The Need to Collaborate: Opportunities for
Human and AI Co-workers
Anja Hornikel, Christian Greiner
a
and Thomas Peisl
b
Department of Business Administration, Munich University of Applied Sciences, Munich, Germany
Keywords: Artificial Intelligence, Collaboration, Grounded Theory, Technology Acceptance, Augmented Intelligence.
Abstract: Researchers studying the development of modern economies and their workforce agree that the work of
professionals may change profoundly in the future. Technology is seen as the main driver of this change,
especially machine learning and artificial intelligence (AI). The biggest benefits can be most likely achieved
by complementary use of human and AI capabilities and intelligence. This qualitative study investigates what
potential collaboration concepts can look like and how collaboration is evaluated by the human. The aim is to
identify potential, beneficial collaboration concepts with AI and to gain a better understanding of the
influencing factors on user acceptance. The results show that the evaluation of potential collaboration appears
to be a process including two phases. In general, many different aspects influence the evaluation of the
collaboration concept in this process, but not all aspects seem to have an effect at the same time. 10 qualitative
interviews are conducted and to narrow the scope of this research the focus lies on academic professionals,
namely knowledge workers such as consultants.
1 INTRODUCTION
The future of work is assumed to change profoundly
for many academic professionals and knowledge
workers, e.g. consultants, especially in the way of
how they will provide services to their customers in
the future. Technology is seen as the main driver of
this development and looking into the future these
professionals need to work differently. One of the
challenges consultants are facing is the economic
problem, that many cannot afford their services. The
services delivered to their customers are perceived as
inefficient, too costly and the appreciation of their
expertise has declined. Questions which arise in this
context are for example if there might be new and
very different ways to organize professional work to
make services more affordable, accessible, and even
increase the quality of the results. A new division of
labour seems necessary and technology, with AI
being one example, can be the key to rethink task
allocation. (Susskind & Susskind, 2017)
The impact which this revolutionary technology
will have on the professions and the way people work
is uncertain and widely discussed. Skilton and
a
https://orcid.org/0000-0002-8184-7128
b
https://orcid.org/0000-0001-5571-2089
Hovsepian (2018) state that fusion is key and that
human and machine intelligence are becoming
increasingly entangled indicating that complementary
use of human and AI capabilities most likely contains
the biggest benefits. Still human professionals will
not be replaced entirely by technology. (Fügener et
al., 2019; Lichtenthaler, 2018; Poortmans et al., 2019)
Regarding these promising prospects, it seems not
surprising that many executives and leaders are
viewing AI as a great opportunity which needs to be
exploited. When trying to implement AI initiatives,
companies struggle rather with human resistance than
with technical difficulties (Schlögl et al., 2019). In
addition, the selection of suitable AI technologies and
specific use cases for the application are challenging
as many AI applications are available, but
complimentary use seems difficult (Bauer & Vocke,
2019).
In this research on collaboration concepts with AI
the focus lies on knowledge workers (e.g. consultants)
and how their work may change due to technology. If
this highly professionalized group of experts can
collaborate successfully with AI and is willing to do so
it could imply that most other professionals could do
Hornikel, A., Greiner, C. and Peisl, T.
The Need to Collaborate: Opportunities for Human and AI Co-workers.
DOI: 10.5220/0010643700003060
In Proceedings of the 5th International Conference on Computer-Human Interaction Research and Applications (CHIRA 2021), pages 139-148
ISBN: 978-989-758-538-8; ISSN: 2184-3244
Copyright
c
2021 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
139
so as well. Looking at consultants’ tasks it can be
agreed that they consist mostly of gathering,
processing, and interpreting data which could also be
done by AI with the potential to achieve even better
results. Therefore, the question is if or when CEOs will
turn to intelligent systems to ask for advice rather than
consultants. (Libert & Beck, 2017)
Futurists predict a third of jobs may be eliminated
through technology, but little research has been
conducted on how employees perceive technological
change (Brougham & Haar, 2018). Academic
research on human-AI collaboration focuses rather on
technical aspects or theoretical frameworks. This
research studying potential collaboration concepts
using AI in white collar jobs is relevant to close the
research gap and contribute to future business success
of professionals.
