Effective knowledge transfer through social network sites

(1)

University of Amsterdam

Faculty of Science

Thesis Master Information Science - BIS

Effective knowledge transfer through

social network sites

Supervisor Dick Heinhuis

Signature:

Second Examiner Tom van Engers

Signature:

Author:

Jan Joost Visser

10174273

(2)

Abstract

Finding new methods for effective knowledge transfer are useful as the maintenance of current systems can be quite costly for companies. This thesis poses the question ’to what extent are social network sites useable as an effective tool for knowledge transfer?’

The first step is to find out which types of social network sites there are and which are used the most in The Netherlands. The literature review finds the best way to predict technology usage and decides upon the use of the Technology Acceptance Model. By adding in factors that impact knowledge transfer the research model for the survey is finalized.

The survey is a self-administered split-user survey in which respondents have to choose between two social network sites with a knowledge transfer task displayed. After choosing which platform they prefer the respondents have to rate statements that are used to measure the concept in the extended technology acceptance model.

The results show that respondents present a very positive attitude to-wards using Twitter for knowledge transfer and, a positive attitude for LinkedIn and somewhat lower results for Facebook. The post-hoc analysis revealed that statistically significant differences exist between the platforms for perceived usefulness and behavioral intention to use, on which Twitter and LinkedIn scored quite high and Facebook quite low. The results war-rant further research into this area as users show a positive attitude towards acceptance of social network sites for knowledge transfer.

(3)

Acknowledgements

First I would like to thank my supervisor Dick Heinhuis for his time and patience during my research, guiding me in the right direction and giving useful tips in how to structure my thesis.

I would also like to thank Johanneke Lamberink for her help in finding the best way to set up and publicize my survey, including helping me to find methods to randomize the survey.

Lastly I would like to thank my friends and relatives for putting up with me while I was working on this thesis.

(4)

List of Figures

1 Summary of different enterprise knowledge sharing types . . . 4

2 Basic concepts for User Acceptance Models . . . 7

3 Visual representation of the Theory of Reasoned Action . . . 7

4 Visual representation of the Theory of Planned Behavior . . . 8

5 Visual representation of the Technology Acceptance Model . . . 8

6 The TAM and three popular extensions . . . 9

7 Theoretical framework . . . 11

8 TAM with significant correlation for Twitter . . . 19

9 TAM with significant correlation for LinkedIn . . . 20

10 TAM with significant correlation for Facebook . . . 21

List of Tables

1 Top 10 leading journals . . . 6

2 Reported sample sizes of studies on the TAM . . . 15

3 Crosstabulation of experiment and platform choice . . . 17

4 Twitter’s Cronbach’s α and Shapiro-Wilk results per concept . . . 18

5 LinkedIn’s Cronbach’s α and Shapiro-Wilk results per concept . . . 19

6 Facebook’s Cronbach’s α and Shapiro-Wilk results per concept . . . 20

7 Cronbach’s α for Likert items on the complete dataset . . . 22

8 Levene’s test for homogeneity of variances. . . 23

9 Mean ranks for concepts with a statistical significant difference . . . 23

Abbreviations

Abbreviation Term

KM Knowledge Management

KMS Knowledge Management Systems SNS Social Network Sites

TAM Technology Acceptance Model PU Perceived Usefulness

PEoU Perceived Ease of Use VI Value of Information BI Behavioral Intention

(6)

1 Introduction

Knowledge management is considered the youngest management discipline that has gained acceptance in the scientific community (Serenko & Bontis, 2013). The field of knowledge management (KM) is a broad field which is rooted in the (support of) creation, transfer and application of knowledge. The knowledge management systems that are currently in use require a lot of money to maintain; companies like BP spent £5 million over 2 years, De Beers spent $2 million over 2 years and Ernst & Young calculates about 6% of its revenues are spent on KM (Knowledge Management FAQ , 2011; Davenport, Long, & Beers, 1998).

Of the functions that knowledge management systems have, knowledge transfer is the most common and imperative activity (Suppiah & Sandhu, 2011). So if new methods for knowledge transfer could be discovered it could be possible for companies to spend less money on knowledge management systems.

With the increase in use of social network sites in the workplace, this possibility might already be in place. Studies show that as much as 75% of employees use social network sites at least once a day in the workplace, while 60% use it several times per day (Ladika, 2012). If it were possible for social network sites to support knowledge management and most specifically knowledge transfer, the benefits are plural.

It would reduce the need for companies to set up or maintain (legacy) knowledge management systems. Moreover it would enable the use of knowledge that is already available in existing networks and time that is already spent on social network sites could be used more productively.

This thesis will take an exploratory look at this issue to assess if users find social network sites to be capable of knowledge transfer. This results in the following research question;

To what extent are social network sites useable as an effective tool for knowl-edge transfer?

To be able to answer this question several steps are necessary, namely finding a way to predict technology use or user behavior, review existing literature on knowledge transfer, combine these fields and conduct an experiment. This research will form the basis for further research in this area, possibly unlocking new potential for social network sites (Yates & Paquette, 2011).

To answer the research question the following subquestions need to be answered; • How can technology use or user behavior for IT systems be predicted?

• How does this prediction relate to the social network sites and knowledge transfer theory?

This study will treat the most used social network sites in the Netherlands as IT systems. The scientific relevance of this thesis is created by bridging the fields of knowl-edge management and use of social network sites. Furthermore it can discover new use for existing systems by applying leading theories to network sites. In addition to the practical implications already mentioned, these also include reducing ’lost time’ spent

(7)

by employees on social network sites and a more efficient use of (professional) networks already in place.

1.1 Structure

This thesis aims to answer the research questions in four sections.

The remainder of the introduction will explore several key terms and issues that are important for this thesis. These include defining the scope of the research, a definition of knowledge and knowledge management and choosing which social network sites are under review.

In Section 2 a literature review will be done, first on prediction of technology use. This will answer the subquestion ’how can technology use or user behavior for IT systems be predicted?’ The next step of the literature review is to link this model to knowledge transfer theory, which will answer the question ’how does this prediction relation to the social network sites and knowledge transfer theory?’ Section 2 ends with the hypotheses that will be used for the survey.

Section 3 contains the methodology and results of the survey. Methodology covers information on procedure, sampling and measurement of the concepts. The result cover all the statistical results of the experiment and gives a conclusion on all statistically significant results, which tie back into the hypotheses.

The final section is the discussion of the results found in the survey. This section explores the proces, limitations, implications of the results and how this research could be used as the basis for further research.

1.2 Scope and assumptions

This research only looks at a new potential use of SNS as a tool for KMS and will take a IT-based approach to the problem. While this research links SNS and KMS together, some aspects will be outside the scope of this research and some assumptions are required. Issues that are outside the scope of this thesis can be divided into knowledge sharing issues and technological and social issues surrounding social network sites.

