The Need to Collaborate: Opportunities for

Human and AI Co-workers

Anja Hornikel, Christian Greiner

and Thomas Peisl

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

https://orcid.org/0000-0002-8184-7128

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

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

Industrial

Revolution: Responding to the Impact of Artificial

Intelligence on Business (1

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