2 CONTEXT
The state of current research in the field of human-AI
collaboration shows different opportunities of
collaboration between human and AI co-workers. In
decision-making for example, Colson (2019)
proposes to evolve from data-driven to AI-driven
decision processes. To fully leverage the value of data
it can be suitable for routine decisions based on
structured data to rely on AI only to eliminate
human’s cognitive bias. Often business decisions do
not solely rely on structured data but also qualitative
insights and additional information. Better decisions
can be made by finding ways to leverage both humans
and AI and create case-specific workflows. For
example, AI can be used to generate different
possibilities based on data and the human can pick the
best alternative using the additional information it has
access to. (Colson, 2019) This approach could be
described as hybrid intelligence (Dellermann et al.,
2019) or augmented intelligence (Rao, 2017). In
terms of collaboration, an international group of
researchers were the first to develop a set of
algorithmic mechanisms which can learn and
collaborate with humans as well as with other
algorithms. (Breazeal, 2003; Crandall et al., 2018;
Dautenhahn, 2007; Kamar et al., 2013) This research
has proven that collaboration with an algorithm is
possible on a level comparable to that with another
human and gives insights on the potential of
intelligent, autonomous systems as teammates.
How potential collaboration could be
implemented in terms of applications was researched
by Bittner et al. (2019). Instead of trying to copy the
human brain, to overcome limitations of AI, the
authors argue that the most valuable approach would
be to combine the capabilities of human and AI agents
to minimize each other’s weaknesses. This view is
supported by Fügener et al. (2019). Their study
showed that collaboration and delegation between
humans and AI can produce results that outperform
humans or AI alone. Huang and Rust (2018) studied
the potential impact of AI on the service industry and
developed a theory of how AI may replace jobs. This
theory supports the conclusions of Fügener et al.
(2019) and Bittner et al. (2019) that the distribution
of tasks is essential for job sharing of humans and AI.
In subsequent research, Huang and Rust (2019)
focused on investigating and proving the emergence
of the Feeling Economy assuming that the importance
of feeling tasks, compared to mechanical and thinking
tasks, will increase. This development will mean that
human workers and AI need to work as a team with a
task allocation matching the task requirements and
respective strengths. AI will take over most thinking
tasks while the human worker focuses on feeling
tasks and interaction with others. Other studies
focused on certain job profiles. Sowa and
Przegalinska explored possible synergies between
(2020) human workers in managerial positions and
AI-powered computer systems while the research
conducted by Wang et al. (2019) aimed at
understanding future impacts automated AI
applications may have on data scientists. In this
research the general opinion was quite optimistic as a
collaborative approach of data science work in the
future using human and AI expertise was seen as most
promising. (Wang et al., 2019) The authors
mentioned above outline how collaboration can be
brought to live in certain industries and jobs, but it
remains unclear how human-AI collaboration
concepts can be applied regarding the tasks of a
knowledge worker.
As the attitude of employees toward technology
and their acceptance of technological systems play an
important role in the adoption of AI systems this
related research stream needs to be included. The
Technology Acceptance Model (TAM) developed by
Davis (1989) is based on prior research by Fishbein
and Ajzen’s (1975) Theory of Reasoned Action (TRA)
and focused on the topic of acceptance and use of
information technologies. Davis based the model on
the assumption that the attitude of a potential user
toward using a certain system is the major determinant
of the actual system use (Davis, 1993). This attitude is
influenced by the perceived usefulness of the system
and its perceived ease of use. The TAM was used as a
foundation for the further development of TAM2, in
which theoretical constructs regarding social
CHIRA 2021 - 5th International Conference on Computer-Human Interaction Research and Applications
140
influences and cognitive aspects were included
preceding perceived usefulness (Venkatesh & Davis,
2000). With the TAM3, Venkatesh and Bala (2008)
focused on the question how managers can support
better acceptance and utilization of new IT systems
(Venkatesh & Bala, 2008). To provide a unified view
on user acceptance Venkatesh et al. (2003) reviewed
and compared eight existing theories which resulted in
the formulation of the Unified Theory of Acceptance
and Use of Technology (UTAUT). (Venkatesh et al.,
2003) The UTAUT (see figure 1) included the
constructs performance expectancy, effort expectancy,
social influence, and facilitating conditions as these
were seen as major determinants of user acceptance
and usage behavior (Venkatesh et al., 2003).
Figure 1: Illustration of UTAUT (own figure based on
Venkatesh et al., 2003).
