Utrecht

University


BACHELOR THESIS
INFORMATION SCIENCES

Chatbot Personality and Customer Satisfaction

Author: Thesis Supervisor:

Hayco de Haan Christof van Nimwegen

External Supervisor: Thesis Supervisor 2:

Joop Snijder Robbert Jan Beun


February 09, 2018
Preface
Before you lies the thesis “Chatbot Personality and Customer Satisfaction” and explores the role
the former has on the latter. It has been written to meet the graduation requirements of my
Information Science Bachelor at Utrecht University, and has been written from November 2017
to January 2018.

This research was conducted for and on behalf of Info Support, a Dutch technology firm, in the
form of an internship. The research question was formulated together with my supervisor from
Info Support, Joop Snijder. One of the hardest challenges while writing this thesis was the novelty
of the research field and the accompanying lack of scientific literature on this topic. As a result,
this thesis took a multidisciplinary approach in an attempt to explore the research question, to
build a better understanding of the problem, and to lay the groundwork for future studies.

I would like to thank Joop Snijder for his guidance and support throughout my internship at Info
Support. Our weekly meetings have always been very open and provided me with valuable
feedback. I also wish to thank the other graduate students and colleagues at Info Support, who
always have been very supportive these last few months.

Furthermore, I would like to thank my head supervisor Christof van Nimwegen for his valuable
input and support throughout my graduation. Despite his air tight schedule, we had weekly
meetings in which he would provide me with constructive feedback on my deliverables and useful
insights. Moreover, I would like to thank Robbert Jan Beun for freeing up time reviewing my thesis
as second supervisor.

Yet, I could not have written this thesis without another strong support group. First of all, I wish
to thank my parents, who supported me with their love and understanding. Furthermore, I also
would like to thank my friends for their enduring support and their interest. Finally, special
thanks to Floortje Kipp for always having my back and for prepping my meals.

I hope you will enjoy reading this thesis.

Hayco de Haan

Veenendaal, February 2018.

Contents
1 Introduction .............................................................................................................................................. 1
2 Methodology ............................................................................................................................................ 3
2.1 Explorative literature study ............................................................................................................... 3
2.2 Survey ............................................................................................................................................... 3
2.3 Validity .............................................................................................................................................. 4
3 Chatbots .................................................................................................................................................. 5
3.1 Development of chatbots .................................................................................................................. 5
3.2 Implementations of Chatbots ............................................................................................................ 6
3.2.1 Chatbots in Enterprises .............................................................................................................. 6
3.2.2 Requirements for enterprise chatbots ........................................................................................ 7
3.3 Potential of Chatbots ...................................................................................................................... 10
4 Personality ............................................................................................................................................. 11
4.1 Taxonomy on Personality ............................................................................................................... 11
4.2 The Five Dimensions ...................................................................................................................... 11
4.3 Anthropomorphism ......................................................................................................................... 12
5 Personality and Customer Satisfaction ................................................................................................. 14
5.1 Customer Satisfaction ..................................................................................................................... 14
5.1.1 Service Quality Models .............................................................................................................. 15
5.2 Personality Influencing Customer Satisfaction ............................................................................... 18
6 Personality in Chatbots: survey ............................................................................................................. 19
6.1 Survey: methodology ...................................................................................................................... 19
6.2 Survey: results .................................................................................................................................. 21
7 Chatbots and Ethics .............................................................................................................................. 24
8 Discussion and conclusion .................................................................................................................... 26
8.1 Limitations and directions for future research ................................................................................... 27
References ................................................................................................................................................. 29
Appendix A ................................................................................................................................................. 34
Appendix B ................................................................................................................................................. 35

1 Introduction
These days it seems that almost every business has adopted its own chatbot to communicate
with either their customers, their own employees or with other businesses. To support this claim
Haptik Inc. (2017), the company behind the application of the same name, recently published a
report stating that there currently exist over 40,000 chatbots across multiple platforms and that
the market size of chatbots can grow from $700 million in 2016 to $3 billion in 2021. Moreover, a
survey by Oracle Inc. (2016) found that 80% of the 800 interviewed businesses were already
using chatbots or planned to implement them into their businesses by 2020. Perhaps the best-
known example of a chatbot implementation is Apple’s Siri1, which is a digital assistant that helps
people with their request. Besides a digital assistant, chatbots also serve other ends such as
accompanying dementia patients (Endurance2), placing online orders (Subway bot3), and offering
financial advice based on spending habits (Erica4). Yet, it appears that most chatbots
implementations are assigned customer service roles, assuming roles traditionally assigned to
humans.
In a report published by Garter (2017), two main reasons are given for the adoption of
chatbots within businesses. The first is increasing customer satisfaction, as chatbots have the
potential to manage customer progress more effectively than humans by using efficient decision
trees. The second is cost reduction, as chatbots will deliver customer satisfaction at significantly
lower cost than human customer service agents. In turn, customer satisfaction leads to customer
retention and eventually increases a business’s profit (Anderson & Mittal, 2000). Besides the
effects of customer satisfaction, factors influencing customer satisfaction have also been
identified including helpful employees, quick service, and service quality (Hokanson, 1995).
Additionally, earlier research that studied the relationship between personality, service quality,
and customer satisfaction, concluded that some personality traits have a significant influence on
customer satisfaction (Ekinci & Dawes, 2009; Hurley, 1998). These studies, however, merely
considered human customer support representatives and its findings therefore cannot be directly
applied to non-human representatives without any further thought. This leads to the formulation
of the following main research question:

What role has the personality of chatbots on perceived customer satisfaction?

This question is then further divided into the following sub questions:

1: How can personality be categorized?

2: What models to customer satisfaction exists and are they influenced by personality?
3: Can earlier related research on customer satisfaction and personality also be applied to
chatbots?

1
https://www.apple.com/ios/siri/
2
http://endurancerobots.com/azbnmaterial/chatbots-for-senior-people-and-patients-with-alzheimer-s-disease/
3
https://www.cnbc.com/2017/04/18/subway-unveils-facebook-chatbot-for-ordering-as-it-looks-to-revamp-digital.html
4
https://www.cnbc.com/2016/10/24/bank-of-america-launches-ai-chatbot-erica--heres-what-it-does.html

1
Since little to no scientific research on this specific topic exists, insights are primarily obtained by
reviewing scientific literature from other domains related to the subject under study. This research
has scientific relevancy, because it unites different aspects from core disciplines of information
science, psychology, and marketing in order to contribute to the gap prevalent in the scientific
knowledge.
This research project is performed for and on behalf of Info Support, a medium sized Dutch
technology company located in Veenendaal, the Netherlands. They are specialized in developing,
managing, and hosting custom software solutions with innovations being one of their core values.
From its clients, Info Support receives an increasing number of requests with respect to chatbots
recently. This is in line with McTaer’s (2016) statement that the year 2016 marked a tipping point
for chatbots as major companies started to invest heavily in the technologies required to develop
sophisticated systems capable of interacting with users in a natural, conversational style. Earlier,
cost reduction and customer satisfaction were identified as important reasons for a business to
replace human representatives with chatbots. Moreover, other research identified that the
personality of human service employees significantly influences customer satisfaction. The
current study is carried out under the assumption that this is also the case for chatbots, however
no current research exists to support this. In other words, these clients benefit from research
concerning the personality of chatbots, emphasizing its practical relevance.
This research does not intend to offer a final and conclusive solution on the research
question. It rather intends to explore the research question, to build a better understanding of the
problem, and to lay the groundwork for future studies.
The remainder of this thesis is structured into seven sections. The next section will
elaborate on the methodology prevalent in this study. The following three sections will clarify the
theoretical background and will discuss the development of chatbots, personality models, and
customer satisfaction theory. In the sixth section, the results of the conducted survey will be
analysed. Section seven will discuss some of the ethical challenges associated with implementing
chatbots within a firm. Finally, the overall results, limitations and future research agenda will be
discussed in section eight.

2
2 Methodology
Since little research has been conducted on this topic, the following study will be explorative by
nature. The research techniques used in this study consist of an explorative literature review
complemented by a short survey distributed among businesses. The goal of the literature review
is to explore available literature in an attempt to identify, evaluate and integrate the findings of
relevant, high-quality studies that address one or more aspects of the research questions (Budgen
& Brereton, 2006). Subsequently, the goal of the survey is then to validate these findings. Both
techniques result in qualitative data which is not uncommon for studies of explorative nature.

2.1 Explorative literature study

In order to identify relevant literature, a keyword-search is performed on the following scholarly
databases: Google Scholar, PubMed, Scopus, ResearchGate, JSTOR, Web of Science, and
Ovid. The types of studies that are primarily selected are studies investigating personality traits,
chatbots or conversational agents, or customer satisfaction. Moreover, the search is limited to
sources published in 1900 and onwards, written in Dutch or English. This review mainly uses
scientific literature (e.g. journal articles, conference proceedings, and books) as sources, however
some insights are also obtained from suggestions made by other credible sources. Furthermore,
the relevancy of the found sources depends on its contribution in answering the current topic
under study and is determined by their scope, objectives, methods and findings. Finally, both
backward and forward searching is performed in order to cover as much relevant literature as
possible (Levy & Ellis, 2006). Backward searching has been applied in order to trace back
research on which the found sources are based and forward searching in order to find more recent
sources.

2.2 Survey
In addition to the literature study, a survey element is added to this research in order to obtain
businesses’ perspectives on this topic. The survey will be sent to sixty different businesses in ten
different industries, including life insurance, public transport, retail, banking, and telecom
organizations. Because of the exploratory nature of this study, guidelines cannot yet be derived
from earlier studies in the formulation of a measure scale. Rather the questions are formulated
using the theory from the literature review and with the research question in mind. In addition, the
questions are independently evaluated by two other parties. Additionally, the NEO-PI-3 test
(McCrae, Costa, Jr, & Martin, 2005) will be used to map the results of the survey with the
according category. The businesses have been selected by means of purposive sampling,
meaning that however the sample represents a large variety of businesses, the results are only
generalizable to a certain extent. Moreover, this survey focusses on the perspectives of
businesses rather than customers, since they probably have a clearer vision with respect to the
preferred personality traits in their customer service representatives, leading to less ambiguity.
See section 6.2 for a more detailed description of the survey.

