Trust and perceived media richness of computer mediated communication mediums : a multi-medium explanatory study of the consumer's perspective

Master Thesis: Business Administration

Specialization: Entrepreneurship & Innovation

Trust and Perceived Media Richness of

Computer Mediated Communication Mediums:

A Multi-Medium Explanatory Study of the

Consumer’s Perspective.

PREFACE

This document represents the conclusion of my master program Business Administration with the specialization in Entrepreneurship and Innovation at the University of Amsterdam. I was engaged in researching and writing this thesis from October 2016 till March 2017.

The project was undertaken in consultation with Independer N.V., where I was working as a project manager on innovations at the time. As a project manager on innovations I was particularly interested in the consumer’s trust in the new communicative digital technology discussed in this research paper.

I would like to thank my supervisor dhr. Prof. dr. Peter van Baalen for his patience, guidance, knowledge and support throughout the whole period. I also would like to thank Bob Guns and all the other colleagues from Independer for helping me and granting me access to all the resources I needed to complete my research. During my time as project manager innovations at Independer I have learned how to turn new emerging technologies into actual value for customers. It truly gave me invaluable experience that most certainly will prove its recurring value many times in my future career.

This document is written by Sebastiaan Lichter who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

ABSTRACT

Purpose – The purpose of this paper is to compare the relationship of media richness and consumers’ trusting intentions between two different computer mediated communication mediums, and to examine the effect of maximizing behavior on this relationship at the individual level.

Design – In this research paper two independent survey questionnaires are performed, where each survey subjects participants to a different medium. Participants viewed example videos prior to communicating their perceptions of media richness and trusting intentions, where each video displayed interactions with the CUI medium or website medium.

Findings – Perceived media richness is positively related to consumers’ trusting intentions for both examined mediums. However, maximizing behavior shows no significant effect on the relationship. Additionally, media richness and trusting intentions are lower for the CUI medium than for the website medium, which may be explained by higher levels of experience with the website medium. Implications – This paper provides strong evidence that e-commerce companies could take advantage by enhancing the media richness of their web-based computer mediated communication channels to increase consumer’s intentions to trust.

Limitations – As the examined communication technology was still in its infancy stage and not commonly used by consumers, measuring trusting behavior with a survey questionnaire was ambitious. While this required trade-offs, the surveys were designed to maintain high levels of validity by simulating real-life situations.

Originality – This paper extends the media richness and e-trust literature by providing initial placement of the new CUI medium in the richness continuum, and investigating whether consumers’ trusting intentions are impacted by media richness in HHE and HAE trust environments.

Keywords: Media richness, Trust, E-commerce, Conversational user interface, Website, Digital innovation, Communication channel, Decision-making, Chat bot.

TABLE OF CONTENTS PREFACE 2 ABSTRACT 3 TABLE OF CONTENTS 5 LIST OF TABLES AND FIGURES 7 1. INTRODUCTION 8 1.1 Problem statement 11 1.2 Research purpose 13 1.2.1 Research goal 13 1.2.2 Research question 13 2. LITERATURE REVIEW 14 2.1 Digital innovation in e-commerce 14 2.1.1 Intelligent Agents 16 2.1.2 Conversational User Interfaces 17 2.2 Media richness 19 2.3 Trust in e-commerce 22 2.4 Decision-making 25 2.5 Theoretical framework 26 2.5.1 Conceptual framework 26 2.5.2 Research hypotheses 30 3. METHODOLOGY 33 3.1 Research philosophy 33 3.2 Research approach 34 3.3 Research strategy 36 3.3.1 Participants 37 3.3.2 Time horizon 38 3.4 Survey instrument 38 3.4.1 Pilot test 39 3.4.2 Scale development 40 3.4.3 Survey procedure 42 3.4.4 Method of analysis 44 3.5 Measures 44 3.5.1 Independent variable 44 3.5.2 Dependent variable 45 3.5.3 Moderator variable 45

3.5.4 Control variables 46 4. DATA COLLECTION 47 4.1 Sample sizes 47 4.2 Data preparation 48 4.2.1 Item labeling and aggregation 49 4.3 Descriptive statistics 49 5. ANALYSIS AND RESULTS 53 5.1 Reliability testing 54 5.2 Construct validity testing 57 5.2.1 Convergent validity 57 5.2.1.1 Factor analysis 57 5.2.1.2 Average Variance Extracted 65 5.2.2 Discriminant validity 65 5.3 Research hypotheses testing 67 5.3.1 Main hypotheses 68 5.3.1.1 Regression analysis one 69 5.3.1.2 Regression analysis two 71 5.3.2 Between subjects hypotheses 72 5.4 Post-hoc exploratory analysis 74 5.5 Results conceptual model 81 6. DISCUSSION 82 6.1 Summary of findings 83 6.2 Theoretical implications 84 6.3 Practical implications 88 6.4 Limitations and future research 89 7. CONCLUSION 92 REFERENCES 94 APPENDICES 105 A. Survey 1 105 B. Survey 2 110 C. Invitation e-mail for participation 115 D. MS plot survey 1 117 E. MS plot survey 2 118

LIST OF TABLES AND FIGURES Tables Table 1. Demographics survey 1 50 Table 2. Demographics survey 2 51 Table 3. Demographics of target population 53 Table 4. Reliability tests of survey 1 56 Table 5. Reliability tests of survey 2 56 Table 6. Confirmatory Factor Analysis 62 Table 7. Exploratory Factor Analysis 64 Table 8. Discriminant validity survey 1 66 Table 9. Discriminant validity survey 2 67 Table 10. Results hierarchical multiple linear regression one 70 Table 11. Results hierarchical multiple linear regression two 72 Table 12. Hypotheses testing 74 Table 13. Generations descriptive of survey 1 75 Table 14. Generations descriptive of survey 2 75 Figures Figure 1. The major application areas of artificial intelligence 17 Figure 2. Communication Media and Information Richness 21 Figure 3. Subclasses of trust 23