The UTAUT was also further developed, resulting
in the UTAUT2 which focused on the consumer
perspective of technology acceptance and use
(Venkatesh et al., 2012). This rather specific context
and point of view of UTAUT2 is not as relevant for
this research as the focus lies on the organizational
perspective, specifically on employees as users of
systems. Therefore, a research gap is identified, as it
has not been described yet how a technology
acceptance model for the human co-working with AI
can look like and which aspects are most important
from the human-centric perspective. Furthermore, the
question remains why or why not would the human
like to co-work with AI? (RQ2)
Hence, the authors propose the following research
questions:
How is the (potential) collaboration with AI
evaluated by the knowledge worker? (RQ1)
Why or why not would the knowledge worker
like to co-(work) with artificially intelligent
systems? (RQ2)
In summary, best practices resulting from the
literature review are evaluated regarding their
applicability for tasks of a knowledge worker and
different collaboration concepts are developed. These
are described in six specific scenarios. Scenarios
make it possible to include the context surrounding a
specific research question and, with that, broadening
the scope of the study (Ramirez et al., 2015). The six
collaboration concepts are the following:
AI as intelligent trend and market research
assistant supports the knowledge worker in
gathering and summarizing information on
any given topic.
The AI virtual tutor joins workshops with
clients to detect and analyze emotions of
participants signaling when further
explanation or a break is needed.
AI takes over process analysis and provides
suggestions of how to improve them.
Smart Sales and Marketing Forecasts can be
provided by AI and the human + AI approach
focusses on including quantitative data and
qualitative input from the human co-worker.
AI functions as agile and autonomous project
manager handling planning, monitoring, and
observing team performance.
AI takes over team management and staffing
by selecting suitable knowledge workers for
each project.
3 METHODOLOGY
Focusing on the exploration and understanding of
how individuals perceive new forms of collaboration
with AI a qualitative research approach is chosen. As
little research has been conducted on human-AI
collaboration the Grounded Theory methodology is
chosen to guide the research process. In addition, the
aim of this research approach is not the verification of
theory but rather the generation of a theory. (Glaser
& Strauss, 1967)
3.1 Research Strategy
To apply the research method described above the
following research approach is chosen. To set the
frame and context of the qualitative research and to
differentiate the collaboration concepts clearly, the
six scenarios are classified and structured resulting in
a working model (see figure 2). The x-axis offers a
classification of the degree of collaboration between
humans and AI following the suggestion of Rao
(2017) and inspired by the conceptual models of
human-machine collaboration developed by
(Dellermann et al., 2019; Simmler & Frischknecht,
2020; Traumer et al., 2017).
The Need to Collaborate: Opportunities for Human and AI Co-workers
141
Figure 2: Working model for AI scenarios in consulting
(own figure).
In this working model, an orientation shall be
offered on how collaboration can gradually evolve.
This shall be achieved by starting with applications
that support human skills on the far left, then moving
toward different modes of collaboration where human
and AI workers have specific deliverable tasks to
achieve a common goal and ending on the right with
the potential substitution of human skills where AI
takes over the tasks formerly performed by a human
worker. The focus of the scenarios is set on the
different forms of potential collaboration and task
allocation. This is described as the most promising
way of applying AI in business regarding synergy
effects and performance increases through
complementary capabilities.
The y-axis pictures the organizational structure
and perspective concerning the business processes in
which AI scenarios may be useful and value-adding.
The three superordinate process categories of
organizational value creation are classified according
to the St. Gallen management model into
management, business and support processes (Rüegg-
Stürm & Grand, 2020). This working model and the
scenarios serve as common ground and tool for the
qualitative interviews, representing a comparable
foundation and starting point for the data collection.
3.2 Data Collection
In preparation for the data collection with 10
qualitative interviews a semi-structured interview
guideline was developed. Answers in these interviews
often provide much deeper and more concrete
insights from the perspective of the person affected
than a standardized survey could.
For every scenario, questions are asked about the
aspects of collaboration, usefulness, trust, control,
and general attitude toward the scenario described.
Questions like would you like to collaborate with AI
in this way? or how will this collaboration with AI
impact performance? are included. In the final part of
the interview, knowledge workers are asked to reflect
on the scenarios presented and to indicate which type
of collaboration concept they would prefer with
regard to the degrees of collaboration (as pictured on
the x-axis of the working model). The interviews are
scheduled for a duration of approximately 60
minutes, are conducted via video call, recorded, and
transcribed.
3.3 Data Analysis
The analysis of the collected data follows Strauss and
Corbin (1990) to generate theory in any form which
explains behavioral patterns and to identify
influencing aspects as well as the relations between
them. Coding as systematic strategy of interpretative
analysis is conducted in different styles of coding
which are open, axial, and selective coding. Open
coding is usually the first approach to data analysis
and is used to intensively analyze the interview
transcription, for example line by line or even word
for word with the aim of exploration and concept
identification which are then labelled with a suitable
code of one or two words. Codes can be based on the
data itself or the scientific knowledge in the research
field. (Strauss & Corbin, 1990)
Figure 3: Coding example (MAXQDA).