3
2.3 Validity
With research, it is important to consider both the concepts of validity and reliability. Reliability is
often referred to as “…the extent to which results are consistent over time and an accurate
representation of the total population under study is referred to as reliability and if the results of a
study can be reproduced under a similar methodology, then the research instrument is considered
to be reliable” (Joppe, 2000, p. 1). However, if a research’s results are reliable this does not
necessarily imply that they are valid as well. Validity refers to “… whether the research truly
measures that which it was intended to measure or how truthful the research results are” (Joppe,
2000, p. 1). However, Golafshani (2003) argues that these components only apply to quantitative
research and not as much to qualitative research. Moreover, she argues that with qualitative
research reliability and validity are conceptualized as trustworthiness, rigor and quality, which can
be increased by means of triangulation. In this study triangulation is achieved by considering
multiple different information sources (data triangulation) as well as using multiple qualitative
methods (methodological triangulation).

4
3 Chatbots
Until now the term chatbot has been loosely used without assigning a clear definition to it.
However, to avoid ambiguity about the meaning of the term, a definition should be assigned to it.
The definition adopted in the remainder of this study, results from synthesizing definitions used in
different scientific works.
A chatbot, or conversational agent, is a software system which exploits natural language
technologies to engage users in information-seeking and task-oriented dialogs (Kerly, Hall, & Bull,
2007; Lester, Branting, & Mott, 2004; Shawar & Atwell, 2007).
Natural Language Processing (NLP) is an area of research that explores the ability of computers
to understand and manipulate natural language (e.g. English, Dutch, or Spanish) text or speech
to do meaningful tasks (Chowdhury, 2003). Such tasks might include translating input to another
language, interpreting the text and compile a summary, or to participate in an ongoing
conversation with a human. Natural language technologies used in early work on chatbots (e.g.
ELIZA) mainly concerned techniques based on textual input. However, as new technologies
rapidly were developed over the last years, input by speech was also assigned a more significant
role.
Moreover, the definition distinguishes two types of dialogs, information-seeking dialogs
and task-oriented dialogs. Information-seeking systems, provide users with relevant information
on their query. For instance, when a customer asks the system for the status of an order they
earlier placed, the system will retrieve this information and present is to the user. Task-oriented
systems, on the other hand, are designed to converse with its users in order to accomplish tasks.
An example of this is online shopping, where users can tell the chatbot to place an order, which
then is automatically executed. The user tells the system what he is looking for along with other
preferences, whereupon the system asks the user for missing information. Once all details have
been processed, the user gives his confirmation and the order is placed. Whereas the information-
seeking system merely provides information about an order, the task-oriented system is more
interactive and allows its users to place an order as well.
The remainder of this section will discuss: the development of chatbots, the different
implementations of chatbots, how chatbots are used in businesses and its requirements, and the
potential of chatbots.

3.1 Development of chatbots

The Imitation Game, sometimes referred to as the Turing test, was coined by Alan Turing (1950)
and was designed to determine whether human behaviour could be imitated by computers. The
original imitation game involves three actors: an interrogator (C), a man or woman (B), and
computer (A). Moreover, the objective of the game is for the interrogator to distinguish A from B.
The computer wins when the interrogator cannot reliably distinguish A from B. In line with the
Turing test, chatbots are developed with the goal to deceive people into thinking they are chatting
with a human being (De Angeli, Johnson, & Coventry, 2001).
Weizenbaum’s ELIZA (1966) is often considered to be one of the first chatbots able to fool
its users. ELIZA inspects the input message and flags important keywords. Subsequently,
identified keywords then lead to a transformation of the user input according to an associated

5
rule, and the resulting sentence is then returned to the user (De Angeli et al., 2001). For instance,
if the input contains the keyword “concentration”, ELIZA’s response could be “have you recently
had enough sleep?”. ELIZA does not understand the reasoning behind the occurring
transformation, but simply matches the identified keywords and provides the user with a standard
response (Shawar & Atwell, 2007). Furthermore, Shawar and Atwell argued that although ELIZA
had its shortcomings (i.e. it often did not understand the users’ input), it nonetheless was the
inspiration for many modern chatbots that aim to fool its users that they are conversing with
another human.

3.2 Implementations of Chatbots

Recently, chatbots are found in a broad range of applications spanning a wide variety of domains.
Chatbots used in education might for example improve a learners critical thinking skill (Goda,
Yamada, Matsukawa, Hata, & Yasunami, 2014). Moreover, Kerry, Ellis and Bull (2009) argue that
conversational agents that extend traditional education systems “are capable of offering bespoke
support for each individual, and to recognise and build upon the strengths, interests and abilities
of individuals in order to foster engaged and independent learners.” Language-learning platform
Duolingo, for instance uses chatbots to get people acquainted with new languages. As users
speak to their chatbots it will learn from this conversation and responds according to the learners’
abilities in order to keep users engaged and allow them to learn a language at their own pace5. In
other words, chatbots are valuable to the educational system as they can be adopted to an
individual’s abilities and learning style. Eventually, this may accelerate the learning process or
lead to improved grades. Another domain in which chatbots are successfully applied is healthcare,
as they help people to lose weight (Kowatsch et al., 2017) and accompany elderly that suffer from
dementia and struggle with loneliness (Shaked, 2017). Notice that the first chatbot, ELIZA, was
also developed with the healthcare in mind, as it simulated real therapy conversations with people
(Weizenbaum, 1966). A more recent example of a healthcare chatbot is Your.MD, which uses
artificial intelligence (AI) to provide patients with the most relevant health information and to
connect them to the right care professional, if needed6. Most noteworthy, however, is the fast
adaption of chatbots in enterprises.

3.2.1 Chatbots in Enterprises

Recently, enterprises are using chatbots at different levels of their business. Fast food chain
Subway, for instance integrated chatbots into their business so that customers could place an
order more easily7. Whereas Marriott International, the hotel chain, allows users to apply for a job
using their chatbot “Marriott Careers”8. According to Lester et al. (2004) there are five major
families of business applications for which chatbots could assume an important role:

5
http://bots.duolingo.com, accessed December 01, 2017
6
https://www.your.md, accessed January 28, 2018
7
https://www.topbots.com/project/subway-food-ordering-bot-review/, accessed February 2, 2018
8
http://news.marriott.com/2017/10/marriott-international-launches-careers-chatbot-facebook-messenger/, accessed
December 01, 2017

6
 • Customer service: answering customers’ general questions about products and services
(e.g. answering questions about how to configure a product)
• Help desk: internally responding to employees’ questions (e.g. questions related to
payslips)
• Website navigation: directing customers relevant portions of complex websites
• Guided selling: helping potential buyers to choose the product or service best fulfilling their
needs and guiding them to a buying decision.
• Technical support: providing users assistance with technical problems (e.g. diagnosing
software problems)

Moreover, he argues that two types of chatbot deployments exist for enterprises. In customer-
facing deployments, chatbots interact directly with the customer in order to help them obtain
answers to their questions. In internal-facing deployments, chatbots are used for matters within
the company such as, training customer sales representatives.
This study primarily focuses on customer faced chatbot deployments, and customer
service applications more specifically. It does so since this type of applications directly interacts
with the customer and they often are a customers’ first point of contact with a business. Therefore,
by focusing on this type of applications probably leads to the clearest results when measuring
customer satisfaction and therefor is most relevant for the subject under study. Furthermore, this
study limits itself to chatbots that use textual inputs since this is the most commonly used sort of
chatbot used by businesses.

3.2.2 Requirements for enterprise chatbots

When implemented within an enterprise a chatbot should gratify two sets of requirements to be
effective, natural language requirements (functional) and enterprise delivery requirements (Lester
et al., 2004). The first set of requirements are the system’s functional requirements and specify
what the system should be able to do. The latter set of requirements are the system’s non-
functional requirements and specify how the system should behave. As mentioned above, a
chatbot is partially defined by its ability to process natural language. For a chatbot to productively
engage in a conversation with users, accurate and efficient natural language processing is
required. The other set of requirements involves the operational effectivity of chatbots in the
enterprise, such as its scalability and reliability. Both sets of requirements will be further
elaborated below according to Lester’s et al. (2004) study.
Natural Language Requirements (functional)
For a conversational agent to respond appropriately to a user’s input, it must 1) interpret the input,
2) determine what actions should be taken in response to the input, and 3) perform the actions,
such as presenting steps to solve a problem or updating a database with a new order.
For instance, if the user’s input were: “I would like to buy a train ticket” the agent must first
determine the meaning of the sentence: the user wants to buy a train ticket. Moreover, the agent
must sense the underlying goals the user is trying to achieve with the sentence. In this case,
although the sentence is formulated as an assertion, it was probably intended as a request to buy
a ticket.

7
 After the agent interpreted the sentence, it should consider how to respond to it
accordingly. This plan of action depends on the current goal of the agent (e.g. offering support or
guide the selling process), the dialog history (previously used sentences by both the agent and
user), and information in databases accessible to the agents (e.g. personal data of customers or
specific information about products). Considering the train ticket scenario, if the agent’s goal is
selling, the user earlier specified the origin and destination of the journey, and the train is not
completely booked, a proper response might be to present a booking form and ask the user to
complete it. Whereas, if the train is completely booked, a proper response would be to offer the
user an alternative choice.
Finally, the agent should execute the previously formulated plan of action. This execution
can manifest itself by means of returning a sentence, presenting information in other modalities
(e.g. video or pictures), and other actions (e.g. logging data to a database). If, for example, the
proper response from the earlier formulated plan of actions was to present a booking form and
ask the user to complete it. The agent should return a sentence such as “Alright! Please fill in the
open fields to complete your booking” along with the form, and log the information into the
database. Figure 3.1 depicts the data flow in a chatbot system. The three steps within the NLP
layer actually consist of several sub-steps, however these will not be further explained as it is
beyond the scope of this study. A more detailed explanation of these sub-steps can be found in
Lester et al. (2004).