Figure 4. Web Trust Model 24

Figure 5. Conceptual Framework 30 Figure 6. Results of the research model 82

1. INTRODUCTION

The rising inclination of companies towards increasing customer experience and reducing operational costs has garnered the growth of the global chat bot market. According to Accenture’s Tech vision report 2017, Artificial Intelligence (AI) is about to become the digital spokesperson for companies. Moving beyond a back-end tool for the enterprise, AI is taking on more sophisticated roles across user interfaces (UI). Currently it’s being used to add frictionless intelligence to make conversational interfaces both simple and smart. This Conversational User Interface (CUI) enables companies to have more personal, yet automated interactions with their customers. According to MindBowser, the e-commerce industry will benefit the most from the CUI technology and is predicted to be most valuable in assisting with customer service, and sales processes. The CUI is an innovative method that enables customers to have more natural and intuitive interactions with computer software that acts on behalf of a company (Huang et al., 2010). This computer software that is commonly synonymized as a chat bot, takes the form of a personal digital assistant that assists customers with doing purchases, answers questions, and gives advice (Accenture, 2017). Gartner predicts that these digital assistants will facilitate 40% of all mobile interactions by 2020, and according to TechEmergence, are expected to be the top consumer application of AI over the next five years. Recent advances in the field of AI, and in particular in the subdomains, Natural Language Processing (NLP) and Machine Learning (ML) have empowered the CUI technology to be of clear commercial value to companies. It now offers access to computer software that is able to understand customers through natural language and can be communicated with in an equally natural way, as when individuals would have

conversations with other humans through conventional messaging apps

(Martinez, 2010).

The topic of such computer-mediated communication (CMC) technologies has piqued interest among researchers due to the fact that it has led to vast changes in the dynamics and forms of communication (Nardi et al, 2000). Similar to other CMC technologies, CUI is facing trusting barriers while in its infancy stage. The degree of trust in this stage reveals to strongly depend on the initial consumers’ assessment of the new communicative technology (McKnight et al., 2002). Previous work by Burgoon et al (2002), Rockmann and Northcraft (2008), and Cho et al (2009), accordingly shows that the perceived capability of new communication media to send rich information is a strong predictor of trusting behavior. According to the media richness theory (Daft et al, 1984), CMC and other forms of communication all have a certain richness that they can provide. Consequently, there have been multiple studies to assess the media richness of different CMC mediums such as Instant Messaging (IM) (Anandarajan et al, 2010), e-mail (Carlson & Zmud, 1999), and websites (Jacob et al, 2010). However, due to its newness, there is no prior research that focuses on the media richness of the CUI medium. In the recent past consumers have solely based they’re trust on the experiences and interactions they had with the company’s website or its human employees. However, a major shift toward the embedment of CUI’s in e-commerce will force companies to assess this new medium’s richness and understand it’s trusting dynamics to determine useful use-cases for implementation. According to a study by McKnight et al (2002), trust in the context of e-commerce plays a central role in helping consumers overcome perceptions of risk and insecurity they feel towards the web-based

vendor. Moreover, technical innovations such as CUI’s powered by AI do not automatically guarantee business or economic success (Teece, 2010), it can only be realized if consumers actually feel comfortable transacting over the new medium with un-familiar vendors (Gefen and Straub, 2004). As companies begin to embed CUI’s into their website as primary point of interaction, guiding users through their services (Bickmore & Cassell, 2001; Heylen et al., 2004), providing information on products, and offering services like insurances (e.g. Lemonade), having insight in consumers’ trusting behavior regarding this new medium becomes an important issue. Timely initiatives on emerging CMC technologies such as this research may help to explain how the perceptions of a medium’s richness might affect the intention to trust the medium.

Therefore, the objective of this research is to reveal mechanisms associated with the perceived media richness (PMR) of two different CMC mediums and its relation to trusting intentions (TI) of these mediums. In this research a CUI and website are the two mediums the model will be testing. This enables the observation of differences in perceived media richness and trusting intentions between the two closely related mediums and provides an answer to the research question and propositions of this research. The approach taken in this research diverges from prior research that examines trust in new digital innovations. Prior research typically examines trust of new digital innovations with a technological or design perspective (Xu et al., 2014; Grodzinsky et al., 2010; Kim & Moon, 1998) or focus more on abstract measurements of trust instead of relating to actual behavior (Luo et al., 2010; Corritore et al., 2005). Others investigate trust by assessing the overall acceptance of a new digital innovation by applying the UTAUT model (Luo et al., 2010; de Ruyter et al.,

2005), without providing more thorough explanations of the consumer’s trusting behavior in an e-commerce context. The perceived media richness of new digital communicative innovation may thus provide crucial insights in the individual behavior of consumers toward trusting the new medium.

Consequently, this research model contributes to the existing literature by extending the media richness theory framework with the integration of the CUI medium, which was absent prior to this research. Additionally, this research contributes to the e-trust literature by generalizing the e-commerce trust measurement framework adapted from McKnight et al (2002). Additionally, this research may pragmatically inform financial service providers and other e-commerce companies how individuals would interact and trust a CUI, and how it differs from the website medium. It could therefore help these companies assess the opportunities and difficulties when using a CUI for business purposes.

This paper is outlined as follows. First, the research problem under examination is addressed. Then descriptions and definitions of crucial concepts are outlined. The theoretical base including media richness, trusting intentions and decision-making behavior is described next. Following with the research methodology and data analysis. Finalizing with, a thorough discussion of the results, theoretical and practical implications, limitations and suggestions for future research, and the conclusion.

1.1. Problem statement

As new digital communicative technologies enter the market, potential benefits of using the new communication medium have to be clear for consumers. These communicative benefits express themselves in the information sending

capabilities of the medium as perceived by consumers (Daft & Lengel, 1984). Previous research has assessed the alleged media richness of different communication media channels and made comparisons between traditional and CMC mediums (Hill et al., 2015). However, few studies have compared different CMC mediums to one another. Others (e.g. Burgoon et al., 2002) revealed that media richness has a strong relationship with intentions to trust the medium. Since the amount of previous research within the field of conversational is still in its exploratory phase, there has been no such assessment of its media richness and its relation to trusting intentions. Despite the lack of research, e-commerce companies are recently replacing their current website medium with the new CUI medium as primary point of interaction. Therefore this research essentially provides these companies, and researchers involved in the field of CMC, with understandings and comparisons of the dynamics between media richness and consumers’ trusting intentions for the website- and the new CUI- medium. Previous work by Schwartz et al (2002), showed that certain individuals demand more information to base their decisions on. It thereby proposes that mediums with richer information sending capabilities better fulfill these individuals’ desire for more information. Additionally, the classification of individuals according to their decision-making behavior has revealed to significantly impact purchase decisions in commerce (Schwartz et al., 2002). Therefore, this research further investigates its effect on e-commerce for two different communication mediums.