To gain more insights into the relationships
between concepts and categories in order to
understand the phenomenon as a whole, a coding
paradigm can serve as analytical tool to support axial
coding around a category. The coding paradigm
according to Strauss and Corbin suggests the
following features: conditions, actions-interactions
and consequences or outcomes.
In order to address the research questions the
authors investigate the relations between categories
and aim to integrate them into one overarching
theory. It is a systematic and concentrated coding
process focusing only on the central category.
MAXQDA is used to operationalise this research.
CHIRA 2021 - 5th International Conference on Computer-Human Interaction Research and Applications
142
4 ANALYSIS
Keeping in mind the research question RQ1 how
(potential) collaboration with AI is evaluated by the
knowledge worker it seems that the answer cannot be
as simple as collaboration is evaluated positively or
negatively, but rather it depends. The research
question RQ2 why or why not would the knowledge
worker like to work with intelligent systems focuses
on the aspects on which the evaluation finally
depends.
The result of the open coding process is a total of
over 50 different codes labelling roughly 800 text
passages in the interview transcriptions. As these 50
codes are located on different levels of abstraction,
they are reviewed to eliminate redundancies and to
regroup similar codes into higher-level concepts or
categories. The review of the 50 initial codes leads to
a consolidation of 12 categories (see figure 4) and
their respective subcategories. The most important
categories with the highest frequency are usefulness
of collaboration, confidence and trust in AI skills and
expected outcome of collaboration.
Figure 4: Consolidation of codes into 12 categories (own
figure).
Following the open coding, the next step is the
axial coding. The aim of this coding process is to
understand the more holistic picture. Derived from
the interview responses and the building of categories
in the open coding the usefulness of collaboration
seems to be the central phenomenon. The answers of
the knowledge workers interviewed imply that they
try to evaluate the usefulness of a potential
collaboration in the first place. Therefore, the central
phenomenon will be described as the evaluation of
usefulness of collaboration.
The last step of coding in Grounded Theory is the
selective coding. The aim of this coding process is the
integration of the developed categories into one
overarching theory. To visualize the results and the
developed theory, a model is created showing how the
evaluation of human-AI collaboration from the
perspective of the human co-worker is performed.
The model is presented in figure 5.
Through in-depth data analysis of the interviews
the answers imply that there seem to be several stages
of evaluating potential collaboration. This would
mean that the overarching theory developed can be
rather seen as process which is why the model
presented is visualized as such. This model developed
in axial and selective coding is based on the
assumption that the whole process starts with the
question how potential collaboration with AI is
evaluated by the human collaborator. And it answers
the RQ1 how (potential) collaboration with AI is
evaluated by the knowledge worker: evaluation can
be seen as a process incorporating two phases.
Phase 1: Personal evaluation. This input is being
processed in the first phase of personal evaluation
focusing on the core concept of evaluating the
usefulness of collaboration. This evaluation of
usefulness may be influenced by multiple aspects as
mentioned before. The core aspects are the central
categories that seem to be more important in this
phase of personal evaluation as the context related
aspects. The core aspects include four concepts which
are part of the answer to RQ2 why or why not would
the knowledge worker like to work with intelligent
systems.
1. The expected outcome of collaboration
centers around the question what is the value
added that can be expected through
collaboration? What are benefits and also
consequences to be expected? As the
interview partners expect a certain effort
needed to realize collaboration with AI, they
would like to know if it is worth this effort
and if there is a return to be expected in some
way. Either in increasing efficiency, saving
resources or benefits and value adding
aspects contributing to improved work
results. The expectancy of a certain outcome
could contribute positively or negatively to
the perceived usefulness of collaboration.
2. The confidence and trust in AI skills is
another core aspect which incorporates the
overall opinion of the interview partners on
the capabilities of AI in a certain scenario.
The question to be answered here could be if
the human collaborator is convinced that AI
can take over a specific task and perform
well. The confidence in AI skills may play
an important role when delegating or
outsourcing tasks to the intelligent system
just like in a human team, one likes to be
certain that team members are capable of
handling the tasks they are responsible for.
Being positive about the usefulness of
The Need to Collaborate: Opportunities for Human and AI Co-workers
143
collaboration seems to be only given if one
is convinced of the skills AI has. Otherwise,
the human collaborator might worry that
collaboration would not be beneficial but
rather complicate things in daily work. This
relates to the expected outcome as well.
3. Operational collaboration is evaluated
regarding the aspect of usefulness in the
sense of how easy it is perceived to
implement into daily work, how interaction
with AI is realized and if the task allocation
is perceived as useful and beneficial. Again,
this latter aspect links operational
collaboration to the concepts of expected
outcome and the confidence in AI skills.