For knowledge sharing the issues outside of scope are the discussion if tacit knowl-edge sharing is possible (Martin, Hatzakis, Lycett, & Macredie, 2005) and how cultural differences that influence knowledge sharing (McDermott & O’Dell, 2001). For techno-logical and social issues the adoption of newly developed technologies (Walczak, 2005), adoption and aversion to using SNS’s and reasons for this (Boyd & Ellison, 2008) and how availability of a SNS impacts knowledge transfer (Wasko & Faraj, 2005) are out-side of scope. Discussion of solutions to these problems are not inout-side the scope of this research.

It is assumed that users are willing and able to share knowledge, but will account for differences in experience with the SNS’s that are taken into account during this research.

(8)

1.3 Defining ’knowledge’

Knowledge is the single most important abstract term applied throughout this paper and it is also a term that is difficult to define. A hierarchical view on knowledge makes a distinction between data, information and knowledge (Gottschalk, 2000). Here data is a symbol, information is processed data and knowledge applies information (Bellinger, Castro, & Mills, 2004). But this distinction contains several issues. Viewing data as raw data or a symbol could be viewed as wrong as all symbols have been given meaning by people and have been influenced before being ”published”. Others view ”processing” and ”applying” as the same terms which calls into question a difference between information and knowledge. For the purpose of this paper knowledge will be viewed as defined by Alavi and Leidner (2001):

”Knowledge is information possessed in the mind of individuals: it is per-sonalized information (which may or may not be new, unique, useful, or accurate) related to facts, procedures, concepts, interpretation, ideas, obser-vations, and judgements (p.109).”

1.3.1 Knowledge management

Knowledge management (KM) is, as an activity, a broad process that encompasses many different aspects including but not limited to knowledge creation, acquisition, assessment, retrieval, transfer, utilization and administrative processes (Suppiah & Sandhu, 2011; Alavi & Leidner, 2001). Knowledge management systems (KMS) are the IT systems that are designed to support any or more of these aspects.

1.4 Social Network Sites

This remainder of the introduction explores the usage numbers, types and possibilities of different Social Network Sites (SNS). Afterwards three sites are explored further because of their nature, availability and how these are used.

Various definitions of social network sites exist like ”online services that support so-cial interactions among users through highly accessible and scalable web-based publishing techniques (p.238)” (Chua & Banerjee, 2013), ”a group of internet-based applications that build on the ideological and technological foundations of Web 2.0, and that allow the creation and exchange of User Generated Content (p.239)” (Chua & Banerjee, 2013) and ”a web-based service that allows individuals to (1) construct a public or semi-public profile within a bounded system, (2) articulate a list of other users with whom they share a connection, and (3) view and traverse their list of connections and those made by others within the system (p.211)”(Boyd & Ellison, 2008).

The problem with many definitions is that they are too broad, as Beer (2008) argues. Existing definitions include YouTube as a social network site because it is possible to create content and a profile, make connections and use connections to exchange and find content. But the main aim of YouTube is to be able to post and view video’s, profiles

(9)

and connections are secondary. For the purpose of this paper the term ’social network sites’ will only include the sites that have connecting people as it’s main goal.

1.4.1 Types of social network sites

Zhao and Chen (2013) explore different enterprise knowledge sharing types in their paper by creating a taxonomy of types and exploring which platforms correspond.

Figure 1 mentions a lot of possible interactions, some of which are irrelevant for the purpose of this research, especially because these types are not what a social network entails. Only social interaction remains as a type, which fits LinkedIn as well because the possibilities that are provided are a lot like Facebook.

Figure 1: Summary of different enterprise knowledge sharing types

In The Netherlands, where the research was done, the most used SNS are in order of most to least Facebook, Youtube, Twitter, LinkedIn and Google Plus (Oosterveer, 2014). This paper will not research YouTube because, for the purpose of this paper, YouTube is not considered a social network site but a video sharing website primarily.

1.4.2 Twitter

Twitter is social network site where users can connect with people, place tweets of 140 characters which can include multimedia, discover what is happening worldwide, per country/region and city and create a personal profile that connects people to other users. Especially the use of hashtags is useful to create a connection or look for information. Any hashtag can be clicked which gives an overview of all (recent) tweets that used that hashtag as well. By indexing the hashtags Twitter gives an overview of ’trending topics’ and for worldwide trends Twitter also explains why it is trending.

Users on Twitter have the possibility to ”retweet”, this forwards a tweet to followers (which they otherwise might have missed). Users can also ”favorite a tweet” by pressing a star icon next to a tweet, this lets the original poster know that either you like their tweet or to store this tweet to find back more easily later. Simulated conversations are possible on Twitter through the @reply function, this can be done by pressing the reply icon next to a tweet, this tweet will start with ”@username” and Twitter will

(10)

show the original tweet that is replied to. Users can also mention another username by placing ”@username” anywhere in the body of a tweet, this is considered a mention. The difference between @reply’s and mentions is that any Twitter user will only see an @reply if they follow both users, while mentions are displayed on the timeline if they only follow the user that places the tweet.

A Twitter user is most likely to be female, having followed some college courses and between the age of 26-34 (Vidyarthi, 2012; Darell, 2011).

1.4.3 Facebook

Facebook is a social network site where users can connect and communicate with friends, place posts, list interests, ’like’ anything, follow public profiles and use application and games. Facebook is primarily aimed at personal use as a way to enrich existing friend-ships, keep in touch with old friends, classmates or colleagues and share experiences.

Facebook users can place a post to their Timeline being a status (text based), photo, place or life event. Any placed post has the ability to link one or more people, add date or location information, include a picture and show the current mood of the user. Users also have the opportunity to change the privacy setting of a post and this gives total control over who can see the post and who can’t.

Users have the opportunity to comment on posts, and these comments can in turn link to other Facebook users. Any post or comment can be ’liked’ by anyone, this expresses enjoyment or support of the content of the post. Users can also share a post, this places the post on the News Feed of their friends with a heading ”User shared OtherUser’s post”, increasing the range of this post. Recently Facebook also introduced the use of hashtags, much like Twitter, which are clickable and can be used to find posts that also include this hashtag.

A Facebook user is most likely to be male, having followed some college courses and between the age of 18-25 (Vidyarthi, 2012; Darell, 2011).

1.4.4 LinkedIn

LinkedIn is a professional network site where users can create a personal profile contain-ing educational history, experience and skills and connect to people. The main aim of LinkedIn is to connect professionals to make them more productive and successful.

The homepage of LinkedIn is somewhat similar to Facebook’s News Feed as it shows posts of connections and recommended posts. Users can comment, like and share posts also very similar to Facebook, but where Facebook only allows picture-files to be linked, LinkedIn accepts any file which can include pictures, documents and presentations.

LinkedIn also has a homepage where updates in your network are placed but users can also share interesting stories and comment on them.

A LinkedIn user is most likely to be male, having at least a bachelors degree and between the age of 26-34 (Vidyarthi, 2012; Darell, 2011).

(11)

2 Literature Review

This section contains the literature review concerning prediction of technology usage and knowledge transfer. The first part of section focusses on technology acceptance to answer the subquestion ’how can technology use or user behavior for IT systems be predicted?’ Afterwards literature on knowledge transfer is explored to tie into technology acceptance to answer the subquestion ’how does this prediction relate to the social network sites and knowledge transfer theory?’ Finally the theoretical framework and hypotheses are given.

