspectives, ultimately providing a nuanced understanding of the impact of
Al on the workforce.
4. Findings
4.1. Context-Dependency
12: “In scholarship, the goal is new knowledge creation. I’m not going to copy and paste exactly
what GPT gives me into a paper because that is not new knowledge. And that’s not the goal of
academia. However, I will copy & paste at [soft skill development startup] because that is not the
goal. The goal is that [clients] walk away having learned a skill. And if GPT can make that easier
for us, sign me up.”
While there were significant differences in the way each interviewee approached the questions, a
common theme was the importance of context-dependency when deciding which tasks to automate
away. Specifically, interviewers 1 and 4 (hereby referred to as II and 14) mentioned that a one-size-
fits-all approach would not be optimal for dynamic work environments, meaning automation
strategies would need to be tailored to suit the specific requirements of different departments and
teams. 12 also noted that the goals of a particular sector (private, public, non-profit, or academia) could
help determine whether Al should be used to augment tasks like writing and content generation.
Moreover, 18, a senior Digital Workforce Transformation consultant, mentioned that the size of a
company is a key factor in their willingness to adopt human-AI augmenting technologies in the first
place. They stated that a small startup with a leader who is tapped into the tech space will be much
more likely to opt into human-AI augmentation in the workforce, as opposed to a company that has
been operating with engrained processes for 50+ years, which may be more reluctant or risk-averse
to testing new technologies in the workplace. Overall, the impact of automation can vary significantly
based on the unique environment, industry, workforce composition, and strategic objectives of a
given company.
4.2. Types of Deskilling
15: “This is like the futurist’s dream, right? The machines take care of all the automated, mindless
work. And then we as humans can be creative and focus on other things. Will that happen in
actuality? I don’t know. It’s not a very black or white thing. The answer will lie somewhere in
between.”
Firstly, most responses indicated that deskilling in areas of manual labour and mechanical data-entry
is largely positive since it frees humans to focus on more meaningful cognitive and emotional work.
Specifically, 15 stated that skills in calculation or summarization will likely be devalued, but this will
only lead to a greater capacity for humans to focus on honing in their analytical and problem-solving
skills. This belief is consistent with the optimistic view of human-AI augmentation on the labour
market, being that it will create new jobs as the demand for nonroutine and nonrepetitive skills
increases [6j. Thus, labour has the potential to ‘upgrade’ from being mechanical to cognitive.
However, when it comes to the more creative manual skills that may be lost in this transition (ie.
artisanal human craftwork), 12 argued that there will always be a market for authenticated time-
consuming human work. They bring up the example of jewelry: buying a ring from someone who
spent 50 hours forging and faceting a gem is completely different than buying something stamped
out in a factory. They believe that there will always be demand for human-crafted language, code,
and prose as we value things just for the sake of being made by humans.
Lastly, 18 noted the importance of the quality of Al tools available in determining how deskilled a
workforce becomes. If workers have high-quality Al tools that can summarize information or obtain
insights more effectively, they are more likely to become deskilled at completing those tasks. This is
simply because if a worker does not have to read an entire document to understand complex
terminology or summarize it, then they are not practicing reading comprehension and
summarization—essentially becoming deskilled in those tasks.
4.3. Key Factors Influencing M a n a g e r i a l Decision-Making
4.3.1. Time and Effort
II: “Use the technology to skill up the people who are in place and say — look, save your brain cells
for the really hard cases. Let the Al handle the easy cases.”
A common theme among responses is the use of human-AI augmenting technologies to improve
individual productivity and efficiency by using Al to save time and effort in the workplace. Al
strategist 15 specifically brought up insights from a project leveraging Al to help banking advisors do
their jobs more effectively by using generative Al to save them time. For example, tools are
implemented to assist with summarizing notes or other long-form content, consolidate sources of
data, and use analytics to generate insightful insights about customer relationships.
They outlined a matrix which can be useful when making decisions about which tasks to automate,
augment, or retain (Figure 1). By using this matrix, organizations can strategically assess their
employee’s tasks and retain human involvement in areas where unique human skills and judgement
are essential for optimal performance.
Figure 1:
Task Automation-Augmentation Matrix
High Effort
Augment Retain
Low High
Time Time
Retain Substitute
Low Effort
The matrix categorizes tasks into four quadrants based on their time consumption and level of effort
required.
A. Low-Time, High-Effort Tasks (Augment): Tasks that fall into this category require significant
effort but can be completed relatively quickly. These tasks are ideal candidates for augmentation
through technology to make them easier for employees to perform.
B. High-Time, High-Effort Tasks (Retain): These are tasks that demand both substantial effort
and time. They are best suited to be retained by human workers as they involve complex decision-
making, creativity, or interpersonal skills that are challenging to replicate for a machine.
C. Low-Time, Low-Effort Tasks (Retain): These tasks are typically quick and easy to accomplish,
requiring minimal cognitive or physical exertion. While low-time, low-effort tasks may not be
intellectually demanding, they can serve as a source of quick wins and tangible accomplishments
for employees.
D. High-Time, Low-Effort Tasks (Substitute): Tasks in this quadrant are characterized by low-
effort requirements but consume a significant amount of time to complete.
