What Does it Mean to Trust Facebook? Examining Technology and Interpersonal Trust Beliefs Nancy K. Lankton Department of Accounting and Information Systems Michigan State University lankton@bus.msu.edu D. Harrison McKnight Department of Accounting and Information Systems Michigan State University mcknight@bus.msu.edu 1 What Does it Mean to Trust Facebook? Examining Technology and Interpersonal Trust Beliefs ABSTRACT Researchers have recently studied technology trust in terms of the technological artifact itself. Two different kinds of trusting beliefs could apply to a website artifact. First, the trusting beliefs may relate to the interpersonal characteristics—competence, integrity, and benevolence. Second, they may relate to corresponding technology characteristics—functionality, reliability, and helpfulness. Since social networking websites like Facebook may demonstrate either interpersonal or technology trust characteristics, researchers may need to carefully choose the beliefs to model. Thus it is important to understand not only the conceptual meaning of these beliefs, but also whether human and technology trust beliefs are distinct. Using data collected from 362 university-student Facebook users, we test two second-order factor structures that represent alternative ways to model the three interpersonal and three technology trust beliefs. We find that the best-fitting measurement model depicts the three conceptually-related pairs of trust beliefs (competence-functionality, integrity-reliability, and benevolence-helpfulness) as three distinct second-order factors. This model outperformed the model splitting trusting beliefs into separate interpersonal and technology second-order factors. The results show people distinguish among three types of conceptually-related trust attributes, and that they trust Facebook as both a technology and a quasi-person. These second-order trust factors can be used in future research to better understand social networking trust and usage continuance intentions. Keywords: Interpersonal trust, technology trust, social networking, websites, measurement, second-order factors Acknowledgements: An earlier version of this paper was presented at AMCIS 2008. We thank the reviewers and editors of DATABASE for their helpful comments on this paper. We also appreciate Fred Rodammer for his help in collecting data for this study. 2 What Does it Mean to Trust Facebook? Examining Technology and Interpersonal Trust Beliefs INTRODUCTION Facebook has grown rapidly as hundreds of millions of users have adopted it to communicate and socialize (Bausch & Han, 2006). Trust may play a role in this meteoric rise. For example, some researchers suggest that social networking users have a generalized trust toward the group of people that visit their sites and read their postings (Kennedy & Sakaguchi, 2009). Other researchers find that people also trust social networking websites (Dwyer, Hiltz & Passerini, 2007; Fogel & Nehmad, 2009; Sledgianowski & Kulviwat, 2009). However, what exactly does it mean to say one trusts Facebook? Does one trust Facebook as a technology (i.e., a website artifact) or as a quasi-person or organization? Trust is often a factor in the use or acceptance of consumer product websites (Gefen, Karahanna & Straub, 2003). When studying trust in product websites, researchers often examine interpersonal trust, i.e., site user trust in the e-vendor (e.g., Bhattacherjee, 2002; Gefen et al., 2003; Kim, 2008). In this context, interpersonal trust means one is willing to depend on the evendor because one believes the e-vendor has such favorable attributes as ability (competence), integrity, and benevolence (Mayer, Davis & Schoorman, 1995). Research has used these interpersonal trust beliefs to represent how users perceive the attributes of the e-vendor. This is pure interpersonal trust, because it involves two humans, a user and an e-vendor. Recently, some empirical information systems research has explored trust in software recommendation agents (e.g., Komiak & Benbasat, 2006; Wang & Benbasat, 2005). These agents are technological artifacts, not humans. This type of trust, called trust in technology, differs from interpersonal trust because it represents a human-to-technology trust relationship 3 rather than a human-to-human trust relationship. Trust in technology means one is willing to depend on the other because one believes the technology has desirable attributes (McKnight, 2005). Although some have said trust can only exist between humans (Friedman, Kahn & Howe, 2000), many researchers now acknowledge that humans can and do trust technology, despite several differences between human-to-human and human-to-technology exchanges (see Lee & See, 2004 and Wang & Benbasat, 2005 for discussions of trust in technology). To-date, trust in technology is an under-explored information systems research domain. Due to the limited amount of technology trust research, it is difficult to answer the question, “What are the appropriate attributes of trust in technology?” Research on trust in software agents has employed interpersonal trust beliefs (i.e., competence, integrity, and benevolence) to represent trust in technology because software agents have some human-like characteristics, such as giving advice and interacting with the user on-screen (Wang & Benbasat, 2005). However, some technical artifacts possess fewer interpersonal characteristics than do software agents. For example, many websites neither give