Figure 3.1. Data flow in a chatbot, adopted from Lester et al. (2004)

Enterprise Delivery Requirements (non-functional)

In contrast with the previous treated set of requirements, this set of requirements is more
concerned with the integration of chatbots into the enterprise and concerns non-functional
requirements. Only chatbots that are scalable, secure, reliable, integrated without much effort and
achieve high performances can be used within large enterprises.
Scalability: Chatbots deployed by enterprises must scale well to accommodate the
exceptional high volumes of inbound contacts typically associated with them. Chatbots must be
designed to handle continuing expanded roll-outs to address increased capacity. Additionally,
since volumes can increase exponentially during periods of crisis, chatbots must be able to
respond dynamically to different levels of demand.

8
 Performance: Chatbot performance is measured in two ways. First, chatbots must be able
to supply a conversational throughput that matches the demand. They must be able to process
hundreds of requests per minute, with peak rates in the thousands. Second, the chatbot must be
able to handle large numbers of simultaneous conversations in a timely matter. For example, if
the chatbot processes all the pending requests consecutively, it meets the first aspect. However,
as it processes the requests one by one, it probably will not be able to ensure adequate response
times and therefor does not meet the second criteria.
Reliability: Chatbots should be able to reliably answer users’ questions considering
software and hardware failures. Failure mechanisms should be in place, such that if a chatbot
server goes down this will not affect the active conversation. This can be accomplished by other
servers taking over pending and new conversations from the defunct server.
Security: Chatbots should meet the same security criteria as the site on which it is
implemented. Nevertheless, as conversations could contain sensitive and critical information,
chatbots should support standard authorization and authentication mechanisms and sent
conversations over encrypted channels.
Integration: Chatbots should integrate neatly with the enterprise infrastructure that is
already in place, represented by different layers. In the application layer, it is key that they easily
integrate with all important business rules. In the data storage level, they should be able to
smoothly integrate with back-office data such as product catalogs and databases holding
information about customer profiles. Lastly, in the presentation layer, chatbots should properly
integrate with content management systems and personalization engines.
Interestingly, however, is the recent development of personality as a chatbot’s
requirement. Enterprise user experience professional Ultan O’Broin (2017) argues that “The job
to be done by your chatbot is vital because it provides the reason for botifying a task in someone’s
life …” However, “… now the style, tone, and attitude of the chatbot —the personality—
encountered along the journey to getting that job done really makes or breaks a great overall user
experience. And a satisfying UX that resonates personally is a powerful strategy for creating more
customers”. Moreover, by designing a personality for a chatbot it becomes more relevant,
trustworthy, and relatable to its users (Thoms, 2017). In other words, personality is also expected
to be an important non-functional requirement for chatbots as it contributes to the overall quality
of the service. In fact, when observing the historical development of software systems, personality
can be regarded as an additional non-functional requirement. With traditional software systems,
it was mainly important that they did what they had to do. For instance, word processors are
expected to process text, and they do not need personality to be able to do this. However, within
modern software systems, such as chatbots, user experience (UX) has been assigned a more
significant role. As aforementioned personality is the tool to improve the UX of a chatbot.
Personality in general, the different aspects to it, and its implications will be discussed
more extensively in section four.

9
3.3 Potential of Chatbots
Chatbots are expected to assume a dominant position into people’s lives, with technology
research and consultancy firm Gartner (2016) even predicting that “By 2020, the average person
will have more conversations with bots than with their spouse.” Moreover, a study examining
chatbots in the retail, e-commerce, banking and healthcare sector prognoses that in 2022
chatbots will be responsible for cost savings of more $8 billion per year, compared to $20 million
this year (Foye, 2017). Elaborating on the business context, chatbots show the potential to
improve the customer service offered by a business as they are not bounded by time and therefor
have the ability to offer customer support 24/7, positively impacting customer satisfaction.
Additionally, chatbots provide businesses with the opportunity to monitor and anticipate to
customer behaviour more closely since the conversations between them and the chatbot can be
analysed more extensively. As the AI behind chatbots will keep improving itself in the near future,
this will probably result in improved analysis of conversations, more sophisticated responses, and
new opportunities for developers to create even better conversational experiences.

10
4 Personality
Personality can be defined as the relatively enduring styles of thinking, feeling, and acting that
characterize an individual (Costa & McCrea, 1995). However, personality is not something
universal and according to Goldberg (1990) the variety of individual differences is nearly
boundless. However, one way to look at personality is that of psychological traits (or personality
traits). Psychological traits are characteristics that describe ways in which people are different
from each other or define ways in which people are similar (Larsen & Buss, 2009). The remainder
of this section will first briefly summarize the development of a shared taxonomy on personality.
Then, the taxonomy adopted in this study will be further elaborated on, describing how it
conceptualizes personality. Finally, this section will end by discussing the habit humans have of
assigning non-human objects a personality.

4.1 Taxonomy on Personality

For over decades researchers have been exploring the concept of personality in order to
formulate a shared taxonomy (Digman, 1990). In their book, Srivastava and John (1999) argued
that a taxonomy on personality should provide a systematic framework for distinguishing,
ordering, and naming types and characteristics of individuals. One of the first attempts to
formulate such a personality framework was done by Thurstone (1934). In his work, he found that
sixty adjectives, that are in common use for describing people, could be accounted for by only
five independent common factors. Almost a decade later, Cattell (1943, 1947, 1948) also
developed a taxonomy on personality, resulting in a complex system consisting of sixteen primary
factors and five second-order factors (Cattell & Mead, 2008). However, most researchers seem
to favor the five-factor model (FFM) on personality, due to their unsuccessful attempts to replicate
Cattell’s findings and their own data supporting the five-dimensional view (e.g. Digman, 1990;
Fiske, 1949; Tupes & Christal, 1961). This thesis also adopts the five-factor model on personality
since its validity is proven by various studies and the reasons described above.

4.2 The Five Dimensions

Within the developed body of literature consensus concerning the number of factors seems to be
reached. However, there is much disagreement over the interpretation of each of those factors.
Norman (1963), for instance, labels the five dimensions as follows:
I) Extraversion or Surgency
II) Agreeableness
III) Conscientiousness
IV) Emotional Stability
V) Culture
However, some others for instance have labelled Dimension IV as neuroticism (Digman,
1990; Eysenck, 1970; McCrae et al., 2005) and Dimension V as intellect (e.g. Digman, 1990;
Fiske, 1949). In the remainder of this study the factor names used in the Neuroticism-
Extraversion-Openness Personality Inventory 3 (NEO PI-3) developed by Costa, McCrea & Martin
(2005) will be adopted — Extraversion (E); Neuroticism (N); Openness to Experience (O);
Agreeableness (A); and Conscientiousness (C) — also known under the acronym OCEAN. This

11
questionnaire was chosen as it is considered as one of the most widely used and researched tool
for the operationalization of the five-factor model (Hoekstra & Filip, 2014).
v Openness to Experience is characterized by having an active imagination, being artistic,
having attention to inner feelings, preferring variety, and being intellectual curious (McCrae
& John, 1992). People who score high on openness have more difficulty ignoring
previously experienced stimuli, exhibit less prejudice, and tend to remember their dreams
more. Facets associated with openness are fantasy, aesthetics, feelings, actions, ideas,
and values (McCrae et al., 2005)
v Conscientiousness implies the desire to take obligations to others seriously and to do a
task well (Thompson, 2008). Conscientious individuals tend to be more passionate and
perseverated for long-term goals, are having more stable and secure relationships, and
have greater job satisfaction and security (Larsen & Buss, 2009). Facets corresponding
to this trait are competence, order, dutifulness, achievement striving, self-discipline, and
deliberation (McCrae et al., 2005)
v Extraversion reflects people’s desire to be with other people and to draw energy from them
(Toegel & Barsoux, 2012). Extraverts frequently engage in social interaction, take the lead
in livening up dull gatherings, and enjoy talking a lot (Larsen & Buss, 2009). Research on
extraversion suggests that extraverts tend to be happier, more involved and more
cooperative than introverts are (Larsen & Buss, 2009). Warmth, gregariousness,
assertiveness, activity, excitement seeking, and positive emotions are facets associated
with extraversion (McCrae et al., 2005).
v Agreeableness is manifested in behavioural characteristics that are perceived as
sympathetic, cooperative, kind, considerate and warm (Thompson, 2008). Agreeable
individuals get along well with others, are well liked, avoid conflict, and prefer professions
in which their likability is an asset (Larsen & Buss, 2009; Toegel & Barsoux, 2012). Facets
that belong to this personality trait are trust, straightforwardness, altruism, compliance,
modesty, and tender-mindedness (McCrae et al., 2005).
v Neuroticism (or emotional instability), results from a lower threshold for activation in the
limbic system, responsible for emotions such as fear, anxiety, anger, and distress (Rusting
& Larsen, 1997). Neurotic individuals tend to overreact to unpleasant events, take longer
to return to a normal state after being upset, are easily irritated, worry about many things,
and seem to be constantly complaining (Larsen & Buss, 2009). Moreover, to identify
persons high on neuroticism, facets like anxiety, angry hostility, depression, self-
consciousness, impulsiveness, and vulnerability are useful (McCrae et al., 2005).