2.1 Literature

Due to the amount of available literature for both knowledge management and technology acceptance the literature review will be focussed on the leading journals in the field and in IT in general. This places a limit on the available literature, but by taking the leading journals in the field it can be assumed that the literature review covers all facets that are accepted in the scientific field.

Rank Knowledge management Information Systems

1 Journal of Knowledge Management IEEE Communications Surveys and Tutorials

2 Knowledge Management Research & Practice

MIS Quarterly

3 International Journal of Knowledge Management

IEEE Wireless Communcations

4 Journal of Intellectual Capital Journal of Information Technology 5 Journal of Information and

Knowl-edge Management

Journal of Cheminformatics

6 The Learning Organization Journal of the American Medical In-formatics Association

7 Journal of Knowledge Management Practice

Journal of Chemical Information and Modeling

8 Knowledge and Process Manage-ment: The Journal of Corporate Transformation

Information Sciences

9 International Journal of Learning and Intellectual Capital

Information & Management

10 Electronic Journal of Knowledge Management

Information Systems Research

Table 1: Top 10 leading journals

According to a study done in 2008, which was updated in 2013, the top ten leading journals for knowledge management (Serenko & Bontis, 2013) and information systems

(12)

(Journal Citation Reports, 2013) can be found in Table 1. This table displays which journals were available for the literature review.

The literature review on technology acceptance will only consider the journals men-tioned in the column information systems. The literature review on knowledge manage-ment will include both columns.

2.2 Predicting technology usage

The first step of the literature review was to find the best method for technology usage prediction for the goal of this paper. To prevent the unnecessary purchase of a new system, it is recommended to research if there is a need for and an intention to use this new system. The basic concepts that are important for user acceptance models are presented in Figure 2

Figure 2: Basic concepts for User Acceptance Models

The most common are the technology acceptance model (TAM), theory of planned behavior (TPB) and the theory of reasoned action (TRA) (Mathieson, 1991; Venkatesh, Morris, Davis, & Davis, 2003).

The TRA, as seen in Figure 3, posits a causal link where ”personal and collective attitudes towards a behavior determine the intention to perform the behavior, and con-sequently the performance of the behavior (p.462)” (Casimir, Ng, & Cheng, 2012). The TRA is the basis for the TPB and the TAM.

Figure 3: Visual representation of the Theory of Reasoned Action

The TPB, as seen in Figure 4, is based on the TRA and posits that behavior is determined by intention, which is in turn predicted by attitude toward the behavior, subjective norms and perceived behavioral control (Mathieson, 1991).

(13)

Figure 4: Visual representation of the Theory of Planned Behavior

The TAM, seen in Figure 5 predicts behavioral intention to use through ”attitude towards the technology” which is determined by perceived ease of use and perceived usefulness. In the final model of the TAM ”attitude towards the technology” has been dropped because it turned out it was not a significant mediating variable (Koufaris, 2002).

Figure 5: Visual representation of the Technology Acceptance Model

Of the three models the TAM is the best fit for what we intend to accomplish, has been validated multiple times for multiple technologies and performs very well in predicting if users intend to use a technology (Koufaris, 2002; Venkatesh, 2000; Adams, Nelson, & Todd, 1992; Davis, 1989; Wixom & Todd, 2005). Although the TAM only performs well in supplying very general information on the users’ opinion about a system (Mathieson, 1991), this is exactly what this research intends to do due to it’s exploratory nature. Other studies have shown that in a second study it is possible to include measures of actual usage in the model (Horton, Buck, Waterson, & Clegg, 2001). These reasons are the basis to use the TAM for this research.

2.2.1 Technology Acceptance Model

The technology acceptance model (TAM) is the most widely applied model of user accep-tance and usage and has been developed because understanding user accepaccep-tance, adop-tion and usage of new systems is important for researchers and practiadop-tioners (Venkatesh, 2000).

(14)

The TAM has been developed by Davis (1989) and posits that perceived ease of use and perceived usefulness determine behavioral intent to use a technology (Koufaris, 2002). The influence of perceived ease of use on perceived usefulness exists due to the fact that all things being equal, the easier a technology is to use the more useful it can be. Extensions of the model are possible and Figure 6 shows them (Wixom & Todd, 2005).

Figure 6: The TAM and three popular extensions

This research will use a slightly modified version of this model, as several studies have proven that ”attitude towards usage” is not a significant mediating variable (Koufaris, 2002; Horton et al., 2001). To ensure significant results during this study, this aspect of the model will be disregarded for the purpose of this research.

Although TAM functions properly in predicting the use of systems, it does not pro-vide specific feedback about the system and possible improvements. It predicts either use or non-use but not how non-use could be prevented if implementation of the system is necessary. Other models, like before mentioned TPB or TRA, are able to do this, which could serve as a basis for further research depending on the results of this study. The TAM can be used in a variety of ways. In applications they might be used to obtain feedback on design approaches ar used after implementation of a system to diagnose problems in user acceptance. Researchers can use the model in understanding factors that influence the success of information systems (Adams et al., 1992)

2.3 Knowledge Transfer

After deciding on the Technology Acceptance Model as the basis for this study the next step is to extend this model with factors that are important for knowledge transfer. This section explores aspects that are important for knowledge transfer before focussing on factors that directly impact knowledge transfer. Both steps are important as knowledge transfer is part of the basis of this research and because it is considered to be the most common and imperative activity in knowledge management (Suppiah & Sandhu, 2011).

(15)

Knowledge transfer is considered to be successful if knowledge is at the right location, at the right time and ready to be used. Or ”the process through which one unit (e.g. group, department or division) is affected by the experience of another (p.61)” (Ko, Kirsch, & King, 2005).

Many factors impact knowledge transfer like motivation, communication-skills and -possibilities, absorptive capacities, nature of the information and the degree of tacitness of knowledge (Ko et al., 2005; Fang, Yang, & Hsu, 2013; Hislop, 2002). Possibilities for knowledge transfer are formal/informal and personal/impersonal, IT systems tend to be impersonal (except for video chat), but the formality depends on the system (Alavi & Leidner, 2001).

2.3.1 Knowledge transfer models

A possible structure for knowledge transfer is a market with senders and receivers. Lin, Geng and Whinston (2005) devised a model for transfer which factors in information (in-)completeness. In this model symmetric transfer happens when sender and receiver have the same amount of information available. In a situation where information com-pleteness is low, signal-jamming occurs but traditional knowledge transfer literature assumes high completeness on both sides. Asymmetric information transfer can be both sender-advantageous and receiver-advantageous.

Another model for knowledge transfer is given by Barachini (2009), the business transaction process. This process assumes that the transfer of (tacit) knowledge isn’t possible. The solution to this problem is to codify knowledge into information, storing it, have transfer take place which the receiver will turn into knowledge on an individual basis. Barachini argues that there are two types of information exchange, Type-1 is the immediate exchange of information in both directions and Type-2 is an unidirectional flow of information.

