AI, Our Assistant and Friend – Challenges and Implications for Human-AI Interaction // KI, mein Freund und Helfer - Herausforderungen und Implikationen für die Mensch-KI-Interaktion

Abstract

New medical procedures for cancer diagnostics, autonomous
driving and productivity improvements across industries and
sectors – artificial intelligence (AI) is gaining ground and
permeates virtually all spheres of life. The use of AI will
significantly disrupt a vast number of jobs and workflows
and generally impact how we interact on a day-to-day basis.
We humans see ourselves facing new challenges due to
increasing interaction with AI in our daily routines. When
designing AI solutions, focus should be not just on
technological enhancement, but equally on the interaction
between humans and AI. AI can only reach its full potential
if this interaction is useful, appealing and applied in the
appropriate context. The challenges and implications of
human-AI interaction and the required measures of companies
and individuals therefore need to be discussed in depth. This
study demonstrates the broad scope of human-AI interaction,
its underlying features and various drivers for the success
and acceptance of human-AI interaction. We summarize our
findings on the future development of human-AI interaction
in 10 theses and illustrate the implications, opportunities,
challenges and action areas for successful human-AI
interaction.
// ***** German Version ***** //
Künstliche Intelligenz durchdringt unser Privat- und Berufsleben immer stärker und ist bereits fester Bestandteil davon. Wir als Menschen sehen uns damit der Herausforderung gegenüber, in unserem Alltag immer öfter mit KI zu interagieren. Wer KI-Lösungen gestaltet, sollte sich daher nicht nur auf ihre technologische Weiterentwicklung fokussieren, es muss auch gleichermaßen um die Interaktion zwischen Mensch und KI gehen.

Basierend auf einer fundierten wissenschaftlichen Recherche und ergänzenden Interviews mit KI-Experten und -Lösungsanbietern vermittelt die vorliegende Studie ein umfangreiches Verständnis der heutigen und zukünftigen Mensch-KI-Interaktion. Darüber hinaus widmet sie sich den korrespondierenden Implikationen.

Im Zuge der Analyse wurden fünf unterschiedliche Interaktionstypen identifiziert, die sich anhand ihrer charakteristischen Interaktionsdimensionen und Bewertungsmerkmale abgrenzen lassen. Diese Interaktionstypen typisieren wir anhand der Begriffe „Schutzengel“, „Heinzelmännchen“, „Informant“, „Kollege“ und „bester Freund“. Sie lassen sich drei unterschiedlichen Gruppen zuordnen: KI als Automat, KI als vielfältiger Helfer und KI als Partner.

Darüber hinaus unterliegen Mensch-KI-Interaktionen einer Vielzahl von Einflussfaktoren. Hierbei hat sich gezeigt, dass die menschliche Erwartungshaltung in der Interaktion mit KI-Lösungen durch deren Transparenz, Personalisierung und Anthropomorphologie beeinflusst wird. Interaktionen, die dieser Erwartungshaltung gerecht werden, schaffen mit der Zeit Akzeptanz und Vertrauen gegenüber einer KI-Lösung. Akzeptanz und Vertrauen wirken sich wiederum auf die Erwartungshaltung und auf die Gestalt zukünftiger Interaktionen aus.

Mit Blick auf die zukünftige Mensch-KI-Interaktion haben wir zehn Thesen formuliert, welche die wesentlichen Veränderungen dieser Interaktion zusammenfassen. Dabei geht es nicht nur um die erfolgreiche Gestaltung heutiger Anwendungsszenarien, ebenso kommt es auf die Vorbereitung zukünftiger Entwicklungen an. Daher müssen Unternehmen in den Bereichen Strategie, Technologie und Organisation mit Bedacht agieren. Die Studie stellt diese Bereiche den relevanten Chancen wie auch Herausforderungen gegenüber und zeigt korrespondierende Handlungsfelder auf.