4.3 Anthropomorphism
Interestingly, such human-like characteristics are not only attributed to humans, but humans also
tend to attribute such characteristics to non-lifelike artefacts. This process of attributing human-
like characteristics to non-lifelike artefacts is called anthropomorphism, or personification, and
helps people to rationalize the artefact’s actions (Duffy, 2003). A great effort in this field of
research has been contributed by computer scientists, who for over a decade have been studying
anthropomorphism and its role in the design of human-robot interaction (HRI) and socially
interactive robots (Duffy, 2003; Fink, 2012; Nowak & Rauh, 2005). Anthropomorphism is not

12
bounded to the technological domain, however, and can also be found in other domains such as
nature (Tam, Lee, & Chao, 2013), animals (Horowitz & Bekoff, 2007), and religion (Barrett & Keil,
1996; Barrett & Richert, 2003). In their search for why people tend to anthropomorphise, Epley,
Waytz and Cacioppo (2007) identified three constructs.
The first construct, elicited agent knowledge, implies that knowledge about humans in
general, or the self in particular, functions as the known and often readily accessible base for
induction about the properties of unknown agents. As knowledge about non-human agents is
acquired, however, knowledge about humans or the self will most likely be substituted (Epley et
al., 2007). In other words, in situations when people have a lack of understanding of the non-
human agent, mainly due to missing information, they tend to supplement this gap with their
general knowledge about humans or specific knowledge about themselves. In sum, elicited agent
knowledge is about the accessibility and applicability of human centred knowledge (Epley et al.,
2007).
The second construct, sociality motivation, describes the need and desire to establish
social connections with other humans. Anthropomorphism enables social satisfaction by
representing non-human agents as sources of humanlike social connection. In short, when people
feel socially disconnected from other humans, they construct sources of connection by creating
humanlike agents out of non-humans through anthropomorphism non-human order to satisfy their
motivation for social connection (Epley et al., 2007; Waytz et al., 2010).
The last construct, effectance motivation, entails a desire for understanding, predicting,
and controlling one’s environment (Waytz et al., 2010). When people face an unknown agent for
the first time, it is likely to assume that they are unfamiliar with its behaviour and motivations. As
aforementioned, in such situations people tend to use their general knowledge about humans or
specific knowledge about themselves to create a better understanding of the agent. In line with
this, anthropomorphism may also increase peoples sense of control over the unknown agents
and making its actions more predictable. In short, effectance motivation entails the motivation to
understand and explain the behaviour of other agents (Epley et al., 2007).
Besides the fact that people anthropomorphize objects in order to make sense of an
otherwise uncertain environment, numerous researchers have also studied other effects that
result from anthropomorphizing nonhuman agents. For instance, Blanchard and Mcnincht (1984)
discovered that anthropomorphism enhances the learning and retention of words among children.
Another study, investigating the effect of brand anthropomorphism, found that people are likely to
take on the behaviours they associate with a brand image when this brand is anthropomorphised
(Aggarwal & McGill, 2011). Yet, research more related to the information science domain found
that anthropomorphizing technology especially influences its credibility and trustworthiness
(Nowak & Rauh, 2005; Waytz, Heafner, & Epley, 2014).

13
5 Personality and Customer Satisfaction
As mentioned before, enterprises are using chatbots at different levels in their business. Yet, this
thesis focuses on customer-facing chatbot applications, and particularly customer support (or
customer service) applications. Goffin & New (2001) identified several reasons for businesses to
employ customer support services, such as achieving higher customer satisfaction, providing a
competitive advantage, and increasing the success rate of new products.
The remainder of this section will discuss the influence of personality from service
employees on customer satisfaction. First, it will cover customer satisfaction in a more general
way along with the antecedents that play an important role in it. Then, once a better understanding
of customer satisfaction has been created, earlier research investigating the influence of
personality on customer satisfaction will be discussed.

5.1 Customer Satisfaction

A popular view on customer satisfaction is the disconfirmation paradigm which explains customer
satisfaction by means of four constructs: expectations, performance, disconfirmation, and
satisfaction. Customers’ expectations are confirmed when a product or service functions as
expected, negatively disconfirmed when the product or service functions dissatisfy expectations,
and positively disconfirmed when the product or service exceeds expectations (Churchill &
Surprenant, 1982). For instance, when a customer buys a new mobile phone he or she has certain
expectations of the phone. However, after receiving the mobile phone, it does not perform as
anticipated and therefore underperforms the customers’ original expectations. As the customers’
expectations are now negatively disconfirmed, this could lead to increased dissatisfaction with
the product. In other words, customer satisfaction is “the number of customers, or percentage of
total customers, whose reported experience with a firm, its products, or its services … exceeds
specified goals” (Farris, Bendle, Pfeifer, & Reibstein, 2010, p. 57). Figure 5.1 depicts the process
of reaching satisfaction within the disconfirmation framework, as described by numerous studies
(e.g. Anderson & Sullivan, 1993; Churchill & Surprenant, 1982; Oliver, 1980; Yi, 1993). The direct
lines from both expectations and performance to satisfaction are dotted because they do not
always exist, but rather depend on the ambiguity of the product or service (Yi, 1993).
As aforementioned, within businesses customer support services play an important role
in achieving higher customer satisfaction goals. Such customer support services are often the
only contact point a customer has with a company and can greatly influence customers’
impression about the company. In their updated IS Success Model, Delone and McLean (2003)
identify three important dimensions that affect customer satisfaction within an information system
(IS): system quality, information quality, and service quality. System quality refers to the quality
of information processing itself, including software and data components, and measures the
technical soundness of the system (e.g. ease of use, whether or not the system contains bugs,
or quality of documentation) (Gorla, Somers, & Wong, 2010). Information quality is concerned
with the quality of the information system outputs such as web pages or reports. Items associated
with information quality are: accuracy, relevance, usability, conciseness, and understandability
(Petter, DeLone, & McLean, 2008). Service quality represents the quality of the support that users
receive from the IT support personnel and is often defined as the degree of discrepancy between
what users expect of a service and the performance of that service (Gorla et al., 2010; Petter et

14
al., 2008). The first two constructs are primarily concerned with the technical aspect of an IS,
whereas the latter construct is more concerned with the relational aspect of an IS. This study
limits itself to the service quality as it focusses on the relational aspect of a customer service and
is believed to be the core criterion for overall customer service, and thus matches best with the
scope of the research (Parasuraman, Berry, & Zeithaml, 1991).

Figure 5.1. The disconfirmation paradigm, adopted from Yi (1993)

5.1.1 Service Quality Models

Numerous conceptualisations of the service quality dimension have been proposed over the
years. Grönroos’ Nordic Model (1984) was one of the first to formulate such a model and was
based on the earlier discussed disconfirmation paradigm. This implicates that service quality
result from comparing expected performance with perceived performance. Moreover, Grönroos
identified two service quality dimensions, functional quality (how the service is delivered) and
technical quality (what the service delivers). However, the Nordic model was limited in the
dimensions to be measured and did not offer techniques to measure either its quality- or functional
dimension. Later, Rust and Oliver (1994) refined Grönroos’ model by adding another dimension
to it. Their three-component model consists of, the service product (comparable to technical
quality), the service delivery (functional quality), and service environment. However, they did not
test their model.
Another prominent model that is also based on the disconfirmation paradigm, is the
SERVQUAL model developed by Parasuraman, Zeithaml and Berry (1985, 1988). They argue
that, irrespective of the type of service, consumers assess service quality using the same generic
criteria, which can be arranged into the following five dimensions:
§ Reliability: the ability to accurately and dependably perform the promised service.
§ Assurance: employees’ ability to express confidence and trust along with their knowledge
and courtesy
§ Tangibles: the presence of physical communication material, equipment, and personnel.
§ Empathy: measure of the care and attention provided to customers.
§ Responsiveness: the readiness to help customers and to offer fast service.
Yet, despite the models’ popularity among academics it was also greatly criticized. For an
extensive review of the SERVQUAL model and its critiques refer to Buttle (1996). However, the
most heard critique is its dimensionality problem. For instance, the model states that the service

15
offered should be reliable, but does not specify what needs to be reliable (Brady & Cronin, 2001).
Responding to SERVQUAL’s shortcomings, Dabholkar, Thorpe and Rentz (1966) proposed a
multidimensional conceptualization of retail service quality, the Retail Service Quality Scale
(RSQS). This model consisted of three levels: 1) overall perceptions of service quality, 2) a
dimensional level consisting of five primary dimensions, and 3) a sub dimensional level consisting
of six sub dimensions (see Figure 5.2). According to Martinez and Martinez (2010) this model
distinguishes itself from previous models as they argue that service quality is not formed by but
defined by several dimensions. However, its limitation lies in the fact that it is specifically designed
for a retail setting and therefore cannot be generalized.

Figure 5.2. Overview of Service Quality Models, adopted from Brady & Cronin (2001)

The last model is a combination of the three-component model from Rust and Oliver
(1994) with the multilevel model proposed by Dabholkar et al. (1996), suggested by Brady and
Cronin (2001). They argue that service quality is determined by three primary dimensions which
are defined by nine sub dimensions. The first primary dimension, interaction quality, refers to the

16
interpersonal interactions that take place between an employee and a customer. Moreover, Brady
and Cronin (2001) indicate that attitude, behaviour, and expertise are the three sub dimensions
that influence customer perceptions of interaction quality. The second primary dimension,
environmental quality, refers to the environment in which the service encounter takes place and
its quality is influenced by ambient conditions (e.g. temperature and music), facility design
(functional or aesthetic layout), and social factors (number, type, and behaviour of people)(Brady
& Cronin Jr, 2001). The last primary dimension, output quality, refers to the actual product the
customer receives when the process has been completed and is associated with valence (degree
to which the service outcome is considered good or bad), tangibles (physical results), and waiting
time (Brady & Cronin Jr, 2001). In turn, these sub dimensions are evaluated by using some of the
SERVQUAL dimensions (see Figure 5.3). Moreover, perceptions of an organization’s
performance on each of the three primary dimensions is formed by combining customers’
evaluation of the sub dimensions. An overall service quality perception is then formed by
aggregating those perceptions.

