The relationship between motives for using a Chatbot and satisfaction with Chatbot characteristics in the Portuguese Millennial population: an exploratory study.

Tim David Rieke

Dissertation

Master in Management

Supervised by Helena Martins

2018

i

Abstract

Even though there is a growing number of studies focusing on Chatbots and artificial

conversations, research lacks of studies analysing Chatbot characteristics and motives for

using this technology. This displays a critical gap in the literature that the present study

addresses. This work thus attempts to analyse the relationship between motives for using a

Chatbot and satisfaction with Chatbot characteristics.

Two questionnaires were developed, one to assess satisfaction with Chatbot characteristics,

according to the Kano model (Kano et al., 1984) and another to assessing motives for using

Chatbots, based on previous qualitative research (Brandtzaeg & Følstad, 2017). Survey

research was directed to Portuguese Millennials (N=258) and statistical analysis indicated

that motives for using Chatbots do not seem to have a clear relationship with satisfaction

with Chatbot characteristics. Furthermore, results suggest that equipping Chatbots with

human-like characteristics, seems to be indifferent to Portuguese Millennials; instead speed

and accessibility of Chatbots seem to be valued, especially when using this technology for

convenience purposes.

A discussion on the possible implications for theory and practice on this topic is presented,

and clues for future research are suggested.

Keywords: Chatbots, Artificial Intelligence, Natural Language Processing, Human-

Computer Interaction

JEL-Codes: D70, O30, M30, O33, O35

ii

Resumo

Embora haja um número crescente de estudos na área dos Chatbots e conversas mediadas

por inteligência artificial, existem ainda poucos estudos que analisem sobre características

dos Chatbots e os motivos para o uso desta tecnologia por parte dos consumidores, o que

revela uma uma lacuna crítica na literatura que o presente estudo pretende abordar. Este

trabalho propõe, assim, analisar a relação entre motivos para usar um Chatbot e satisfação

com as características do mesmo.

Dois questionários foram desenvolvidos, um para avaliar a satisfação com as características

do chatbot de acordo com o modelo de Kano (Kano et al., 1984) e outro para aferir as

motivações para o uso de chatbots com base em investigação qualitativa anterior (Brandtzaeg

& Følstad, 2017). A presente investigação contou com uma amostra de 258 milenniais

portugueses, tendo os resultados indicado que os motivos para usar Chatbots não parecem

ter uma relação clara na satisfação com as características do Chatbot. Além disso, os

resultados demonstraram que equipar os Chatbots com características semelhantes às

humanas, parece ser indiferente aos Millennials portugueses; em vez disso, a velocidade e a

acessibilidade do Chatbots parecem ser valorizadas, especialmente ao usar essa tecnologia

para fins de conveniência.

Uma discussão sobre as possíveis implicações para a teoria e prática sobre este tópico são

apresentadas e pistas para investigação futura são sugeridas.

Palavras-chave: Inteligência Artificial, Natural Language Processing, Chatbots, Human-

Computer Interaction

JEL-Codes: D70, O30, M30, O33, O35

iii

Table of Contents

Abstract ............................................................................................................................. i

Resumo ............................................................................................................................ ii

Table of Contents ........................................................................................................... iii

List of Figures .................................................................................................................. v

List of Tables .................................................................................................................. vi

Introduction ..................................................................................................................... 1

Chapter 1. Chatbots ......................................................................................................... 4

1.1. Human-Computer Interaction.......................................................................... 4

1.2. Definition of Chatbot ........................................................................................ 6

1.3. Evolution and Revolution ................................................................................. 7

1.4. Types and Classification of Chatbots .............................................................. 9

1.5. Voice-Based vs. Text-Based ........................................................................... 12

1.6. Major Use Case and Role in Society .............................................................. 13

Chapter 2. Chatbots and Artificial Intelligence ........................................................... 14

2.1. Artificial Intelligence ....................................................................................... 14

2.2. Machine Learning ........................................................................................... 16

2.3. Natural Language Processing ........................................................................ 17

Chapter 3. Satisfaction with Chatbot Characteristics ................................................. 18

3.1. Development of Satisfaction ........................................................................... 18

3.2. Capturing Satisfaction with Chatbot Characteristics ................................... 20

Chapter 4. Chatbot Characteristics .............................................................................. 23

4.1. Emotions .......................................................................................................... 23

4.2. Personality ....................................................................................................... 26

4.3. Conversational Abilities .................................................................................. 28

iv

4.4. Efficiency.......................................................................................................... 30

Chapter 5. Motives for Using a Chatbot ...................................................................... 31

5.1. Productivity ...................................................................................................... 31

5.2. Entertainment .................................................................................................. 33

5.3. Social and Relational ....................................................................................... 34

5.4. Novelty and Curiosity ...................................................................................... 35

Chapter 6. Hypothesis Development ........................................................................... 36

Chapter 7. Methodology - Survey ................................................................................. 37

7.1. Survey Fundaments ......................................................................................... 37

7.2. Questionnaire Development ........................................................................... 41

7.3. Questionnaire Evaluation ............................................................................... 44

7.4. Data Collection Procedures ............................................................................ 46

Chapter 8. Results .......................................................................................................... 47

8.1. Sample ............................................................................................................... 47

8.2. Instrument Validity and Reliability ................................................................ 50

8.3. Descriptive Analysis ........................................................................................ 54

8.4. Discussion ........................................................................................................ 61

Chapter 9. Limitations and Further Research ............................................................. 63

Conclusions .................................................................................................................... 64

References ...................................................................................................................... 66

Attachments ................................................................................................................... 79

v

List of Figures

Figure 1: Chatbot Classification according to Kassibgi ............................................................ 10

Figure 2: Kano Model ................................................................................................................... 21

Figure 3: Hypotheses Development ............................................................................................ 36

Figure 4: Gender Composition of the Survey ........................................................................... 47

Figure 5: Distribution of respondents’ age................................................................................. 48

Figure 6: Employment Status of Respondents .......................................................................... 49

Figure 7: Respondents Familiarity with Chatbots ..................................................................... 54

Figure 8: Respondents Usage of Chatbots in the Past ............................................................. 55

Figure 9: Respondents Daily Usage of Chatbots ...................................................................... 55

Figure 10: Respondents Curiosity ................................................................................................ 56

Figure 11: Respondents Motives for Using a Chatbot .............................................................. 56

Figure 12: Relevance of Chatbot Accessibility for Respondents ............................................ 58

Figure 13: Relevance of Chatbot Speed for Respondents ........................................................ 59

Figure 14: Relevance of Chatbot Sadness for Respondents .................................................... 59

Figure 15: Speed of a Chatbot in Relation with Respondents Motives ................................. 60

Figure 16: Accessibility of a Chatbot in Relation with Respondents Motives ...................... 60

vi

List of Tables

Table 1: Chatbot Classification according to Mason ................................................................ 11

Table 2: Definitions of Artificial Intelligence ............................................................................ 15

Table 3: Identified Chatbot Characteristics................................................................................ 38

Table 4: Kano Evaluation Table .................................................................................................. 39

Table 5: Identified Chatbot Characteristics................................................................................ 39

Table 6: Questionnaire 2nd Part - Motives for Using a Chatbot .............................................. 42

Table 7: Questionnaire 3rd Part - Satisfaction with Chatbot Characteristics ......................... 43

Table 8: Kano Frequency Distribution ....................................................................................... 44

Table 9: Academic Level of the Respondents ........................................................................... 48

Table 10: Cronbach Alpha of Motive Dimensions ................................................................... 50

Table 11: KMO & Bartlett's Test ................................................................................................. 51

Table 12: Factor loadings from Principal-Components Analysis for the Motives for using

Chatbot ............................................................................................................................................. 51

Table 13: Kano-Method Evaluation ............................................................................................ 57

1

Introduction

Artificial Intelligence (AI) is on the rise, and it must be seen as an important IT pillar that

will determine the competitiveness of companies shortly in what is commonly referred to as

the 4th industrial revolution (Kaplan, 2016). In general, three major technological trends now

have come together and have led to the advancements in AI: big data, affordable high-

performance computers and machine learning systems. According to this development, it

must be stated that AI is not just the future anymore it is part of our present (Domingos,

2015; Hirschberg & Manning, 2015; Kaplan, 2016).

AI has a long tradition in human history. Already in the Middle Ages, people created

machines that were supposed to give the impression of aliveness through technical

mechanisms (Mijwel, 2015). History of Artificial Intelligence., 2016). Today, cogitation and

intelligence are the criteria for machines to be perceived "human-like" (Noble, 2013;

Arrabales, Ledezma, & Sanchis, 2013). To what extent this similarity is pronounced, is

illustrated by the naturalness and the developments in the field of Human-Computer

Interaction. One of the first programs in this area was ELIZA, a system developed in the

1970s by Joseph Weizenbaum, which can be perceived as the first chatbot in history (Bassett,

2018).

Chatbots are dialogue systems that communicate with a user through natural language and a

user interface (Dale, 2016). Nevertheless, the technological options at that time were not yet

mature enough to develop a "human-like" acting system. Recently, these systems have

experienced a new upswing (Yang, 2012). Researchers, entrepreneurs and individuals are

using these digital helpers and see potential in improving the flow of information both inside

and outside the company through the use of Chatbots (Turban et al., 2018).

Previously, customers who wanted to connect with a company either had to fill out forms

or call hotlines with often long queues. This type of communication can often be one-sided,

annoying and slow for the customer (Thomas & McSharry, 2015). On the other hand,

communication with friends and colleagues is increasingly taking place via messaging

platforms (Muldowney, 2017). The popularity of messaging and bot systems is steadily

increasing. Since 2015, more people are using applications for communication in social

networks. That is almost three billion people a day worldwide. In Europe and the US, the

2

platforms WhatsApp and Facebook Messenger are mainly used, while in Asia WeChat and

Line dominate (Gartner, 2015).

Consequently, a breakthrough into a new communication paradigm can be observed.

Communication and interaction are increasingly controlled and determined by algorithms.

Bots and messaging systems are vigorously debated and often have to serve as megatrends

of the next few years. Ostensibly, it is all about new communication interfaces that bring

efficiency and convenience benefits as the logical next evolutionary stage (BMBF, 2016).

This is driven, in particular, by the advances in Artificial Intelligence, which make it possible

to create learning algorithms and chatbots that can automate communication while being

perceived "human-like" by the users (Quarteroni, 2018).

A quick connection to the new communication paradigm can, on the one hand, result in

more efficient work processes, higher customer loyalty, an increase in sales and thus a

competitive advantage for companies (Braun, 2013). For customers, especially the increase

in convenience and productivity is crucial as accessing information or obtain help can be

done in minutes (Brandtzaeg & Følstad, 2017). If corporations overslept the trend, it might

happen that they will not be considered by the customer in the future when choosing services

or products. The customers are disappointed, and the brand's image can be damaged

(Boobier, 2018; Daugherty & Wilson, 2018). For corporations, it is thus essential to

understand the requirements to respond to the desired treatments or outcomes. In addition

to a reputable company and quality products, this primarily concerns the interaction between

companies and their customers, including the corresponding service and thus the satisfaction

(Fleming & Asplund, 2007; Morgan, 2009; McColl-Kennedy & Smith, 2006).

This research aims to analyse the relationship between motives for using a chatbot and

satisfaction with chatbot characteristics. Moreover, thus, to answer in what ways do

motives for using a Chatbot affect the satisfaction with chatbot characteristics? If and

how do the Chatbot characteristics that are valued by users vary with the motives for

using them? Likewise, this interlinkage represents a gap found in the literature. Even though

there is a growing number of researches focusing on Chatbots and artificial conversations,

the present literature review did not find of any studies analysing the influence of various

characteristics on satisfaction.

3

This paper proceeds with a literature review about Chatbots, Satisfaction with Chatbot

characteristics and the motives for using a Chatbot. During the second part, the dissertation

introduces the methodology, which is chosen to verify established hypotheses for their

veracity. The work ends with the analysis of the collected data and the regarding conclusion.

4

Chapter 1. Chatbots

The following chapter begins with a clarification and definition of Human-Computer

Interaction and Chatbots, followed by an overview of historical developments in this area.

It continues with a quick Overview of how Chatbots have developed during the last decades,

what types of Chatbots exist, how they could be classified and what the major use cases in

society are. The goal of this chapter is to provide the fundamentals of Chatbots and to

understand the relevance of characteristics of Chatbots determined in Chapter 4.

1.1. Human-Computer Interaction

The history of Human-Computer Interaction (HCI) is a story in which human needs are

increasingly reflected in the machines (Myers, 1998). While in the beginning, due to technical

restrictions, the user-friendliness and usability could not be taken into consideration and the

functionality was in the foreground, these inbuilt limitations have diminished thanks to

technological advancements. Keywords such as " user-centred" design highlight that the user

no longer has to align themselves with the machine but that rather it is the machine that must

meet the needs of the user (Cockton et al., 2016).

Consequently, usability can be considered as one of the key concepts of HCI. According to

ISO (2013), usability is defined as follows:

“extent to which a product can be used by specified users to achieve specified goals with

effectiveness, efficiency, and satisfaction in a specified context of use” (ISO/TS 20282-

2:2013(en), 2013, p. 1).

Zadrozny et al. (2000) stated that to improve HCI it must be made possible for users “to

express their interest, wishes, or queries directly and naturally, by speaking, typing and

pointing” (p. 117). This statement can rely on to the fact that people intend to use natural

language to interact with computers, similarly to what they do when interacting with other

human beings (Ogden, 1988).

This view is taken up through the broader definition of HCI by Preece et al. (1994) who state

that HCI concerns and targets at “a wide variety of different kind of people and not just

technical specialists as in the past, so it is important to design HCI that supports the needs,

knowledge and skills of the intended users” (p. 6).

5

A decisive step towards a user-friendly, intuitive operation of the computer was made in the

mid-1970s with the transition to the graphical user interface. These innovations have been

incorporated into PC's and thus been crucial to the wide distribution and use of these devices

(Dourish & Bell, 2011). The resulting term "user interface or interface" refers to the interface

between humans and computers. This interface includes hardware that allows users to

interact with the computer for instance, screen, keyboard, microphone. Since the interface is

the part of a computer that allows humans to interact with the computer, it is central to

usability and thus to the user experience (Stone et al., 2005).

According to Ciechanowski et al. (2018a), it is significant to understand that “interactive

interfaces mediate the redistribution of cognitive tasks between humans and machines”(p.

1). This can be applied to the subject of this paper and thus to the following chapter.

Chatbots are of great relevance in HCI, as their purpose is to interact with users, using natural

language (Muldowney, 2017; Shevat, 2017).

6

1.2. Definition of Chatbot

Chatbots, also called chatterbots, belong to the category of software agents as so-called

interface agents or conversational agents. Chatbots allow humans to interact with the

computer based on natural language, this can be text-based or voice-based (Muldowney,

2017; Shevat, 2017)

Mayo (2018) proposed a new definition of a Chatbot, in which he describes it as:

“an application, often available via messaging platforms and using some form of intelligence,

that interacts with a user via a conversational user interface” (p. 3).

Shevat (2017) strengthens this definition and delineates Chatbots as a “new way to expose

software services through a conversational interface” (p. 2).

The definition of Mayo (2018) can be analysed more thoroughly and shows that a Chatbot

is an application and thus a program that was scripted by a developer. Basically, a Chatbot

processes the input of a user and responds accordingly (Shevat, 2017).

Also, the mention of the availability via messaging platforms is of relevance where most of

the modern Chatbots can be found (Muldowney, 2017). Alesanco et al. (2018) illustrate the

relevance of messaging platforms, as they are used by most of the people to interact with

other users and Mayo (2018) also uses the term intelligence in his definition, to refer to

progress in the field of Artificial Intelligence. The concerning changes in AI have led to the

evolvement of Chatbots considerably too (Alpaydin, 2016). This development is explained

in more detail in Chapters 1.3 and 1.4.