2.3.2 Factors impacting knowledge transfer

Many factors impact the transfer of knowledge and they can be divided into three cate-gories, technological, organizational and individual factors (Paroutis & Saleh, 2009).

Technological factors include richness of transmission channels, timing and speed. Richness of transmission channels is a term commonly used in ’media richness theory’ which describes the degree of richness as ”the ability of information to change under-standing within a time interval”. Thus richer media are characterized by the use of responsive feedback, multiple cues and appropriate use of language while lean media do not have these characteristics (Dennis & Kinney, 1998). Timing and speed are techno-logical aspects that the system could provide for the users.

Organizational factors include shared understanding, budget and culture. Shared understanding is the extent to which the work values, norms, philosophy and prior work experience are similar (Ko et al., 2005). The most often mentioned barrier for successful knowledge transfer is culture (Martin et al., 2005; McDermott & O’Dell, 2001; Suppiah

(16)

& Sandhu, 2011). Societal and organizational culture both are limiting factors, especially when employees are worried about competitive advantage. However other studies show that organizational culture can be a positive factor for knowledge transfer.

Personal factors include absorptive capacity, perceived value of the source knowl-edge and motivational disposition. The absorptive capacity of the receiver is the ability not only to acquire and assimilate but also to use knowledge, whereas the motivational disposition is the willingness to acquire knowledge from the source. Both factors can greatly influence knowledge transfer activities. Other factors that impact knowledge transfer are the nature of the transfer, whether it is between groups, individuals, people who never meet face to face or who might just have a lot in common, all these factors impact the role which information technology can play (Hislop, 2002).

2.4 Theoretical framework and hypotheses

By combining the TAM and its popular extensions with the theory on knowledge transfer, the graphical representation of the research model can be found in Figure 7. In this figure perceived usefulness, perceived ease of use and behavioral intention to use are directly taken from the TAM. Of the three factors mention in section 2.3.3 only the personal factor of ’perceived value of the source knowledge’ will be taken into account, which is perceived to impact the perceived usefulness of the platform under review.

The rest of the factors which are mentioned are either outside the scope of this research, not relevant due to the exploratory nature of this research or can be taken into consideration for further research (and this research is only interested into a very general view on using SNSs for knowledge transfer). As mentioned in the introduction, the model also accounts for previous experience with the SNS under review, this is done to prevent contamination of the results (a user with no experience in a SNS will find this SNS harder to use, impacting the results).

Figure 7: Theoretical framework

From the model presented in Figure 7 the following hypotheses are found.

• H1 There is a positive relation between experience with the SNS and perceived ease of use.

(17)

• H2 There is a positive relation between value of information and perceived useful-ness.

• H3 There is a positive relation between perceived ease of use and perceived use-fulness.

• H4 There is a positive relation between perceived ease of use and behavioral in-tention to use.

• H5 There is a positive relation between perceived usefulness and behavioral inten-tion to use.

Hypotheses are formulated for each variable to test during the research. Value of the information is hypothesized to be highest on LinkedIn due to the professional nature of the platform. Experience is hypothesized to have the highest impact on Facebook due to the capabilities of the comments. Perceived ease of use is hypothesized to be highest on Twitter due to the lay-out and reach of the platform. Perceived usefulness is hypothesized to be highest on LinkedIn do to the professional nature of the platform. Behavioral intention to use is hypothesized to be highest on Facebook due to the wide-spread nature of the platform.

For complete testing the hypotheses are worded to be tested with a split user exper-iment.

H6a Value of the Information will be perceived higher on Facebook than on Twitter. H6b Value of the Information will be perceived higher on LinkedIn than on Facebook. H6c Value of the Information will be perceived higher on LinkedIn than on Twitter.

H7a Experience will have an higher impact on Ease of Use on Facebook than on Twitter.

H7b Experience will have an higher impact on Ease of Use on Facebook than on LinkedIn.

H7c Experience will have an higher impact on Ease of Use on LinkedIn than on Twitter.

H8a Perceived Ease of Use will be higher on Twitter than on Facebook. H8b Perceived Ease of Use will be higher on LinkedIn than on Facebook. H8c Perceived Ease of Use will be higher on Twitter than on LinkedIn.

H9a Perceived Usefulness will be higher on Twitter than on Facebook. H9b Perceived Usefulness will be higher on LinkedIn than on Facebook. H9c Perceived Usefulness will be higher on LinkedIn than on Twitter.

H10a Behavioral Intention to Use will be higher on Facebook than on Twitter. H10b Behavioral Intention to Use will be higher on Facebook than on LinkedIn. H10c Behavioral Intention to Use will be higher on LinkedIn than on Twitter.

(18)

3 Methodology and Results

This section covers the experiment used to test the hypotheses. The first half, method-ology, explores the creation of the survey that was used. The second half, results, gives the statistical results of the survey in two parts. Part one explores hypotheses 1 through 5 divided up for each SNS, while the second part explores hypotheses 6 through 10 by way of a post-hoc analysis on all the data.

3.1 Methodology

This section explores the creation and results of the survey that were used to validate the model and test the hypotheses. To do this a web-based self-administered split-user questionnaire will be used.

3.1.1 Procedure

The survey is comprised of three parts, demographics, making a choice and rating state-ments related to the concepts of the TAM. Respondents are greeted on an online welcome page and click trough to the start of the survey, this url is created randomly through the use of PHP. <?php // C r e a t e t h e a r r a y $ l i n k s = array ( ) ; $ l i n k s [ 0 ] = ’<a h r e f =”s u r v e y 1 . html”>h e r e </a> ’ ; $ l i n k s [ 1 ] = ’<a h r e f =”s u r v e y 2 . html”>h e r e </a> ’ ; $ l i n k s [ 2 ] = ’<a h r e f =”s u r v e y 3 . html”>h e r e </a> ’ ; // Count l i n k s $num = count ( $ l i n k s ) ; // Randomize o r d e r

$random = rand ( 0 , $num− 1 ) ; // P r i n t random l i n k

echo $ l i n k s [ $random ] ; ?>

In this code-snippet ’survey1.html’ links to Facebook-Twitter, ’survey2.html’ to Twitter-LinkedIn and the last link to Facebook-LinkedIn, but users only saw a hyperlink with the text ’here’ to make it completely random for each respondent.

The demographics consist of gender, age and highest finished education and will asses experience with all three SNSs based on frequency of use. On the second page users will see a picture containing an information transfer task with the two social network sites for that experiment. Here users have to make a choice on which platform they prefer for information transfer. On the final two pages users are asked to rate statements on a 7-point Likert scale concerning the concepts that are under review.

(19)

The information transfer tasks that respondents are asked to choose from are based on the question ”which video on demand service currently available in The Netherlands is best?” The information presented in the pictures are the same, but based on the structure that is available to that platform (likes or @mentions). By ensuring that the information is given is the same for all three sites contamination of the results will be prevented, wether the possibilities of the platform impact the results will be reviewed in the discussion in Section 4.