Full text

Think beyond
tomorrow
AI, our assistant and friend –
challenges and implications for
human-AI interaction
Executive summary
In cooperation with

2 | Think beyond tomorrow Executive summary
AI, our assistant and friend – challenges and
implications for human-AI interaction
New medical procedures for cancer diagnostics, autonomous
driving and productivity improvements across industries and
sectors – artificial intelligence (AI) is gaining ground and
permeates virtually all spheres of life. The use of AI will
significantly disrupt a vast number of jobs and workflows
and generally impact how we interact on a day-to-day basis.
We humans see ourselves facing new challenges due to
increasing interaction with AI in our daily routines. When
designing AI solutions, focus should be not just on
technological enhancement, but equally on the interaction
between humans and AI. AI can only reach its full potential
if this interaction is useful, appealing and applied in the
appropriate context. The challenges and implications of
human-AI interaction and the required measures of companies
and individuals therefore need to be discussed in depth. This
study demonstrates the broad scope of human-AI interaction,
its underlying features and various drivers for the success
and acceptance of human-AI interaction. We summarize our
findings on the future development of human-AI interaction
in 10 theses and illustrate the implications, opportunities,
challenges and action areas for successful human-AI
interaction.
The characteristic interaction types of AI use cases
Humans interact in a variety of ways with AI and across a
broad range of applications. To describe human-AI interaction,
we refer to established communications theories in
combination with characteristics and features from other
forms of interaction (human-human, human-computer,
human-robot). At the same time, AI requires closer analysis;
it is technically complex, the different AI capabilities are
manifold and the number of use cases correspondingly large.
We have identified five specific features of AI and humans’
interaction with AI. In sum, they distinguish human-AI
interaction from interaction with other technologies:
• ►TheautonomyofAIagentswiththeircapacityfor
autonomous learning and ability to make decisions and act
on their own
• ►Theinterdependenceofthecontentofhuman-AIinteraction:
integrated analysis of the interaction history and context
which is taken into account in future interactions
• ►Theopportunitiesforamoreadvanced
anthropomorphology due to a human-like appearance or
the human-like abilities and behavior of AI agents
• ►Newandmoreintuitiveinteractionchannels
• ►Emotionalandsocialintelligenceasafoundationforsome
forms of human-AI interaction

Think beyond tomorrow Executive summary | 3
To map the specific patterns of interaction between humans
and AI precisely and comprehensively, we distinguish
between nine different interaction dimensions:
An analysis of different AI use cases in everyday life shows
that interactions, despite potentially similar features in the
nine dimensions, can be classified based on the criteria of
freedom of action and reciprocal engagement. For example,
interaction often starts with the same trigger and leads to a
similar outcome but differs in terms of the degree of freedom
of action and reciprocal engagement. In this study, we have
identified and classified five characteristic interaction types
of current or potential AI use cases using the terms guardian
angel, pixie, informant, colleague and best friend. We group
these interaction types in three different groups. The first
group describes AI as an automaton which oversees,
protects and if necessary supports the actions of humans
like a guardian angel. The second group, AI as a versatile
assistant, relates to interactions in which AI supports the
work of humans in the background, supplies humans with
desired information or helps to produce outcomes in close
interaction. Finally, the third group bundles use cases in
which AI as a partner interacts in the role of best friend due
to a high degree of personalization and social elements.
Human
AI
Number and dependency
of interactions
Action direction
and channel
Interaction transparency
Interaction environment
Action frequency Interaction outcomeInteraction trigger
Interaction Interaction
Action
Low High
Acting dependently Acting autonomously
Reciprocal engagement
Freedom of action
Guardian
angel
AI as an automaton
Best
friend
AI as a partner
Colleague
Informant
Pixie
AI as a versatile assistant

4 | Think beyond tomorrow Executive summary
Key factors affecting human-AI interaction
Numerous factors have an impact on human-AI interaction.
Understanding and taking human expectations on board is
key to instilling acceptance of and trust in AI-powered
solutions, and ultimately the success of human-AI interaction.
In this context, it is vital to correctly select the form of AI
features such as transparency, anthropomorphology and
personalization.
The term transparency describes the extent to which a
person is aware that they are interacting with AI and knows
the process and its outcomes. Knowledge of what an AI
solution can achieve, how it achieves this and the quality
of the outcome increases human understanding of and
ultimately the success of human-AI interaction. However,
the appropriate degree of transparency varies depending
on the situation and context of the interaction.
Whether or not anthropomorphology – the attribution of
human characteristics to AI – should be emphasized depends
highly on the end user and context of the interaction. A high
degree of anthropomorphology can be a fun factor and have
a positive effect on user experience; however, it should not
be used solely as a means to an end or to generate attention.
The more AI is personalized, the more individually it can
interact with humans. This means, for example, that users
only receive interaction content and outcomes that are
relevant to them. In addition, a highly personalized AI
solution can reduce the number and frequency of actions,
minimizing the time a user spends interacting with AI. In
many contexts, personalized AI solutions are therefore
potentially better able to meet human expectations.
Consequently, the relevance of the various factors varies
overall depending on the type of interaction and interaction
context. Interactions that meet expectations foster
acceptance and trust in an AI solution over time. In turn,
acceptance and trust have an impact on expectations and
the form of future interactions. The characteristics and
background of a person that shape their individual
expectations of AI also play a significant role.
Human
AI
Interaction Interaction
Action
Background
Media, marketing, culture,
country, education, age,
interests ...
Expectations
AnthropomorphologyTransparency Personalization
Outcome
Positive vs. negative
Acceptance
Interaction context
Professional vs. private life,
interaction types, application
area/use case ...
Trust