The last part of the definition related to interaction via Conversational User Interfaces also

needs a more detailed explanation. The classical Human to Human conversation can take

place both text and voice based. When users interact with Chatbots, the conversation is

mostly text-based (Mayo, 2018), even though there are Chatbots or personal assistants like

Siri who work with voice recognition (Husnjak, Perakovic, & Jovovic, 2014).

In the present work, the term "Chatbot" is used for all programs that enable natural language

interaction. No matter if these are personified, whether the input works via keyboard,

microphone or touchscreen. In addition to the language, additional elements of the

communication are interpreted and integrated into the communication.

7

1.3. Evolution and Revolution

From an early stage, the human-machine interaction was an issue in research and later also

in the market. Dialogue systems attempted to make the interaction between the system and

human more "human-like". A Chatbot as a dialogue system is nothing new here. Already in

the 1960s, first attempts were made to computer-simulate a speaking person. More and more

Chatbots have come and gone on the market (Brandtzaeg & Følstad, 2017).

A prominent figure then and now was the computer scientist Alan Turing (Graham-

Cumming, 2012), who came up with an idea on how to determine whether a computer would

have a mind that was equivalent to humans. In the course of this test, a human questioner

with a conversation partner unknown to him leads a conversation. One person is a human,

the other a machine. If, after the intensive interview, the questioner cannot explicitly state

which of them is the machine, the machine has passed the Turing test (Rapaport, 2006).

Turing, widely considered one of the fathers of computer science, is in the origin of what

scientists and experts still refer to as the “Turing test” when they talk about smart bots

(Levesque, 2017).

In this period of euphoria, the development of the first Chatbot ELIZA by Joseph

Weizenbaum took place (Braun, 2013). This Chatbot simulated conversation “by pre-setting

text outputs to be triggered by specific text inputs” (Muldowney, 2017, p. 4). Although

Weizenbaum was unable to pass the Turing test with his development, ELIZA was the first

computer program that could faithfully fake a human being and thus gained considerable

fame (Khan & Das, 2017).

Other Chatbot examples like JabberWacky or ALICE did as well not pass the Turing Test,

but received significant awards in various contests (Janarthanam, 2017). To script the

knowledge and conversational content, techniques such as Artificial Intelligence Markup

Language (AIML) and ChatScript were developed. However, these developments have been

considered by experts not to be any real progress in AI or concerning building useful

conversational assistants (Ciechanowski et al., 2018a; Sameera and Woods, 2015).

The intelligent service Siri, released in 2011, was designed as the user's personal assistant.

The assistant developed by Apple, aimed at performing tasks such as making calls, reading

messages, and setting alarms and reminders. This development is of great importance in the

recent past of conversational interfaces. Another milestone of the AI set in the same year

8

with IBM's Watson, which can answer open questions in real time through natural language

processing (Janarthanam, 2017; Shevat, 2017).

The discussed selection of Chatbots is exemplary and not exhaustive. It serves as an overview

and should point out the essential lines in the evolution of dialogue systems or Chatbots.

9

1.4. Types and Classification of Chatbots

Chatbots can be used for different purposes. In addition to the distinction between B2B and

B2C bots, team and personal bots, there are various differentiations in this field (Radziwill

& Benton, 2017).

However, despite the diversity, all bots have one thing in common: they are text-based or

voice-based systems that use Natural Language Processing to communicate with their users

based on pre-defined rules or Artificial Intelligence. In the literature, there seems to be no

general distinction or definition of Chatbot types (Janarthanam, 2017; Shevat, 2017; Khan &

Das, 2017). However, it can be stated that mainly two different types of Chatbots exist:

• Rule-based-/Scripted-/Sequential bots

• Intelligent (AI-based) bots

The first type offers guided communication, using an existing set of pre-formulated rules

and answers. Rule-based bots have an informative character. Since no artificial intelligence is

used here, an open dialogue with them is not possible or at least only very limited. The

programming effort is comparatively low (Janarthanam, 2017; Grigorev et al., 2018). Unlike

the rule-based bots, the second type uses Artificial Intelligence techniques such as Machine

Learning and NLP to understand enquiries. These bots learn the language much like a child

and can create cross-references and recognise meaningful connections. They find answers to

open questions and understand customer concerns, even without them having to be

programmed exactly like that (Janarthanam, 2017).

According to Kassibgi (2017), Chatbots can be classified along different axes and categorises

Chatbots on the basis of the technique these Chatbots use to classify a given input and

generate meaningful output. The classification of Kassibgi (2017) is displayed in Figure 1. In

his analysis, he differentiates between the three categories "Pattern Recognition",

"Algorithms" and "Neural Networks".

Chatbots assigned to the pattern recognition category process input and compare it to a list

of predefined patterns. If the given input matches a pattern, the predefined answer is selected

and delivered as the required output. As these Chatbots only process inputs that exactly

match a pattern, the programmer must therefore define all possible input patterns during the

implementation phase (Shawar & Atwell, 2007).

10

Chatbots that are belonging to the "Algorithms" category, on the other hand, rely on

probabilistic methods such as Hidden Markov Chain (Ramesh et al., 2017) or Naive Bayes

(Kamphaug et al., 2018) to classify any input. These Chatbots are thus able to deal with inputs

that the developer has not explicitly programmed. Hence, these systems are more flexible

regarding input than the first category. However, they have a rigid set of input classes and

associated outputs (Shevat, 2017).

The ability to learn new input classes and outputs are theoretically accompanied by systems

belonging to the third category "Neural Networks"; these Chatbots will work out an answer

from a question using connections made from repetitive iterations obtained through training

data. The iterations are optimised to allow the neural network to generate the responses with

higher accuracy (Kassibgi, 2017).

From the perspective of the author, this classification not only differentiates three techniques

but also visualises their evolution in the field of input-output alignment.

Figure 1: Chatbot Classification according to Kassibgi

Source: Author’s elaboration, according to Kassibgi (2017)

Another approach to classify Chatbots was undertaken by IBM, considering the function of

a Chatbot as a classification category (Mason, 2017). The categories of the concerning

classification are briefly described in the following Table 1.

11

Table 1: Chatbot Classification according to Mason

Category Description

Support

Support-Chatbots dominate individual domains (e.g., User Help Desk). These Chatbots must be able to guide the user and help him in his concerns. This requires an awareness of contexts. For example, a question should be given concrete, contextually appropriate answer without providing further information.

Skills

Skills-Chatbots act on commands given by the user. Therefore, Chatbots hardly need an awareness of contexts. The user commands the Chatbot where, when, and what to do (e.g., "Turn on the light in the kitchen"). These Chatbots are currently mostly in "smart home" applications.

Assistant

The Assistant-Chatbot represents a middle ground between a support and a skills chatbot. These Chatbots are aware of several topics and can therefore be of help to a user in various situations. In addition to the function of a helper, they also conduct conversations with the user. This is to teach the Chatbot to have a dialogue. A well-known example of this is Siri Apple

Source: Author’s elaboration, according to Mason (2017)

According to the author, the abovementioned classifications are neither comprehensive nor

fully mature. Considering the classification of Kassibgi (2017), the naming of a category with

the term "Algorithm" seems rather unfortunate, since the two other categories also work

with algorithms. More appropriate would be terms that reflect the learning ability of the

systems, or the ability to deal with not explicitly defined input. The classification provided

by IBM (Mason, 2017), does not seem disjoint, as the "Assistant" category is a hybrid of the

other two categories.

12

1.5. Voice-Based vs. Text-Based

A question that is still controversial among the observers of the development of Chatbots is

the type of communication. Will speech or writing dominate? Hoare (2014), states that

communication with multiple parties is possible when using text, and that text can be

indexed, searched and translated, as well as tags and notes. In addition, summaries and

corrections can be made (Hoare, 2014).

Also, Jonathan Libov (2015), who works as a venture capital investor for Union Square,

prefers text to language. He points out that the comfort of writing is more important than

the convenience of speaking. Libov argues that text-based communication is more

comfortable because it saves time and brings fun. While speaking does not require so much

effort and is, therefore, more convenient, text-based interaction, in turn, is flexible and

personal. According to Libov (2015), Natural Language Processing is not yet good enough

to rely on oral communication alone. Instead, innovations in text-based communication

allow for faster responses. These new features can extract the selections made in a message,

allowing for the user does not to have to write the answer himself, just select it.

Advocates of voice-based communication emphasise that language can be more natural and

faster. Especially for in-house applications, for example, to regulate light or music, linguistic

instructions seem more natural and easier, according to van Doorn and Duivestein (2016).

Speech recognition is becoming increasingly accurate and works in some devices even at a

distance, such as Amazon's Echo. In fact, the four currently largest personal assistants are

voice-based: Siri, Now / Home, Cortana and Echo (Messina, 2016). Messina (2016) points

out that you cannot give instructions by text while driving and one does not want to take

notes in a lecture via the microphone: ultimately, there seems to be a need for both text- and

voice-based Chatbots.

13

1.6. Major Use Case and Role in Society

There are many possible use cases of Chatbots. In the media industry, they can be used, to

transmit news or sports results. The travel industry applies them for hotel and flight

bookings. Also, the banking sector integrates Chatbots to monitor accounts and transactions,

paying bills or executing transfers (Gentsch, 2018; Gladysh, 2018; Mindbowser, 2017).

Generally considered, Chatbots take place where support and service are required.

Customers want to obtain quick answers to simple questions or to carry out standardised,

simple processes. Thus, Chatbots are primarily used as an inbound touchpoint to answer

consumer questions about products, businesses and campaigns (Chakrabarti & Luger, 2015).

Almost every company offers customer service by phone or e-mail. However, this kind of

service is either overloaded or complicated to obtain from the customer's point of view.

Besides that, Service-staff loses valuable consulting time as they have to verify customer

numbers or other data first. Increasingly, outbound scenarios arise in which Chatbots actively

communicate with the customer according to defined rules and events (Gentsch, 2018).

Furthermore, the problem-solving skills of employees vary, due to various reasons for

example, experience of the employee or undermanning in the service centre. With a Chatbot,

many routine queries such as account balance and executed transactions can be quickly

answered. The integrated learning algorithms of the Chatbot continuously improve the scope

and quality of the knowledge base, which means that the customer's problem is less and less

likely to be routed to an employee (Gentsch, 2018).

To sum up, Chatbots are predominantly used with the primary goals of increasing efficiency,

reducing human chat agents, or empowering the ability to deal with a large number of

individual customer queries. According to these developments, Mou and Xu (2017) perceive

Chatbots as a promising alternative compared to traditional customer service.

Various statistics show that these developments will continue; according to a Survey

conducted by Mindbowser (2017), 95% of the consumers believe customer service is going

to be the principal beneficiary of Chatbots. Surveys predict that by 2020, 80% of companies

would like to use Chatbots in customer service. The market research institute Gartner

predicts that by 2021, more than 50 % of “enterprises will spend more per annum on bots

and Chatbot creation than traditional mobile app development” (Gartner, 2017, p.1).

14

Chapter 2. Chatbots and Artificial Intelligence

Chatbots are currently heavily charged with the performance attribute Artificial Intelligence

(AI) (Mrkalj, 2018). However, most bots are currently still implemented trivially. As a rule, a

database is scanned for specific keywords, by which predefined texts or text modules are

automatically controlled. More intelligent systems automatically detect textual findings that

are relevant from the Internet and then compile them into text (Janarthanam, 2017). Due to

these different manifestations, there are always conceptual ambiguities. Chatbots and

Artificial Intelligence are often mentioned in the same breath, and indeed there are many

links between the two technological developments (Janarthanam, 2017). However, Chatbots

and AI are not synonymous. Chatbots are an end product or application, while Artificial

Intelligence is an underlying technique that works in the background (Mayo, 2018). The

following chapters serve to define, conceptualise and understand the terms Artificial

Intelligence, Machine Learning, Natural Language Processing and thus the functionality of

Chatbots.

2.1. Artificial Intelligence

So far, human intelligence has not been defined accurately and demarcated, and there is a

wealth of different definitions and theories in the literature (Legg & Hutter, 2007). It is

perceived as a central skill of individuals and groups in both professional and private life

(Franken, 2010). Intelligence is generally defined as a combination of processes and

individual abilities (Legg & Hutter, 2007). There are several areas that include intelligent

thinking, for example: verbal-linguistic intelligence, interpersonal and intrapersonal

intelligence, logical-mathematical intelligence or spatial-visual intelligence (Gardner, 1996).

Since it is not yet clearly defined what human intelligence is, there is also no clear distinction

or definition for artificial intelligence as well.

In general, there are two main categories and four different approaches to define Artificial

Intelligence. The first category includes systems that are thinking and acting humanely. This

comprises intelligent systems able to perform activities like solving and recognising new

problems or making intelligent decisions. The counterparts are systems that are thinking and

acting rationally and thus choose and make decisions based on gathered knowledge (van de

Gevel & Noussair, 2013). Russel and Norwig (2010) collected and presented eight different

definitions according to the main categories mentioned above.

15

Table 2: Definitions of Artificial Intelligence

Thinking Humanly

“The exciting new effort to make computers

think…machines with minds, in the full and literal

sense.”(Haugeland, 1978)

“[The automation of] activities that we associate with

human thinking, activities such as decision-making,

problem solving, learning . . .” (Bellman, 1978)

Thinking Rationally

“The study of mental faculties through

the use of computational models.”

(Charniak and McDermott, 1985)

“The study of the computations that

make it possible to perceive, reason, and

act.” (Winston, 1992)

Acting Humanly

“The art of creating machines that perform functions

that require intelligence when performed by people.”

(Kurzweil, 1990)

“The study of how to make computers do things at

which, at the moment, people are

better.” (Rich & Knight, 1991)

Acting Rationally

“Computational Intelligence is the study

of the design of intelligent agents.”

(Poole et al., 1998)

“AI . . . is concerned with intelligent

behaviour in artifacts.” (Nilsson, 1998)

Source: Russel and Norvig (2010, p. 2)

It can be stated, that Artificial Intelligence covers various sub-disciplines such as Machine

Learning, Computer Vision or Natural Language Processing. According to this matter,

Artificial Intelligence must be seen as an interdisciplinary field, that is based and composed

of its subfields and the regarding applications and systems (Corea, 2017).

16

2.2. Machine Learning

Machine Learning, according to Copeland (2016) can be seen as a new approach to achieving

Artificial Intelligence. According to Raschka (2016), this sub-discipline is about self-learning

algorithms that extract knowledge from data to make specific predictions. The requirement

of human intervention for the manual derivation of rules and the development of models

based on the analysis of large amounts of data is therefore unnecessary.

Lantz (2013) or Suthaharan (2015) mainly present three factors, that influenced the

advancements in Machine Learning: (1) Big Data; (2) Computing Power; (3) Statistical

Methods. We live in a decade where vast amounts of data are available and above all recorded

digitally. Whereas in the past, data generated by computers was perceived as a secondary

product of digital technologies, it is now perceived as a significant resource (Alpaydin, 2016).

As a consequence, to this growing amount of data, high-performance computers had to be

developed and became a compelling necessity. The changes in Computing Power and Big

data can be seen as the stimulating factors for the development of new statistical methods to

analyse the concerning amounts of data and thus for a cycle of ongoing advancements

(Lantz, 2013).

There are mainly three different types of machine learning: supervised learning, unsupervised

learning and semi-supervised learning. The differences between these three tasks are briefly

described in the following paragraphs.

In the context of supervised learning, the system is already provided with example data. It

receives input variables and an output variable; both are used in an algorithm to learn the

concerning mapping function. The goal is to learn from the given labelled data and the

mapping function to classify new input data and predict the regarding outputs (Alpaydin,

2016; Cord & Cunningham 2014; Pacheco, 2015).