advice nor interact with users. Therefore, while interpersonal trust applies to software agents, it may only partially apply to websites. For example, people interface with other people on Facebook, but they neither obtain advice directly from Facebook itself nor interact with Facebook as a person or quasi-person. While trust in social networking websites research has generally examined interpersonal trust attributes (Dwyer et al., 2007; Fogel & Nehmad, 2009; Sledgianowski & Kulviwat, 2009), people may trust Facebook in other ways. For example, McKnight (2005) explains that people may trust a technology because it provides specific functionality, operates reliably, and is helpful to its users. Thus people may be willing to depend on Facebook (or any technology) because it has these technology-related attributes that make it trustworthy (McKnight, 2005). 4 We propose three technology-related trust beliefs that parallel the three most commonly used interpersonal trust beliefs. We suggest that the technology trust belief functionality is analogous to the interpersonal trust belief competence, in that they both refer to users’ beliefs about what the other can do for them. Similarly, we introduce reliability as a technology trust belief similar to the interpersonal trust belief integrity because they both refer to users’ beliefs that the other will do what we expect they will do. We suggest helpfulness as a technology trust belief that parallels the interpersonal trust belief benevolence in that they both relate to beliefs that the other provides responsive aid. This paper tests empirically whether interpersonal trust beliefs are separate and distinct from technology trust beliefs. The paper also examines how well the pairs of conceptual attributes above (e.g., benevolence and helpfulness) hold together as distinct attribute pairs. We gather data from students who use Facebook, a social networking website. Many social networking websites have grown in popularity among university students. These sites allow their users to create profiles and personal networks. It is possible, even though people do not interact with Facebook as a “person,” that they may still attribute human characteristics to it, as in prior research (Dwyer et al., 2007; Fogel & Nehmad, 2009; Sledgianowski & Kulviwat, 2009), and as Reeves and Nass (1996) have found with various technologies. However, we believe social networking websites represent technologies about which users may perceive both human-like and technology-like trust characteristics, and thus form both interpersonal and technology trusting beliefs. Thus our research contributes by examining both types of trusting beliefs as they relate to Facebook. In examining interpersonal and technology trusting beliefs, we develop hypotheses relating to their factor structure, and test two alternative second-order factor structures by 5 comparing measurement model results. We also contribute by assessing the second-order factors’ nomological validity or whether the factors behave as they should within a well-established theoretical framework (McKnight, Choudhury & Kacmar, 2002a; Straub, Boudreau & Gefen, 2004). Using trust theory, we analyze the trust second-order factors first as consequents of reputation, privacy concern, and ease of use, and then as predictors of trusting intention and continuance intention. These relationships have been established with interpersonal trusting beliefs in other contexts (e.g., Lowry, Vance & Moody et al., 2008; Gefen et al., 2003). However, to our knowledge prior trust-social networking research has not yet examined such relationships, nor have these relationships been used to validate alternative interpersonal and technology trusting belief factor structures. THEORY AND HYPOTHESES DEVELOPMENT Technology Trust Beliefs Researchers in various fields have investigated technology trust. For example, human computer interface researchers have examined trust in automation by testing the extent to which human operators will trust automated control of systems such as semi-automatic pasteurization plants with optional manual control (e.g., Muir & Moray, 1996; see Lee & See, 2004 for a review). In the social sciences, researchers have examined trust in the technological artifact of online environments (Komiak & Benbasat, 2006; Lee & Turban, 2001; Wang & Benbasat, 2005), and in various business information systems (Lippert, 2001, 2007; Lippert & Swiercz, 2005). While trust in technology research is just beginning, scholars across these contexts appear to consistently find trust in technology exists and is composed of multiple beliefs. Some trust beliefs relate to the human-like characteristics of technology. For example, Wang & Benbasat 6 (2005) apply the three most common interpersonal trust beliefs—competence, integrity, and benevolence—to their study of Internet recommendation agents. However, other researchers use trust beliefs that relate more to the technology-like characteristics of technology including its functionality and reliability (Lippert, 2001; Muir & Moray, 1996). Choosing which trust beliefs to use may depend on the extent to which the technology possesses human-like characteristics. For example, the software agents Wang & Benbasat (2005) studied have more human-like characteristics than Muir and Moray’s (1996) automated systems. Recommendation agents “interact” with users and provide them advice on particular products. By contrast, software systems like Microsoft