4. The fourth core aspect is the perception of
self and own skills the knowledge worker
has. Regarding the evaluation of usefulness
of collaboration, it seems to play a role for
the interview partners if AI is taking over
tasks that are associated with human-only
skills or if they accept and recognize own
limitations. Human-only skills or tasks that
the knowledge worker expect themselves to
perform well are unlikely to be outsourced
to AI as this is not perceived as a useful
action. While if the limitations of own skills
are recognized or the tasks are considered as
unsatisfying, the interview partners seem
more willing or even pleased to allocate
these tasks to AI. This quite subjective
perception and judgement appears to
influence the evaluation of usefulness.
As context related aspects, the concepts of
professional environment, general attitude and
knowledge and prior experience with AI can have an
influence on evaluating usefulness as well and
contribute as well to the answer of RQ2 why or why
not would the knowledge worker like to work with
intelligent systems.
For the aspect of professional environment
especially, the area of work, project settings and
prerequisites regarding the organizational structure
seem to have an influence on the perceived usefulness
of collaboration. The suggested collaboration
scenario needs to be relevant and familiar to the
interview partner and solve a known problem.
Otherwise, the perceived usefulness may be lower as
the knowledge worker does not understand how this
collaboration should have a beneficial impact. The
general attitude including the affinity and openness
toward technology may have an influence in the sense
that interview partners confirming to be open toward
new technologies may have a more positive opinion
on the usefulness of collaboration. Former
experiences with AI or subject matter knowledge
regarding AI could have a positive and negative
influence on the perceived usefulness. Relating also
to the expected outcome and confidence in AI skills
as interviewees may know if the suggested
collaboration could actually work well and if the
expected results are realistic. This evaluation of
usefulness results in an opinion on the usefulness of
collaboration based on the discussed influencing
aspects.
Phase 2: Consideration of external parameters. In
the second phase of evaluating potential
collaboration, external parameters are considered.
These external parameters complete the answer of
RQ2 why or why not would the knowledge worker like
to work with intelligent systems. The formed opinion
in the first phase could be influenced by certain
gateway conditions which are the concepts of
influencers, confidence in technical setup and the
autonomy level of AI. Interpreting the answers of the
interview partners it seems that although an opinion
on the usefulness of collaboration has been
established these gateway conditions could still
impact this opinion and even act as showstoppers or
no go’s for the knowledge workers. Influencers like
clients or managers could prevent knowledge workers
from pursuing collaboration if they voice concerns or
dislike. The confidence in the technical setup of the
collaboration especially regarding transparency and
data security, seems to be a prerequisite. Otherwise, a
formerly positive opinion on the usefulness could not
be sufficient to maintain the intention to collaborate.
Autonomy levels of AI are often defined by the
developing company and programmers creating the
technological setting of the collaboration. In case
these autonomy levels are non-negotiable or not
adaptable to a potential collaborators’ wishes, it
might lead to a rejection of the collaboration concept
as a whole.
The two phases of this evaluation process start
with the core concept of usefulness as the answers
imply that if usefulness is not seen as positive,
interview partners seem not to think about the
external parameters very deeply. If the usefulness is
confirmed, the gateway conditions appear to come
into play contributing to the decision if collaboration
will be pursued. Therefore, the outcome of this
procedural concept of evaluating collaboration with
AI conducted by the human collaborator leads to the
intention to collaborate or to not collaborate. And it is
important to emphasize that not all influencing
concepts and aspects seem to have an effect at the
same time but rather sequentially.
Connecting the results of this research with the
related work and specifically the UTAUT (see figure
1), the procedural manner of evaluating collaboration
with technology (in this case AI) has not been
CHIRA 2021 - 5th International Conference on Computer-Human Interaction Research and Applications
144
Figure 5: Funnel model of evaluating Human-AI collaboration (own figure).
emphasized. The concepts of the UTAUT model
performance expectancy, effort expectancy, social
influence and facilitating conditions could be similar
in meaning to the mentioned concepts of expected
outcome of collaboration, operational collaboration,
professional environment/influencers, and
confidence in technical setup. Although the
congruence of these concepts would need to be
analysed in more detail to draw a final conclusion.
The newly discovered core concepts of confidence
and trust in AI skills and perception of self and own
skills have not been mentioned in the related work yet
and add to the existing body of knowledge. In
addition, the gateway condition autonomy level of AI
is a new concept as well.
Hence, the following funnel model is proposed as
presented in figure 5 and answers the research
questions RQ1 how (potential) collaboration with AI
is evaluated by the knowledge worker and RQ2 why
or why not would the knowledge worker like to work
with intelligent systems.