Unsupervised learning, on the other hand, allows to build a model based only on given inputs

independently. (Pacheco, 2015, Suthaharan, 2015). Both methods can also be combined in

so-called semi-supervised learning. In this case, both labelled and unlabelled data is used for

training. In contrast to the other two approaches, the learning algorithm receives only a part

of labelled data (Abney, 2007).

17

2.3. Natural Language Processing

Since human speech is unstructured and constantly changing in both written and spoken

form, its analysis is challenging but not impossible. In addition to the analysis and extraction

of knowledge, another approach is increasingly used. Through semi- and fully structured

data, texts can be generated. Both the text analysis and the text production can be realised

with the methods of the Natural Language Processing (Chowdhury, 2003; Kumar, 2011).

The following definition strengthens this position.

Liddy (2001) describes Natural Language Processing as

“a theoretically motivated range of computational techniques for analysing and representing

naturally occurring texts at one or more levels of linguistic analysis to achieve "human-like"

language processing for a range of tasks or applications” (p. 1).

Both in optical character and speech recognition, the usage of a language model, which

includes contextual information helps substantially. The extensive research on programmed

rules in computational linguistics showed that the best language model is based on learning

it from a considerable amount of example data (Alpaydin, 2016). Consequently, Machine

Learning provides the basis for Natural Language Processing systems to understand

extensive nuances that exist in Human Language and thus “learn to respond in a way that a

particular audience is likely to comprehend” (Marr, 2016, p. 2).

Nowadays the development of Natural Language Processing systems can be displayed both

in scientific research and practical technology. Therefore, examples are consumer products

and the concerning applications like Apple’s Siri or Skype Translator (Hirschberg and

Manning, 2015). Hirschberg and Manning (2015) described various purposes of Natural

Language Processing. Firstly, it can aim at supporting human to human communication in

the form of Machine translation. Secondly, Natural Language Processing can contribute to

the communication between humans and machines or even benefit both machines and

humans. Applying Machine Learning techniques, the system can analyse and learn from the

vast amount of data, in this case human language content.

18

Chapter 3. Satisfaction with Chatbot Characteristics

Satisfaction with products or services is one of the most important components of the

intangible assets of a company, its value is mainly determined by the actual income from an

existing or future customer relationship. In addition, companies use further potentials

including repurchases or cross-buying. Besides that, the satisfaction with products or services

can influence the acquisition of new customers by recommending the products or services

and can be pivotal in the development of new products (Caruana et al., 2015; Haislip &

Richardson, 2017; Heo & Song 2007).

3.1. Development of Satisfaction

The development of satisfaction concerning Chatbot Characteristics can be explained in

several ways. In science, the Confirmation/Disconfirmation-Paradigm (C/D-Paradigm) has

primarily prevailed (Homburg & Stock-Homburg, 2006). Thus, satisfaction arises when the

perceived performance of product use is compared with the expectations of the user

(Churchill & Surprenant, 1982).

If the perceived performance corresponds to the expectations (confirmation), this leads to

satisfaction. If the perceived performance exceeds the expectations (positive

disconfirmation), the result is an unusually high level of satisfaction. Dissatisfaction, on the

other hand, results if the perceived performance clearly does not meet the expectations

(negative disconfirmation). According to this approach, satisfaction should arise in

confirmation and positive disconfirmation (Krueger, 2016)

Other authors assume that only indifference arises when confirming the performance,

satisfaction is therefore only formed on positive disconfirmation (Hill, 1986). Furthermore,

it is assumed that the boundary between satisfaction and dissatisfaction is not characterised

by a score but as a tolerance zone. If the comparison of perceived performance and

expectations is within this range, the performance is considered satisfactory. With a very

strong positive disconfirmation, the customers are enthusiastic.

Perceived performance is the level of proficiency perceived by the customer. A performance

that is objectively equal can be perceived differently by different users. Essential sources of

expectations are the personal needs of, his past experiences, verbal recommendations by

acquaintances and promises concerning the product (Zeithaml et al., 1992).

19

The C/D-Paradigm is the basic model for explaining the development of customer

satisfaction. A number of psychological theories provide detailed approaches for a more

detailed explanation. These include the Two-Factor theory of customer satisfaction, which

explains the emergence of different satisfaction levels depending on the type of performance.

Satisfaction and dissatisfaction, according to this theory, are two independent dimensions,

meaning they cannot be considered as opposite poles of one dimension (Mowrer, 1960).

Satisfaction can also be regarded as a feeling; Accordingly, the importance of emotions has

been demonstrated in satisfaction research (Wirtz & Bateson, 1999). Satisfaction can thus be

defined as an attitude toward an object that includes the following aspects (Homburg et al.,

1999):

• The Cognitive Component: that is the formation of an opinion about an object, e.g. about

a product or service

• The Emotional Component: that are the feelings that occur when evaluating the

respective objects.

The relative impact of cognitive and emotional components on satisfaction with product

criteria may change over time. In a study by Homburg et al. (2006), the influence of the

cognitive component increased over time, while the influence of the emotional component

decreased. According to this, satisfaction should be considered as a dynamic construct.

Satisfaction and dissatisfaction are each triggered by different factors. The so-called hygiene

factors are responsible for dissatisfaction. If they are not fulfilled, dissatisfaction occurs. If

these factors are fulfilled, then satisfaction is not created, but only a neutral state, which is

called non-dissatisfaction. Satisfaction arises through the so-called motivators. If

expectations of motivators are not met, people experience a neutral state of non-satisfaction

(Mowrer, 1960; Herzberg et al., 1959).

20

3.2. Capturing Satisfaction with Chatbot Characteristics

The following section aims to present the key aspects of the Kano model and to analyse its

performance in terms of explaining the satisfaction with Chatbot Characteristics. This serves

as the groundwork when it comes to classifying Chatbot characteristics. The classification is

required to analyse how relevant motives to use a Chatbot affect the satisfaction with the

corresponding characteristics.

There are various methods for measuring and capturing satisfaction, which can be systemised

according to different criteria. Often a distinction is made between the type of measurement,

i.e. objective or subjective and the orientation of the measurement content (Bienstock

Mentzer, & Kahn, 2015; Bruhn, 2016).

The Kano model, named after its developer, is a feature-oriented process and refers to

product, service or interaction characteristics, judged by the user (Nascimento et al. 2012; Li-

Li, Lian-Feng, & Qin-Ying, 2011; Bi & Wang 2013) This model is based on the two-factor

theory and tries to determine the attributes that drive satisfaction (Griffin & Hauser, 1993;

Li-Li et al., 2011). The concerning attributes are classified according to their impact on

satisfaction. There are three types of attributes that cause different levels of satisfaction

(Kano et al., 1984; Sauerwein et al., 1996):

• Must-be attributes: People perceive these factors as the minimum criteria that must be

fulfilled by a product, service or interaction. As a result, dissatisfaction is created in

the event of non-fulfilment and neutral attitude upon fulfilment. Consequently, the

satisfaction cannot be significantly improved if the performance level of a "Must-be"

attribute continues to increase beyond the level expected by the customer or user

(Hussain, Mkpojiogu, & Kamal 2015; Matzler et al., 1996).

• One-dimensional attributes: These attributes are also referred to as expected criteria and

are thus explicitly requested. There is a linear relationship between the level of

confirmation and satisfaction. This means that as the degree of fulfilment increases,

satisfaction increases and vice versa. In highly competitive markets, only products or

services that have high levels of performance attributes have a chance. As a result,

performance attributes also play a prominent role in corporate communications. In

many product areas, this attributes category serves above all to differentiate it from

the competition (Hussain et al., 2015; Matzler et al., 1996).

21

• Attractive attributes: These attributes exert the most significant influence on the

satisfaction of the people. "Attractive" attributes are not expected and therefore not

explicitly formulated and demanded. If a corporation succeeds in equipping its

product with these attributes, this measure leads to positive disconfirmation and thus

to a high level of satisfaction. Otherwise, there will be no sense of dissatisfaction

when all the "Must-be" and "One-dimensional" attributes have the level of

performance defined by the customer (Hussain et al., 2015; Matzler et al., 1996).

The following Figure 2 illustrates the relationship between the degree of fulfilment for the

three attribute categories and the resulting level of customer satisfaction. The timeline should

make it clear that a degradation of the attribute categories is possible. For example, an

innovative problem-solving approach that is initially perceived as an "Attractive" attribute

can quickly degenerate into a "Must-be" attribute when adapted or copied in a short time by

many of the corporation's competitors (Witell & Fundin, 2005; Loefgren, Witell, &

Gustafsson, 2011).

Figure 2: Kano Model

Source: Huang (2017)

In addition, "Indifferent" and "Reverse" attributes are also of relevance in the Kano model.

In the case of "Indifferent" attributes, there is no influence on the satisfaction, regardless of

the fulfilment or non-fulfilment of the product characteristic. The "Reverse" attributes

22

describe a category indicating that respondents do not desire the requested product feature

and, in the presence, even leads to dissatisfaction (Kano et al. 1984, Loefgren & Witell, 2008).

One explanation for the current relevance of the Kano model is the recognition that a high

degree of compliance with different attributes does not necessarily lead to a high level of

satisfaction since the type of service requirement also influences perceived service quality

and thus satisfaction (Sauerwein, 2000).

According to Mikulic and Prebezac (2011), a vital advantage of the Kano model is the

avoidance of a strict and linear view on the impact of product attributes on customer

satisfaction. As a result, it is possible to identify specific attributes that can cause satisfaction

or dissatisfaction.

The Kano model is thus chosen as the research method for this work and will be explained and analysed in more detail in Chapter 7.

23

Chapter 4. Chatbot Characteristics

To investigate which characteristics of a Chatbot are of relevance relating to motives for

using a Chatbot, experiences, the concerning behaviour of humans using these systems and

technical developments in these areas need to be carried out. This is of relevance in order to

develop hypotheses, respective survey fundaments and thus to address the research question.

The author is aware that no clear distinction can be defined for the listed characteristics. The

term personality, for example, is based not only on an avatar but also on the emotions and

conversational abilities displayed.

4.1. Emotions

As early as 1977, Weizenbaum used his program ELIZA to observe that people build an

emotional relationship with the computer and attribute human characteristics to it. In this

regard, Reeves and Nass (1996) for the first time published broad-based investigations of

this phenomenon, setting a milestone in human-computer interaction in 1996. The

realisation that humans adopt computers with astonishing consistency as social beings have

added weight to social components of human-computer interaction. Previously, purely

cognitive criteria and the image of the computer as a pure tool for purposeful completion of

tasks were in the foreground (Reeves & Nass, 1996).

As presented in previous chapters, a Chatbot uses natural language to interact with its

opponent. It must be mentioned here, that the connection between language and emotion is

still widely discussed today. While common sense suggests that language has nothing to do

with emotion, it is evident that statements from other people affect our emotions. Likewise,

humans use their words to describe their own emotions or those of others. Accordingly, it

is believed that “this is the extent of the relationship between language and

emotion”(Lindquist, Maccormack, & Shablack, 2015, p. 1-2).

In various studies, this view is taken up and emotions are perceived as physical types, which

essentially differ from linguistic or conceptual processing (Shariff & Tracy, 2011; Fontaine

et al., 2013). On the other hand, the number of studies that give the language a much greater

role in connection with emotion is growing (Lindquist et al., 2015). Not least because of this

development, the relevance of emotion when interacting with Chatbots based on natural

language is once again clarified.

24

It can also be stated that emotions play an essential role in everyday life of a human being.

These emotions can be expressed during an interaction in a variety of ways. This can be

speech, gestures or text (Robinson & El Kaliouby, 2009). Since people interact with

Chatbots, the subject of emotion is thus highly relevant in this area. As described in the

previous chapters, the interaction with a Chatbot through the Conversational User Interfaces

can be both text-based or voice-based.

Through the described developments of the concepts of the NLP, for example, Emotion

Detection from a textual source can be implemented in a Chatbot (Hardik et al., 2018).

Overall, advances in technology enable Chatbots to recognise and express emotions, which

in turn paves the way for improved human-computer interaction (Robinson & El Kaliouby,

2009).

In this regard, however, it should be noted that at some moments people are unable to

recognise or communicate their own emotions. Besides that, emotions are also a major

challenge for machines because they first need a "proper position for emotion modelling"

and secondly, they “need advanced natural language processing” in order to develop the

emotion models (Hardik et al., 2018, p. 1).

Researches see the implementation of mirroring as a way to add emotion to a Chatbot

(McTear, 2016; Muldowney, 2017). Mirroring is central in all activities in which interpersonal

communication plays an important role. It is primarily used to connect with other people.

Mirroring can be described as the alignment of an emotional response between two

conversation partners (Iacoboni, 2008). In his study, McTear (2016) explains that mirroring

in Chatbots can take place through the choice of vocabulary, gestures or facial expressions.

An example would be so-called smileys or Emojis. Furthermore, he claims that Chatbots

equipped with the ability to mirror are perceived as emphatic.

The approach of equipping Chatbots with Emotion is taken up by other authors as well.

They argue that emotions have social and cognitive functions that are essential to an

intelligent system (Damasio, 2006; Muldowney, 2017). Empathy in the sense of being able

to see one situation, one problem, one action from the situation of the other concerned is

one of them. This social and cognitive function promotes and generates prosocial behaviour,

25

conveys communication skills, positively affects the length of the relationship, and reduces

aggressive behaviour (Segal et al., 2017).

With respect to the Kano method and based on this analysis, the author identified three

Emotion-related characteristics that are to be investigated in connection with motives for

using a Chatbot:

• Happiness

• Sadness

• Empathy

26

4.2. Personality

Another trait that needs to be considered is the perception of users regarding the human

body as a channel of communication, represented by an avatar in a Chatbot. There are several

definitions of the term avatar. Bahorsky, Graber and Mason, S. (1988) describe it as “a

pictorial representation of a human in a chat environment” (p.8), whereas Loos (2003)

characterises avatars as “a representation of the user as an animated character in virtual

worlds” (p. 17).

Studies and researches show that "human-like" behaviours associated with computer

technology have an impact on the perception of the user. Due to language production,

alternating conversation and reciprocal responding, users tend to personalise these

technologies (Nass & Moon 2000; Nass et al., 1995).

In their cutting-edge paper, Gillespie and Corti (2016) created situations in which a

participant had to have a conversation with a human, whose words were dictated by a

Chatbot. In this so-called speech shadowing, the human part repeats the words given by the

Chatbot in real-time and speaks them out loud towards the participant. As the human

interacts as the avatar using the Chatbot as a source, they are perceived as hybrid agents

(echoborg). The authors concluded that most of the participants that interacted with an

"echoborg", unlike those interacting with a simple text-interface, did not perceive their

conversation as artificial or robotic.

These findings strengthen the hypothesis that the human body or an avatar significantly

influences the interactions of human-chatbot communication. In a more recent study,

Ciechanowski et al. (2018b) attempted to analyse the impacts of avatars that are similar to

humans. For this purpose, they compared human-chatbot interaction with and without these

avatars. Their results highlight that for the participant's interaction with the Chatbot without

an avatar was more pleasant than the conversation with the enhanced Chatbot. Ciechanowski

et al. (2018b) suggest that according to the results, Chatbot “should not be designed to

pretend to be human” (p. 213).

To successfully conduct counselling sessions, dialogue systems need more than just specialist

knowledge. Social and non-verbal factors are crucial for the success of communication

between man and machine. Regardless of how an avatar is portrayed, this figure must be able

to mimic human communication behaviour. This will give the human being the necessary

27

degree of trust for receiving and conducting a conversation (Kuppevelt, Dybkjær, & Bernsen,

2005).

Another integral part of a personality and interpersonal interaction is humor. Humor can

play a vital role in a social relationship. It can serve to regulate a conversation or disguise

disagreements. In addition, research has shown that laughter can cause topic shifts in a

conversation or problem-solving. Furthermore, criticism and frustration can be slowed down

and prevented. This can, for example, be of great importance in a service conversation with

a Chatbot and create trust (Nijholt, 2003)

It should be noted here that the mentioned arguments regarding humour concern

interpersonal interaction. However, as described in previous chapters, technical

developments allow Chatbots to display humour through graphic, animation, and speech

synthesis technologies during an interaction.