1.2. Research purpose

The overall purpose of this quantitative research is to examine the relationship of perceived media richness and trusting intentions for two CMC mediums, and investigate the effect of maxiziming behavior on this relationship. This research endeavors to, validate the previously mentioned relationships proposed from the literature, and generalize the findings of previous research by studying the consumers’ perspective in an e-commerce context.

1.2.1. Research goal

In this research the dynamics between the perceived media richness, trusting intentions and the governing factor of the maximizer scale are examined. The aim of this study is to first, validate the relationship between media richness and trusting intentions for both mediums. Second, examine the effect of the maximizer scale on this relationship. Third, observe differences in trusting intentions between the two mediums and reveal how this is related to the measured lower or higher perceived media richness of the mediums respectively. And finally, extend the media richness literature with an initial assessment of the CUI medium’s richness. 1.2.2 Research question The research goals result into the following research question: “Does a higher perceived media richness lead to a higher intention to trust the CMC medium and how does maximizing behavior influence this relationship?”

As companies engage in the era of AI and dive into the upcoming trend of using the CUI for business, questions arise on how to utilize this new digital technology and how consumers would behave. Therefore, companies need to understand if consumers trust this new technology and if it’s prepared for implementation in the core services that companies have to offer. It is of tremendous importance for companies to analyze if and how the CUI medium can be leveraged to deliver value to the company and its customers. However, there is little value if consumers don’t trust a new communication technology. This research contributes to this understanding by examining the differences in trusting intentions between an e-commerce website and the upcoming CUI medium.

2. LITERATURE REVIEW

In section 2.1, relevant and necessary literature is discussed to gain background knowledge that is essential for this research. In section 2.2 and 2.3, the underlying theories and models for research into media richness and trusting intentions are presented. Next, in section 2.4, theories underlying the moderating factor examined in this research are discussed. Finally, in section 2.5 the conceptual framework including the proposed hypotheses is presented. 2.1 Digital innovation in e-commerce Digital innovation has been a widely studied topic in the recent years (Austin et al., 2012; Gregory et al., 2014; Tumbas et al., 2015) and it naturally follows that there have emerged multiple different definitions of (digital) innovation since it comes in many forms. Freeman and Soete (2012) argue that innovation is a two-sided phenomenon where both recognition of a potential market for a new

product or process and technical knowledge are involved. Zmud (1982) agrees by distinguishing between product innovations “the introduction of new products or services that shift or expand an organization’s domain” and process innovations “the introduction of new methods, procedures or responsibilities within existing domains”. “New” seems to be the recurring factor for these definitions, which is also supported by Rogers (1995) who describes innovation as an idea, practice or object that is perceived as new by an individual or other unit of adoption. Also Becker and Whisler (1967) define innovation as “the first or early use of an idea by one of a set of organizations with similar goals”. Innovations can exist in a “physical” form such as a car, airplane, personal computer and other material products or exist in a “digital” form. The latter is becoming increasingly important in the last few decades. Digital innovations are evolving with a rapid pace due to the adaptability and flexibility of digital technologies (Tiwana, et al., 2010; Yoo et al., 2012). These digital technologies are combinations of information, computing, communication and connectivity technologies (Bharadwaj et al., 2013). More recently, digital technology expanded beyond internal dimensions, penetrating firms’ product and service offerings (Yoo et al., 2012). This was also the case a decade ago when a physical store’s products and service offerings was being digitized and moved to online shopping websites. E-commerce defined as “commercial activities conducted on the internet” (Hake, 1999) has since evolved, expanding the frontiers of digital innovation (Nylén & Holmström, 2014). Henfridsson et al., (2014) and Yoo et al., (2012) give an explanation for this expansion of digital innovation by highlighting that the unique properties of digital technology enable new types of innovation processes. Implying that new digital technology such as e-commerce

opens up new possibilities for innovations and spurs its development. As a result, e-commerce companies are now offering more than convenience and cheaper products by investing in digital innovations such as AI powered agents to expand their business into other domains (Nylén & Holmström, 2014). IDC (http://www.idc.com) assumes that the dramatic growth in e-commerce will accelerate the demand for these agents.

2.1.1 Intelligent Agents

These recent investments in AI technology are driving new digital innovations in e-commerce such as recommendation agents (Xiao & Benbasat, 2007), negotiation agents (Huang et al., 2010), and service agents (Krallmann et al., 2003), which have become an increasingly vital element in the competitive landscape of the financial service segment (Suoranta et al., 2005). AI technology’s capacity to innovate in business and society is key but not sufficient by itself to make the leap from enthusiastic early adopters to the mainstream market (Benjamins, 2006). In general, technology users are expecting increasingly more from the intelligent agents with whom they interact. Such expectations include intelligence, cognition, and interaction through the use of natural language (Ford et al., 2001). Primarily, intelligent agents are running in the background without having to interact with customers directly, providing a company with smart algorithms to smoothen out their business processes (He et al., 2003). Lately, service agents are being utilized at the front desk of e-commerce companies as the so-called chat bots, directly interacting with their customers. Due to the advances in NLP and ML, chat bots are now mature enough to serve customers by communicating in natural language (Shawar &

Atwell, 2005). NLP is considered a subdomain of AI and has over the years become an interesting area of study in computational statistics and text data mining (Martinez, 2010). NLP includes approaches that use computers to analyze, determine semantics, and translate between natural languages (Martinez, 2010). Another subdomain of AI is ML, which is the major success factor in the ongoing digital transformations across industries (Alpaydin, 2014). The goal of ML is to learn a computer to perform a task by studying example data or past experience to solve a problem (Alpaydin, 2014). The combination of NLP and ML that is highlighted in figure 1 result into a powerful intelligent software agent that is able to naturally interact with humans and learn from its past conversations to increase its performance.