From this matrix, we see that deskilling via augmentation or substitution is most likely to occur
with high-time, low-efforts tasks (such as data entry, basic administrative tasks, or updating a
comprehensive spreadsheet) and low-time, high-effort tasks (such as writing a detailed project
proposal or developing a complex financial model).
4.3.2. Managerial Control
13: “I’m not anti-technology - I think technology has an important role to play in improving
human society. But so often, under the kind of economic and social system we live in, technologies
are developed by companies or by the military in order to serve particular interests. And those
interests are not aligned with improving the rights of workers and the power workers have in
society.”
The idealistic promise of technology was to reduce working hours and free up humans to focus on
more creative work while raising living standards and average wealth [8]. However, some
interviewees noted that technology is instead being implemented to stifle worker’s rights.
Specifically, 13 described the rollout of algorithmic management, wherein a growing number of
workers are subject to algorithms informing them how to do their jobs rather than forming a close
relationship with a manager. They brought up the example of Uber, a company that claimed to be
rolling out an innovative and disruptive approach to transportation, but whose real innovation was
to attack and decimate the rights that workers had in delivering that service. They mentioned that at
the core of Uber’s model, beyond the ability to bypass traditional regulation, is an algorithmic system
that is shaping and determining how drivers work. Responses from 16 and 17 also indicated that
human-AI augmenting technologies are being employed not only to reduce costs, but to increase
control over workers.
4.4. Blurred Lines Between Augmentation and Substitution
4.4.1. Augmentation as a First Step Towards Substitution
A common theme among interviews was the lack of clear boundaries between augmentation and
substitution, as augmentation can often be the first step towards a full-on substitution strategy. 17
brought up the example of an Al-based tool that was brought in to augment a team of HR managers
at JP Morgan Chase in order to identify predictors of a job candidate’s potential job performance.
After a full year of interaction between the Al tool and the human experts with the goal of mitigating
biases in predictions, JP Morgan Chase decided to fully automate away candidate assessment— and
lay off a number of managers in the process. The bank’s justification for removing humans from this
activity was to increase the level of fairness and standardization among candidate assessments while
making the process more efficient. In this example, an Al model developed to augment human work
ended up being considered sufficiently robust enough to work autonomously without the assistance
of a human, thereby showing how the augmentation of certain tasks can eventually lead to the
automation of whole jobs altogether.
15: “When it comes to augmentation, the top 1% of your ■workforce is already performing really
well because they're doing certain things. What Al does is that it helps the average part of your
workforce do the things that your top 1% is already doing, but more easily.”
Moreover, as Al augments human workers to make them more productive and efficient, fewer
workers may be needed as the same level of output can be achieved using fewer inputs. 15 notes that
the bank they consult for does not aim to lay people off, but rather to implement technologies and
cut the bottom tier of performers due to natural churn in the workplace. So, as a company loses the
bottom rung of performers, they will notice that even if they do not replace the workers, they are still
able to retain the same level of output as before. However, a company may not hire back for the
positions they cut as a result of implementing AI-augmenting technology, indicating the first signs
of structural technological unemployment.
4.4.2. Augmentation Eroding the Satisfaction of Human Work
Another complication in human-AI augmentation arises when Al erodes the satisfaction of human
work for the sake of improved productivity and efficiency. Specifically, 13 brought up a hypothetical
example of a graphic designer who has spent years building up their skills in illustration and
typography, to the point of being confident that they are an excellent designer. However, in the
precarious world of graphic design, companies augment human work by generating a design using
Al and hiring a human to edit it or enhance it to a level that is presentable to the wider public [10],
Thuus, companies end up devaluing certain skills by using Al to churn out something that is nearly
good enough to use commercially, and later hiring a human to fix the details before they can properly
present it. Under this model, technologies may not primarily be deployed to replace humans, but to
change (and often erode) the way human work is done.
5. Recommendations
In terms of recommendations for employers to reduce the negative effects of deskilling in the
workplace, two main suggestions were brought up in interviews: 1) to increase the diversity of the
teams developing and deploying Al, and 2) to include employees in the decision-making process of
human-AI augmentation projects.
Firstly, there is a need to bring in more diverse perspectives when deciding to automate away
certain tasks to Al. 13 brought up how the people developing Al tools are engineers and computer
scientists who hold a particular view of the world and a particular idea of what skills are important.
Conversely, the types of people who are making art and writing fiction full-time are not the types of
people who typically sit at the heads of tech companies and help make decisions about what their
tools are going to do. Despite recent efforts at increasing diversity in the corporate world, there is
still a lack of representation of marginalized groups in the development and deployment of Al [11],
Thus, 18 mentioned the importance of having a team that is diverse “in all senses of the word” to
mitigate some of the biases and blind spots that are in place when employing technologies developed
by a typical group of software engineers.