Figure 5.3. The Hierarchical Service Quality Model, adopted from Brady and Cronin (2001)

The conceptualization of service quality used in this study is the hierarchical model
proposed by Brady and Cronin (2001) because it has been used in numerous studies that found
strong support for this model (e.g. Akter, D’Ambra, & Ray, 2013; Wu & Ko, 2013; Zhao, Lu, Zhang,
& Chau, 2012). Additionally, Brady and Cronin (2001) argue that neither perspective (two-
dimensional or multi-dimensional) is wrong, but is incomplete without the other instead. Therefore,
it can be seen as an improvement on both perspectives. Finally, the focus of this study is on
interaction quality rather than the other two dimensions since the interpersonal interactions that
take place during service delivery often influence service quality perceptions the most (Ekinci &
Dawes, 2009).

17
5.2 Personality Influencing Customer Satisfaction
So far, little research has been published concerning the effects of front-line employee personality
on service quality and thus on customer satisfaction. However, it is argued that front-line service
employees play an important role in developing high satisfaction among customers as employees’
attitudes, skills, behaviours, and personalities influence customers perceptions of service quality
(Bowen & Schneider, 1985; Ekinci, Dawes, & Massey, 2008). Moreover, Ekinci and Dawes (2009)
suggest that customer-oriented service behaviours (e.g. responding, smiling, and helping) are
likely to be influenced by the personality traits of service employees.
Hurley (1998) was one of the first to explore the effect of personality on service quality
among front-line employees. The results of his research indicated that service quality is influenced
by personality and that superior service providers show higher levels of extraversion and
agreeableness. Additionally, Liao and Chuang (2004) investigated the impact of four personality
dimensions (extraversion, conscientiousness, agreeableness, and neuroticism) on service
performance (employees’ behaviour) and customer outcomes (e.g. customer satisfaction and
customer loyalty). They found that conscientiousness and extraversion had a significant positive
relationship with service performance and positively influenced customer satisfaction. They
excluded the openness to experience dimension from their study as there was no convincing
evidence to support a relationship between openness to experience and service performance.
More recently, Ekinci and Dawes (2009) examined the relationship between frontline
service employee personality traits, interaction quality, and customer satisfaction. They proposed
a model in which the effects of the five earlier described personality dimensions – openness to
experience, conscientiousness, extroversion, agreeableness, neuroticism (section 4.2) – on
customer satisfaction was mediated by interaction quality. Additionally, they argued that employee
personality traits can be conceived as direct determinants of interaction quality since they appear
to be related to employee attitudes and customer orientation, both of which have been identified
as determinants of interaction quality. The results of their research showed that customer
satisfaction is indeed influenced by the personality of employees. However, they only found
support for four of the five traits. Extraversion, conscientiousness, and agreeableness were found
to have a significant impact on interaction quality, whereas openness to experience was found an
important (direct) predictor of customer satisfaction.
By comparing these three studies, it becomes apparent that they show overlapping results
for three of the five aforementioned personality dimensions. Hurley (1998) identified extraversion
and agreeableness as important traits for superior service providers, Liao and Chuang (2004)
found that conscientiousness and extraversion positively influenced customer satisfaction, and
the most extensive and recent research on this topic from Ekinci and Dawes (2009) concluded
that extraversion, agreeableness, and conscientiousness were strong predictors on interaction
quality and thus customer satisfaction. In other words, there is scientific evidence that the
personality traits of service employees are likely to be key determinants of customer satisfaction,
extraversion, agreeableness, and conscientiousness in particular.

18
6 Personality in Chatbots: survey
In the previous section, some studies that investigated the influence of a service employees’
personality on perceived customer satisfaction have been outlined. These studies found that
certain personality traits of customer support representatives are more likely to influence customer
satisfaction than others. Moreover, as mentioned before personality is not exclusively attributed
to humans, but is also attributed to non-human artefacts in order to rationalize its actions.
Additionally, earlier research indicated that anthropomorphism plays a role in the design of
socially interactive robots (Duffy, 2003; Fink, 2012; Nowak & Rauh, 2005). More recently, some
contributions have been made with respect to anthropomorphising chatbots that, for instance,
suggest that humanizing chatbots affects their trustworthiness (Seeger & Heinzl, 2018).
Nevertheless, research on the personality of a chatbot remains scarce in the scientific literature.
Despite the fact that this topic is currently scientifically misrepresented, some internet sites
have been writing about assigning a personality to chatbots. Chatbot Magazine is probably the
most popular source for information about chatbots and have recently also published some
articles about the personality of a chatbot. For instance, one of their articles gave a superficial
description about how people could design a chatbot’s personality, whereas another article
reasoned why chatbots benefit from having a personality (Shinde, 2016; Zilnik, 2016). Yet, other
sites also share Chatbot Magazine’s interest in this topic. Technology firm Xandra, for instance,
published a guide to developing chatbot personalities, including why they are so important and
an enumeration of things to considers when building a chatbot with personality (Thoms, 2017).
Award-winning chatbot platform ‘The Personality Forge’ even lets people build their own chatbot
with personality using their AI engine, which integrates memories, emotions, knowledge of
hundreds of thousands of words, sentence structure, unmatched pattern-matching capabilities,
9
and a customized scripting language .
To summarize, there is evidence that the personality of human customer support
representatives has an impact on customer satisfaction and that personality is also attributed to
non-human artefacts, such as chatbots. Therefore, it is plausible to assume that the personality
of chatbot customer support representative also influences customer satisfaction.
The remainder of this section will discuss the results of the survey as well as the survey
itself, that was send to businesses in order to validate the aforementioned hypothesis by capturing
their perspectives.

6.1 Survey: methodology

Participants
As mentioned in section 2.2, this survey was distributed among sixty businesses employing a
customer support. These businesses were selected by means of purposive sampling and they
were all located in the Netherlands. The businesses were chosen so that they covered a wide
variety of industries and included businesses active in telecom, retail, finance, energy, aviation,
consumer electronics, insurances, and public transport. It was chosen to distribute this survey
among businesses rather than customers, as businesses probably maintain a communication
policy that informs their employees about how they should behave in interaction with the

9
https://www.personalityforge.com, accessed January 25, 2018

19
customer. This should lead to less ambiguity. Only five businesses have been included in the
results, as the other fifty-five businesses either did not complete the survey or did not respond.

Materials
The survey consists of 16 items including 3 items about demographics, 4 items about customer
support, and 9 items about personality traits (see Appendix A for an overview of all items).
Because of the exploratory nature of this study, guidelines cannot yet be derived from earlier
studies in the formulation of a measure scale. Rather the questions are inspired by the theory
originating from the literature review and with the research question in mind. In addition, the
questions are independently evaluated by two other parties. Depending on the answers given by
participants, participants have to complete paths of different lengths, ranging from seven items to
ten items (see Appendix A). Table 6.1 gives an overview of the seven most important items of the
survey, the demographic items were left out as they were not filled out by all participants and
therefore are less meaningful. Also, it should be noted that in the last question the neuroticism
dimension was reversed to better fit with the other dimensions and to avoid confusion. The NEO
PI-3 (Costa & McCrae, 2005) was used to map specific personality traits to the corresponding
dimension. The NEO PI-3 is a 240-item questionnaire, with a 5-point Likert scale assessing an
individual's’ personality in terms of the big five personality dimensions (e.g. O: Our ideas of right
and wrong may not be right for everyone in the world; C: I’m picky about how jobs should be
done; E: I act forcefully and energetically; A: Human need is more important than economics; N:
I feel awkward around people).

Table 6.1

Sample of questions from the survey.

1. What is the main reason to employ a customer support service within your company?
2. What kind of customer support representatives does your business employ?
3. What kind(s) of channels for customer support service(s) does your business employ?
4. Which personality traits does your company value in your human customer support
representatives?
5. Which personality traits does your company value in your chatbot customer support
representatives?
6. In the (hypothetical) case that your company uses chatbots as customer support
representatives, would it then value other traits for chatbots than for its current
representatives?
7. Rank (drag & drop) the following five personality dimensions on its importance,
considering your customer support representatives

20
Procedure
The businesses were contacted by either e-mail or Facebook messenger and the invitation
included a letter (as can be seen in Appendix B). This letter contained the request to participate,
information about the relevance, the tasks the businesses would have to perform, the workload,
as well as the confidentiality and anonymity of the study. If businesses did not respond, a reminder
was sent after a week. The questionnaire was accessible by means of a link that was also included
in the letter. Before the actual questionnaire was shown, a brief introduction appeared with some
information about the research, the anonymity and the confidentiality.

Analysis
It is expected that the results of this survey replicate those of the previous literature study, which
means that extraversion, agreeableness, and conscientiousness will be identified as the most
important personality dimensions influencing customer satisfaction. Moreover, it is expected that
businesses will value the same personality traits for their chatbot representatives as they do for
their human representatives.