Figure 1. The major application areas of artificial intelligence

2.1.2 Conversational User Interfaces

Sycara et al (2003) precisely defined intelligent software agents as programs acting on behalf of their human users to perform complex information-gathering

tasks. Others consider these agents as combining aspects of perception, interpretation of natural language, learning, and decision-making (Schleiffer, 2005). There is no generally accepted definition of the concept of an agent, due to the fact that each definition grew directly out of its respective application area (Schleiffer, 2005) (figure 1). This also holds for the definition of a chat bot, which also grew directly out of the natural language application area. Shawar and Atwell (2005) defined chat bots as “machine conversation systems that interact with human users via natural conversational language”. Users primarily interact with chat bots to engage in small talk. However, a variety of new chat bot architectures and technologies have arisen recently, each simulating natural language more accurately and at the same time performing complex information gathering tasks (Shawar & Atwell, 2005). These developments have upgraded the chat bot from solely engaging in small talk to an intelligent software agent that acts as a natural interface application (e.g. Amazon’s Alexa, Google’s Assistant, Microsoft’s Cortana, Apple’s Siri) that humans can access to gather complex information. Conversational user interfaces are the actual front-end medium or portal for human users that consists of a synchronous one-to-one online interface comparable to an instant messenger (Hill et al., 2015) for users to directly have a dialogue with an intelligent agent. Users type or say a statement or question and then the intelligent agent on the back-end will send its quick response, mostly serving as a customer service agent in e-commerce (Krallmann et al., 2003). Additionally, CUIs enable interaction with users in meaningful ways by offering suggestions or completing transactions (Huang et al., 2010). Concluding that conversational user interfaces are a form of CMC that allow human users to directly have dialogues in natural language with intelligent

software (McTear, 2002). In this research the focus is on conversational user interfaces that only support typed text for a chat bot to respond to. This is due the fact that speech will draw too much attention on itself as a technology rather than having a focus on the inherent characteristics of the CUI as a media channel. Moreover, it suggests other complications (McTear, 2002) that are out of the scope of this research.

2.2 Media richness

The media richness theory (MRT) established by Daft & Lengel, (1984; 1986) and Daft et al., (1987) describes organizational communication channels as possessing a set of objective characteristics that determine the capacity of each channel to carry rich information, with rich information being more capable than lean information of reducing ambiguity in messages. Similarly, the social presence theory suggests that media channels that are capable of carrying richer information allow the sender to be more “present” with the receiver, reinforcing effective communication (Lombard & Ditton, 1997). All communication channels (e.g. letters, posters, e-mail, telephone, and face-to-face) possess implicit characteristics that relate to objective richness capacities (Daft & Lengel, 1984). Lengel (1983) defined richness as the potential information carrying capacity of data. Therefore, media richness refers to a communication channel’s ability to transmit messages that communicate rich information (Daft & Lengel, 1984). According to Lengel (1983), communication media vary in the richness of information processed. Moreover, on the base of these differences, channels can be organized along a continuum (e.g. Lengel, 1983, Daft & Lengel 1984, Rice, 1992) (figure 2). Each medium on the media richness continuum has a different

place in the hierarchy and scholars have been adding media to it since Lengel (1983) first proposed the continuum. The explanation for the hierarchical nature is that each medium differs in its (1) feedback capability, (2) utilization of the channel, (3) source and (4) language (Daft & Lengel, 1984). In the CMC literature, MRT assesses the capacity of communication media to facilitate shared understanding (Robert & Dennis, 2005). Compared with face-to-face communication that scores the highest in richness, CMC media are viewed as poor for transmitting messages (Hill et al., 2015). Others have found CMC to be able to communicate emotion as well or better than face-to-face communication (Derks et al., 2008). Although CMC has been compared to other traditional communication media, few studies have compared different forms of CMC to one another (Hill et al., 2015). MRT proposes that shared understanding can be improved when correct communication media are chosen that best fit the information processing needs of the task at hand (Daft & Lengel, 1986). Matching the correct media with the information needs has encouraged many scholars to study media selection in different contexts (e.g. Trevino et al., 2000; Carlson & Davis, 1998; Simon & Peppas, 2004). Although scholars mainly focused on media selection based on MRT, they also noted variances among users of the same medium (e.g. Huang & Yen, 2003; Markus, 1994; Liu et al., 2009). Carlson and Zmud, (1999) addressed this variance by proposing the Channel Expansion Theory (CET) and showed that user’s perception of the richness of a communication medium varied. According to the MRT a CUI proposes to be observed as richer and more synchronous than precious forms of CMC such as email (Daft & Lengel, 1984). However, individual users can have different perceptions about the capability of mediums such as CUIs in transmitting shared

understandings. This is the same phenomenon as Carlson and Zmud (1999) found about email and is referred to as perceived media richness. Based on the work of Carlson and Davis (1998), perceived media richness is defined as the degree to which a person believes a certain communication medium has the capability to transmit information based on the characteristics it possesses. Assessing the perceived media richness of CMC mediums is key to shed light on the consumer’s perspective on media richness. Therefore, perceived media richness is used in the conceptual framework of this research. Carlson and Zmud (1999) argued that there is no reason to expect that CET is only applicable to e-mail. Moreover, the perceived media richness scale is sufficiently generic that it could be easily adapted for any other communication channel. Therefore, examination of the scale across a variety of channels such as CUIs and communication contexts as e-commerce could provide additional validation for both (Carlson & Zmud, 1999). Additionally, Carlson and Zmud (1999) suggested that perceived media richness could be applied to other dependent variables such as trust.

2.3 Trust in e-commerce

In the past decades, trust has been widely studied by scholars that defined trust according to their specific disciplinary view (Taddeo, 2009). Psychologists define trust as a tendency to trust others (Rotter, 1971), and social psychologists define it as certain cognition about the trustee (Rempel et al., 1985). Some scholars have viewed trust as a unitary concept (Rotter, 1971), but most agree that trust is multidimensional (Grodzinsky et al, 2010; McKnight et al., 1998; 2002). Taddeo (2009) has analyzed several definitions of trust that have been developed over the last few decades. Following his analysis, trust is defined as the following set of statements: (1) trust is a relation between the trustor and the trustee, (2) trust is a decision by the trustor to delegate to the trustee some aspect of importance to the trustee in achieving a goal, (3) the less information the trustor has about the trustee, the higher the risk and more trust is required. Among technologists, some scholars have argued whether trust can be similarly applied to relationships with online entities. Friedman (2000) argued that, “people trust people, not technology”. Others contradict and suggest that people tend to anthropomorphize their interactions with information systems (Nass et al., 1994), and that no distinction is necessary between human-human trust and human-system trust (Marsh & Dibben, 2005). More recently, Grodzinsky et al., (2010) made a distinction between trust in physical- and electronic environments and organized trust according to six principles of trust and three features of e-trust interactions. In this framework four different kinds of trust relations have been recognized: (1) human-to-human; (2) human-to-artificial agent (AA); (3) AA-to-human and (4) AA-to-AA. This results into eight subclasses

of trust where the participants and interaction environment determine the subclass (figure 3). Many variants of the above trust types exist, including a significant number of HHE trust in e-commerce research (McKnight et al., 2002; Chen & Dhillon, 2003; Kim et al., 2008). In particular, McKnight et al (2002) contributed to the e-commerce trust literature by developing a widely applicable trust measurement model for e-commerce. They suggested that the adoption of e-commerce by consumers is not only affected by the perceptions of the technology (e.g. Technology Acceptance Model by Davis et al., 1989), but also by the trust in the web-vendor. McKnight et al (2002) also emphasized that trust is especially critical in the e-commerce environment. Bhattacherjee (2002) agrees by stating that a lack of trust in web-vendors can discourage adoption of e-commerce.