With respect to the Kano method and based on this analysis, the author identified two

Personality-related characteristics that are to be investigated in connection with motives for

using a Chatbot:

• Humour

• Avatar

28

4.3. Conversational Abilities

As previously mentioned, a growing body of literature has examined artificial conversations

regarding human-chatbot interaction (Ciechanowski et al., 2018a; Radziwill & Benton, 2017).

Hill et al. (2015) for instance conducted a comparison between human-human online

conversations and human-chatbot interactions and concluded that the interactions with a

Chatbot had longer durations with shorter messages than the conversations between

humans. Furthermore, their study points out that the communication with Chatbots “lacked

much of the richness of vocabulary […]and exhibited greater profanity” (Hill et al., 2015, p.

245).

Conversation is probably the most common way of communicating and interacting, both in

private and work context. Several studies present conversation as an essential Social skill and

thus reinforce the importance of conversational abilities (Knight, 2016). According to Turkle

(2016), conversation is key to improve creativity, relationships, and productivity, thereby

strengthening the value of social skills.

It is necessary to state that, early systems worked exclusively with pattern recognition and

could just simulate a dialogue. Today's solutions, on the other hand, are sharply focused on

the assistant's character, with the fundamental goal of assisting the user in his case and

providing him with the necessary information through "human-like" interaction,

respectively, to carry out the processes he needs (Janarthanam, 2017).

However, there is room left for improvement. Chakrabarti and Luger (2015) stress that the

majority of Chatbots are more feasible for question-answer type dialogues, reverting to

several question-answer pairs. They criticise that Chatbots “are unable to hold a longer

conversation, understand the conversation, gauge whether the conversation is going in the

desired direction, and act on it” (Chakrabarti & Luger, 2015, p. 6879).

The current generation of Chatbots has noticeably improved conversational skills due to the

many advances in Natural Language Processing. Past Chatbot versions were mostly set to a

classic Q & A type dialogues (Shah et al., 2016).

Nowadays these developments allow a naturally fluid, colloquial dialogue so that the user can

interact with the system similar as with a human. If necessary, the Intelligent Chatbot receives

human reinforcement. This ensures that every utterance is understood, regardless of the

29

accent, pronunciation errors or word deviations. With these technological possibilities,

language can be correctly understood with its complexities such as corrections, slips,

paraphrases or dialect (Gentsch, 2018).

A suitable hypothesis regarding Chatbots conversational abilities is held in the work of

Cassell and Tartaro (2007) suggesting that “the goal of human-agent interaction […] should

not be a believable agent; it should be a believable interaction between a human and agent

in a given context” (p. 407).

With respect to the Kano method and based on this analysis, the author identified two

characteristics related to Conversational abilities that are to be investigated in connection

with motives for using a Chatbot:

• Context Awareness

• Common speech

30

4.4. Efficiency

Ostensibly, Chatbots are about new communication interfaces, which are bringing the next

evolutionary step to efficiency and convenience benefits (Gentsch, 2018). A significant

advantage of using Chatbots is that companies can offer their services where most users are,

such as in the messaging and social network apps (Muldowney, 2017). Facebook Messenger,

for example, registered a total of 1.3 billion active users per month in July 2018 (We Are

Social, 2018). Because of this, text-based Chatbots are more often used for communicating

with the user. Also, the presence of mobile devices provides a lower entry barrier for the

user when it comes to using a text-based Chatbot (Muldowney, 2017).

Furthermore, the demands on the speed and competence of information have increased

continuously in recent years (Peppard & Ward, 2016). Real-time service, as a customer's

claim, has become a commonplace and indispensable part of effective service (Buttle &

Maklan, 2015). As mentioned online and mobile communication has become a matter of

course. Moreover, with that, additional expectations go along with the communication

(Hennig-Thurau et al., 2010).

The performance, types and application fields of Chatbots were examined in detail in

Chapters 1 and 2. With respect to the Kano method and based on this analysis, the author

identified three Efficiency-related characteristics that are to be investigated in connection

with motives for using a Chatbot:

• Speed

• Accessibility

• Text vs. Voice

31

Chapter 5. Motives for Using a Chatbot

From a business perspective, Chatbots are primarily intended to be an important

communication channel for customers (Mou & Xu, 2017; Gentsch, 2018). Nowadays people

seem to prefer to write a text messages rather than calling or sending an e-mail (Battestini,

Setlur, & Sohn, 2010). This is also a generational issue because the so-called Millennials now

communicate almost exclusively via chats (Howe, 2015).

While the previous chapters highlight the benefits and acceptance of Chatbots, there are also

opinions and assessments that this acceptance is less substantial than anticipated. In their

ground-breaking study, Brandtzaeg and Følstad (2017) link this development primarily to the

needs or motives of the users, which in their opinion are too often ignored in the design of

Chatbots. This opinion is confirmed by Malhotra, Galletta and Kirsch (2008). They conclude

that the development of new interactive technologies such as a Chatbot “necessitates better

understanding of how users’ endogenous motivations influence their attitudes and

intentions, as well as related beliefs, including perceived ease of use and usefulness”

(Malhotra et al., 2008, p. 293). The following chapter aims to systematically examine possible

motives when using chatbots in order to link them with Chatbot Characteristics and measure

their influence. The study "Why People Use Chatbots" from Brandtzaeg and Følstad (2017)

serves as the fundament of this analysis.

5.1. Productivity

In the analysis of the Brandtzaeg and Følstad (2017) study, especially the relevance of the

motives in terms of productivity was highlighted. The frequency of reporting productivity as

one of the main reasons for using Chatbots was 68%. For the participants, speed, simplicity,

ease of use and comfort were the key factors. The mentioned reasons for this were related

with saving time or the elimination of waiting times. As well as the ease of use in general.

Furthermore, the researchers found out that, as expected, the simplicity of obtaining

information is a motive for using Chatbots. Other participants, albeit at a lower frequency,

preferred to ask a Chatbot than a real person instead. Also, the motive to adapt the Chatbot

to one's own needs was another analysed purpose in terms of productivity. Accordingly, the

identified productivity-related motives and purposes of the underlying study were taken up

and included in the questionnaire:

• Obtain assistance or information

32

• Receive quick answers

• Ease of Use

• Rather ask a Chatbot than a Human being

• Tailor a Chatbot to specific needs

• Convenience

A further example that confirms the motive of productivity is the Chatbot FAQ Chat,

designed by Shawar, Atwell and Roberts (2005). The researchers developed this Chatbot with

the intention of providing search results from the FAQs of a university with comparable

results as those provided by search engines. The goal was to show that Chatbots can be a

suitable alternative to traditional search engines. The result of the study confirmed the

researchers' hypotheses. Most users considered the solution as entirely positive and above all

as an attractive, new alternative to retrieve information from FAQ using natural language

questions (Shawar et al., 2005).

33

5.2. Entertainment

Another important motive is the purpose to use Chatbots for entertainment. In addition to

gaining information, people use Chatbots to chat and thereby banish boredom and entertain

themselves (Menal, 2017).

This is also illustrated by the original idea and goal of developing Chatbots. The first attempt

was as mentioned before in this paper, the Chatbot ELIZA, created by Weizenbaum. The

goal was to imitate the interpersonal conversation and thereby to entertain the user. It was

developed to imitate a psychotherapist in a treatment interview (Khan & Das, 2017). The

principle was relatively simple and based on the so-called keyword Matching. This technique

checks for the existence of a keyword. Based on the stored rules corresponding answers are

assigned and retrieved (Shawar & Atwell, 2007).

The purpose of entertainment through and with chatbots was also identified in the study of

Brandtzaeg and Følstad (2017) as another important motive for using Chatbots. A significant

number of respondents explained the value of using chatbots for entertainment as one of

their essential motives. The group of respondents considers chatbots to be satisfying and

entertaining. Furthermore, chatbots were also seen as a way to lose time and negatively fight

boredom. Accordingly, the identified entertainment-related motives and purposes of the

underlying study were taken up and included in the questionnaire:

• Overcome boredom

• Humour of Chatbots

• Entertainment

This reinforces the relevance of entertainment and fun as important aspects in terms of social

relationships and the relationship between Chatbots and humans (Brandtzaeg and Følstad,

2017). It can be seen in this regard that interactive systems that provide a sense of community

and support pleasant social interactions create a better experience for the user (Monk, 2000).

Especially since chatbots are designed to be more "human-like" than traditional interactive

systems this is important. According to Brandtzaeg and Følstad (2017), it is important to

emphasise that entertainment-related motives do not necessarily exclude or suppress other

motives and purposes such as productivity. Based on the results, people mostly want to do

their work productively, but do it in a fun and social manner.

34

5.3. Social and Relational

Various studies have been conducted in relation to the paradigm "Computer are Social

Actors". A key finding of this is that, similar to the communication with other humans,

people exhibit social traits when interacting with a machine. This retains valid even if they

are aware that they are interacting with a machine. (Nass & Moon, 2002; Reeves & Nass,

1996). In this regard, Nass and Moon (2002) argue that social responses to these

conversational agents are often automatic. As a result, a mindless process takes place in which

users focus on social cues rather than other agent features.

Also, Brandtzaeg and Følstad (2017) considered and analysed the potential social and

relational purpose of Chatbots. One important finding that emerged was that participants'

answers indicate social and relational motives, based on their experiences in interacting with

Chatbots. In this regard, the participants of the study in Chatbots see a way to avoid

loneliness or to satisfy the craving for socialisation. Other participants also indicate to use

chatbot with the intention to train and improve their conversational skills. Regarding

Chatbots in the field of language practice, there are opinions of experts who confirm the

potential positive benefit. Fryer and Carpenter (2006) for example state that “chatbots could

provide a means of language practice for students anytime and virtually” (p. 8). Accordingly,

the identified social and relational-related motives and purposes of the underlying study were

taken up and included in the questionnaire:

• Feeling lonely

• Talk to someone

• Train conversational skills

35

5.4. Novelty and Curiosity

Motives related to Curiosity and Novelty are particularly relevant to the selected target group

of the Portuguese Millennials. Since Chatbots can be seen as a new and innovative

technology, they are increasingly used by so-called innovators and early adopters (Brandtzaeg

& Følstad, 2017). With regard to new technologies, these are increasingly found in the

millennial generation (Blackburn, 2011).

In the study by Brandtzaeg and Følstad (2017), motives associated with Curiosity and

Novelty fall into the fourth main category. Above all, the participants' answers concern the

curiosity to discover something new, to test the skills and related boundaries as well as the

Chatbot in the development phase.

The relevance of curiosity regarding information search behaviour has already been

addressed several times. For example, McQuail (1987) noted that satisfying curiosity is a

significant factor in relation to motives using media. However, the study of McQuail

concerns older media like TV, rather than innovative and interactive systems such as

Chatbots.

Accordingly, the identified social and relational-related motives and purposes of the

underlying study were taken up and included in the questionnaire:

• Try something new

• Test out Chatbot skills

• Explore new technologies

• Curiosity

• Fascination

36

Chapter 6. Hypothesis Development

This thesis aims to analyse the relationship between motives for using a Chatbot and the

satisfaction with Chatbot characteristics. Based on the theoretical groundwork and

systematic literature review, five suitable hypotheses could be derived for the research

question. The variables in the survey, which are to be defined as the independent variables,

are motives for using Chatbots. These motives concern the areas of Productivity,

Entertainment, Social & Relational and Novelty & Curiosity. The dependent variables,

therefore, include the Satisfaction with Chatbot Characteristics: Emotion, Personality,

Conversational Abilities and Efficiency.

Figure 3: Hypotheses Development

Source: Author’s elaboration

H1 - Emotion: For people using Chatbots for entertainment purposes, Chatbot display of

emotion is more important than to people using a Chatbot for the purposes of productivity

H2 - Personality: For people using Chatbots for productivity purposes, an avatar is less

important than to people using a Chatbot for the motives of curiosity

H3 - Conversational Abilities: Characteristics linked to Conversational Abilities are more

relevant, when people use Chatbots for social relational purposes, as opposed to motives of

curiosity.

H4 - Efficiency: Characteristics linked to Efficiency are more relevant, when people use

chatbots for productivity purposes, as opposed to motives of entertainment

37

Chapter 7. Methodology - Survey

In the previous chapter, hypotheses have been set up, which must be scrutinised with

supporting data for their veracity. Therefore, a questionnaire is used for quantitative data

collection and analysis. The questionnaire was made available online to a group of Portuguese

Millennials. The survey intends to explore in what ways motives for using a Chatbot affect

the satisfaction with Chatbot characteristics and if and how chatbot characteristics that are

valued by users vary with the motives for using them. In the following, survey fundaments

and the development of the concerning questionnaire are presented. The chapter will finish

with the principles of the evaluation and interpretation process as well as the procedures of

data collection.

7.1. Survey Fundaments

According to Matzler et al. (1996), the Kano model plays a vital role in understanding which

attributes of a product have more than proportional influence on satisfaction as well as

characteristics that are an absolute must in the eyes of users.

The starting point for the Kano method is the identification of all relevant

attributes/characteristics for the product or service of interest (Berger et al., 1993). The

principal points discussed in Chapter 1 are intended to provide a systematic and in-depth

analysis of all issues, applications, and the entire environment of Chatbots. Essential factors

of Chatbots regarding the satisfaction and Chatbot characteristics that might be of influence

have been addressed and analysed in Chapters 3 and 4.

In this way, possible starting points for improvements of a Chatbot were identified, which

in turn could lead to an increase or decrease in satisfaction. As a result of the analysis, a total

of 10 different factors for a Chatbot were identified, which are presented in Table 3.

38

Table 3: Identified Chatbot Characteristics

➢ Emotion •Happiness •Empathy •Sadness

➢ Personality •Humour •Avatar

➢ Conversational Abilities •Context Awareness •Common speech

➢ Efficiency •Speed •Accessibility •Text vs. Voice

Source: Author’s elaboration

For each of the listed chatbot characteristics, two statements are developed, one "functional"

(positive) and one "dysfunctional" (negative). The functional one captures the reaction of

when the respective characteristic is present. The dysfunctional statement, on the other hand,

captures the reaction in the case of non-existence. (Sauerwein et al., 1996).

An example of a functional and dysfunctional statement related to a Chatbots is:

• Functional statement: The Chatbot is aware of context during your conversation

• Dysfunctional statement: The Chatbot is NOT aware of context during your

conversation

The answers options for these two statements are presented through a five-level rating scale,

which has the following levels (Sauerwein et al., 1996):

• 1 - I dislike it very much; 2 - I dislike it somewhat; 3 - I am neutral; 4 - I like it somewhat; 5 - I

like it very much

Subsequently, the answers concerning these two statements are entered in the Kano

evaluation table (Table 4). For example, if a subject answers the functional statement in the

above example with "I like it very much" and the dysfunctional with "I am neutral" the field

in the first row and the third column must be selected from the evaluation table.

39

Table 4: Kano Evaluation Table

Source: Author’s elaboration, according to Matzler et al. (1996)

The letters A, O, M, I, R and Q denote into which category the answer to these two

statements falls. The meaning of the categories is described in detail in Chapter 3.2. The

measuring of motives for using a Chatbot is made through self-developed questionnaire

based on the motives that have been analysed in Chapter 5. As a result of the analysis, a total

of 10 different factors for a Chatbot were identified. Besides the four main motive categories,

three other motives have been identified, that cannot be assigned to one of the main

categories. The motives are listed in Table 5. The concerning questionnaire is presented and

described in Chapter 7.2.