Secondly, Al governance researcher 16 mentioned the importance of consulting workers before
rolling out an augmentation strategy in the workplace. They recommend holding town hall meetings
with one’s workforce and asking employees how Al has changed their day-to-day experiences at
work, as well as their fears surrounding Al. An effective strategy 16 has seen is not a top-down or
bottom-up approach, but somewhere in the middle where companies can have a conversation with
employees while genuinely listening to their concerns. Although 18 highlights that employee
consultation is a lengthy process, it allows organizations to gain insights into the employee’s unique
needs, challenges, and concerns regarding Al implementation in the workplace to ensure the skills
they value are being retained.
6. Conclusion
To conclude, the integration of human-AI augmenting technologies in the workplace has led to an
ambiguous landscape of challenges concerning deskilling, technological unemployment, and Al
ethics. While it is common for managers to push for an augmentation strategy rather than a
substitution strategy, this paper has argued that deskilling can also occur as a result of human-AI
augmentation.
Through conducting eight semi-structured interviews with industry professors and Al
researchers, two key factors influencing managerial decision-making around automation were
uncovered: time/effort (translating to greater productivity, efficiency, and profit) and managerial
control. When it comes to augmenting human labour, multiple types of deskilling can occur, from
manual to cognitive to emotional. As well, the blurred lines between augmentation and substitution
can mean that employers use augmentation as a first step towards a full-on augmentation strategy.
Finally, workers should be wary that the conditions of their work may be remade by a company
implementing Al in the workplace to potentially reduce their pay, the rights they have at work, and
the power they have in the workplace. While augmentation is peddled as the be-all-end-all solution
to societal problems around automation, we must question the notion that implementing workplace
technology is an inherent good, and instead think about whether these technologies are being rolled
out in a pro-worker way.
Declaration on Generative Al
The author has not employed any Generative Al tools.
References
[1] Raisch, Sebastian, and Sebastian Krakowski. 2021. “Artificial Intelligence and Management: The
Automation-Augmentation Paradox: Academy of Management Review.” Academy of
Management Review 46 (1): 192-210. https://doi.org/10.5465/amr.2018.0072.
[2] Susskind, Daniel. 2021. “Technological Unemployment.” In The Oxford Handbook of Al
Governance, edited by Justin B. Bullock, Yu-Che Chen, Johannes Himmelreich, Valerie M.
Hudson, Anton Korinek, Matthew M. Young, and Baobao Zhang, 0. Oxford University Press.
https://doi.org/10.1093/oxfordhb/9780197579329.013.42.
[3] Tschang, Feichin Ted, and Esteve Almirall. 2021. “Artificial Intelligence as Augmenting
Automation: Implications for Employment.” Academy of Management Perspectives 35 (4): 642-
59. https://doi.org/10.5465/amp.2019.0062.
[4] Kim, Tae Wan, and Alan Scheller-Wolf. 2022. “Technological Unemployment, Meaning in Life,
Purpose of Business, and the Future of Stakeholders.” In Business and the Ethical Implications of
Technology, edited by Kirsten Martin, Katie Shilton, and Jeffery Smith, 13-31. Cham: Springer
Nature Switzerland. https://doi.org/10.1007/978-3-031-18794-Q 2.
[5] Vallor, Shannon. 2015. “Moral Deskilling and Upskilling in a New Machine Age: Reflections on
the Ambiguous Future of Character.” Philosophy & Technology 28 (1): 107-24.
https://doi.org/10.1007/sl3347-014-Q156-9.
[6] Huang, Ming-Hui, Roland Rust, and Vojislav Maksimovic. 2019. “The Feeling Economy:
Managing in the Next Generation of Artificial Intelligence (Al).” California Management Review
61 (4): 43-65. https://doi.org/10.1177/0008125619863436.
[7] Berdahl, Loleen, and Jason J. Roy. 2021. “Explorations: Conducting Empirical Research in
Canadian Political Science.” Oxford University Press.
[8] Hughes, Carl, and Alan Southern. 2019. “The World of Work and the Crisis of Capitalism: Marx
and the Fourth Industrial Revolution.” Journal of Classical Sociology 19 (1): 59-71.
https://doi.org/10.1177/1468795X18810577
[9] Bernard Marr. 2023. A Short History of ChatGPT: How We Got To Where We Are Today. Forbes.
https://www.forbes.com/sites/bernardmarr/2023/05/19/a-short-history-of-chatgpt-how-we-got-
to-where-we-are-today/?sh=286bl544674f
[10] Woodruff, Allison, Renee Shelby, Patrick Gage Kelley, Steven Rousso-Schindler, Jamila Smith-
Loud, and Lauren Wilcox. 2024. “How Knowledge Workers Think Generative Al Will (Not)
Transform Their Industries.” In Proceedings of the CHI Conference on Human Factors in
Computing Systems, 1-26. Honolulu HI USA: ACM. https://doi.org/10.1145/3613904.3642700.
[11] Collett, Clementine, and Sarah Dillon. 2019. “Al and Gender: Four Proposals for Future
Research.” Apollo - University of Cambridge Repository. https://doi.org/10.17863/CAM.41459.
[12] Keynes, John Maynard. 1930. “Economic Possibilities for Our Grandchildren.” In Essays in
Persuasion, edited by John Maynard Keynes, 321-32. London: Palgrave Macmillan UK.
https://doi.org/10.1007/978-l-349-59072-8 25.