Figure 3. Subclasses of trust

The trust measurement model for e-commerce (Mcknight et al., 2002) is composed of four trust constructs: disposition to trust, institution-based trust,

trusting beliefs, and trusting intentions (figure 4). McKnight et al (1998; 2002) defined trusting intentions as the “intention to engage in trust related behaviors with a specific web-vendor”. This research builds on the integrative trust measurement model for e-commerce proposed by McKnight et al (2002) and therefore includes the trusting intentions construct in the theoretical framework. The trusting intentions construct consists of four measurable sub-constructs: willingness to depend, follow advice, give information, and make purchases. Trusting intentions was found to be a direct predictor of actual trusting behavior (McKnight et al., 2002), which is one of the relevant reasons why it was included in this research. Another reason to use the model of McKnight et al (2002) is that it has proven to be extremely relevant and capable of measuring trust in different e-commerce environments. Moreover, in theory its measurement-level subconstructs facilitate measuring trusting intentions for HHE and HAE -trusting types, which correspond with the website- and CUI- mediums respectively. A more thorough explanation about the measurement tools is provided in section 2.5.

2.4 Decision-making

The last half-century of behavioral decision research (e.g. Edwards, 1954; Bernoulli, 1954; von Neumann & Morgenstern, 1944) has described behavior in terms of utility maximization. There has been both consensus and dissensus of the norm among scholars in this research domain (e.g. Baron, 2000; Plous, 1993). The movement that disagrees with the norm includes “satisficing” as the opposite behavioral decision-making norm, choosing an alternative that is good enough, rather than “maximizing”, selecting the option with the highest utility (Simon, 1978). Simon’s (1978) work on both norms formed the basis for Schwartz et al (2002) to develop a scale that measures the degree to which individuals report trying to “maximize”, rather than “satisfice” and was validated across a number of survey and experimental studies. The scale includes items that measure alternative searching (Nenkov et al., 2008) such as “No matter how satisfied I am with my job, it’s only right for me to be on the lookout for better opportunities”. Other items capture ways in which one might set high standards for themselves or find it difficult to make a decision (Nenkov et al., 2008). In modern life, people are overflowed with different options (e.g. music, jobs, food, websites, apps and products) to choose from what makes maximizing not an effortless task (Schwartz, 2004). Schwartz et al (2002) discovered that people who tend to implement a maximizing strategy, experience less happiness, optimism, self-esteem and satisfaction. The explanation of this phenomenon by Schwartz et al (2002) states that individuals who are self reported “maximizers” may have less constructive decision making styles that causes regret and counterfactual thinking about what might have been a better choice to make. De

Bruin et al (2007) agree and found that maximizers appear to be poorer decision makers. They rely more on external information sources (Iyengar et al., 2006), which lead them to further questioning their choices. It follows that this maximizing behavior actually undermines the very satisfaction that maximizers are trying to achieve (Schwartz et al., 2002; Schwartz, 2004). To increase satisfaction and happiness regarding purchases, people might in general be better off with constrained and limited choice than unconstrained choice (Schwartz, 2004). Following the literature, maximizing would predict reports of more product comparison, information gathering, social comparison, and counterfactual thinking in an e-commerce context. In this research the maximizer scale of Schwartz et al (2002) is used as a moderator on the relation ship between media richness and trusting intentions. In section 2.5, a more thorough explanation is presented.

2.5 Theoretical framework

In the previous sections, relevant literature including its models that form the theoretical basis for this research has been discussed. Therefore, theories and models of media richness, trust and decision-making contribute to the theoretical framework in this research. In section 2.5.1, the conceptual model that gives a visualized representation of the performed research is discussed and in sector 2.5.2, the corresponding hypotheses are formulated.

2.5.1 Conceptual framework

As discussed in section 2.3, a variety of scholars have defined trust according to their view of the world. Most agree that trust is a multidimensional concept

(McKnight et al., 2002) and that multiple trust types exist (Grodzinsky et al., 2010). McKnight et al (2002) developed a multidimensional trust framework that consists of measurable sub-constructs. The trust framework has been applied to a wide range of different studies in the e-commerce domain, including the adoption of online banking (e.g. Dimitriadis & Kyrezis, 2010; Naimat, 2013), and website credibility (e.g. Corritore et al., 2005). The model is originally used to explain the multidimensionality of trust in different e-commerce settings, but the sub-construct “Trusting Intentions” can also be employed to predict actual adoption (McKnight et al., 2002). Actual adoption is impractical to include in this research due to the fact that the chat bot technology and CUI medium are still in its infancy, and thus use cases are scarce. However, “trusting intentions” has proved to be a reliable and measurable predictor of adoption (McKnight et al., 2002). Moreover, multiple efforts (e.g. Dimitradis & Kyrezis, 2010) have also successfully combined the behavioral component of trust: “Trusting Intentions” with the TAM model of Davis et al (1989) and have shown the predictive power of TI on adoption. “Trusting Intentions” is composed of four measurable sub-constructs: Willingness to Depend (WD), Follow Advice (FA), Give Information (GI), and Make Purchases (MP). As previously discussed, multiple scholars have used the model on the HHE trust subclass although it’s applicable to other types of electronic mediated trust as well (McKnight et al., 2002), which makes it an exceptionally appropriate model for this research to build on. Therefore, this research extends the findings of previous research by using the models’ sub-construct “Trusting Intentions” for both the website and CUI medium in an e-commerce setting, which are associated with HHE and HAE subclasses of trust respectively. In response to their findings, McKnight et al (2002) suggested