This survey is not able to have a business-related question for two main reasons. Firstly there haven’t been any experiments that resemble this research, if there were this could have functioned as a jumping-off point for setting up this experiment. The sources that were researched for this problem included the sources already available dur-ing the literature review and searchdur-ing online for terms ”knowledge transfer research”, ”knowledge transfer task”, ”knowledge transfer test” and ”knowledge transfer experi-ment”. Neither option provided an experiment. Secondly to get a high sample it is necessary to think of a knowledge transfer task that can be understood by as many people as possible. Deciding to go for a VOD-service is based on the fact that it’s a very actual topic with many different opinions based on many different aspects (money, accessibility and what the platform offers).

3.1.2 Sampling

The literature review supplied many surveys and studies which applied the TAM and of all these studies the sample-size was reported. Because these studies managed to find significant results in their study, the sample sizes can be used as a basis for tho study. A report on the samples is given in Table 2, as can be seen the mean of these studies was 120 and the average was 236. This study aimed for a sample of 200 which is in line with the reviewed literature.

Because the population for this study is worldwide, there will be no specific sampling to reach certain individuals. People will be asked to participate for this study through the social network sites under review (primarily) and be distributed through e-mail (secondary). By using SNSs as a primary method to reach people it will ensure at least some familiarity with at least one platform, which ensures a basic level of usability. E-mail will be used as a method to reach groups of users more easily and to follow-up on people that might have missed the survey or didn’t have time when they were first presented with it.

(20)

Title Sample size Predicting User Intentions: Comparing the Technology

Ac-ceptance Model with the Theory of Planned Behavior

149

Assessing IT Usage: The Role of Prior Experience 786 Applying the Technology Acceptance Model and Flow

The-ory to Online Consumer Behavior

280

Determinants of Perceived Ease of Use: Integrating Con-trol, Intrinsic Motivation, and Emotion into the Technology Acceptance Model

70 & 160

Perceived Usefulness, Ease of Use and Usage of Information Technology: A Replication

68 & 116 & 73

Perceived Usefulness, Perceived Ease of Use, and User Ac-ceptance of Information Technology

120

A Theoretical Integration of User Satisfaction and Technol-ogy Acceptance

465

Table 2: Reported sample sizes of studies on the TAM

3.1.3 Measurement of concepts

Various studies have done research into which statements best measure perceived use-fulness (PU), perceived ease of use (PEOU) and behavioral intention to use (BI). For PU and PEOU the statements have been validated and proven reliable by Davis (1989) and Adams et al. (1992) so they will be used for this experiment as well.

For PU the statements will be;

• Using this platform allows me to find information more quickly. • Using this platform is an effective way to find information. • Using this platform makes finding information easier. • I find this platform to be useful for finding information. For PEoU the statements will be;

• I find this platform to be easy to learn.

• I find it easy to get this platform to do what I want it to do.

• My interaction with this platform is clear and understandable. It does not require a lot of my mental effort.

• I find this platform to be flexible.

• I find it easy to become skillful with this platform. • I find this platform to be easy to use.

The statements for behavioral intention to use have been used and validated by Venkatesh (2000) and will be used for this experiment as well.

• Assuming i had access to the system I intend to use it.

• Given that i had access to the system i predict that i will use it. For value of the information the questions will be;

(21)

• I find the information received to be useful. • I find I can make an informed decision.

All statements concerning the task have to be rated on a 7-point Likert scale of which only the endpoints were anchored. The lowest option was anchored ’strongly disagree’, options 2 through 6 were numbered and the final options was anchored ’strongly agree’. Studies show that a 7-point Likert scale shows the highest test-retest reliability and are more likely to show higher reliability than other number of options (Finstad, 2010; Wakita, Ueshima, & Noguchi, 2012). The reason to anchor only the extreme options is based on the fact that this study is expected to have mostly respondents for who English is not the native language and this reduced difficulty with interpreting exact meaning or translation. The fact that this might result in a slightly higher standard deviation (Weng, 2004) it is more important that a lot of people are able to fill in the survey.

During this study the the Likert items will be treated as interval ratio for a number of reasons, even though this is a controversial point. The most important reason is based on the existing literature on technology acceptance, all reviewed articles have treated the answers on the Likert items as interval ratio so this study will not deviate from this standard. Additionally there are sources available which are consistent with the assumption that it is possible to use parametric statistics with Likert data (Norman, 2010; Pell, 2005),

3.2 Results

This subsection explores the results of the statistical analysis of the results which will be explored based on the hypotheses given at the end of Section 2 and this subsection can be roughly divided into three parts.

The first part gives some general results on the data. Afterwards hypotheses 1 through 5 explore the relationships of the different concepts that can be found in the ex-tended TAM-model for the purpose of this research. These hypotheses will be explored for each SNS separately and will take into account the assumptions that are required to test for a relationship between concepts. Finally hypotheses 6 through 10 are ex-plored through a post-hoc analysis on the complete data, these hypotheses explore the differences between the SNS’s so require a look at all the data at the same time.

3.2.1 General results

The survey was finished by 192 people and of which 5 needed to be filtered out due to being assigned a experiment with two SNS’s that they did not have an account on. Of the remaining 187 respondents 80 were male and 107 were female, approximately 35% of the respondents were aged in the range 20-29 and both age ranges 30-39 and 40-49 had approximately 24% of the respondents. Most respondents (88) have finished a bachelor’s degree, 43 are in possession of a master’s degree, 34 have finished trade/technical/voca-tional training, with the rest lower or higher.

(22)

Table 3 displays how many respondents saw which experiment and for what platform they choose.

Platform choice

Facebook Twitter LinkedIn

Experiment

Facebook v. LinkedIn 41 0 17 58 Facebook v. Twitter 55 11 0 66 Twitter v. LinkedIn 0 20 43 63 96 31 60 187

Table 3: Crosstabulation of experiment and platform choice

3.2.2 Results per Social Network Site

The first step of the analysis was to divide the data into three separate datasets, based on the platform that respondents choose. This division is used to test hypotheses 1 through 5 which were;

• H1 There is a positive relation between experience with the SNS and perceived ease of use,

• H2 There is a positive relation between value of information and perceived useful-ness,

• H3 There is a positive relation between perceived ease of use and perceived use-fulness,

• H4 There is a positive relation between perceived ease of use and behavioral in-tention to use and

• H5 There is a positive relation between perceived usefulness and behavioral inten-tion to use.

For each platform the steps were to assess the internal reliability of the Likert items, check the assumptions that are required to calculate correlations and, if all requirements are met, calculate the correlations.

To check if it is possible to convert the separate Likert items into a Likert scale, Cronbach’s α is used to measure internal reliability. Because this study needs to make a decision based on groups (versus individuals), the threshold value that is maintained for Cronbach’s α is 0.8 (Kuijpers, van den Ark, & Croon, 2013). If this value is reached the Likert items will be combined into a Likert scale.

If the different Likert items are combined into Likert scales, these can be used to calculate the correlation between the concepts, which in turn are used to test hypotheses 1, 2, 3 4 and 5.