6.2 Survey: results

When analysing the results of the survey, it becomes apparent that businesses most often (80%)
maintain an internal customer support service in order to provide customers with information and
help them with their questions about the provided services. One of these businesses specifically
identified a positive Netto Promoter Score (NPS), resulting from an increase in customer
satisfaction, as a reason for having customer support. Only one of the participated businesses
stated to employ both human and chatbot customer support representatives, whereas the other
businesses only employ human representatives. Within the sample all businesses maintain a
telephone communication channel as well as a social media presence, whereas mail was used
by all but one and communicating by chat being the least common as only three businesses offer
this option. To the question which personality traits businesses deemed most important for their
human representatives, resulted in varied answers (see Table 6.3). However, empathetic ability
was mentioned most often as personality trait valued in human customer support representatives.
After mapping the given traits to its accompanying dimension, it also became apparent that most
of these traits could be placed within the agreeableness dimension, whereas only one trait
corresponded to the extraversion dimension (see Table 6.2). Only one business replied to the
question about which personality traits they deemed important for chatbot representatives, as
they were the only business in the sample to have chatbots as customer support representatives.
However, the other businesses had to answer an alternative question about the hypothetical case
in which they would have chatbot representatives. On both questions businesses answered that
they do not value different personality traits for chatbots as they do for humans. Business 1, which
currently implements a chatbot, did add “proactive thinking in finding solutions” as a valued trait
for chatbots (see Table 6.3). However, they later stated that they also deem this an important trait
for human representatives, but according to them this was self-evident. The last question, that
asked the participants to rank the different personality dimensions on their importance, also
resulted in varying outcomes. Nevertheless, extraversion was by all businesses identified as the
least important dimension, whereas openness to experience and agreeableness were most often
placed within the top three.

21
Table 6.2

Mapping of identified personality traits into the accompanying dimension

Big Five dimension Outcomes

Openness Proactive thinking (actions) (1x)
Resourceful (ideas) (1x)
Knowledge (ideas) (1x)
Conscientiousness Client- and solution oriented (competence) (2x)
Accurate (competence) (1x)
Independent (self-discipline) (1x)
Extraversion Accessibility (gregariousness) (1x)

Agreeableness Empathic ability (tender-mindedness) (3x)

Friendliness (altruism) (1x)
Respectful (straightforwardness) (1x)
Helpfulness (altruism) (1x)
Neuroticism -

22
Table 6.3
Survey Results

23
7 Chatbots and Ethics
Chatbots are a promising technology that offer organizations a lot of new opportunities, such as
improving learners critical thinking skills, enriching game interaction with players and helping
people lose weight. However, as with any new technology it has ethical challenges and
implications that need to be considered to warrant that it is used in a responsible way.
An example of a chatbot that was not used in a responsible way was Microsoft’s chatbot
“Tay”. Tay was an experiment that was meant to converse with people on Twitter by learning from
their input. At first the conversations were kind and friendly but soon changed to offensive and
racist. This change in mood was caused by users who were supplying the chatbot with input filled
with offensive and racist content.
Another important consideration in the use of chatbots is transparency. First of all, it is
important to let customers know when they are communicating with a chatbot rather than an
actual human. Bridget Botelho (2017), from technology marketing company TechTarget,
explained that “consumers expect a level of trust, and when they find out a company is using a
machine to interact with them, they could feel betrayed and may even turn against the brand …”
This phenomenon is similar to the “uncanny valley effect”, which proposes that as robots become
more human, people’s emotional response to those robots first increased, but then sharply
declined (Mori, 1970). Just as with robots, people have high expectations of chatbots when they
present themselves as humans. However, when a chatbot not exactly behaves like a real human
being, these expectations are then replaced with distrust. Moreover, ownership of information and
privacy should be clearly communicated with customers. If a chatbot assembles a shopping list
based on earlier orders and user preferences, does it then belong to the chatbot or the user? Can
user information retrieved from conversations with chatbots be sold to third parties? If so, should
the user then be informed about this? According to Trips Reddy (2017), senior content manager
at IBM, businesses that want to implement their own chatbot should address these issues and
should be transparent in communicating their terms of service, privacy policy, and whether or not
the user is conversing with a chatbot.
Yet, probably one of the biggest challenges concerning chatbots lies in their great potential
to replace human beings in their job. Researchers predict that all human jobs will be automated
within 120 years and that there is a 50% probability that this could even happen within the next
45 years (Grace, Salvatier, Dafoe, Zhang, & Evans, 2017). Nevertheless, it is argued that the
highest efficiency can be achieved when chatbots and humans work together (von Malitz, 2016).
Chatbots can be used to answer simple queries in a far more efficient and quickly manner than
an actual human, whereas humans can take over from chatbots when the situations become more
complex. For instance, chatbots are able to conduct the initial interaction with customers and
record the customers’ information and details on the incident, after which it is forwarded to the
most qualified human agent. Additionally, history shows that technological advances, such as
chatbots, do not result higher unemployment rates, but rather boosts employment by creating
jobs in new sectors (Stewart, De, & Cole, 2015). In other words, as chatbots are still in its infancy
they are not expected to take over jobs from humans for the foreseeable future, rather it is
suggested that both human and machine will need to work together and complement each other
in order to attain the highest efficiency. Moreover, even if chatbots eventually take over jobs from

24
humans, history shows that such technological revolutions in the end always result in higher
employment rates.

25
8 Discussion and conclusion
The aim of this research project was to investigate what role the personality of a chatbot has on
the perceived customer satisfaction. In order to do so, first a literature study was performed after
which the findings of this study were tested in a short experimental research.
Since current scientific research on this topic is scarce, the literature study was mainly
conducted on the individual components that made up the main research question. Literature on
chatbots was reviewed and used for assigning it with a clear definition, an overview of its
development, stating requirements for enterprise implementations, and ascribing its potential. To
answer the first research sub-question about how personality can be categorised, literature on
personality was searched for theories about how personality can be defined and measured. This
resulted in the adoption of the five-factor model, categorizing personality in five dimensions:
openness to experience, conscientiousness, extraversion, agreeableness, and neuroticism. Yet,
another interesting finding from this review was that it identified people’s habit of human
characteristics, such as personality, to non-human artefacts. At last, in order to answer the second
research sub-question, marketing literature was reviewed to identify existing models to customer
satisfaction and to identify earlier research that studied the relationship between personality and
customer satisfaction. Interestingly, earlier research studying this relationship indeed existed and
identified that the extraversion, agreeableness and conscientiousness dimensions had the
greatest impact on customer satisfaction. However, these findings apply to human
representatives and therefore cannot be directly applied to chatbots without any further
consideration.
In order to answer the last research sub-question about whether this earlier research on
customer satisfaction and personality can also be applied to chatbots, a short experimental
research, in the form of a survey, was conducted. From the literature study, it was hypothesized
that this would be the case as people would humanize chatbots and thus ascribe a personality to
it. The findings of the survey indeed showed that the participating businesses did not value
different personality traits for human customer support representatives than for chatbot
representatives. This is in accordance with Fong, Nourbakhsh and Dautenhahn’s (2003)
statement that “the common, underlying assumption is that humans prefer to interact with
machines in the same way that they interact with other people.” Therefore, it appears that the
aforementioned research can also be applied to chatbots. However, the results did not exactly
replicate those of the previous studies regarding the personality dimensions. Extraversion,
agreeableness and conscientiousness were by earlier research identified as most important
dimensions influencing customer satisfaction. Yet, in this study businesses did not consider all
dimensions as important and even ranked extraversion as the least important dimension.
Overall, based on both earlier research and the results evident in this research project, it
is assumed that the personality of a chatbot influences customer satisfaction. Since chatbots are
an effective addition to a firm’s customer support and the personality of such a chatbot is likely to
influence customer satisfaction, businesses should explore their options on how chatbots with a
personality can be successfully exploited within their firm. Nevertheless, additional research
needs to be performed to further investigate how customer satisfaction is exactly being influenced
by a chatbot’s different personality dimensions.

26
8.1 Limitations and directions for future research
As does any research project, this study has some limitations. A key one is the use of purposive
sampling as sampling technique. This sampling technique is considered to be effective when time
available is limited and only a limited number of people can serve as primary data source.
However, it is a non-probability sampling method and therefor the resulting research findings
cannot be generalized. Another factor that hinders generalization of this study’s findings is the
small sample size. Accordingly, future research should consider adopting a more comprehensive
sampling design and using a larger sample so that the sample will be more representative for the
entire population.
Since no scientific research existed on this topic, guidelines and scales for measuring the
relationship between the personality of chatbot and customer satisfaction still need to be
developed. The items present in this study resulted from the previous literature review and were
drafted with the research question in mind. Nevertheless, the instrument has yet to be tested on
its validity and reliability. Therefore, for future research it would be relevant to develop an
appropriate, valid instrument that measures the relationship between a chatbot’s personality and
customer satisfaction.
It should also be mentioned that a measurement error might have occurred in the results.
The item that asked the participants to rank the personality dimensions on their importance did
not specifically specify whether the number one referred to the most important dimension and
number five to the least important dimension, or vice versa. Although there is enough reason to
assume that this did not happen, future research will need to demonstrate whether the results
from this item are indeed reliable.
According to Ekinci and Dawes (2009) marketing research uses personality traits to study
a variety of behaviours from two perspectives: personality psychology and social psychology. The
former perspective is from the point of view from employees’ and businesses, whereas the latter
perspective adopts a consumers’ point of view. This study employed the latter perspective as it
investigated the influence of business perceptions of frontline employee personality traits on
customer satisfaction. That is, businesses were asked for the personality traits they valued most
in their customer support representatives. However, for future research it is recommended to
adopt the former perspective, and thus capture the customer’s perspective. In the end, this may
lead to more relevant findings as it directly involves the subject from which customer satisfaction
originates.
Another consideration in developing and implementing chatbots with a personality is its
adaptivity. In this research, personality was examined through the scope of personality
dimensions and which dimensions in general were most likely to influence customer satisfaction.
However, this does not have to mean that the same dimensions are equally applicable in every
situation, rather it may be that in one situation other dimensions are of more importance than in
another scenario. For instance, conscientiousness and agreeableness might be the most
important dimensions for a chatbot representing a financial organization, whereas extraversion
and openness may be of greater importance for a chatbot interacting with younger people.
Therefore, further research should be conducted on the adaptivity of a chatbots’ personality and
how it affects customer satisfaction.
Another research avenue could be to investigate how personality can be manifested in
chatbots. Earlier research has found that a person’s personality is reflected in the way they write,

27
speak, and behave has found that the style in which a text is written reflects the personality of the
author (Ball & Breese, 2000; Luyckx & Daelemans, 2008; Polzehl, 2014). For instance, extravert
people tend to use more words and pronouns in their messages, whereas neurotic peoples’
messages contain more negative emotion words and acronyms (Holtgraves, 2011). More specific
to the human computer interaction (HCI) domain, it was found that robots are also able to express
personalities through gestures (Kim, Kwak, & Kim, 2008). Future research could examine whether
this also applies to chatbots and may identify additional ways for a chatbot to reflect its personality.
In line with this view, the following research questions could be investigated by future
studies: (1) How do customer’s perceptions of a customer support chatbot influence perceived
customer satisfaction? (2) Do customers assess chatbot personality differently in different
contexts (i.e. industries, cultures, or age)? (3) How can personality effectively be established in
chatbots?