multiple future research directions that are relevant for this research. First, they suggested to “test their measures in the context of other web-based tasks where sharing information and making purchases are more directly supported”. Second, “the items should be changed to measure or manipulate structural perceptions of advice-giving sites on the Web”. Third, “move beyond the domain of initial trust to trust in on-going relationships”. Following the future research suggestions of McKnight et al (2002), this research endeavors to fill in most of these gaps by using the model for two different media channels, which in turn are used to (in)-directly communicate with a financial advice-giving web-vendor. These two different media channels in this research are assessed by its “Perceived Media Richness” (Carlson & Zmud, 1999). The underlying MRT developed by Daft & Lengel (1984) has been widely applied to multiple traditional communication mediums (e.g. face-to-face, telephone, and letters), but less to CMC mediums (e.g. websites, instant messaging) (Carlson & Zmud, 1999). Some studies in the CMC domain have examined the effect of media richness on affective- and cognitive-based trust (e.g. Rockmann & Northcraft, 2008) and others have found a strong relationship between media richness and trusting intentions in a web-presentation setting (Cho et al., 2009). Remarkably, there have not been many scholars who compared the PMR and TI relationship between different CMC mediums, let alone in an e-commerce setting. This research addresses these discussed shortcomings by comparing the PMR of two different CMC mediums and to find its effect on the intention to trust the CMC medium. The PMR variable consists of six measurable items that are applicable to communication mediums in general including CMC mediums (Carlson & Zmud, 1999). Moreover, PMR reflects the consumer’s interpretation of the

richness of a medium. Following the literature (e.g. Daft & Lengel, 1984; Carlson & Zmud, 1999), the CUI medium will presumably be positioned higher in the media richness hierarchy than the website medium due to its proposed higher synchronicity/interactivity and richer use of language. However, PMR interestingly allows the actual measurement of media richness as it is viewed by consumers and enables this research to validate the proposed theory. Carlson and Zmud (1999) have also suggested future research to examine the scale across other CMC mediums and to find causal relationships with other dependent variables such as trust. Moreover, if an individual perceives a medium to be capable to fulfill his/her information needs and therefore facilitates shared understanding (Robert & Dennis, 2005; Carlson & Zmud, 1999) it may be convincing that the individual will trust this medium (Rockmann & Northcraft, 2008). Therefore, this research expects to find the independent variable “Perceived Media Richness” to positively affect the dependent “Trusting

Intentions” variable.

This research includes the influence of the moderator “Maximizer Scale” on the relationship between “Perceived Media Richness” and “Trusting Intentions”. The maximizer scale (MS) of Schwartz et al (2002) suggests that individuals who tend to maximize rely on many information sources and try to maximize the utility. Individuals who tend to maximize thus have higher information needs than others to base their decisions on. As mentioned, PMR is the degree that individuals perceive the medium to be capable of fulfilling these exact information needs. This implies that individuals with a tendency to maximize would find PMR a more substantial factor to trust the medium and thus the relation between PMR and TI might reveal to be stronger to maximizers. Due to

the many options of products in e-commerce it is extremely relevant and remains a problematic issue that influences people’s behavior on a daily basis (Schwartz et al., 2002). Therefore, this research includes this behavior-influencing tendency to maximize in the conceptual framework as a moderator that positively affects the strength of the relationship between PMR and TI. For a visual representation of this section see the conceptual framework below (figure 5). The conceptual framework is also used for between-subject analysis, examining the effects of both mediums on the PMR and TI variables. A more thorough explanation is presented in section 3. Figure 5. Conceptual Framework 2.5.2 Research hypotheses In the previous sections, all relevant theoretical models have been described that form the basis for the conceptual model that is visualized in section 2.5.1. The

conceptual model of this research consists of the proposed positive relationship between Trusting Intentions (TI) and Perceived Media Richness (PMR) with the positively moderating effect of the Maximizer scale (MS). Both PMR and MS are considered as independent variables influencing the dependent TI variable. The hypotheses presented first are categorized per independent variable and the hypotheses regarding the comparison of variables’ value between subjects are presented next.

Perceived Media Richness (PMR) is defined as “the degree to which a person believes a certain communication medium has the capability to transmit information based on the characteristics it possesses” (Carlson & Zmud, 1999), resulting into perceiving media as rich or lean. At the heart of the MRT, trust is an essential but implicit construct. As mentioned before, the emphasis of MRT is on the match between ambiguity of a message and communication medium (Cho et al., 2009). Van Koert (2003) relates to this by stating, “the aspect of ambiguity of a message is, of course, relevant with respect to the issue of trust”. Furthermore, he states that communicating through richer media that supports responsive feedback and reduce ambiguity may lead to increased trust. In their experimental study, Burgoon et al (2002) have shown that media richness is positively associated with levels of trust. It is therefore expected to find similar results for the two different CMC mediums examined in this research, as previous scholars have found for other communication mediums (e.g. Burgoon et al., 2002). The two hypotheses below propose the relationship to be positive for each survey with a different medium as subject.

Hypothesis 1 (H1): A higher perceived media richness (PMR) of a CUI positively affects the intention to trust (TI) a CUI.

Hypothesis 2 (H2): A higher perceived media richness (PMR) of a website positively affects the intention to trust (TI) a website.

Maximizer scale (MS) is defined as “the degree to which a person has the tendency to optimize when making decisions” and it is characterized by increased information-seeking and social comparison (Schwartz et al., 2002). It follows that a richer medium, providing more and richer information will be of greater influence to the information-seeking individual for trust related intentions. It is therefore expected that PMR is of greater importance as an influencing factor on trust for individuals that have a tendency to maximize. Therefore, the two hypotheses below propose MS to positively moderate the relationship between PMR and TI for each survey with a different medium as subject.

Hypothesis 3 (H3): A higher maximizer scale (MS) will positively influence the strength of the relationship between the PMR and TI of a CUI.

Hypothesis 4 (H4): A higher maximizer scale (MS) will positively influence the strength of the relationship between the PMR and TI of a website.

According to the MRT, the characteristics (e.g. feedback, emotional tone, and use of language) of both communication mediums may result into the CUI being higher perceived in media richness than a website (Daft & Lengel, 1984; Carlson & Zmud, 1999). As Burgoon et al (2002) similarly found, in this research it is expected that if a CUI is perceived higher in media richness it will also have a

higher value of Trusting Intentions. Therefore, both subjects’ average PMR and TI are examined to be able to conclude if a CUI is a richer medium than a website and if it therefore has a greater degree of users’ trust. These findings will generate insight for e-commerce companies when embedding CUIs as communication medium on their websites. The hypotheses regarding the between-subject value of the PMR and TI variables are proposed below.

Hypothesis 5 (H5): The PMR of a CUI is higher than the PMR of a website.

Hypothesis 6 (H6): The intention to trust (TI) a CUI is higher than the intention to trust (TI) a website.