Hypothesis 1 is a relationship between a scale measured at the ordinal level and one at the interval level. Correlation between these type of scales can be calculated with Spearman’s ρ which, besides the type at which the scales are measured, requires a monotonic relationship between the scales which can be showed through a scatterplot

(23)

(Spearman’s Rank-Order Correlation using SPSS Statistics, 2013).

In hypotheses 2 through 5 the scales are measured at the interval level, for which the correlation can be calculated with Pearson’s r. Pearson’s r, besides the type at which the scales are measured, has three assumptions; a linear relation between the scales, no or a low amount of significant outliers and the data should be normally distributed (Pearson’s Product-Moment Correlation using SPSS , 2013). Significant outliers are shown through the use of box-plots (Checking for Outliers, 2008) and the Shapiro-Wilk test is used to check if data is distributed normally.

Exploration of the assumptions and the correlations are done per platform.

The strength of the correlation is based on Dancey and Reidy (2007), where 1 denotes a perfect relation, anything over .7 denotes a strong relation, over .4 a moderate relation and everything else a weak relation.

Twitter 31 People choose Twitter as the preferential platform for information transfer of which 20 were male, and the highest frequency was reported in the groups 20-29 years old and ’in possession of a bachelor’s degree’. Seven people don’t actively use their Twitter-account, although 18 use it daily or multiple times per day.

For Twitter the highest internal reliability was reported, with all values above 0.9, which can be seen in Table 4. Because the required value is met all Likert items are combined into Likert scales.

Concept Cronbach’s α Shapiro-Wilk VI .916 .062 PU .933 .159 PEoU .956 .066 BI .966 .086

Table 4: Twitter’s Cronbach’s α and Shapiro-Wilk results per concept

The assumption required to calculate Spearman’s ρ for hypothesis 1 is a monotonic relationship between experience and perceived ease of use. The scatterplot showed a non-monotonic relation between the concepts so H1 is rejected for Twitter.

To test hypotheses 2, 3, 4 and 5 the assumptions for Pearson’s r need to be met. A box plot of the data in Twitter gave 4 significant outliers, which is enough for a calculation of Pearson’s r, there was a linear relation and as the column for Shapiro-Wilk in Table 4 shows, the data is distributed normally.

All correlations were reported significant with one strong correlations, from PEoU to BI (r = .782, n = 31, p < .01) and two moderate strength correlations from VI to PI (r = .692, n = 31, p < .01) and PU to BI (r = .451, n = 31, p < .05) and one weak correlation from PEoU to PU (r = .381, n = 31, p < .05). A graphical representation of these correlations can be found in Figure 8.

(24)

Figure 8: TAM with significant correlation for Twitter

The results for Twitter have the overall highest correlations and highest Cronbach’s α values. With a high internal reliability and high correlations the results imply that users find Twitter especially useful for knowledge transfer. User satisfaction of Twitter for knowledge transfer is especially susceptible for high value of information. This is probably because Twitter provides the possibility to use hashtags and @mentions, which means that more information does not require any other action. Especially in this survey where the VOD-services were @mentioned so the official profile could be found easily and other platforms do not provide this capability.

LinkedIn 60 People choose LinkedIn as their preferential platform of which 31 were female and the highest frequencies were reported in the age 40-49 years old and ’in possession of a master’s degree’. About half the respondents (27) use their account a few times per week and four don’t actively user their accounts.

Cronbach’s α of the Likert items for LinkedIn met the threshold, which can bee seen in Figure 5 and were combined into Likert scales.

Table 5: LinkedIn’s Cronbach’s α and Shapiro-Wilk results per concept

A scatterplot showed a non-monotonic relation between experience and perceived ease of use which violates one of the assumptions needed for Spearman’s ρ. H1 is re-jected for LinkedIn.

For testing hypotheses 2 through 5 the assumptions to calculate the Pearson r cor-relation were explored first. All assumptions were met as there were just 2 significant

(25)

outliers, the scatterplot showed a linear relation between the concepts and all data was normally distributed, as can be seen in the column for Shapiro-Wilk in Table 5.

Calculating Pearson’s r gave statistically significant results of moderate strength for two correlations, from PEoU to PU (r = .488, n = 60, p < .01) and from PU to BI (r = .419, n = 60, p < .01) and one weak correlation from VI to PU (r = .299, n = 60, p < .05). The correlation from PEoU to BI did not give a statistically significant result. A graphical representation of the significant correlations can be found in Figure 9.

Figure 9: TAM with significant correlation for LinkedIn

These calculation result in the rejection of H4 and the acceptation of H2, H3 and H5 for LinkedIn.

The results for LinkedIn imply that the platform is not easy to use, but users do perceive it to be useful. And while there is a moderate positive correlation between perceived usefulness and behavioral intention (”all thing being the same, the easier a platform is to use, the more useful it can be”) there is no significant correlation between perceived ease of use and behavioral intention.

Facebook Of the 96 respondents that choose Facebook 56 were female and the highest frequencies were reported in the age 20-29 and ’in possession of a bachelor’s degree’. Users who choose Facebook as the preferential platform for information transfer were most active with 68 users using Facebook multiple times per month and 19 using it daily. All Likert-items met the threshold for Cronbach’s α, as can be seen in Table 6, and were combined into Likert-scales.

(26)

To test hypothesis 1 through Spearman’s ρ a monotonic relation between experience and PEoU is required, but a scatterplot showed a non-monotonic relation. This results in the rejection of H1 for Facebook.

Testing the required assumptions for calculating Pearson’s r did not result in any issues, there was 1 significant outlier and a linear relation between the concepts. But as the column Shapiro-Wilk in Table 6 shows that two concepts are not distributed normally, PU and BI. These results make the use of Pearson’s correlation impossible so a non-parametric correlation was required.

The alternative is to use Spearman’s ρ which is requires a monotonic relation between the concepts, but no normality. All scatter-plots showed a monotonic relationship, so it was possible to use Spearman’s ρ to calculate the correlations. The calculations gave statistically significant results for all correlations between the concepts. Two of which are of moderate strength, from VI to PU (rs(94) = .417, p < .01) and PU to BI (rs(94) =

.513, p < .01), and two weak relations from PEoU to PU (rs(94) = .318, p < .01) and

PEoU to BI (rs(94) = .305, p < .01).

A graphical representation of the significant correlations can be found in Figure 10.

Figure 10: TAM with significant correlation for Facebook

The results conclude in the acceptation of H2, H3, H4 and H5 for Facebook.

The results for Facebook show positive correlations for all concepts. Compared to Twitter, which reported the same amount of statistically significant correlations, the values are lower for Facebook, but they do show positive correlations. The highest correlation was reported from perceived usefulness to behavioral intention, which means that users will tend to use Facebook for knowledge transfer more because they think it is useful than because it is easy to use. The other correlation of moderate strength was from value of information to usefulness, which also implies that Facebook can benefit from a high value of information for user acceptance for knowledge transfer.

(27)

3.2.3 The complete dataset

The final step of calculating the results was to do a check on the complete data. This is necessary to test the final hypotheses repeated here.

• H6 Value of the Information will be perceived highest on LinkedIn and lowest on Twitter.