28
References
Aggarwal, P., & McGill, A. L. (2011). When brands seem human, do humans act like brands? Automatic
behavioral priming effects of brand anthropomorphism. Journal of Consumer Research, 39(2), 307–
323.
Akter, S., D’Ambra, J., & Ray, P. (2013). Development and validation of an instrument to measure user
perceived service quality of mHealth. Information & Management, 50(4), 181–195.
Anderson, E. W., & Mittal, V. (2000). Strengthening the satisfaction-profit chain. Journal of Service
Research, 3(2), 107–120.
Anderson, E. W., & Sullivan, M. W. (1993). The antecedents and consequences of customer satisfaction
for firms. Marketing Science, 12(2), 125–143.
Ball, G., & Breese, J. (2000). Relating personality and behavior: posture and gestures. In Affective
interactions (pp. 196–203). Springer.
Barrett, J. L., & Keil, F. C. (1996). Conceptualizing a Nonnatural Entity: Anthropomorphism in God
Concepts. Cognitive Psychology, 31(3), 219–247. https://doi.org/10.1006/cogp.1996.0017
Barrett, J. L., & Richert, R. A. (2003). Anthropomorphism or Preparedness? Exploring Children’s God
Concepts. Review of Religious Research, 44(3), 300–312. Retrieved from
http://www.jstor.org/stable/3512389
Blanchard, J., & Mcnincth, G. (1984). The effects of anthropomorphism on word learning. The Journal of
Educational Research, 78(2), 105–110.
Botelho, B. (2017). Chatbot technology raises ethical questions. Retrieved January 15, 2018, from
http://searchcrm.techtarget.com/opinion/Chatbot-technology-raises-ethical-questions
Bowen, D. E., & Schneider, B. (1985). Boundary-spanning-role employees and the service encounter:
Some guidelines for management and research. The Service Encounter, 127, 148.
Brady, M. K., & Cronin Jr, J. J. (2001). Some new thoughts on conceptualizing perceived service quality: a
hierarchical approach. Journal of Marketing, 65(3), 34–49.
Budgen, D., & Brereton, P. (2006). Performing systematic literature reviews in software engineering. In
Proceedings of the 28th international conference on Software engineering (pp. 1051–1052). ACM.
Buttle, F. (1996). SERVQUAL: review, critique, research agenda. European Journal of Marketing, 30(1), 8–
32. https://doi.org/10.1108/03090569610105762
Cattell, H. E. P., & Mead, A. D. (2008). The Sixteen Personality Factor Questionnaire (16PF). In The SAGE
Handbook of Personality Theory and Assessment (pp. 135–159). https://doi.org/DOI:
10.4135/9781849200479
Cattell, R. B. (1943). The description of personality: basic traits resolved into clusters. The Journal of
Abnormal and Social Psychology, 38(4), 476–506. https://doi.org/10.1037/h0054116
Cattell, R. B. (1947). Confirmation and clarification of primary personality factors. Psychometrika, 12(3),
197–220. https://doi.org/10.1007/BF02289253
Cattell, R. B. (1948). The Primary Personality Factors in Women Compared With Those in Men. British
Journal of Statistical Psychology, 1(2), 114–130. https://doi.org/10.1111/j.2044-
8317.1948.tb00231.x
Chowdhury, G. G. (2003). Natural language processing. Annual Review of Information Science and
Technology, 37(1), 51–89.
Churchill Jr, G. A., & Surprenant, C. (1982). An investigation into the determinants of customer
satisfaction. Journal of Marketing Research, 491–504.
De Angeli, A., Johnson, G. I., & Coventry, L. (2001). The unfriendly user: exploring social reactions to
chatterbots. In Proceedings of The International Conference on Affective Human Factors Design,
London (pp. 467–474). London.
Delone, W. H., & McLean, E. R. (2003). The DeLone and McLean model of information systems success: a

29
 ten-year update. Journal of Management Information Systems, 19(4), 9–30.
Digman, J. M. (1990). Personality Structure: Emergence of the Five-Factor Model. Annual Review of
Psychology, 41(1), 417–440. https://doi.org/10.1146/annurev.ps.41.020190.002221
Duffy, B. R. (2003). Anthropomorphism and the social robot. Robotics and Autonomous Systems, 42(3–
4), 177–190. https://doi.org/10.1016/S0921-8890(02)00374-3
Ekinci, Y., & Dawes, P. L. (2009). Consumer perceptions of frontline service employee personality traits,
interaction quality, and consumer satisfaction. The Service Industries Journal, 29(4), 503–521.
Ekinci, Y., Dawes, P. L., & Massey, G. R. (2008). An extended model of the antecedents and
consequences of consumer satisfaction for hospitality services. European Journal of Marketing,
42(1/2), 35–68.
Epley, N., Waytz, A., & Cacioppo, J. T. (2007). On seeing human: A three-factor theory of
anthropomorphism. Psychological Review, 114(4), 864–886. https://doi.org/10.1037/0033-
295X.114.4.864
Eysenck, H. J. (1970). The structure of human personality (3rd ed.). London: Taylor & Francis Ltd.
Retrieved from https://books.google.nl/books?id=q1dqAAAAMAAJ
Farris, P. W., Bendle, N., Pfeifer, P., & Reibstein, D. (2010). Marketing metrics: The definitive guide to
measuring marketing performance (Second). Pearson Education.
Fink, J. (2012). Anthropomorphism and human likeness in the design of robots and human-robot
interaction. Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial
Intelligence and Lecture Notes in Bioinformatics), 7621 LNAI, 199–208.
https://doi.org/10.1007/978-3-642-34103-8_20
Fiske, D. W. (1949). Consistency of the factorial structures of personality ratings from different sources.
The Journal of Abnormal and Social Psychology, 44(3), 329–344. https://doi.org/10.1037/h0057198
Fong, T., Nourbakhsh, I., & Dautenhahn, K. (2003). A survey of socially interactive robots. Robotics and
Autonomous Systems, 42(3), 143–166.
Foye, L. (2017). Chatbots: Retail, e Commerce, Banking & Healthcare 2017-2022. Basingstoke, United
Kingdom.
Goda, Y., Yamada, M., Matsukawa, H., Hata, K., & Yasunami, S. (2014). Conversation with a Chatbot
before an Online EFL Group Discussion and the Effects on Critical Thinking. The Journal of
Information and Systems in Education, 13(1), 1–7.
Goffin, K., & New, C. (2001). Customer support and new product development-An exploratory study.
International Journal of Operations & Production Management, 21(3), 275–301.
Goldberg, L. R. (1990). An alternative “description of personality”: The Big-Five factor structure. Journal
of Personality and Social Psychology, 59(6), 1216–1229. https://doi.org/10.1037/0022-
3514.59.6.1216
Gorla, N., Somers, T. M., & Wong, B. (2010). Organizational impact of system quality, information
quality, and service quality. The Journal of Strategic Information Systems, 19(3), 207–228.
https://doi.org/https://doi.org/10.1016/j.jsis.2010.05.001
Grace, K., Salvatier, J., Dafoe, A., Zhang, B., & Evans, O. (2017). When will AI exceed human
performance. Evidence from AI Experts. CoRR.
Grönroos, C. (1984). A service quality model and its marketing implications. European Journal of
Marketing, 18(4), 36–44.
Hoekstra, H., & Filip, D. F. (2014). NEO-PI-3: Persoonlijkheidsvragenlijst. Amsterdam: Hogrefe Uitgevers.
Hokanson, S. (1995). The deeper you analyze, the more you satisfy customers. Marketing News, 29(1),
16.
Holtgraves, T. (2011). Text messaging, personality, and the social context. Journal of Research in
Personality, 45(1), 92–99.
Horowitz, A. C., & Bekoff, M. (2007). Naturalizing anthropomorphism: Behavioral prompts to our