5 FURTHER RESEARCH
Based on the data analysis conducted in this research,
four propositions for further research are suggested.
Firstly, the data indicates that the evaluation of
potential collaboration seems to be conducted in a
procedural manner starting with the evaluation of
usefulness and considering external parameters only
in the second phase. Therefore, it is suggested to
validate the following propositions P1 and P2 for
verification by qualitative and quantitative means:
Potential collaborators evaluate the
usefulness of collaboration first before
considering other aspects. (P1)
Consideration of external parameters (e.g.,
data security) occurs later in the evaluation
process. (P2)
Secondly, the responses show that several
aspects influence the perceived usefulness of
collaboration. Considered as especially interesting is
the concept of self-perception. It would be interesting
to know if this concept has such a high influence on
usefulness that it could be a showstopper. Proposition
P3 relates to this idea:
Perceived usefulness depends mostly on the
individual self-perception of own skills of
the potential collaborator. (P3)
Thirdly, the gateway conditions seem to play
such an important role that they could minimize the
intention to collaborate although collaboration is
perceived as useful. Therefore, the researcher
suggests proposition P4 for verification:
The intention to collaborate depends on the
evaluation of the gateway conditions. (P4)
As a first step following this research it would be
interesting to interview leading experts and
researchers on the matter of human-AI collaboration
to hear their opinion on the findings and results
discovered. In general, it might be interesting to
investigate human-AI collaboration in other white-
collar jobs looking at different groups of academic
professionals. To extend the research to other
countries could offer interesting insights how the
cultural background may influence the perception and
The Need to Collaborate: Opportunities for Human and AI Co-workers
145
evaluation of usefulness. To study the strength of
influencing factors and their cause-and-effect
relations more deeply and to verify these relations
quantitatively could be another interesting future
research subject. A longitudinal study that includes
information on human-AI collaboration from the
current period when collaboration concepts are not
yet widely used and to investigate differences over
time would offer be an interesting research approach
as well.
6 REFLECTION
The research results show that further complementary
aspects influencing human-AI collaboration were
discovered which have not been discussed by former
research.
Looking at the working model and the
collaboration scenarios developed, it should be
considered that these may constrain the view of the
interview partners and the research itself. They served
as common ground to make responses comparable as
most interview partners did not have experience in
collaborating with AI in a professional context. The
selection of interview partners was done according to
certain criteria and the researchers focused on
selecting a heterogenous group of interview partners
from diverse backgrounds, sectors, and areas of
expertise. Ideally for the Grounded Theory approach,
theoretical sampling would be desirable where the
choice of interview partners depends on the obtained
results from prior interviews.
A theoretical saturation could be noticed by the
researchers during the coding process as with an
increasing number of interviews conducted less new
codes emerged from the responses. The sample of ten
interviewees is quite small and focused on the
professional group of consultants, all living in
Germany. Therefore, the findings are not
representative, and generalizability is limited. The
different levels of experience and knowledge about
AI of the interviewees could be seen as limitation as
this may have influenced the responses
7 CONCLUSIONS
The aim of this research was to identify potential
human-AI collaboration concepts focusing on the job
profile of a knowledge worker, e.g. a consultant.
Additionally, insights and findings should be
generated how these collaboration concepts are
evaluated and perceived by knowledge workers as
well as what influences their willingness to
collaborate with AI. The review and analysis of
current research regarding collaboration of human
and AI co-workers showed that the collaboration with
AI is possible in multiple ways across sectors and that
combining human and AI capabilities can be
valuable.
Nevertheless, the size of the benefit depends on
the specific collaboration concept. The division of
work and task allocation would change but
knowledge workers evaluate this, depending on the
scenario, as desirable and reasonable. The intention to
collaborate with AI and, with that, the acceptance of
intelligent systems seems to be influenced by a
diverse range of factors. Surprisingly was the finding
that the evaluation of collaboration concepts appears
to be a process for the interview partners including
two phases of evaluation. This process is visualized
as framework (see section 4) showing that not all of
the influencing aspects have an effect at the same
time.
If usefulness of the suggested collaboration
scenario is not recognized or not seen as big enough,
collaboration will not be pursued, and the second
phase of evaluation is not entered. In case the
evaluation of usefulness is assessed positively the
consideration of external parameters comes into play.
These gateway conditions should not be
underestimated as they might be contradictory to the
positive opinion on usefulness. Additionally, if they
are weighed heavily by the individual, these
conditions can be an obstacle and even showstopper
for the decision and intention to collaborate with AI.