Table 5: Identified Chatbot Characteristics

➢ Productivity

Ease of Use; Rather ask a Chatbot than a Human being;

Receive quick answers; Ease of Use; Rather ask a Chatbot;

Tailor a Chatbot to specific needs; Convenience

➢ Entertainment Overcome boredom; Humour of Chatbots;

Entertainment

➢ Social and Relational Feeling lonely; Talk to someone; Train conversational

skills

➢ Novelty and Curiosity Try something new; Test out Chatbot skills; Explore new

technologies; Curiosity; Fascination

➢ Other

Ease of communication with a Chatbot about important

issues; Provision of automatic responses; Default method

of Customer support

Source: Author’s elaboration

Chatbot Characteristic

Dysfunctional (negative) Statement

I like it very

much

I like it somewhat

I am neutral

I dislike it somewhat

I dislike it very much

Functional

(positive)

Statement

I like it very much

Q A A A O

I like it somewhat

R I I I M

I am neutral R I I I M

I dislike it somewhat

R I I I M

I dislike it very much

R R R R Q

A(Attractive); O(ne-dimensional); M(Must-be); I(Indifferent); R(Reverse); Q(Questionable)

40

Another fundament of this survey was the target group Portuguese Millennials. To generate

a better understanding of why Millennials are of significant relevance when analysing the

subject Chatbot, the term is briefly analysed in the following.

There are several terms for the generation that spent its youth in the first decade of the new

millennium. For example, Millennial Generation, Nexus Generation or Gen Y (Ng &

McGinnis Johnson, 2015). For this work the term "Millennial" is used. Furthermore, a

birthday boundary for millennials is chosen. According to various authors Millennials are

born between 1980 and 2000 (Howe & Strauss, 2000).

This generation approaches live differently in many areas than older generations. One of the

main reasons for this is that the Millennials, as the first age group, grew up with the Internet

from an early age, and thus knows and uses many areas and processes in digital form only.

Through this digitization, millennials have been confronted with completely new

technologies of communication and interaction (Howe & Strauss, 2000).

This results in a completely different communication behaviour, which older generations

take over only gradually (Dabija, Brandusa, & Tipi, 2018). An equally common term for them

is therefore "Digital Natives". The technologically driven lifestyle promotes completely new

behaviours concerning the community, but in particular a new self-concept in dealing with

modern technical innovations (Jones & Shao, 2011)

41

7.2. Questionnaire Development

The aim of the study is to analyse the relationship between motives for using a Chatbot and

the satisfaction with the identified Chatbot characteristics. Therefore, the concerning

characteristics need to be to classified according to the extent of their influence on the

satisfaction. Furthermore, motives for using a Chatbot need to be captured and tested. As a

result, the construction of the regarding questionnaire, which has the following structure, is

necessary.

• Part 1- Demographic Questions & Experiences with a Chatbot

• Part 2- Motives for using a Chatbot

• Part 3- Satisfaction with Chatbot Characteristics

Attention is paid to a logical sequence of questions starting with general questions and ending

with a specific assessment. The questions are easy to understand. The specific sequence and

the division into three parts makes it easier for respondents to answer the questions (Doering

& Bortz, 2016). The questionnaire mainly consists of closed questions or predefined answers,

which has the advantage that the respondents' answers can be compared more easily (Hyman

& Sierra, 2016).

Due to the small number of questions, the questions concerning demographic and personal

information are provided in the first part of the questionnaire. Subsequently, it should be

determined on the basis of five questions, whether the test person has ever been in contact

with a Chatbot and what the general attitudes and experiences are. The answers are provided

in five-level Multi-item scale according to Likert (Bertram, 2007; Joshi et al., 2015). This

should help to analyse the preferences of the respondents and to confirm or reject the

hypotheses.

The second part has the purpose of capturing the motives for using a Chatbot. The answers

are provided in a five-level Multi-item scale according to Likert in order to generate higher

information content and differentiated measured values and to increase reliability (Bertram,

2007; Joshi et al., 2015). The statements are provided in the following table.

42

Table 6: Questionnaire 2nd Part - Motives for Using a Chatbot

Purpose/Motive ITEM I use/have used/would use a Chatbot…

Productivity

1 …to obtain assistance or information

2 …to get quick answers

3 …because they are easy to use

4 …when I have questions that might seem simple/stupid to another human

5 …because I can tailor a Chatbot to my specific needs, the more I interact with it

6 …because it is convenient

Entertainment

7 …when I’m bored

8 …when they have a sense of humor

9 …because it is entertaining

Social & Relational

10 …when I feel lonely

11 …because I like the sense of talking to someone when I use a Chatbot

12 …to train my conversational skills

Novelty & Curiosity

13 …to try something new

14 …to test out their skills

15 …to explore this new technology

16 …Because they are new and intriguing

17 …Out of curiosity

Other

18 … because it is easier to talk to a Chatbot than to talk to

people about important issues

19 … because it can provide automatic responses when others are not available

20 … Because it is was the default method of a certain customer support.

Source: Author’s elaboration

In the third and final part of the questionnaire, satisfaction with Chatbot characteristics is

captured with the goal to classify Chatbot attributes. Therefore, the Kano method was

chosen.

Mikulic and Prebezac (2011) tested different methods to classify product attributes for

strengths and weaknesses. The results of their study are based on the validity and reliability,

the informational value of the results and the technical capabilities of the methods tested.

One key finding was that the Kano questionnaire is one of “the only approaches […] capable

of classifying Kano attributes in the design stage of a product/service” (Mikulic & Prebezac,

2011, p. 44). Furthermore, participants of a survey do not need to have any experience with

43

the attributes (Mikulic & Prebezac, 2011). Both findings are highly relevant for this work, as

the concerning Chatbot does not exist and the attributes are hypothesized as well.

Also, in terms of the number of attributes to be tested, the Kano method sets no limits. The

approach of hypothesized provision/non-provision-based mode of attributes chosen for the

questionnaire is perceived as valid and highly reliable for assessing the kano model (Mikulic

& Prebezac, 2011).

Table 7: Questionnaire 3rd Part - Satisfaction with Chatbot Characteristics

Purpose/Motive Statement

Emotion

The Chatbot is (NOT) able to express happiness during your

interaction

The Chatbot is (NOT) empathetic during your interaction

The Chatbot does (NOT) express sadness during your interaction

Personality The Chatbot does (NOT) have a sense of humor

The Chatbot is (NOT) embodied by an Avatar

Conversational

Abilities

The Chatbot is (NOT) aware of context during your conversation

The Chatbot is (NOT) able to process common speech

Efficiency

The Chatbot is (NOT) fast when processing your

inquiries/questions

The Chatbot is (NOT) easy to access

The communication with a Chatbot is (NOT) Text-based

Source: Author’s elaboration, according to Sauerwein et al. (1996)

44

7.3. Questionnaire Evaluation

By relating the answers of the functional and the dysfunctional statements, it is possible to

determine the category of the corresponding characteristic. The frequency of the categories

occurring in each cell of the Kano-Scorecard is noted. After that, the cell frequencies

belonging to the same category are added to the evaluation table, and the result is entered in

a frequency table in the form as it is presented in Table 8. The frequency table indicates the

number of times respondents chose a particular category.

The following table contains the fictitious frequency distribution of 100 subjects who

answered the statement mentioned in the example of a Chatbot displaying context awareness

during a conversation.

Table 8: Kano Frequency Distribution

Characteristic A O M I R Q Total Category

Context Awareness 45 (45%)

10 (10%)

25 (25%)

15 (15%)

4 (4%)

1 (1%)

100 (100%)

A

Source: Author’s elaboration, according to Matzler et al. (1996)

On the basis of this frequency table, it must then be determined whether the requested

chatbot characteristic represents a "Must-be", "One-Dimensional" or "Attractive" attributes.

There are various possibilities for determining the category. The most common evaluation

and interpretation are based on the most frequent choice, that is, the response category that

has been chosen most frequently determines what type of request represents the queried

product characteristic. In the example, the most frequent category is "A", the requested

chatbot characteristic thus represents an "Attractive" attribute.

In some cases, the described way of determining the category may not be adequate. This

applies if the two categories have very similar frequency or even the same frequency. In such

cases, the following decision rule is used in this work:

• M> O> A> I

This decision rule is prescriptive, stating that "Must-be" attributes are more important than

"One-dimensional" attributes, "One-dimensional" attributes are more important than

"Attractive" attributes, and "Attractive" attributes are more important than Indifferent

attributes (Sauerwein, 2000; Huang, 2017).

45

A further figure plays a role in the evaluation the Kano-Method, this is the satisfaction

coefficient. This figure can be subdivided into the coefficient of satisfaction (SI) and the

coefficient of dissatisfaction (DI). The coefficients indicate how greatly the satisfaction can

be increased by the presence of a product feature or to what extent this merely avoids

dissatisfaction (Matzler et al., 1996). The following are the formulas for computing SI and

DI (Berger et al., 1993):

• SI = (A + O)/ (A + O + M + I)

• DI = (- 1)(O + M)/ (A + O + M + I)

The values of these coefficients range from zero to one for SI and from zero to minus one

for DI. Values close to one and minus one, indicate that a product feature is important for

satisfaction or avoidance of dissatisfaction. Coefficients close to zero, on the other hand, are

regarded as unimportant by the satisfaction rating or avoidance of dissatisfaction (Matzler et

al., 1996).

46

7.4. Data Collection Procedures

For investigation of the hypotheses and research questions, a quantitative survey in the form

of a questionnaire was chosen. To collect a larger sample in a timely and cost-effective

manner, an online survey was conducted. The questionnaire (see Attachments) was created

with the service SurveyMonkey and made available online to the subjects. SurveyMonkey

enables an attractive, external design of the questionnaire, an anonymous data collection and

a transfer of the data into the evaluation programme.

To test and evaluate the questionnaire, a pre-test with talked discussion with a group of ten

Portuguese Millennials was conducted before the main survey. This procedure aimed to find

out if the questions are understandable for the subjects, if there are technical problems, useful

comments and how long it takes to answer the questionnaire. As a result of the pre-test,

some statements were adjusted for a better understanding of the items and four new

questions concerning the participants’ experiences made with a Chatbot were included.

A convenience sample was gathered, since the survey was sent to university undergraduates,

master-level students and alumni of the faculties of Economics and Engineering of the

University of Porto, in Portugal and shared in social media (Facebook and LinkedIn

platforms mainly). Only the Portuguese version of the survey was emailed and shared so that

Portuguese could answer the survey with negligible linguistic barriers and in order to

preselect and filter the target group.

Structure-wise the survey was composed of 52 items. The first six items are related to

demographic information; four items concern the participants’ experiences with a Chatbot,

20 items are about the participants’ motives of using a Chatbot and 20 items refer to the

satisfaction with Chatbot characteristics. The survey also included two open questions, with

the intention of allowing the participants comments, as well as email address, in case interest

in receiving a summary of the survey results.

47

Chapter 8. Results

In the following chapter, the relationship between motives for using Chatbots, and the

satisfaction with Chatbots characteristics is analysed using IBM SPSS Statistics in its 25th

version. In Chapter 8.1, the sample of this empirical research is explained in more detail,

afterwards the relevant data and the individual dimensions are graphically displayed,

descriptively evaluated and interpreted.

8.1. Sample

A total of 317 respondents agreed to participate in the survey whereof 45 questionnaires

were not completed; 14 questionnaires were excluded, as, despite the preventive measures

undertaken, these respondents did not match with the target group of Portuguese millennials.

An adequate sample of 258 responses remained, allowing for the intended statistical

procedures.

Figure 4: Gender Composition of the Survey

As displayed in Figure 4 the number of female and male respondents was almost equal. Only

one respondent preferred not to answer this question.

As the target group of this study were Portuguese Millennials (subjects currently aged

between 18 and 38 years old). The age distribution played a significant role regarding the

analysis. Most of the respondents were under 26 years old, totalizing over half of the sample

(M = 25.36; SD = 4.37; Min = 18; Max = 38).

13151%

12649%

Gender

Female Male

48

Figure 5: Distribution of respondents’ age

When the subjects were questioned on their academic level the majority answered that they

hold a Bachelor degree or equivalent and there is a considering number of participants, who

already have completed their master degree and doctorate as is observable in table 9.

Table 9: Academic Level of the Respondents

Academic level Frequency Percentage

Doctoral or equivalent 5 2%

Master or equivalent 75 29%

Bachelor or equivalent 136 53%

High school 37 14%

Middle school 5 2%

Total 258 100%

An interesting analysis point is also present in the distribution of the Employment Status

field of the respondents displayed in Figure 6. Almost half of participants are employed for

wages. As the survey was sent to some faculties of University of Porto, the presence of 39%

working students is a comprehensible result.

148

78

32

0

20

40

60

80

100

120

140

160

18-25 26-30 31-38

Res

po

nden

ts

Age Range

Age group

49

Figure 6: Employment Status of Respondents

10139%

208%

2911%

10842%

Employment Status

Working student Unable to work

Self-employed Employed for wages

50

8.2. Instrument Validity and Reliability

The questionnaire of Motives for using a Chatbot developed for this work was subjected to

internal reliability tests and validation procedures. For this purpose, Cronbach’s alpha and

exploratory factor analysis were applied. The psychometric properties of the instrument are

covered in the following paragraphs.

Considering that the Kano model does not yield a quantitative, but rather a qualitative scale

that is not clearly translated into numeric fashion, these procedures were not statistically

viable, which considering the trove of literature previously presented and the specific

rationale of the model an acceptable option.

Reliability

In order to determine reliability, Cronbach’s Alpha (α) was calculated for the questionnaire

part relating to motives for using a Chatbot. Even though there are different studies on the

acceptable values of Cronbach’s Alpha a value higher than 0.700 denominates the scale as

suitable for analysis in statistical models (Cortina, 1993; Bland & Altman, 1997).

The value of Cronbach’s Alpha for each dimension of the concerning questionnaire part is

presented in the following table. The scale displays very good reliability, with a mark higher

than 0.700. The results for this scale are of great relevance, as it was self-developed based on

the reviewed literature.

Table 10: Cronbach Alpha of Motive Dimensions

Dimension No. of Items Cronbach’s Alpha

Convenience 5 .932

Social 6 .809

Exploring 4 .793

51

Validity

Exploratory factor analysis using principal components analysis with a varimax rotation was

used to test the instrument validity of the questionnaire part including the motives for using

a Chatbot. To assess the factorability of each set of items, the KMO and measure of sampling

adequacy and Bartlett’s test of sphericity was used. The scores of all items concerning Kaiser-

Meyer-Olkin (KMO) were above the commonly recommended value of .6. Reliability results

are presented in table 10 and exploratory factor analysis results are presented in table 11.

Table 11: KMO & Bartlett's Test

Kaiser-Meyer-Olkin Measure of Sampling Adequacy. .840

Bartlett's Test of Sphericity

Approx. Chi-Square 2216.778

df 120

Sig. .000

Table 12: Factor loadings from Principal-Components Analysis for the Motives for using Chatbot

Item-Statement

Factor loading

Communality

1 2 3

15 - to explore this new technology .920 0.863

17 - out of curiosity .880 0.804

16 - because they are new and intriguing

.866 0.824

14 - to test out their skills .824 0.708

13 - to try something new .786 0.746

52

2 - to get quick answers .824 0.684

1 - to obtain assistance or information

.769 0.595

3 - because they are easy to use .727 0.539

20 - because it is was the default method of a certain customer support.

.691 0.480

6 - because it is convenient .658 0.437

19 - because it can provide automatic responses when others are not available

.605 0.435

11 - because I like the sense of talking to someone when I use a Chatbot

.828 0.723

10 - when I feel lonely .819 0.711

12 - to train my conversational skills .809 0.670

18 - because it is easier to talk to a Chatbot than to talk to people about important issues

.547 0.328

As can be observed, the factor structures presented in Table 11 are not exactly coincident

with the developed scale that included five dimensions (Productivity, Entertainment, Social

& Relational, Novelty & Curiosity and Other). All items from the originally formulated

"Entertainment" dimension were dropped due to low factorability and one item of the

originally formulated scale of "Productivity" (4 - …when I have questions that might seem

simple/stupid to another human) was also dropped due to low communalities. Hence, as it

is observable in table 11, 3 dimensions emerged that although not exactly the same as

originally formulated are still quite consistent with Furthermore, the formulated rationale.