3. METHODOLOGY

The methodology section outlines the research philosophy, approach, strategy, and methods used to answer the research question and examine the hypotheses. Furthermore, this section outlines the sample and measurement instrument that were utilized in this research.

3.1 Research philosophy

According to Guba & Lincoln (1982), the philosophical paradigm within a research is of utmost importance as it is the “basic belief system that guides the investigation”. This research was based on a positivist research philosophy. A positivist philosophy assumes that there is one universal truth in existence, which could be discovered by a non-participant, observant researcher (Saunders & Lewis, 2012). This research is therefore based on the law of cause and effect. The principle of cause and effect suggests that proposed theories can be tested,

and as a result, a proposed theory may be confirmed or not (Saunders & Lewis, 2012). Therefore, this research assumed that all variables included in the conceptual framework were objective artifacts of truth that could be measured and tested (Saunders & Lewis, 2012).

3.2 Research approach

By virtue of the positivist research philosophy the purpose of this research was to test existing literature to find causal relations between variables. This corresponds with the purpose of explanatory research, where problems or phenomena are studied in order to investigate and establish causal relationships among variables, mostly concerning the “why” and “how” questions (Saunders & Lewis, 2012). Differently, descriptive research has a stronger focus on answering the “what” questions and exploratory research is about discovering general information about a topic (Saunders & Lewis, 2012). This research first conducted a literature review in order to specify the research question to be examined. Considering the research question, that concerns the nature of the causal relationship between PMR and TI with an additional focus on the influencing factor of MS on this relationship, it naturally follows that the purpose of this research is primarily explanatory, which is also known as causal research. Explanatory research can in its essence be quantitative or qualitative (Saunders & Lewis, 2012). Quantitative represents data collection techniques that generates or uses numerical data, while qualitative represents data collection techniques that generates or uses non-numerical data (Saunders & Lewis, 2012). Qualitative research is primarily employed for exploratory research where it is used to gain an understanding of underlying reasons by observing or

interviewing participants (Saunders & Lewis, 2012). Differently, quantitative research is used to quantify a problem into quantitative values that can be attached to variables, which can be subjected to statistical tests in order to get a clearer understanding of the relationship (Saunders & Lewis, 2012). Therefore, quantitative is a better fit for explanatory research as discussed above. In this research, data is collected from a sample of consumers to test hypotheses and generate an outcome that could be generalized to a specific larger population. Therefore, the variables under examination in this research are quantified values that have been tested using a quantitative research method to find causal

relationships.

Thence, the literature was used to identify theories and to develop a clear theoretical position prior to the collection of data. This is known as a deductive approach (Saunders & Lewis, 2012). Whereas, an inductive approach, is research that endeavors to develop theory after data collection and thus is the inverse of deduction. This research’s deductive approach consists of reviewing existing literature, and empirically testing the propositions derived from the literature. The thoroughly selected secondary data needed for reviewing was retrieved from the Digital University library, Google scholar, and different web sources. Aligned with a deductive approach, measurement items for the three main variables were adapted, and modified from the original validated surveys used in the reviewed literature (McKnight et al., 2002; Daft & Lengel, 1984; Schwartz et al., 2002). The measurement instrument is discussed more thorough in section 3.4.

3.3 Research strategy

According to Saunders and Lewis (2012) it is essential for every researcher to have a clear research strategy. In their work, they outlined different kinds of research strategies: the experiment, survey, case study, action study, grounded theory, ethnography, and archival research. A research strategy is guided by the research question, objectives, existing knowledge, resources, and philosophical view of the researcher (Saunders & Lewis, 2012). In this research a survey strategy was used, which is primarily associated with a deductive approach. Additionally, a survey strategy allows the collection of data about the same variables from a large number of people what makes it possible to generate findings that are representative of the larger population (Saunders & Lewis, 2012). Another advantage of this strategy is that the researcher is more in control of the process. However, a mentionable drawback is that the data collected is unlikely to be as detailed as those collected by using other research strategies (Saunders & Lewis, 2012). Although research strategies such as a survey come with limitations such as retrospective and uncontrollable biases, multiple studies (e.g. Carlson & Zmud 1999; Anandarajan et al, 2010) have shown that they provide valuable insights for understanding the concepts of media richness and trust. Due to the investigation of generalizable, causal relationships between quantified variables in this research and the existence of validated measurement items designed for these variables, the most appropriate strategy for data collection is the survey questionnaire.

3.3.1 Participants

This research concerns the concepts of media richness and trust for two different communication mediums in an e-commerce setting, where the particular web-vendor sells-, and gives advice about, financial products. Therefore, data is collected through two online questionnaires held under customers of the financial service provider and web-vendor Independer N.V. The questionnaires have been shared through an e-mail campaign, randomly selecting participants for each questionnaire that enables further generalization of the results for the larger population. The target population is defined as all customers of Independer N.V, who are adults with a Dutch nationality, capable and legitimate to make their own choices about purchasing financial products, thus the inclusion criteria consists of age being > 17 and nationality being Dutch. Therefore the representative population is a widely defined group of individuals between the age of 18 and 100 with a Dutch nationality. This choice was made due to the wide adoption of e-commerce by all ages and to be able to perform post-hoc clustering techniques on age. Thereby, gaining insight in outcome differences between age generations such as Millennials and Baby boomers. The respondents all differ in their degree of experience concerning the medium that was under assessment. This posed no issue since intentional trusting behavior rather than actual trusting behavior was under examination, which was due to the infancy of the CUI communication medium. The actual sample characteristics are discussed in section 3.6. Since the researcher was an employee, he had the privilege to have access to the customer base of Independer N.V, for spreading the online questionnaires.

3.3.2 Time horizon

According to Saunders and Lewis (2012), there are two main time dimensions for a research design: cross-sectional, and longitudinal. In this research, a cross-sectional research design was applied, where data from participants was collected at only one period in time. Contrarily, a longitudinal research design enables to study change and development over time. However, the latter was not achievable due to imposed time constraints. Furthermore, a cross-sectional research design would primarily employ the survey strategy and produce quantitative data to explain the relationship between certain variables (Saunders & Lewis, 2012), which confirm with the goals of this research.