The diversity of influences that affect the
evaluation of human-AI collaboration concepts,
leading to the intention to collaborate and lastly to the
actual collaboration itself shows how complex a
successful implementation of AI initiatives may be.
The human collaborator plays a crucial part in this
implementation and should be onboarded and
integrated in the AI initiative early in the process.
REFERENCES
Bauer, W., & Vocke, C. (2019). Work in the Age of
Artificial Intelligence Challenges and Potentials for
the Design of New Forms of Human-Machine
Interaction. In (pp. 493–501). Springer, Cham.
https://doi.org/10.1007/978-3-030-20154-8_45
Bittner, E. A. C., Oeste-Reiß, S., Ebel, P. A., & Söllner, M.
(2019). Mensch-Maschine-Kollaboration: Grundlagen,
Gestaltungsherausforderungen und Potenziale für
verschiedene Anwendungsdomänen. HMD Praxis
Der Wirtschaftsinformatik, 56(1), 34–49.
https://doi.org/10.1365/s40702-018-00487-1
CHIRA 2021 - 5th International Conference on Computer-Human Interaction Research and Applications
146
Breazeal, C. (2003). Toward sociable robots. Robotics
and Autonomous Systems, 42(3-4), 167–175.
https://doi.org/10.1016/S0921-8890(02)00373-1
Brougham, D., & Haar, J. (2018). Smart Technology,
Artificial Intelligence, Robotics, and Algorithms
(STARA): Employees’ perceptions of our future
workplace. Journal of Management & Organization,
24(2), 239–257. https://doi.org/10.1017/jmo.2016.55
Colson, E. (2019, July 8). What AI-driven decision
making looks like. https://hbr.org/2019/07/what-ai-
driven-decision-making-looks-like#
Crandall, J. W., Oudah, M., Tennom, Ishowo-Oloko, F.,
Abdallah, S., Bonnefon, J.‑F., Cebrian, M., Shariff, A.,
Goodrich, M. A., & Rahwan, I. (2018). Cooperating
with machines. Nature Communications, 9(1), 233.
https://doi.org/10.1038/s41467-017-02597-8
Dautenhahn, K. (2007). Socially intelligent robots:
Dimensions of human-robot interaction. Philosophical
Transactions of the Royal Society of London.
Series B, Biological Sciences, 362(1480), 679–704.
https://doi.org/10.1098/rstb.2006.2004
Davis, F. D. (1989). Perceived Usefulness, Perceived
Ease of Use, and User Acceptance of Information
Technology. MIS Quarterly, 13(3), 319.
https://doi.org/10.2307/249008
Davis, F. D. (1993). User acceptance of information
technology: system characteristics, user perceptions
and behavioral impacts. International Journal
of Man-Machine Studies, 38(3), 475–487.
https://doi.org/10.1006/imms.1993.1022
Dellermann, D., Ebel, P., Söllner, M., & Leimeister, J. M.
(2019). Hybrid Intelligence. Business &
Information Systems Engineering, 61(5), 637–643.
https://doi.org/10.1007/s12599-019-00595-2
Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention
and behavior: An introduction to theory and research.
Addison-Wiley Publishing Company.
Fügener, A., Grahl, J., Gupta, A., & Ketter, W. (2019).
Collaboration and Delegation Between Humans and
AI: An Experimental Investigation of the Future of
Work. ERIM Report Series Research in
Management(ERS-2019-003-LIS).
Glaser, B. G., & Strauss, A. L. (1967). The discovery of
grounded theory: Strategies for qualitative research.
AldineTransaction.
Huang, M.‑H., & Rust, R. T. (2018). Artificial Intelligence
in Service. Journal of Service Research, 21(2), 155–
172. https://doi.org/10.1177/1094670517752459
Huang, M.‑H., Rust, R., & Maksimovic, V. (2019). The
Feeling Economy: Managing in the Next Generation of
Artificial Intelligence (AI). California Management
Review, 61(4), 43–65. https://doi.org/10.1177/
0008125619863436
Kamar, E., Gal, Y., & Grosz, B. J. (2013). Modeling
information exchange opportunities for effective human–
computer teamwork. Artificial Intelligence, 195, 528–
550. https://doi.org/10.1016/j.artint.2012.11.007
Libert, B., & Beck, M. (2017, July 24). AI May Soon
Replace Even the Most Elite Consultants.
https://hbr.org/2017/07/ai-may-soon-replace-even-the-
most-elite-consultants
Lichtenthaler, U. (2018). Substitute or Synthesis: The
Interplay between Human and Artificial Intelligence.