Hence, "Productivity" dimension was renamed, as other items factored in the same

dimension that more than work itself were focused on convenience. The Social & Relational

and Curiosity & Novelty dimensions were renamed as well. Regarding the Social dimension

53

this is due to the fact, that statements concerning relations do not exist. The Exploring

dimension does not necessarily relate with curiosity and novelty. According to these findings

the following three dimensions can be defined:

• Convenience

• Social

• Exploring

54

8.3. Descriptive Analysis

After the general demographic questions at the beginning of the survey, the subjects were

introduced to the topic and the overall acceptance, experience, motives and the satisfaction

with Chatbots was assessed. The results are presented in the following paragraphs.

Experience with Chatbots

The following figures present the percentage distribution of the 258 respondents for each

statement, according to their level of agreement. As mentioned in Chapter 7.1, a five-point

Likert scale was applied in the questionnaire. Here “1” stands for strong disagreement of the

respondent towards the statement, and “5” for strong agreement and thus for the opposite.

The introductory question asked, whether the individual test subjects are already familiar

with the subject Chatbot. Over half of the respondents claimed to agree with this statement.

Only 10% stated, that they do not know much about this topic.

Figure 7: Respondents Familiarity with Chatbots

Figure 8 depicts that the sample group displays a very different level of agreement on their

usage of Chatbots during the past. More than 30% claimed they had used Chatbots rarely or

not at all. On the other hand, more than half of the participants claimed to have used

Chatbots in the past.

10%

16%18%

36%

29%

0%

5%

10%

15%

20%

25%

30%

35%

40%

1 2 3 4 5

Per

cen

tage

of

Res

po

nd

ents

Level of Agreement

"I’m familiar with the concept of Chatbots"

55

Figure 8: Respondents Usage of Chatbots in the Past

Figure 9: Respondents Daily Usage of Chatbots

20%

13% 14%

34%

20%

0%

5%

10%

15%

20%

25%

30%

35%

40%

1 2 3 4 5

Per

cen

tage

of

Res

po

nd

ents

Level of Agreement

"I have used chatbots in the past"

44%

29%

19%

6%3%

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

50%

1 2 3 4 5

Per

cen

tage

of

Res

po

nd

ents

Level of Agreement

"I use chatbots in my daily life"

56

Figure 10: Respondents Curiosity

The curiosity about Chatbots is present in the majority of participants: almost half of all

respondents considered Chatbots as a very interesting topic, although most participants don’t

have a notion of using Chatbots in a daily basis (cf. Figure 9).

Motives for Using a Chatbot

The level of overall motives of the entire sample group to use Chatbots in general are

presented in Figure 11, where the motive for using a Chatbot for convenience, displays the

highest value and social motivations display the lowest.

Figure 11: Respondents Motives for Using a Chatbot

8%

17%

30%

35%

10%

0%

5%

10%

15%

20%

25%

30%

35%

40%

1 2 3 4 5

Per

cen

tage

of

Res

po

nd

ents

Level of Agreement

"I’m curious about the topic of chatbots"

2,9612

3,6628

1,8721

0

0,5

1

1,5

2

2,5

3

3,5

4

Exploring Convenience Social

Lev

el o

f A

gree

men

t

Motive

Motives for using a Chatbot

57

Satisfaction with Chatbot Characteristics & Coefficient of Satisfaction

As described in detail in Chapter 7.3, the evaluation of the Kano method is an integral part

of assessing satisfaction with Chatbot characteristics. Furthermore, the results serve as the

dependent variables for, testing the established hypotheses. The following table presents the

frequency distribution of the classification defined on the basis of the Kano evaluation tables.

Furthermore, the final category and the satisfaction coefficient for each characteristic was

computed.

Table 13: Kano-Method Evaluation

Characteristic A O M I R Q SI/SD Total Dominant Category

Happiness 15 0 14 201 15 13 SI= 0,065 SD= -0,060

258 (100%)

I

Empathy 16 3 14 201 14 10 SI = 0,081 SD= -0.072

258 (100%)

I

Sadness 2 1 3 195 45 12 SI= 0,014 SD= -0,019

258 (100%)

I

Humor 17 12 11 186 25 7 SI= 0.128 SD= -0.101

258 (100%)

I

Avatar 11 3 21 206 11 6 SI= 0,058 SD= -0,095

258 (100%)

I

Context Awareness 21 20 29 166 10 12 SI= 0,173 SD= -0.207

258 (100%)

I

Common Speech 26 21 20 174 11 6 SI= 0,195 SD= -0,170

258 (100%)

I

Speed 35 69 27 121 1 5 SI= 0,412 SD= -0,380

258 (100%)

I

Accessibility 29 65 30 123 3 8 SI= 0,380 SD= -0,384

258 (100%)

I

Text Based 15 11 18 194 9 11 SI= 0,109 SD= -0,121

258 (100%)

I

Note: SD/SI (Coefficient of satisfaction and coefficient of dissatisfaction)

58

The evaluation displays that all analysed characteristics could be classified as indifferent-

attributes and thus have no direct impact on satisfaction or dissatisfaction, however,

although, the satisfaction coefficient of almost all characteristics shows no noteworthy

deviations, "Speed" and "Accessibility" can be analysed deeper. As Figure 14 depicts, over

60 participants perceive "Accessibility" of a Chatbot as a "One-Dimensional" attribute and

almost 30 describe the characteristic as an "Attractive" attribute. Further relevance of this

feature is displayed through 30 respondents, who perceive "Accessibility" as a "Must-be"

attribute.

Figure 12: Relevance of Chatbot Accessibility for Respondents

The characteristic "Speed", which also belongs to the Efficiency dimension defined in

Chapter 4.4 displays similar results. Over 30 respondents perceive the speed performance of

a Chatbot as an "Attractive" attribute. Almost 70 participants rated this characteristic as an

"One-Dimensional" attribute. Further 30 respondents perceive this feature as a "Must-be"

attribute. The frequency distribution is displayed below in Figure 13.

29

65

30

123

38

0

20

40

60

80

100

120

140

A O M I R Q

Num

ber

of

Res

po

nd

ets

Kano Category

Accessibility

59

Figure 13: Relevance of Chatbot Speed for Respondents

A Further result, that presents the relevance of the Emotion dimension of Chatbot

Characteristics, is presented in Figure 14. Less than five respondents valued the presence of

the characteristic sadness as an "One-Dimensional" or "Attractive" attribute. Rather it is

perceived as a "Reverse" attribute by almost 50 respondents.

Figure 14: Relevance of Chatbot Sadness for Respondents

35

69

27

121

1 5

0

20

40

60

80

100

120

140

A O M I R Q

Num

ber

of

Res

po

nen

ts

Kano Category

Speed

2 1 3

195

45

12

0

50

100

150

200

250

A O M I R Q

Num

ber

of

Res

po

nd

ents

Kano Category

Sadness

60

Motives for Using a Chatbot & Satisfaction with Chatbot Characteristics

The following Figure represents the 3 motive dimensions in relation with the Characteristic

dimension Efficiency and the concerning attribute Speed. All three motive dimensions

display similar results. Excluding the results for social motives, when analysing the One-

dimensional category.

Figure 15: Speed of a Chatbot in Relation with Respondents Motives

Similar results are displayed for the Characteristic Dimension Efficiency with the concerning

Characteristic Accessibility. Almost 40% of the respondents with convenience motives

perceive Accessibility as a "One-Dimensional" or as an "Attractive" attribute.

Figure 16: Accessibility of a Chatbot in Relation with Respondents Motives

Q R I M A O

Exploring 1,7 46,8 8,1 13,9 29,5

Convenience 0,8 0,4 47,7 8,4 13,8 28,9

Social 1,6 69,4 1,6 14,5 12,9

0,0

10,0

20,0

30,0

40,0

50,0

60,0

70,0

80,0

Per

cen

tage

of

Res

po

nd

ents

Kano Category

Motives & Speed

Q R I M A O

Exploring 1,7 0,6 49,1 11,0 12,1 25,4

Convenience 1,3 1,3 47,7 11,7 11,7 26,4

Social 1,6 58,1 4,8 16,1 19,4

0,0

10,0

20,0

30,0

40,0

50,0

60,0

70,0

Per

cen

tage

of

Res

po

nd

ents

Kano Category

Motives & Accessibility

61

8.4. Discussion

This work has dealt with the research question in what ways motives for using a Chatbot

may affect the satisfaction with Chatbot characteristics and if and how Chatbot

characteristics that are valued by users vary with the motives for using them. The questions

asked could be only partially answered by a descriptive analysis. According to Neill’s (2008)

statistical decision tree and considering the nature of the scales there are no more statistical

tests that can be conducted with this sample. All of the determined Chatbot characteristic

were perceived as "Indifferent" attributes by the respondents, thus making it impossible to

test the originally formulated hypotheses.

With regards to the experience with Chatbots, the majority of the respondents (54%)

reported having already had experienced interaction with a Chatbot. Although the majority

of participants seem to rarely use Chatbots, the curiosity about the topic is present and

research indicates it might be increasing (Brandtzaeg & Følstad, 2017). This finding is also

reinforced by the study's findings regarding the motives to use Chatbots: Exploring could be

identified as one of the three key dimensions regarding the motives to use a Chatbot.

Contrary to other authors and developers who highlight the relevance of avatars, emotions,

personality in in Chatbots and describe these characteristics as essential when developing

Chatbots (Janarthanam, 2017; Nass & Moon 2000; Nass et al., 1995) and even claim that the

goal must be to equip a Chatbot with human attributes, to be perceived "human-like" by the

user and thus to generate a successful interaction (Gentsch, 2018; Kuppevelt, et al., 2005), it

was interesting to realize that Portuguese Millennials, do not seem to consciently value these,

but rather consider speed and accessibility as important features, delegating all other

characteristics as indifferent in the Kano model.

Furthermore, this survey revealed that Characteristic dimensions like "Emotion" or

"Personality" do not seem to have any relevance to Portuguese Millennials. Characteristics

like happiness, sadness or empathy do not cause an increase in satisfaction, instead they partly

might have a reverse effect on satisfaction. Instead the sample group valued Efficiency

related Characteristics as important. This is confirmed through the analysis of the

relationship between the motives and characteristics. All three motive dimensions display a

positive effect on satisfaction in relation with Characteristics Accessibility and Speed.

Another key finding is the abstinence of Entertainment as a motive to use Chatbots. While

62

in the study by Brandtzaeg and Følstad (2017) the dimension "Entertainment" was assessed

as an important motive, this dimension did not seem to make sense in the Portuguese

Millennials population; instead, 3 new dimensions could be defined that can be used to

capture motives for using a Chatbot (Convenience, Exploring and Social), where the motive

to use Chatbots for convenience purposes thereby displays the highest value.

63

Chapter 9. Limitations and Further Research

In this work, results and tendencies could be carried out by the descriptive analysis and thus

new insights were gathered. However, the hypotheses could not be tested, based on the

results of the survey, that presupposed that different characteristics would yield different

values and instead subjects classified all characteristics as indifferent to them. The Kano

model was chosen because it is widely used in research for decision making and to assess

products and the concerning features, but perhaps future research could benefit from a

model that allows for the computation of levels of satisfaction with characteristics in a more

forward, numeric fashion.

With respect to the survey-results and considering the Kano model as a qualitative model it

can be stated, that no further statistical tests could be done, except the frequency analysis

presented in Chapter 8.3.

Further research is needed to eliminate the limitations of this work. It can also be stated that,

the acceptance, experience, satisfaction with Chatbots has not been thoroughly researched

in general and the developed CSE-Scale could be validated in other segments of population

and other countries. In this study the attitude towards chatbots, especially with regard to the

motives of Portuguese of millennials, was considered to be of use. For this reason, a study

with a larger sample, representative of the population, would be interesting to analyse. In the

present study, questions regarding a fictitious Chatbot and the corresponding characteristics

were described and queried; it would therefore be interesting to replicate the present study

based on a specific Chatbot and its users.

64

Conclusions

Conclusions are derived from the survey analysis and the concerning discussions, the current

state of knowledge and the presented use of Chatbots in this work. In addition, implications

for practice are named.

Based on the research within this work it can be stated that Chatbot Characteristics and

motives for using this technology have been overlooked and only partly investigated in

literature. The current literature regarding Chatbots is mainly concerned with technical

approaches and focuses less on perceptions of the user and the regarding Chatbot

characteristics, and there is still a gap in studies that address the implications of this

technology and customer reactions to it. However, in general and during the last years the

topic Chatbot started to raise interest again for experts, researcher and companies (Gentsch,

2018). This can be traced to the technological developments in the area of Artificial

Intelligence and Natural Language Processing.

Therefore, this work aimed at verifying and analysing the relationship between Chatbot

characteristics and motives for using them. According to that, the research on the topic of

this Master's thesis has been divided into three phases. In the first phase, the subject Chatbot

was developed and reviewed on the basis of relevant, past and present knowledge. The

second phase defined the development of relevant Chatbot characteristics, as well as

potential motives to use Chatbots. The outcomes of the second research phase are based on

existing and topic-related studies. In the third phase, the causal connection between

characteristics and motives was to be pointed out through a questionnaire-based survey. It

could be shown that characteristics such as efficiency and accessibility are of particular

relevance to the selected target group.

In general, Portuguese Millennials seem to regard almost all of the indicated characteristics

as indifferent. One reason for this could be that the target group has little knowledge of

Chatbots and their preparation. This is coherent with several studies that presented findings

concerning the perception of Chatbots. According to these studies, a lot of the time people

do not realize that they are speaking or texting with a CB or a virtual agent. This might be a

reason for the low awareness of this topic in the general population.

65

It seems like the target group considers the Chatbot as a technology which serves to obtain

information quickly and efficiently and do not seem to value personality, emotions or an

avatar characteristics in this technology – features that usually contribute to increase costs

associated with it in terms of value spent in research and development and time to conclude

projects; rather, our inquired users seem to be operating in a different paradigm where the

features that really matter are speed and accessibility, contrary to a HCI "humanoid"

approach.

66

References

ABNEY, S. (2007). Semisupervised Learning for Computational Linguistics. CRC Press. ALESANCO, Á., SANCHO, J., GILABERTE, Y., ABARCA, E., GARCÍA, J. (2018). Bots

in messaging platforms, a new paradigm in healthcare delivery: application to custom prescription in dermatology. In: Eskola H., Väisänen O., Viik J., Hyttinen J. (eds) EMBEC & NBC 2017. EMBEC 2017, NBC 2017. IFMBE Proceedings, 65. Singapore: Springer.

ALPAYDIN, E. (2016). Machine learning: The new AI. Cambridge, MA: MIT Press. ARRABALES, R., LEDEZMA, A., & SANCHIS, A. (2013). Characterizing and Assessing

Human-Like Behavior in Cognitive Architectures. Berlin: Heidelberg. BAHORSKY, R., GRABER, J. & MASON, S. (1988). Official Internet Dictionary: A

Comprehensive Reference for Professionals Reviews. Portland: ABS Consulting. BASSETT, C. (2018). The computational therapeutic: exploring Weizenbaum’s ELIZA as a history of

the present. AI & SOCIETY. DOI: 10.1007/s00146-018-0825-9 BATTESTINI, A., SETLUR, V., SOHN, T. (2010). A large scale study of text-messaging use.

ACM International Conference Proceeding Series. 229-238. doi:10.1145/1851600.1851638.