Think beyond tomorrow Executive summary | 5
Theses on the future
development of human-AI interaction
Based on our understanding of typical human-AI interaction
processes, this study formulates 10 theses for its future
development. These theses encompass both the successful
design of today’s application scenarios as well as adequate
preparation for the future potential of human-AI interaction.
Thesis 1
With the ongoing development of AI, personalization, social elements, task diversity and AI’s understanding of context
in interactions with humans are increasing.
The first thesis relates to a fundamental change in AI and
is therefore depicted as an umbrella thesis overarching
the other nine theses. These are divided into three
segments covering the changes in terms of the roles and
responsibilities of AI (What), the process of human-AI
interaction (How) and the implications for the successful
design of future AI applications (So What).
Trust in human-AI
interaction must be gained
through repeated positive
outcomes and/or by establishing
a social bond
What How So what
Interactions between
humans and AI forms the basis
for merging their respective
unique capabilities
Like humans, AI has scope
to act and make decisions
to different extents
3
Human-AI interactions
are becoming more direct
and therefore largely
independentofspecic
interaction channels
5
Look and feel of
AI no longer need to rely
so heavily on looking and
functioning like a human
Thesis 8
The user experience with
AI develops into an overarching
and continuous user journey
6
The interaction types are
developing toward two extremes:
AI as an automaton and
AI as a partner
4
AI adapts to human
expectations by providing
content- and context-driven
services
7
Ethicsandmoralityarekey
components of human-AI
interaction and require
data- and value-driven
learning
10
Thesis
Thesis
Thesis
Thesis
Thesis
Thesis
Thesis
Thesis
2
9

6 | Think beyond tomorrow Executive summary
Action areas and implications
for human-AI interaction in companies
AI solutions have a long way to go before they can actually
take over the work of humans and pose any immediate risk
to their jobs. However, AI will change many professions in the
near future. AI offers companies numerous opportunities, but
also creates challenges and a need to act. Thus, companies
must take steps to be able to use AI in the future and remain
competitive going forward. The opportunities, challenges
and action areas identified cover the business activities of
companies, technology advancement and the related change
in the human work environment. Companies must therefore
act wisely when it comes to strategy, technology and
organization.
Strategy Technology Organization
• Exploitoptimizationpotential
in the value chain
• Boost productivity
• Develop personalized products
and services
• Useefcienttechnology
• Exploitrapidinnovationcycles
• Boost the autonomy of AI
• Assume repetitive and arduous
tasks
• Free up capacities for creative and
complex tasks
• Increase job satisfaction
• Clinging to old business models
• No standard solutions
• No regulatory framework for
AI solutions
• Backward IT landscapes
• Lack of user-friendly interfaces
• No transparency
• Knowledge gaps in terms of AI
capabilities and applications
• Rejection of and aversion to AI
• Utopian expectations regarding
AI capabilities
• EstablishrequisiteAIexpertise
in the company
• Identifydomain-specic
optimization potential
• Help develop the regulations
for the use of AI systems
• Involve all employees in the
development of user interfaces
• Develop AI solutions with the
greatest possible transparency
• Collect environmental data
to develop use case solutions
(data = revenue stream)
• Introduce AI workshops to
experiment with AI solutions
• Iterative rollouts of AI solutions
• Group-specicchange
management
Challenges OpportunitiesAction areas