Research-Technology Management, 61(5), 12–14.
https://doi.org/10.1080/08956308.2018.1495962
Poortmans, P. M. P., Takanen, S., Marta, G. N., Meattini,
I., & Kaidar-Person, O. (2019). Winter is over: The use of
Artificial Intelligence to individualise radiation therapy for
breast cancer. Breast (Edinburgh, Scotland), 49, 194–200.
https://doi.org/10.1016/j.breast.2019.11.011
Ramirez, R., Mukherjee, M., Vezzoli, S., & Kramer, A. M.
(2015). Scenarios as a scholarly methodology to
produce “interesting research”. Futures, 71, 70–87.
https://doi.org/10.1016/j.futures.2015.06.006
Rao, A. (2017, May 10). A Strategist’s Guide to Artificial
Intelligence: As the conceptual side of computer science
becomes practical and relevant to business, companies
must decide what type of AI role they should play.
Strategy + Business, 2017, Issue 87, pp. 1–10.
https://www.strategy-business.com/article/A-Strategists-
Guide-to-Artificial-Intelligence
Rüegg-Stürm, J., & Grand, S. (2020). Das St. Galler
Management-Modell: Management in einer komplexen
Welt (2., überarbeitete Auflage). Haupt Verlag.
Schlögl, S., Postulka, C., Bernsteiner, R., & Ploder, C. (Eds.)
(2019). Artificial Intelligence Tool Penetration in
Business: Adoption, Challenges and Fears. Knowledge
Management in Organizations: Vol. 1027. Springer.
https://ezproxy.bib.fh-muenchen.de:2058/chapter/10.10
07%2F978-3-030-21451-7_22
Simmler, M., & Frischknecht, R. (2020). A taxonomy of
human–machine collaboration: Capturing automation
and technical autonomy. AI & SOCIETY, 1–12.
https://doi.org/10.1007/s00146-020-01004-z
Skilton, M., & Hovsepian, F. (2018). The 4
th
Industrial
Revolution: Responding to the Impact of Artificial
Intelligence on Business (1
st
edition 2018). Springer
International Publishing.
Sowa, K., & Przegalinska, A. (2020). Digital Coworker:
Human-AI Collaboration in Work Environment, on the
Example of Virtual Assistants for Management
Professions. In Przegalinska A., Grippa F., Gloor P. (Ed.),
Springer Proceedings in Complexity. Digital
Transformation of Collaboration (pp. 179–201). Springer.
https://doi.org/10.1007/978-3-030-48993-9_13
Strauss, A., & Corbin, J. (1990). Basisc of qualitative
research: Grounded theory procedures and techniques.
SAGE.
Susskind, R. E., & Susskind, D. (2017). The future of the
professions: How technology will transform the work of
human experts. Oxford University Press.
Traumer, F., Oeste-Reii, S., & Leimeister, J. M. (2017).
Towards a Future Reallocation of Work between Humans
and Machines Taxonomy of Tasks and Interaction Types
in the Context of Machine Learning. In Thirty Eigth
International Conference on Information Systems (ICIS).
https://doi.org/10.2139/ssrn.3159131
The Need to Collaborate: Opportunities for Human and AI Co-workers
147
Venkatesh, Morris, & Davis (2003). User Acceptance of
Information Technology: Toward a Unified View. MIS
Quarterly, 27(3), 425. https://doi.org/10.2307/30036540
Venkatesh, Thong, & Xu (2012). Consumer Acceptance and
Use of Information Technology: Extending the Unified
Theory of Acceptance and Use of Technology. MIS
Quarterly, 36(1), 157. https://doi.org/10.2307/41410412
Venkatesh, V., & Bala, H. (2008). Technology
Acceptance Model 3 and a Research Agenda on
Interventions. Decision Sciences, 39(2), 273–315.
https://doi.org/10.1111/j.1540-5915.2008.00192.x
Venkatesh, V., & Davis, F. D. (2000). A Theoretical
Extension of the Technology Acceptance Model: Four
Longitudinal Field Studies. Management Science, 46(2),
186–204. https://doi.org/10.1287/MNSC.46.2.186.11926
(Management Science, 46(2), 186-204).
Wang, D., Weisz, J. D., Muller, M., Ram, P., Geyer, W.,
Dugan, C., Tausczik, Y., Samulowitz, H., &
Gray, A. (2019). Human-AI Collaboration in Data
Science: Exploring Data Scientists’ Perceptions of
Automated AI. Proceedings of the ACM on
Human-Computer Interaction, 3(CSCW), 1–24.
https://doi.org/10.1145/3359313.
CHIRA 2021 - 5th International Conference on Computer-Human Interaction Research and Applications
148