BELLMAN, R. E. (1978). An Introduction to Artificial Intelligence: Can Computers Think?. Boyd

& Fraser Publishing Company. BERTRAM, D. (2007). Likert Scales… are the meaning of life: CPSC 681-Topic

Report. Retrieved from http://poincare.matf.bg.ac.rs/~kristina/topic-dane-likert.pdf. Accessed on 13.07.2018

BIENSTOCK C.C., MENTZER J.T., KAHN K.B. (2015). How are Service Firms Measuring

and Managing Service Quality/Customer Satisfaction?. In Wilson E.J., Hair J.F. (eds) Proceedings of the 1996 Academy of Marketing Science (AMS) Annual Conference. Developments in Marketing Science: Proceedings of the Academy of Marketing Science. Cham: Springer.

BI, J. & WANG, X. (2013). Integrating Kano’s model and quality function deployment to evaluate

exposition—from the perspective of exhibitors. 2013 International conference on quality, risks, maintenance, and safety engineering (QR2MSE), Chendu, 15–18 July 2013, IEEE. doi:10.1109/QR2MSE.2013.6625967

BLACKBURN, H. (2011). Millennials and the adoption of new technologies in libraries through the

diffusion of innovations process. Library Hi Tech, 29 (4), 663-77. BLAND, J. M., & ALTMAN, D. G. (1997). Statistics notes: Cronbach’s alpha. BMJ. 314, 572. BMBF (2016). Bringing technology to the people: Research programme on human-machine interaction.

Demographic Change Division; Human-Machine-Interaction. Bonn

67

BOOBIER, T. (2018). Advanced Analytics and AI: Impact, Implementation, and the Future of Work.

Wiley. BRANDTZAEG, P.B., & FØLSTAD, A. (2017). Why People Use Chatbots. In Kompatsiaris

I. et al. (eds) Internet Science. INSCI 2017. Lecture Notes in Computer Science, 10673. Cham: Springer.

BRAUN, A. (2013). Chatbots in der Kundenkommunikation: Springer Berlin Heidelberg. BRUHN, M. (2016). Qualitätsmanagement für Dienstleistungen. Grundlagen, Konzepte, Methoden.

Berlin: Springer. BUTTLE, F., & MAKLAN, S. (2015). Customer relationship management: Concepts and technologies.

Abingdon, Oxon: Routledge. CARUANA, A., RAMASASHAN, B., & KRENTLER, K.A. (2015). Corporate Reputation,

Customer Satisfaction, & Customer Loyalty: What is the Relationship?. In Spotts H. (eds) Assessing the Different Roles of Marketing Theory and Practice in the Jaws of Economic Uncertainty. Developments in Marketing Science: Proceedings of the Academy of Marketing Science. Cham: Springer.

CASSELL, J., & TARTARO, A. (2007). Intersubjectivity in Human–Agent Interaction. Interaction

Studies, 8, 391-410.doi: 10.1075/is.8.3.05cas CHAKRABARTI, C., & LUGER, G. F. (2015). Artificial conversations for customer service chatter

bots: Architecture, algorithms, and evaluation metrics. Expert Systems with Applications, 42(20), 6878-6897. DOI: 10.1016/j.eswa.2015.04.067

CHARNIAK, E., & MCDERMOTT, D. V. (1985). Introduction to artificial intelligence. Reading

- Mass.a.o.: Addison-Wesley. CHURCHILL, JR., & SURPRENANT, C. (1982). An Investigation into the Determinants of

Customer Satisfaction. Journal of Marketing Research (JMR), 19, 491-504. doi:10.2307/3151722.

CHOWDHURY, G. G. (2003). Natural language processing. Annual review of information

science and technology, 37(1), 51–89. CIECHANOWSKI, L., PRZEGALINSKA, A., MAGNUSKI, M., & GLOOR, P. (2018).

In the shades of the uncanny valley: An experimental study of human–chatbot interaction. Future Generation Computer Systems. doi:10.1016/j.future.2018.01.055

CIECHANOWSKI L., PRZEGALINSKA A., & WEGNER K. (2018). The Necessity of New

Paradigms in Measuring Human-Chatbot Interaction. In: Hoffman M. (eds) Advances in Cross-Cultural Decision Making. AHFE 2017. Advances in Intelligent Systems and Computing, 610. Cham: Springer.

68

COCKTON, G., LARUSDOTTIR, M., GREGORY, P., & CAJANDER, A. (2016). Integrating User-Centred Design in Agile Development. Springer International Publishing.

COPELAND, M. (2016). What’s the Difference Between Artificial Intelligence, Machine Learning, and

Deep Learning?. Retrieved from https://blogs.nvidia.com/blog/2016/07/29/whats-difference-artificial-intelligence-machine-learning-deep-learning-ai/. Accessed on 01.06.2018

COREA, F. (2017). Artificial Intelligence and Exponential Technologies: Business Models Evolution and

New Investment Opportunities. Springer International Publishing. CORD, M., & CUNNINGHAM, P. (2014). Machine Learning Techniques for Multimedia Case

Studies on Organization and Retrieval. Berlin: Springer Berlin. CORTINA, J. (1993). What is Coefficient Alpha? An Examination of Theory and Applications.

Journal of Applied Psychology, 78(1), 98-104. CRESWELL, J. W., & CLARK, V. L. P. (2017). Designing and Conducting Mixed Methods

Research. SAGE Publications. DABIJA, D., BRANDUSA, B., & TIPI, N. (2018). Generation X versus Millennials communication

behaviour on social media when purchasing food versus tourist services. E+M Ekonomie a Management, 21, 191-205. doi:10.15240/tul/001/2018-1-013.

DALE, R. (2016). The return of the chatbots. Natural Language Engineering, 22(5), 811-817. DAMASIO, A. R. (2006). Descartes error: Emotion, reason and the human brain. London: Vintage

Books. DAUGHERTY, P. R., & WILSON, H. J. (2018). Human + Machine: Reimagining Work in the

Age of AI. Harvard Business Review Press. DOERING, N., & BORTZ, J. (2016). Forschungsmethoden und Evaluation in den Sozial- und

Humanwissenschaften (5th ed.). Berlin Heidelberg: Springer-Verlag. DOMINGOS, P. (2015). The Master Algorithm: How the Quest for the Ultimate Learning Machine

Will Remake Our World. Basic Books. DOSHI, S., PAWAR, B., SHELAR, G., & KULKARNI, S. (2017). Artificial Intelligence Chatbot

in Android System using Open Source Program-O. IJARCCE, 6, 816-821. doi:10.17148/IJARCCE.2017.64151.

DOURISH, P., & BELL, G. (2011). Divining a Digital Future: Mess and Mythology in Ubiquitous

Computing: MIT Press. FLEMING, J. H., & ASPLUND, J. (2007). Human sigma: Managing the employee-customer

encounter. New York: Gallup.

69

FONTAINE J. R. J., SCHERER K. R., & SORIANO C. (2013). A paradigm for a multidisciplinary investigation of the meaning of emotion terms. In Components of Emotional Meaning: A Sourcebook, eds Fontaine J. R. J., Scherer K. R., Soriano C. Oxford: Oxford University Press.

FRANKEN S. (2010). Menschliche Intelligenz(en). In Verhaltensorientierte Führung. Gabler GARDNER, H. (1996). Multiple intelligences. Developing Museum Exhibitions for Lifelong Learning

Gillingham. Kent: Group for Education in Museums. GARTNER, INC. (2017). Gartner Top Strategic Predictions for 2018 and Beyond. Retrieved from.

https://www.gartner.com/smarterwithgartner/gartner-top-strategic-predictions-for-2018-and-beyond/. Accessed on 01.06.2018

GENTSCH, P. (2018). Best Practices. In: Künstliche Intelligenz für Sales, Marketing und

Service. Wiesbaden: Springer Gabler. GILLESPIE, A., & CORTI, K. (2016). The Body That Speaks: Recombining Bodies and Speech

Sources in Unscripted Face-to-Face Communication. Front. Psychol. doi:10.3389/fpsyg.2016.01300

GLADYSH, M. (2018). 5 industries that benefit from chatbots already. BotsCrew. Retrieved from https://botscrew.com/industries-benefit-chatbots. Accessed on 28.03.18

GRAHAM-CUMMING, J. (2012). Alan Turing: Intelligence & life. New Scientist, 214(2867),

vi-vii. doi:10.1016/S0262-4079(12)61378-5 GRIFFIN, A., & HAUSER, J. R. (1993). The Voice of the Customer. Marketing Science, 12(1),

1-27. doi:10.1287/mksc.12.1.1 GRIGOREV, A., SHANMUGAMANI, R., BOSCHETTI, A., MASSARON, L., &

THAKUR, A. (2018). TensorFlow Deep Learning Projects: 10 real-world projects on computer vision, machine translation, chatbots, and reinforcement learning. Packt Publishing.

HAISLIP, J. Z., & RICHARDSON, V. J. (2017). The effect of Customer Relationship Management

systems on firm performance. International Journal of Accounting Information Systems, 27, 16-29. doi:10.1016/j.accinf.2017.09.003

HAUGELAND, J. (1989). Artificial Intelligence: The Very Idea. MIT Press. HARDIK, S., D GOSAI, D., & GOHIL, H. (2018). A Review on a Emotion Detection and

Recognization from Text Using Natural Language Processing. HEO, J., PARK, S., & SONG, C. (2007). A Study on the Improving Product Usability Applying the

Kano’s Model of Customer Satisfaction. In Human-Computer Interaction: Interaction Design and Usability, 482-489. doi: 10.1007/978-3-540-73105-4_53

HERZBERG, F., MAUSNER, B. & SNYDERMAN, B. B. (1959). The motivation to work.

New Brunswick: Transaction.

70

HILL, D. J. (1986). Satisfaction and Consumer Services. In NA - Advances in Consumer Research,

13, eds. Richard J. Lutz, Provo, UT: Association for Consumer Research, 311-315. HILL, J., RANDOLPH FORD, W., & FARRERAS, I. G. (2015). Real conversations with

artificial intelligence: A comparison between human–human online conversations and human–chatbot conversations. Computers in Human Behavior, 49, 245-250. doi:10.1016/j.chb.2015.02.026

HIRSCHBERG, J., & MANNING, C. (2015). Advances in natural language processing. Science

(New York, N.Y.). 349, 261-266. doi:10.1126/science.aaa8685. HOARE, G. (2014). Always bet on text. Livejournal. Retrieved from

http://graydon.livejournal.com/196162.html. Accessed on 03.03.18 HOMBURG, C., KOSCHATE, N. & HOYER, W. D. (2006). The role of cognitionband affect in

the formation of customer satisfaction: A dynamic perspective. Journal of Marketing, 70, 21–31.

HOMBURG, C., GIERING, A., & HENTSCHEL, F. (1999). Der Zusammenhang zwischen

Kundenzufriedenheit und Kundenbindung. Mannheim HOWE, N. (2015). Why Millennials Are Texting More And Talking Less. Forbes. Retrieved from

https://www.forbes.com/sites/neilhowe/2015/07/15/why-millennials-are-texting-more-and-talking-less/#681d28145975. Accessed on 03.03.18

HOWE, N., & STRAUSS, W. (2000). Millennials Rising: The Next Great Generation. New

York: Vintage. HUANG, J.W. (2017). Application of Kano Model in Requirements Analysis of Y Company’s

Consulting Project. American Journal of Industrial and Business Management, 7, 910-918. doi:10.4236/ajibm.2017.77064.

HUSNJAK, S., PERAKOVIC, D., & JOVOVIC, I. (2014). Possibilities of Using Speech

Recognition Systems of Smart Terminal Devices in Traffic Environment. Procedia Engineering, 69, 778-787. doi:10.1016/j.proeng.2014.03.054

HUSSAIN, A., MKPOJIOGU, E. O. C., & KAMAL, F.M. (2015). Eliciting user satisfying

requirements for an e-health awareness system using kano model. In Xiaodong Z (ed) Recent advances in computer science. Proceedings of the 14th WSEAS International Conference on Computer and Computational Science (ACACOS’15), Kuala Lumpur. WSEAS Press.

HYMAN, M., & SIERRA, J. (2016). Open- versus close-ended survey questions. NMSU Business

Outlook. 14. IACOBONI, M. (2008). Mirroring People. New York: Farrar, Straus & Giroux

71

JAMES, E. A., & SLATER, T. H. 2013. Writing Your Doctoral Dissertation or Thesis Faster: A Proven Map to Success, SAGE Publications.

JANARTHANAM, S. (2017). Hands-On Chatbots and Conversational UI Development: Build

chatbots and voice user interfaces with Chatfuel, Dialogflow, Microsoft Bot Framework, Twilio, and Alexa Skills. Packt Publishing.

JOHNSON, R. B., & ONWUEGBUZIE, A. J. (2004). Mixed methods research: A research

paradigm whose time has come. Educational researcher, 33(7), 14-26. JONES, C., & SHAO, B. (2011). The net generation and digital natives: Implications for

higher education. York, UK: Higher Education Academy. JOSHI, A., KALE, S., CHANDEL, S., & PAL, D. (2015). Likert Scale: Explored and Explained.

British Journal of Applied Science & Technology, 7, 396-403. doi:10.9734/BJAST/2015/14975.

KAMPHAUG, Å., GRANMO, OC., GOODWIN, M., & ZADOROZHNY, V.I. (2018).

Towards Open Domain Chatbots—A GRU Architecture for Data Driven Conversations. In: Diplaris S., Satsiou A., Følstad A., Vafopoulos M., Vilarinho T. (eds) Internet Science. INSCI 2017. Lecture Notes in Computer Science, vol 10750. Cham: Springer.

KANO, N., SERAKU N., TAKAHASHI, F., & TSUJI, S. (1984). Attractive Quality and Must-

be Quality, Hinshitsu: The Journal of the Japanese Society for Quality Control, 39-48. KAPLAN, J. (2016). Artificial intelligence. Oxford: Oxford University Press. KASSIBGI, G. (2017). Soul of the machine: How chatbots work. Retrieved from

https://medium.com/@gk_/howchat-bots-work-dfff656a35e2. Accessed on 03.03.18

KHAN, R., & DAS, A. (2017). Build Better Chatbots. A Complete Guide to Getting Started

with Chatbots: Apress. KNIGHT, M. (2016). The Importance of Conversation. Business and Professional

Communication Quarterly, 79(3), 267-269. doi:10.1177/2329490616665823 KRUEGER, F. (2016). Study II: The Confirmation/Disconfirmation-Paradigm in a Cross-Cultural

Perspective – A Study across Countries. In: The Influence of Culture and Personality on Customer Satisfaction. International Management Studies. Wiesbaden: Springer Gabler.

KUMAR, E. (2011). Natural language processing. New Delhi: I.K. International Publishing

House. KUPPEVELT, J. V., DYBKJÆR, L., & BERNSEN, N. O. (2005). Advances in natural

multimodal dialogue systems. Dordrecht: Springer.

72

KURZWEIL, R. (1990). The age of intelligent machines. Cambridge, Massachusetts etc.: MIT Press.

LANTZ, B. (2013). Machine learning with R: Learn how to use R to apply powerful machine learning

methods and gain an insight into real-world applications. Birmingham: Packt Publishing. LATTEMANN, C., & ROBRA-BISSANTZ, S. (2017). Digital Customer Experience. HMD

Praxis der Wirtschaftsinformatik, 54(5), 637-638. LEGG, S., & HUTTER, M. (2007). A Collection of Definitions of Intelligence. Advances in

Artificial General Intelligence: Concepts, Architectures and Algorithms, 157. doi:10.1207/s15327051hci0301_2.

LEVESQUE, H. J. (2017). Common Sense, the Turing Test, and the Quest for Real AI:

Reflections on Natural and Artificial Intelligence: MIT Press. LIBOV, J. (2015). Futures of text. Retrieved from http://whoo.ps/2015/02/23/futures-of-

text. Accessed on 01.06.2018 LIDDY, E.D. (2001). Natural Language Processing. In Encyclopedia of Library and Information

Science (2nd ed.). NY: Marcel Decker, Inc. LI-LI, Z., LIAN-FENG, H., & QIN-YING, S. (2011). Research on requirement for high-quality

model of extreme programming. 2011 International Conference on Information Management, Innovation Management and Industrial Engineering (ICIII), Shenzhen, IEEE, 26–27 Nov 2011.