Think beyond tomorrow Executive summary | 7
About the study
The full text of the study is available in German only at
www.ey.com/de_de/ai/wie-menschlich-kann-kuenstliche-
intelligenz-sein
The study is based on broad-based literature research and
interviews conducted by the authors on the current situation
at companies. Twenty-five in-depth interviews were
conducted with renowned AI experts and users. The
interview partners come from different industries and
work in technology companies, AI start-ups or research.
We extend our special thanks to our interview partners for
their valuable thoughts and opinions. We would also like to
thank our colleagues Simon Blöthner, Michael Glahn, Nadine
Kaiser, Jens Keuter, Mikail Kibar and Karin Sahr for their
support in drafting the study and their comments and
suggestions.
About the authors
Shape the future interaction with AI successfully with us and
exploit its potential at your company. We have an in-depth
understanding of the application possibilities and their
impactonbusiness.Expertsandexperiencedpractitioners
from the Project Group Business & Information Systems
EngineeringoftheFraunhoferFITandEYwillworkwithyou
to find the right answers for your organization.
Dr. Yilmaz Alan
PartneratEY(Advisory,Munich)
Tel. +49 160 93915751
yilmaz.alan@de.ey.com
Silvana Hinsen
Senior Consultant for Technology
TransformationatEYandPhDstudentat
the University of Bayreuth
Tel. +49 160 939 25330
silvana.hinsen@de.ey.com
Patrick Beisel
Consultant for
TechnologyTransformationatEY
Tel. +49 160 93915809
patrick.beisel@de.ey.com
Prof. Dr. Nils Urbach
Professor of Information Systems and
Strategic IT Management at the University
of Bayreuth and Fraunhofer FIT
Tel. +49 921 55 4712
nils.urbach@m-rc.de
Jan Jöhnk
Research assistant at Fraunhofer FIT and
PhD student at the University of Bayreuth
Tel. +49 921 55 4714
jan.joehnk@m-rc.de
Malte Weißert
Student assistant at Fraunhofer FIT and
business administration (M.Sc.) student at
the University of Bayreuth
Tel. +49 921 55 4710
malte.weissert@m-rc.de
Stephan Blumenthal
StudentemployeeatEY(Technology
Transformation) and information systems
(M.Sc.) student at the University of Cologne
Tel. +49 221 2779 14743
stephan.blumenthal@de.ey.com
Peter Hofmann
Research assistant at Fraunhofer FIT and
PhD student at the University of Bayreuth
Tel. +49 921 55 4716
peter.hofmann@m-rc.de

EY | Assurance | Tax | Transactions | Advisory
About the global EY organization
TheglobalEYorganizationisaleaderinassurance,tax,transactionand
advisory services. We leverage our experience, knowledge and services
tohelpbuildtrustandcondenceinthecapitalmarketsandineconomies
the world over. We are ideally equipped for this task — with well trained
employees, strong teams, excellent services and outstanding client
relations. Our global purpose is to drive progress and make a difference
by building a better working world — for our people, for our clients and
for our communities.
TheglobalEYorganizationreferstoallmemberrmsofErnst&Young
GlobalLimited(EYG).EachEYGmemberrmisaseparatelegal
entity and has no liability for another such entity’s acts or omissions.
Ernst&YoungGlobalLimited,aUKcompanylimitedbyguarantee,
does not provide services to clients. For more information, please visit
ey.com.
InGermany,EYhas20locations.Inthispublication,“EY”and“we”
refertoallGermanmemberrmsofErnst&YoungGlobalLimited.
©2020Ernst&YoungGmbH
Wirtschaftsprüfungsgesellschaft
All Rights Reserved.
GSA Agency
BKR 2002-332
EDNone
InlinewithEY’scommitmenttominimizeitsenvironmentalimpactthisdocumenthasbeenprinted
CO2 neutral and on FSC®-certiedpaperthatconsistsof60%recycledbers.
This publication contains information in summary form and is therefore intended for general guidance
only. Although prepared with utmost care this publication is not intended to be a substitute for
detailed research or the exercise of professional judgment. Therefore no liability for correctness,
completeness and/or currentness will be assumed. It is solely the responsibility of the readers
to decide whether and in what form the information made available is relevant for their purposes.
NeitherErnst&YoungGmbHWirtschaftsprüfungsgesellschaftnoranyothermemberoftheglobal
EYorganizationcanacceptanyresponsibility.Onanyspecicmatter,referenceshouldbemade
to the appropriate advisor.
ey.com/de