• H7 Experience will have the highest impact on Ease of Use for Facebook and lowest for Twitter.

• H8 Perceived Ease of Use will be highest on Twitter and lowest on Facebook. • H9 Perceived Usefulness will be highest on Twitter and lowest on Facebook. • H10 Behavioral Intention to Use will be highest on Facebook and lowest on

Twit-ter.

The first step of the analysis on the complete dataset is to see if Cronbach’s α meets the required value of 0.8, the results of this analysis are found in Table 7. Because the required value was met all Likert items were combined into Likert scales.

Concept Cronbach’s α VI .869 PU .949 PEoU .909 BI .950

Table 7: Cronbach’s α for Likert items on the complete dataset

Non-parametric test Similar studies showed that an ANOVA is the usual non-parametric post-hoc analysis for data of this nature (Kim, Sohn, & Choi, 2011; Straub, Keil, & Brenner, 1997).

There are several assumptions that need to be met to be able to perform an ANOVA. These are measuring the dependent variable at interval or ratio level, having the inde-pendent variable consist of two or more categorial, indeinde-pendent groups, independence of observations, having no or hardly any significant outliers, that the dependent variable should be approximately normally distributed and finally there needs to be homogene-ity of variances (One-way ANOVA in SPSS , 2013). This study complies with the type of dependent and independent variables, but the rest of the assumptions need to be measured.

The assumption of independence of observations is met as there are no respondents that filled in the test twice. There are 7 significant outliers which were tested by the use of box plots (Checking for Outliers, 2008), as this is only 3,7% of the total population this will not impact the use of ANOVA. Homogeniety of variances was tested with Levene’s test for homogeneity which resulted in significant results for all concepts, the results can be found in Table 8. Because test for normality were necessary for the tests on the separated data these results can be found in Tables 4, 5 and 6. Although two concepts on Facebook violate the assumption of normality require for an ANOVA but research has been done into the robustness of an ANOVA and ANOVA is robust against this

(28)

assumption (Driscoll, 1996; Schmider, Ziegler, Danay, Beyer, & Buhner, 2010; Norman, 2010)

Concept Significance for Levene’s test

VI .880

PU .424

PEoU .625

BI .756

Table 8: Levene’s test for homogeneity of variances.

Because all required assumptions are met it is possible to do an one-way ANOVA with a Tukey post-hoc analysis with platform choice as the dependent variable. This resulted with two concepts which gave a statistically significant difference, PU (F (2, 184) = 6.964, p = .001) and BI (F (2, 184) = 5.385, p = .005) and two concepts which didn’t, VI (F (2, 184) = 1.790, p = .170) and PEoU (F (2, 184) = .458, p = .634).

Tukey’s post-hoc analysis revealed that for perceived usefulness users on Facebook were less positive (2.95 ± 1.16) than users that choose Twitter (3.65 ± 1.32, p = .020) and LinkedIn (3.62 ± 1.31, p = .004).

To further explore the differences in PU and BI for the platforms a Kruskal-Wallis test was done for the purpose of the mean ranks (McDonald, 2014). For this test only the mean ranks are reported in Table 9.

Mean Rank PU Mean Rank BI Twitter 109.90 96.82

LinkedIn 107.91 110.48 Facebook 80.17 82.79

Table 9: Mean ranks for concepts with a statistical significant difference

Linking to the hypotheses By combining all the results from the statistical tests the results are linked back to the hypotheses.

Hypothesis 6 mentioned that value of information will be perceived highest on LinkedIn and lowest on Twitter. As the one-way ANOVA did not give a statistically significant difference for platform choice and value of information H6a, H6b and H6c are rejected. Since no significant correlation (Spearman’s r) was found between experience with the social network site and perceived ease of use H7a, H7b and H7c are rejected. Hypoth-esis 8 mentioned that perceived ease of use will be highest on Twitter and lowest on Facebook. As the one-way ANOVA did not give a statistically significant difference for platform choice and value of information H8a, H8b and H8c are rejected.

Hypothesis 9 mentioned that perceived usefulness will be highest on LinkedIn and lowest on Facebook. The one-way Anova showed a statistical difference for the concepts based on platform choice. Because PU was assessed highest on Twitter and lowest on

(29)

Facebook H9a and H9b are accepted and H9c is rejected. The reason why Twitter scored higher than LinkedIn will be explored in the final section.

Hypothesis 10 mentioned that behavioral intention to use will be highest on Facebook and lowest on Twitter. The one-way Anova showed a statistically significant difference for the concepts based on platform choice, but the hypotheses are rejected because LinkedIn scored highest and Facebook scored lowest, which means that H10a, H10b and H10c are rejected. Why the hypotheses were rejected even though a statistically significant difference was found will be explored in the final section.

(30)

4 Discussion

This section looks back at the steps that were taken during the course of this research from the method to the statistical calculations and the hypotheses.

This thesis followed three steps, exploring the capabilities of the most used social networks in the Netherlands, finding the best model for predicting technology usage for the purpose of this research, extend this with factors that impact knowledge transfer and setting up a survey to test the model.

The research model and hypotheses are based on the TAM, its popular extensions and the theory of knowledge transfer. The study was made to measure the concepts in the TAM and find out which platforms are able to be used for knowledge transfer, and to what extent. A post-hoc analysis was used to show statistically significant differences between the platforms.

The study was able to show several key differences between the platforms as well as explore which platform had the highest values of user acceptance for knowledge transfer. The value of these results is also done in this final section.

4.1 Limitations

Most limitations of this research have been mentioned in the introduction under scope. These assumptions include taking an IT-based approach to the problem, not taking a psychological look at issues surrounding knowledge transfer and adoption of new tech-nologies and that users are able and willing to share knowledge.

These assumption do need to be taken into account when discussing the results, especially because it impacts the conclusions that can be reached. All results are placed inside the limitations of this study and this needs to be taken into account when further research is done into this area.

Especially the aspect of adoption and aversion of using SNS’s for new purposes. It is highly plausible that users do not wish to use the SNS’s they use for their personal reasons for professional goals.

But the biggest limitation that any research into SNS’s runs into is that of privacy. Because it is impossible to ask full access to someone’s personal social network profile there will always be a factor of uncertainty into research of this kind. This exploratory research does not take this issue into account, but if more specific research into this area is done then this aspect will become more and more important for researchers to account for.

Some of the hypotheses of this study did not yield statistically significant results. Hypothesis 1 and 7 were both rejected as there was no statistically significant relation reported between experience with the SNS and perceived ease of use for any platform. There is a high probability that this was due to a design flaw in the research as the questionnaire reduced ’experience’ to ’average use’, which is a too simple assumption for measuring experience. A more complex check on experience (frequency of use, amount of

(31)

use, how long they have an account and how they use the platform) might yield different results and possibly uncover a significant relation between the concepts.

Hypothesis 6 and 8 were both rejected too, which imply that users, in this survey, did not base their choice for the platform on value of information or perceived ease of use. Using a different model for user acceptance might yield significant results or give insight into why these concepts are not important for platform choice.