LINDQUIST, K. A., MACCORMACK, J. K., & SHABLACK, H. (2015). The role of language

in emotion: predictions from psychological constructionism. Frontiers in Psychology, 6, 444. doi:10.3389/fpsyg.2015.00444

LOOS, P. (2003). Avatar. In Lexikon Electronic Business, Thomas Schildhauer, ed. Munich:

Oldenbourg, 16–19. LOEFGREN, M. & WITELL, L. (2008). Two decades of using Kano’s Theory of Attractive Quality:

A Literature Review. The Quality Mgmt. J., 1(1), 59-76. LOEFGREN, M., WITELL, L., & GUSTAFSSON, A. (2011). Theory of attractive quality and

life cycles of quality attributes. The TQM Journal, 23 (2), 235-246, DOI: 10.1108/17542731111110267

MALHOTRA, Y., GALLETTA, D., & KIRSCH, L. (2008). How endogenous motivations influence

user intentions: beyond the dichotomy of extrinsic and intrinsic user motivations. J. Manag. Inform. Syst. 25(1), 267–300.

MARR, B. (2016). What Is The Difference Between Artificial Intelligence And Machine Learning?.

Retrieved from https://www.forbes.com/sites/bernardmarr/2016/12/06/what-is-the-difference-between-artificial-intelligence-and-machine-learning/#4f8310162742. Accessed on 23.11.2017

73

MASON, M. (2017) AI for the Enterprise: 3 types of business chatbots you can build. IBM. Retrieved

from https://www.ibm.com/blogs/watson/2017/12/3-types-of-business-chatbots-you-can-build/. Accessed on 24.08.2018

MATZLER, K., HINTERHUBER, HH. BAILOM, F. & SAUERMEIN, E. (1996). How to

delight your customer. J Product Brand Manag 5(2), 6–18. MAYO, J. (2017). Programming the Microsoft Bot Framework: A Multiplatform Approach to Building

Chatbots. Microsoft Press. MCCOLL-KENNEDY, J.R., & SMITH, A.K, (2006). Customer emotions in service failure and

recovery encounters. In Zerbe, W.J., Ashkanasy, N.M. and Haertel, C.E.J. (Eds), Individual and Organizational Perspectives on Emotion Management and Display, 237-68. Stamford: JAI Press.

MCQUAIL, D. (1987). Mass Communication Theory: An Introduction, 2nd edn. Sage, London. MCTEAR, M.F. (2017). The Rise of the Conversational Interface: A New Kid on the Block?. In:

Quesada J., Martín Mateos FJ., López Soto T. (eds) Future and Emerging Trends in Language Technology. Machine Learning and Big Data. FETLT 2016. Lecture Notes in Computer Science, 10341. Cham: Springer.

MEFFERT, H., BURMANN, C., & KIRCHGEORG, M. (2014). Marketing: Grundlagen

marktorientierter Unternehmensführung Konzepte - Instrumente - Praxisbeispiele: Springer Fachmedien Wiesbaden.

MENAL, D. (2017). A Tool of Conversation: Chatbot. International Journal of Computer

Sciences and Engineering, 5, 158-161. MESSINA, C. (2016). Thoughts on the UX of Bots. Medium. Retrieved from

https://medium.com/chris-messina/ux-of-bots-e565fb7c4d4e#.fubh935eb. Accessed on 23.11.2017

MIJWEL, M. (2015). History of Artificial Intelligence. University of Baghdad. MIKULIC, J., & PREBEZAC, D. (2011). A critical review of techniques for classifying quality

attributes in the Kano model, Managing Service Quality: An International Journal. 21 (1), 46-66. doi:10.1108/09604521111100243

MINDBOWSER (2017). Chatbot survey 2017 - Current state of chatbots and their outlook in 2017.

Retrieved from http://mindbowser.com/chatbot-market-survey-2017/. Accessed on 25.05. 2018

MONK, A. (2000). User-centred design: the home use challenge. In Sloane, A., van Rijn, F. (eds.)

Home Informatics and Telematics. Information Technology and Society, 181–190, Boston: Kluwer Academic Publishers.

MORGAN, R. L. (2009). Calming upset customers: Stay in control. In Any situation. Axzo.

74

MOU, Y., & XU, K. (2017). The media inequality: Comparing the initial human-human and human-

AI social interactions. Computers in Human Behavior, 72, 432-440. doi:10.1016/j.chb.2017.02.067

MOWRER, O. H. (1960). Learning theory and behavior. Hoboken, NJ, US: John Wiley & Sons

Inc. MRKALJ, M. (2018). Chatbots and AI: The Key Event Tech Trends for 2018. Retrieved from

https://chatbotsmagazine.com/chatbots-and-ai-the-key-event-tech-trends-for-2018-7b45cbc723a. Accessed on 14.05.2018

MULDOWNEY, O. (2017). Chatbots: An Introduction And Easy Guide To Making Your Own.

Dublin, Ireland: Curses & Magic. MURALI, S., PUGAZHENDHI, S., & MURALIDHARAN, C. (2016). Modelling and

Investigating the relationship of after sales service quality with customer satisfaction, retention and loyalty – A case study of home appliances business. Journal of Retailing and Consumer Services, 30, 67-83. doi:10.1016/j.jretconser.2016.01.001

MYERS, B. (1998). A Brief History of Human Computer Interaction Technology. ACM interactions,

5. NASS, C., & MOON, Y. (2002). Machines and Mindlessness: Social Responses to Computers. Journal

of Social Issues, 56: 81-103. doi:10.1111/0022-4537.00153 NASS, C., & MOON, Y., FOGG, B., REEVES, B. & DRYER, C. (1995). Can computer

personalities be human personalities?. Int. J. Hum.-Comput. Stud., 43, 223-239. doi:10.1145/223355.223538

NASCIMENTO, P. AGUAS, R., SCHNEIDER, D., & DE SOUZA, J. (2012). An approach

to requirements categorization using Kano’s model and crowds. Proceedings of 2012 IEEE International Conference on Computer Supported Cooperative Work in Design (CSCWD), Wuhan, IEEE, 23–25 May 2012, 387–392. DOI: 10.1109/CSCWD.2012.6221847

NEILL, J. (2008) from: Howell, D. C. (2008). Fundamental statistics for the behavioral sciences (6th

ed.). Belmont, CA: Wadsworth. NG, E., & MCGINNIS JOHNSON, J. (2015). Millennials: Who are they, how are they

different, and why should we care?. The Multi-generational and Aging Workforce: Challenges and Opportunities, 121-137. doi:10.4337/9781783476589

NIJHOLT, A. (2003). Humour and Embodied Conversational Agents. (CTIT Technical reports

series; No. 2003-03). Enschede: Centre for Telematics and Information Technology (CTIT).

NILSSON, N. J. (1998). Artificial Intelligence a new synthesis. San Francisco, Calif.: Morgan

Kaufmann Publishers.

75

NOBLE, D. F. (2013). The Religion of Technology: The Divinity of Man and the Spirit of Invention: Knopf Doubleday Publishing Group.

OGDEN, W. C. (1988). Using natural language interfaces. In M. HELANDER, Ed. Handbook of Human-Computer Interaction, 281-299. Amsterdam: Elsevier.

OLIVER, R. (1980). A Cognitive Model of the Antecedents and Consequences of Satisfaction Decisions.

Journal of Marketing Research, 17 (4), 460-69. ORACLE CORPORATION (2016). Can Virtual Experiences Replace Reality?. The future

role for humans in delivering customer experience. Retrieved from https://www.oracle.com/webfolder/s/delivery_production/docs/FY16h1/doc35/CXResearchVirtualExperiences.pdf. Accessed on 01.06.2018

PEPPARD, J., & WARD, J. (2016). The strategic management of information systems: Building a

digital strategy. Chichester: John Wiley & Sons. POOLE, D. L., MACKWORTH, A. K., & GOEBEL, R. (1998). Computational intelligence:

a logical approach. New York: Oxford University Press. PREECE, J., ROGERS, Y., SHARP, H., BENYON, D., HOLLAND, S., & CAREY, T.

(1994). Human-Computer Interaction. Essex, UK. Addison-Wesley Longman Ltd. QUARTERONI, S. (2018). Natural Language Processing for Industry. Informatik-

Spektrum, 41(2), 105-112. doi:10.1007/s00287-018-1094-1 RADZIWILL, N. M., & BENTON, M. C. (2017). Evaluating Quality of Chatbots and Intelligent

Conversational Agents. arXiv preprint arXiv:1704.04579. RAMESH, K., RAVISHANKARAN, S., JOSHI, A., & CHANDRASEKARAN, K. (2017).

A Survey of Design Techniques for Conversational Agents. In: Kaushik S., Gupta D., Kharb L., Chahal D. (eds) Information, Communication and Computing Technology. ICICCT 2017. Communications in Computer and Information Science, vol 750. Springer, Singapore.

RAPAPORT, W. (2006). Turing Test A2 - Brown, Keith. In Encyclopedia of Language &

Linguistics (2nd ed.). 151-159. Oxford: Elsevier. RASCHKA, S. (2016). Machine Learning und Data Science mit Python. Frechen: MITP. REEVES, B., & NASS, C. (1996). The Media Equation: How People Treat Computers, Television,

and New Media Like Real People and Places. New York, NY, USA: Cambridge University Press.

RICH, E., & KNIGHT, K. (1991). Artificial intelligence (2nd ed.). New York etc.: McGraw-

Hill. ROBINSON, P., & EL KALIOUBY, R. (2009). Computation of emotions in man and

machines. Philosophical Transactions of the Royal Society B: Biological Sciences, 364(1535), 3441–3447. doi:10.1098/rstb.2009.0198

76

RUSSELL, S. J., & NORVIG, P. (2010). Artificial intelligence: A modern approach. Boston, MA:

Pearson education. SAMEERA, A., & WOODS, J. (2015). Survey on Chatbot Design Techniques in Speech Conversation

Systems. International Journal of Advanced Computer Science and Applications. SAUERWEIN, E. (2000). Das Kano-Modell der Kundenzufriedenheit. Reliabilität und Validität einer

Methode zur Klassifizierung von Produkteigenschaften. Wiesbaden: Springer Fachmedien. SAUERWEIN, E., BAILOM, F., MATZLER, K., & H. HINTERHUBER, H. (1996). The

Kano Model: How to Delight Your Customers. International Working Seminar on Production Economics, 1.

SEGAL, E. A., GERDES, K. E., LIETZ, C. A., WAGAMAN, M. A., & GEIGER, J. M.

(2017). Assessing Empathy. Columbia University Press. SHARIFF A. F., TRACY J. L. (2011). What are Emotion Expressions for? Curr. Dir. Psychol.

Sci. 20, 395–399. doi:10.1177/0963721411424739 SHAWAR, B. A., & ATWELL, E. (2007). Chatbots: are they really useful? In LDV Forum, 22,

29-49. SHAWAR, B. A., ATWELL, E., & ROBERTS, A. (2005). FAQChat as an Information

Retrieval System. In Vetulani, Z. (ed.), Human Language Technologies as a Challenge 2nd Language and Technology Conference Wydawnictwo Poznanskie, Poznan, Poland, 274-278.

SHAH, H., WARWICK, K., VALLVERDÚ, J., & WU, D. (2016). Can machines talk?

Comparison of Eliza with modern dialogue systems. Computers in Human Behavior, 58, 278–295. doi:10.1016/j.chb.2016.01.004

SHEVAT, A. (2017). Designing Bots. Creating Conversational Experiences. O'Reilly Media. SPENCE, P. R., WESTERMAN, D., EDWARDS, C., & EDWARDS, A. (2014). Welcoming

Our Robot Overlords: Initial Expectations About Interaction With a Robot. Communication Research Reports, 31(3), 272-280. doi:10.1080/08824096.2014.924337

STAKE, R. E. (2000). Case Studies. In N. K. Denzin, & Y. S. Lincoln (Eds.), Handbook of

Qualitative Research, 435-453. Thousand Oaks, CA: Sage. STANDARDIZATION, I. O. F. (2013). ISO/TS 20282-2:2013(en) Usability of consumer

products and products for public use - Part 2: Summative test method. Retrieved from https://www.iso.org/obp/ui/#iso:std:iso:ts:20282:-2:ed-2:v1:en. Accessed on 28.08.18

STONE, D., JARRETT, C., WOODROFFE, M., & MINOCHA, S. (2005). User interface

design and evaluation. Elsevier.

77

SUTHAHARAN, S. (2016). Machine learning models and algorithms for big data classification: Thinking with examples for effective learning. New York: Springer.

THOMAS, R., & MCSHARRY, P. (2015). Big Data Revolution: What farmers, doctors and insurance

agents teach us about discovering big data patterns. United Kingdom: John Wiley & Sons. HENNIG-THURAU, T., MALTHOUSE, E., FRIEGE, C., GENSLER, S., LOBSCHAT,

L., RANGASWAMY, A., & SKIERA, B. (2010). The Impact of New Media on Customer Relationships: From Bowling to Pinball. Journal of Service Research, 13 3, 311-330. doi:10.1177/1094670510375460

TURKLE, S. (2016). Reclaiming conversation: The power of talk in the digital age. New York, NY:

Penguin. TURBAN, E., OUTLAND, J., KING, D., LEE, J. K., LIANG, T. P., & TURBAN, D. C.

(2018). Intelligent (Smart) E-Commerce. In Electronic Commerce 2018. 249-283. Cham: Springer.

VAN DE GEVEL, A. J., & NOUSSAIR, C. N. (2013). The Nexus between Artificial Intelligence

and Economics. Berlin, Heidelberg: Springer Berlin Heidelberg. VAN DOORN, M. & DUIVESTEIN, S. (2016). The Bot Effect: ‚Friending your brand ‘. Report.

Applied Innovation Exchange, SogetiLabs. WE ARE SOCIAL. (2018) Most popular mobile messaging apps worldwide as of july 2018,

based on number of monthly active users (in millions). In Statista - The Statistics Portal. Retrieved from https://www.statista.com/statistics/258749/most-popular-global-mobile-messenger-apps/. Accessed on 28.08.18

WINSTON, P. H. (1992). Artificial intelligence (3rd ed.). Reading, Mass.: Addison-Wesley. WIRTZ, J., & BATESON, J. (1999). Consumer satisfaction with services: Integrating the environment

perspective in services marketing into the traditional disconfirmation paradigm. Journal of Business Research, 44 (1), 55-66. doi:10.1016/S0148-2963(97)00178-1

WITELL, L., & FUNDIN, A. (2005). Dynamics of service attributes: A test of Kano's theory of

attractive quality. International Journal of Service Industry Management, 16, 152-168. doi: 10.1108/09564230510592289.

YANG, X. S. (2012). Artificial Intelligence, Evolutionary Computing and Metaheuristics: In the

Footsteps of Alan Turing: Springer Berlin Heidelberg. YIN, R.K. (1984). Case Study Research: Design and Methods. Beverly Hills, Calif: Sage

Publications. ZADROZNY, W., BUDZIKOWSKA, M., J. CHAI, KAMBHATLA, N., LEVESQUE, S.,

& NICOLOV, N. (2000). Natural Language Dialogue for Personalized Interaction. Communications of the ACM (CACM) 43, 8, 116-120.

78

ZEITHAML, V. A., PARASURAMAN, A. & BERRY, L. (1992). Qualitätsservice. Was Ihre Kunden erwarten – was Sie leisten müssen. Frankfurt a.M.: Campus.

79

Attachments

Questionnaire – English Version

80

81

82

83

84

85

86

87

88