30
 humanizing of animals. Anthrozoos, 20(1), 23–35. https://doi.org/10.2752/089279307780216650
Hurley, R. F. (1998). Customer service behavior in retail settings: A study of the effect of service provider
personality. Journal of the Academy of Marketing Science, 26(2), 115–127.
Joppe, M. (2000). The Research Process. Retrieved May 28, 2007, from
http://www.ryerson.ca:80/~mjoppe/rp.htm
Kerly, A., Hall, P., & Bull, S. (2007). Bringing chatbots into education: Towards natural language
negotiation of open learner models. Knowledge-Based Systems, 20(2), 177–185.
Kerry, A., Ellis, R., & Bull, S. (2009). Conversational agents in E-Learning. Applications and Innovations in
Intelligent Systems XVI, 169–182.
Kim, H., Kwak, S. S., & Kim, M. (2008). Personality design of sociable robots by control of gesture design
factors. In Robot and Human Interactive Communication, 2008. RO-MAN 2008. The 17th IEEE
International Symposium on (pp. 494–499). IEEE.
Kowatsch, T., Nißen, M., Shih, C.-H. I., Rüegger, D., Volland, D., Filler, A., … Büchter, D. (2017). Text-
based Healthcare Chatbots Supporting Patient and Health Professional Teams: Preliminary Results
of a Randomized Controlled Trial on Childhood Obesity. In Proceedings of the Persuasive Embodied
Agents for Behavior Change (PEACH2017) Workshop, co-located with the 17th International
Conference on Intelligent Virtual Agents (IVA 2017).
Larsen, R. J., & Buss, D. M. (2009). Personality Psychology: Domains of Knowledge About Human Nature
(4th ed.). McGraw Hill Higher Education.
Lester, J., Branting, K., & Mott, B. (2004). Conversational agents. In M. P. Singh (Ed.), The Practical
Handbook of Internet Computing (1st ed., pp. 220–240). New York: New York: Chapman & Hall.
Levy, Y., & Ellis, T. J. (2006). A systems approach to conduct an effective literature review in support of
information systems research. Informing Science, 9.
Liao, H., & Chuang, A. (2004). A multilevel investigation of factors influencing employee service
performance and customer outcomes. Academy of Management Journal, 47(1), 41–58.
Luyckx, K., & Daelemans, W. (2008). Using syntactic features to predict author personality from text.
Proceedings of Digital Humanities, 2008, 146–149.
Martinez, J. A., & Martinez, L. (2010). Some insights on conceptualizing and measuring service quality.
Journal of Retailing and Consumer Services, 17(1), 29–42.
McCrae, R. R., Costa, Jr, P. T., & Martin, T. A. (2005). The NEO–PI–3: A more readable revised NEO
personality inventory. Journal of Personality Assessment, 84(3), 261–270.
McCrae, R. R., & John, P. O. (1992). An introduction to the Five-Factor Model and its applications.
Journal of Personality, 60(2), 175–215. https://doi.org/10.1111/j.1467-6494.1992.tb00970.x
McTear, M. F. (2016). The Rise of the Conversational Interface: A New Kid on the Block? In International
Workshop on Future and Emerging Trends in Language Technology (pp. 38–49). Springer.
Mori, M. (1970). The uncanny valley. Energy, 7(4), 33–35.
Norman, W. T. (1963). Toward an adequate taxonomy of personality attributes: Replicated factor
structure in peer nomination personality ratings. The Journal of Abnormal and Social Psychology,
66(6), 574–583. https://doi.org/10.1037/h0040291
Nowak, K. L., & Rauh, C. (2005). The Influence of the Avatar on Online Perceptions of
Anthropomorphism, Androgyny, Credibility, Homophily, and Attraction. Journal of Computer-
Mediated Communication, 11(1), 153–178. https://doi.org/10.1111/j.1083-6101.2006.tb00308.x
O’Broin, U. (2017). Personality Brings Life to Chatbot User Experience. Retrieved January 10, 2018, from
https://chatbotsmagazine.com/avoiding-a-clash-of-personalities-chatbot-design-is-no-different-
3f0bcd30defd
Oliver, R. L. (1980). A cognitive model of the antecedents and consequences of satisfaction decisions.
Journal of Marketing Research, 17(4), 460–469.
Parasuraman, A., Berry, L. L., & Zeithaml, V. A. (1991). Refinement and reassessment of the SERVQUAL

31
 scale. Journal of Retailing, 67(4), 420.
Parasuraman, A., Zeithaml, V. A., & Berry, L. L. (1985). A Conceptual Model of Service Quality and Its
Implications for Future Research. Journal of Marketing, 49(4), 41–50.
https://doi.org/10.2307/1251430
Parasuraman, A., Zeithaml, V. A., & Berry, L. L. (1988). SERVQUAL: A multiple-item scale for measuring
consumer perceptions of service quality. Journal of Retailing.
Petter, S., DeLone, W., & McLean, E. (2008). Measuring information systems success: Models,
dimensions, measures, and interrelationships. European Journal of Information Systems, 17(3),
236–263. https://doi.org/10.1057/ejis.2008.15
Polzehl, T. (2014). Personality in speech: assessment and automatic classification. Springer.
Reddy, T. (2017). The code of ethics for AI and chatbots that every brand should follow. Retrieved
January 15, 2018, from https://www.ibm.com/blogs/watson/2017/10/the-code-of-ethics-for-ai-
and-chatbots-that-every-brand-should-follow/
Rust, R., & Oliver, R. (1994). Service Quality: New Directions in Theory and Practice. Thousand Oaks,
California: Sage Publications. https://doi.org/10.4135/9781452229102
Rusting, C. L., & Larsen, R. J. (1997). Extraversion, neuroticism, and susceptibility to positive and negative
affect: A test of two theoretical models. Personality and Individual Differences, 22(5), 607–612.
https://doi.org/10.1016/S0191-8869(96)00246-2
Seeger, A.-M., & Heinzl, A. (2018). Human Versus Machine: Contingency Factors of Anthropomorphism
as a Trust-Inducing Design Strategy for Conversational Agents. In Information Systems and
Neuroscience (pp. 129–139). Springer.
Shaked, N. A. (2017). Avatars and virtual agents–relationship interfaces for the elderly. Healthcare
Technology Letters, 4(3), 83–87.
Shawar, B. A., & Atwell, E. (2007). Chatbots: are they really useful? In LDV Forum (Vol. 22, pp. 29–49).
Shinde, S. (2016). 4 reasons why chatbots need a personality, not a vending machine brain.
Srivastava, S., & John, P. O. (1999). Handbook of Personality: Theory and Research. In A. L. Pervin & P. O.
John (Eds.), Handbook of Personality: Theory and Research (2nd ed., pp. 102–138). Elsevier.
Retrieved from http://books.google.co.uk/books?id=olgW-du4RBcC
Stewart, I., De, D., & Cole, A. (2015). Technology and People: The great job-creating machine. Deloitte,
London: UK.
Tam, K. P., Lee, S. L., & Chao, M. M. (2013). Saving Mr. Nature: Anthropomorphism enhances
connectedness to and protectiveness toward nature. Journal of Experimental Social Psychology,
49(3), 514–521. https://doi.org/10.1016/j.jesp.2013.02.001
Thompson, E. R. (2008). Development and validation of an international English big-five mini-markers.
Personality and Individual Differences, 45(6), 542–548.
Thoms, J. (2017). A guide to developing bot personalities. Retrieved January 10, 2018, from
https://www.xandra.com/blog/a-guide-to-bot-personalities
Thurstone, L. L. (1934). The vectors of mind. Psychological Review, 41(1), 1–32.
https://doi.org/10.1037/h0075959
Toegel, G., & Barsoux, J.-L. (2012). How to become a better leader. MIT Sloan Management Review,
53(3), 51–60.
Tupes, E. ., & Christal, R. . (1961). Recurrent Personality Factors Based on Trait Ratings. Lackland Air
Force.
Turing, A. M. (1950). Computing machinery and intelligence. Mind, 59(236), 433–460.
von Malitz, B. (2016). Why chatbots won’t replace humans. Retrieved January 15, 2018, from
https://memeburn.com/2016/04/chatbots-wont-replace-humans/
Waytz, A., Heafner, J., & Epley, N. (2014). The mind in the machine: Anthropomorphism increases trust
in an autonomous vehicle. Journal of Experimental Social Psychology, 52, 113–117.

32
Waytz, A., Morewedge, C. K., Epley, N., Monteleone, G., Gao, J.-H., & Cacioppo, J. T. (2010). Making
sense by making sentient: Effectance motivation increases anthropomorphism. Journal of
Personality and Social Psychology, 99(3), 410–435. https://doi.org/10.1037/a0020240
Weizenbaum, J. (1966). ELIZA---a computer program for the study of natural language communication
between man and machine. Communications of the ACM, 9(1), 36–45.
https://doi.org/10.1145/365153.365168
Wu, H.-C., & Ko, Y. J. (2013). Assessment of service quality in the hotel industry. Journal of Quality
Assurance in Hospitality & Tourism, 14(3), 218–244.
Yi, Y. (1993). The determinants of consumer satisfaction: the moderating role of ambiguity. ACR North
American Advances, 20, 502–506.
Zhao, L., Lu, Y., Zhang, L., & Chau, P. Y. K. (2012). Assessing the effects of service quality and justice on
customer satisfaction and the continuance intention of mobile value-added services: An empirical
test of a multidimensional model. Decision Support Systems, 52(3), 645–656.
Zilnik, A. (2016). Designing a Chatbot’s Personality. Retrieved January 25, 2018, from
https://chatbotsmagazine.com/designing-a-chatbots-personality-52dcf1f4df7d

33
Appendix A
A hierarchical overview of the items and the different paths

34
Appendix B
Invitation letter send to businesses

Dear Sir or Madam,

By sending this mail, I would like to ask for your participation in my bachelor thesis. My
research will focus on the personality of chatbots and its influence on customer satisfaction. As
you might know, implementing chatbots within companies lead to an increase in the
effectiveness of your customer support services and a decrease costs associated with those
services.

Moreover, customer satisfaction may be influenced by personality traits of customer support

representatives. The right personality could result in an increase in customer satisfaction. With
this in mind, I would like to ask you to fill out the questionnaire accessible through the following
URL:

https://qtrial2017q4az1.az1.qualtrics.com/jfe/form/SV_bDfbebprXocCnqJ

This survey will ask for some general questions about your company and the customer support
services your company employs. The survey only takes 5 minutes to complete and its results will
be kept in confidentiality.

The survey is not only aimed businesses that are using chatbots as customer support
representatives, but is also at businesses that have human, or other kinds of, customer support
representatives.

I look forward to seeing your response. If you have any questions in the meantime, you can
contact me via h.dehaan@students.uu.nl.

Yours faithfully,

Hayco de Haan

Supervisor: Christof van Nimwegen (Faculty of Science, Information Sciences, University

Utrecht)

35