4.2 Technology acceptance

The basis of the research was the TAM with its extension and related to personal and technological factors that impact knowledge transfer. Taking the TAM as the basis for the research was good for the ability to take proven methods and apply these to a new exploratory research as is shown by the high internal validity reported for the Likert items.

The other important aspect was the fact that the TAM is very good at predicting user acceptance at a very base level. It either shows that users are willing to use a new technology or if they aren’t. This is reflected in the hypotheses that were rejected, the model does not present explanation for these rejections.

Other models, like the Theory of Planned Behavior, are able to provide insight into why a certain technology might not be accepted. Because the TAM was able to show some significant results in this study, a new research with one of the other models is warranted to provide further insight into user acceptance for using SNS’s for knowledge transfer.

4.3 The survey

By taking a look at the general results of the survey and the general statistics which can be found in section 1.3.2 until 1.3.4 the sampling of this survey appears to be in line with the expected values. This shows in the education level of the respondents, which was highest on LinkedIn, and the age of the users, which was youngest on Facebook. Because the sample of users appears to have a likening to the general population of users this benefits the strength of the results found. Some interesting facts did come to light as younger users where more likely to prefer Twitter for knowledge transfer than the other two platforms.

The reason to use a web-based self-administered survey was to be able to reach a high response rate in a relatively short time. This was successful as there were 193 responses in about two weeks. But there could have been better methods to get valid results for a study of this nature.

This research specifically could benefit greatly from the possibility to conduct semi-structured interview after filling in the survey. The follow-up questions could be used to find out which concepts, possibilities and restrictions have the biggest impact on successful knowledge sharing and how or if SNS’s might be able to facilitate it. It would also enable the researcher to test if the respondents are willing to use SNS’s for new purposes besides the ones that they are already using it at the moment.

(32)

An issue that will always arise in any research concerning social network sites is the aspect of privacy. There are ways to assess if users think SNS’s are useable as a knowledge transfer tool, but a ’perfect study’ will require full access to the account used for knowledge transfer. Without this full control, which is not desirable either, there is always a factor of uncertainty for the conclusions linked to research concerning social network sites.

4.4 Results

The results of the study show promising results for both the scientific field and for companies.

By combining the fields of knowledge transfer and social network sites through a user acceptance model there are some very promising results. Users seem to have a positive attitude towards using social network sites as a knowledge transfer tool. As previously mentioned, by using different approaches to this model through different models or taking a look at more specific elements of social network sites, this possibility needs to be explored further.

Companies could really benefit from the fact that social network sites are so widely available, users know the limitations but more importantly the capabilities of the plat-form. And if they are willing to use their network for knowledge management the companies can benefit hugely from the networks that are maintained in a natural way.

4.4.1 Twitter

The highest promise in user acceptance appears to be for Twitter. While Twitter was the platform that was chosen least, the respondents that did choose it are more positive about the capabilities of the platform for knowledge transfer and the highest internal reliability was reported for this platform.

It scores highest on perceived usefulness and second highest on behavioral intention to use. Combined with the high correlations reported from value of information to perceived usefulness and from perceived ease of use to behavioral intention to use it appears like users are very positive about the capabilities of the platform and the platform can perform great when the quality of the information provided is high.

4.4.2 Facebook

Facebook appears to show the lowest promise in the ability for the platform to support knowledge transfer. It was the platform that was chosen most but scored lowest on both perceived usefulness and behavioral intention to use.

While two reported correlations where of moderate strength, from value of informa-tion to perceived usefulness and from perceived usefulness to behavioral inteninforma-tion to use, combining these two aspects is worrisome. A moderate strength correlation from one concept to the other while both concepts score the lowest across the platforms does not show good promise for using Facebook as a tool for knowledge transfer in this study.

(33)

4.4.3 LinkedIn

In the results of this study, LinkedIn appears to end up somewhere in the middle of the pack with some interesting results. The platform scored highest on behavioral intention to use, second highest on perceived usefulness (but a close second), but also did not provide statistically significant results from perceived ease of use to behavioral intention to use.

From these results it is implied that respondents that choose LinkedIn are very willing to use the platform for knowledge transfer, but this is not rooted in the ease of use of the platform. This also implies that LinkedIn derives it’s succes from perceived usefulness and nothing else. If this is true LinkedIn could be a very useful platform for knowledge transfer, especially if it were able to make the site more easily accessible.

4.4.4 Statistical differences between the platforms

Hypothesis 9 predicted that perceived usefulness will be highest on LinkedIn and low-est on Facebook. The one-way ANOVA showed a statistically significant difference for perceived usefulness on the different platforms, and Tukey’s post-hoc analysis showed that PU was rated highest on Twitter and lowest on Facebook. The mean rank that the Kruskal-Wallis test showed is the same as the result of the one-way ANOVA but the mean ranks for Twitter (109.90) and LinkedIn (107.91) were quite close and much higher than Facebook (80.17). Because the difference between Twitter and LinkedIn is that close it is difficult to know where the difference comes from. It might possibly be the effect of two important aspects of the platforms, where on Twitter more information was just one click away (from the use of hashtags or @-mentions) the professional nature of LinkedIn might influence how users experience the usefulness of anything displayed on that platform.

Hypothesis 10 predicted that behavioral intention to use will be highest on Facebook and lowest on Twitter. While the one-way ANOVA showed a statistically significant difference for behavioral intention on the different platforms, but Tukey’s post-hoc anal-ysis was unable to show the highest or lowest scoring platform. The mean ranks from the Kruskal-Wallis test showed the highest mean rank for LinkedIn (110.48), lowest for Facebook (82.79) and the middle value belonged to Twitter (96.82). While a significant difference was found the prediction of high values for Facebook and low values for Twit-ter did not turn out to be correct. It is inTwit-teresting to note that for both aspects with a statistically significant difference Facebook scored the lowest, which might be due to the fact that the respondents use this platform so much that they also run into the lim-itations of the platform. Another reason might be that Facebook is the most ’personal’ platform of the three, meaning that Twitter is mostly for short bursts of information or fun, LinkedIn is used for professional connections but Facebook can be used for al-most everything else. The reason that LinkedIn scored highest on behavioral intention is probably due to the professional nature of the platform.

(34)

4.5 Suggestions

This exploratory study showed a statistical difference between some aspects of the plat-forms and positive feedback into the capabilities of the social network sites for the purpose of knowledge transfer. Because of these results further research is warranted to find out how SNS are usable as a knowledge transfer tool.

This could be done by taking a qualitative approach to the problem, which could provide further insights into how users perceive the usability of the different platforms. But also taking a more specific look at the capabilities and limitations of the platforms of each platform and doing a comparative study on these could provide valuable insights into how platforms are already able to support knowledge transfer and possibly how simple changes could enable this in the future.

Also by taking a less strict technological view of the issue at hand but combining it with sociological views, the results of these studies will be broader and more applicable for companies in the future. Those studies could also look into the quality and efficiency of knowledge transfer through social network sites. All these, and more, aspects are important for successful knowledge transfer.

But as this study has showed, there is a basis for further exploration into the use of social network sites as a knowledge transfer tool.