3.4 Survey instrument

In order to compare two different communication mediums, two different surveys were held. Both surveys were composed of the exact same question items, but differ in the communication medium the participants were exposed to. In both surveys, the participants were exposed to a video that showed an example of the communication medium under examination. After exposure, participants were asked to base their answers solely on the content of the video they have watched. In survey 1 this video1 _{showed an example of the CUI}

medium and in survey 2 the video2 _{showed an example of the website medium.}

Both videos have shown an interaction with the respective medium from a consumer’s perspective in an e-commerce setting. To ensure that participants based their answers on the perceived characteristics of the channel (Daft &

1 _{CUI example video: https://www.youtube.com/watch?v=us_2tbntxZQ&t=2s} 2 _{Website example video: https://www.youtube.com/watch?v=OBbNlMYHiwo}

Lengel, 1984), the contents such as colors, advised product, and text were generally kept the same in both examples. In the following sections, the CUI medium survey and website medium survey are referred to as survey 1 and survey 2 respectively.

3.4.1 Pilot test

A pilot test has been conducted prior to the actual research to ensure that respondents will understand the meaning of the questions, follow the survey instructions, and finish the survey within the expected time frame. Furthermore, the survey has been pilot tested to find erroneous items and to measure the reliability of the item constructs with the Cronbach’s Alpha indicator (Cronbach & Meehl, 1955). To ensure the latter, a pilot test has been conducted for both surveys by randomly selecting a considerable amount of respondents from the same population as used in the actual research: survey 1 (N = 129; Response rate = 17,8%), and survey 2 (N = 105; Response rate = 14,3%). In this pilot test all respondents were asked to give feedback through e-mail. Feedback from both groups indicated that in certain web browsers the videos were not always shown properly in its size, and some question items adapted from Schwartz et al (2002) were too complicated for respondents to correctly interpret. The issue with the videos was resolved by hosting them at a different provider. The latter implication was also supported by a low Cronbach’s Alpha for the MS variable in both survey 1 and 2: (α = 0.579) and (α = 0.584) respectively. Considering that item deletion did not show an increase in the overall Cronbach’s Alpha, and that feedback indicated that the problem might originate from semantics of these specific items, the question items were revised under supervision of an

experienced linguist to ensure better clarity and content of these question items. However, the other construct items had good-high Cronbach’s Alphas for both surveys generally between (α = 0.739) and (α = 0.963), also these items were not subjected to negative feedback from respondents. Question items about “topic experience” were not adapted from measurement items of other studies. Therefore, this scale was measured by asking a question in four similar yet different ways. The results of the pilot test revealed high Cronbach’s Alphas for this construct in survey 1 and 2: (α = 0.922) and (α = 0.831) respectively. Also, the pilot testers experienced that the Likert-scale questions were appropriate for the majority of the items. Therefore, every item except the items relating to age, gender, educational level, and nationality were confirmed as appropriate Likert-scale questions. 3.4.2 Scale development

Since none of the items were deleted after the initial pilot, both final surveys consisted of 33 items. Since the population is Dutch, all adapted items were translated from English to Dutch by the help of an experience linguist. The total of items were then categorized into 8 blocks; Demographics, Maximizer Scale, Topic Experience, Perceived Media Richness, Willingness to Depend, Follow Advice, Give Information, and Make Purchases. This comprehensive list of question items can be explained by the fact that the survey attempts to; (1) examine the PMR for two communication mediums, (2) examine TI, which consist of multiple sub-constructs, (3) validate the relationship between PMR and TI by taking into account Topic Experience as control variable, and finally, (4) to examine the moderating effect of MS on this relationship. The demographics of

respondents were collected using single-item tests for age, gender, educational level, and nationality. The Maximizer Scale was measured using a 6-item scale adapted from Nenkov et al (2008) who analyzed the original 13-item scale of Schwartz et al (2002) and concluded that a 6-item MS scale performs best and should be used by future researchers. Topic Experience was not adapted from previous studies, and was measured using a 4-item scale, where each item asked the valid question in a slightly different way. Therefore, this scale satisfies with a minimum of three items (Saunders et al., 2009). Also the Cronbach’s Alpha’s in the pilot test have shown high reliability for this scale as mentioned in the previous section. Perceived Media Richness was measured using a 4-item scale adapted from Carlson and Zmud’s (1999) work, which based the scale on four criteria set forth by Daft and Lengel (1984). These four criteria are: (1) feedback capability, (2) utilization of the channel, (3) source and (4) language. The Trusting Intentions construct was measured by measuring the items in four sub-constructs adapted from McKnight et al (2002): Willingness to Depend, Follow Advice, Give Information, and Make Purchases. Willingness to Depend was measured using a 4-item scale, Follow Advice with a 6-item scale, Give Information with a 3-item scale, and Make Purchases with a 2-item scale. The latter scale originally consisted of three items, but one of the items wasn’t used due to its impracticality in the context of this research. True-score theory illustrates that more items lead to better construct representation (Marsh et al., 1998). If, however, off-design circumstances dictate that the scale has only two corresponding items, then the most appropriate reliability coefficient is the Spearman-Brown statistic (Eisinga et al., 2013). Therefore a split-half reliability test has shown a Spearman-Brown reliability coefficient of (rs = 0.868) and (rs =

0.702) for survey 1 and 2 respectively, which indicated a good-high reliability of the scale and thus the Make Purchase scale was included. The majority of items were adapted from previous studies that have successfully used a 7-point Likert-scale to test the question-items. For the self-developed Topic Experience scale, feedback has indicated that a 7-point Likert-scale was in this case also appropriate. Therefore, all items except age, gender, educational level, and nationality were measured with a 7-point Likert-scale (Likert, 1932) ranging from: (1 = totally disagree, 2 = disagree, 3 = partially disagree, 4 = neutral, 5= partially agree, 6 = agree and 7 = totally agree). The choice to include a neutral option was made to reduce bias by freeing respondents from having to select a category that does not truly represent their neutral opinion. Moreover, some scholars (e.g. Mattell & Jacoby, 1972) have shown that higher number of points on a Likert-scale may serve to decrease the usage of a neutral middle value. Thus, considering the large sample size and potentially a large number of respondents for both surveys, this choice would not have risked a too large portion of the collected data being inconclusive. Therefore, a decrease in bias was found more important than an increase in inconclusive data being collected. A full overview of items used in both surveys may be found in Appendix A and B.

3.4.3 Survey procedure

To investigate the intention to trust two different mediums, the surveys have been spread amongst potential users of these mediums, whom could be anyone potentially buying products online. Executive support to spread the surveys among the customers, of the web-vendor in question, was obtained. For both the pilot test as the actual research, the surveys have been randomly distributed to