Access provide little advice and interact little with the user, especially in the conversational manner that people do. Thus, technology trust beliefs may be more appropriate for Access than are interpersonal trust beliefs. Social networking websites represent a technology in which the distinction between human and technology characteristics is less clear. These technologies may demonstrate some human-like trusting characteristics such that users may develop competence and integrity beliefs. For example, Dwyer et al. (2007) find that users generally agree with the statement “I trust that Facebook will not use my personal information for any other purpose.” This statement reflects Facebook’s integrity in terms of safeguarding private information, an attribute we associate with people. Other researchers find that statements like “I feel that this website is honest” reflect trust in social networking websites (Sledgianowski & Kulviwat, 2009). Again, honesty is an attribute we usually apply to people. Social networking sites may also demonstrate technology-like trusting characteristics that elicit beliefs such as “Facebook is very reliable and consistent to use.” Therefore, researchers may apply both interpersonal and technology trust beliefs to understand users’ trust in Facebook. Thus it is important to determine the extent to which 7 Facebook users relate better to technology trust or interpersonal trust. To our knowledge, no research to-date has done this. In this research, we propose three trust-in-technology beliefs that are related, yet distinct from the three most commonly used interpersonal trust beliefs, which are competence, integrity, and benevolence (Gefen et al., 2003). We test whether the six beliefs are distinct from each other and investigate alternative factor structures for them. The following paragraphs explain the three proposed technology trusting beliefs (see Table 1). [Insert Table 1 Here] Functionality Belief Functionality means the degree to which an individual believes the technology will have the functions or features needed to accomplish one’s task(s) (McKnight, 2005) (see Table 1). Functionality originates conceptually from the interpersonal trust competence belief that represents an individual’s belief that a trustee has the ability, skills, and expertise to perform effectively (Mayer et al., 1995). While individuals demonstrate competence by performing a task well or by giving good advice, technology demonstrates ‘competence’ by performing a function well or by providing system features the user needs in order to perform a task. Thus trust in the competence of technology generally refers to the technology’s ‘functional’ capability to perform a task (McKnight, 2005). Similar trusting beliefs have been used in technology contexts including trust in automation (Muir and Moray, 1996). Reliability Belief Reliability is defined as the degree to which an individual believes the technology will continually operate properly, or will operate in a consistent, flawless manner (McKnight, 2005) 8 (see Table 1). This technology trust belief has its conceptual foundation in the integrity belief of interpersonal trust that represents the trustor’s perceptions that the trustee adheres to a set of principles that the trustor finds acceptable (Mayer et al., 1995, p. 719). A person may demonstrate reliability by keeping commitments and telling the truth. Technologies cannot demonstrate honesty or a moral conscience by keeping promises or commitments. However, every technology comes with the implicit promise that it will work reliably and consistently. Therefore, a technology demonstrates integrity by being reliable or by consistently doing what it implicitly promises to do every time the technology is used. Showing its human roots, reliability has been used by interpersonal trust researchers as an interpersonal trust belief (Rempel, Holmes & Zanna, 1985). Reliability has also been used before in technology trust studies (Lippert, 2001; Muir & Moray, 1996) (see Table 1). Helpfulness Belief Helpfulness is defined as the degree to which an individual believes the technology will provide adequate and responsive help, usually through a help function (see Table 1). Helpfulness is based on the benevolence belief from interpersonal trust and trust in online environments (Mayer et al., 1995; Gefen et al., 2003). The benevolent trustee cares and acts in the trustor’s interest (Wang & Benbasat, 2005; Mayer et al., 1995). We assume technology is not helpful in terms of volition or moral agency (i.e., it cannot consciously care about its user). In fact, that would constitute unwarranted personification of the technology (McKnight, 2005). Instead, we presume that technology demonstrates its helpfulness through help functions that aid goal attainment. Individuals who perceive that a technology can provide the help needed will perceive fewer risks and uncertainties associated with technology use. 9 In summary, we propose three technology trust beliefs —functionality, reliability, and helpfulness—that are based on three interpersonal trust beliefs–competence, integrity, and benevolence, respectively. Our hypotheses are based on two alternative ways of modeling these beliefs using second-order factors. A second-order or multi-dimensional factor is a theoretically meaningful, overall abstraction of interrelated dimensions or first-order factors (Law, Wong & Mobley, 1998). A common latent or reflective second-order factor exists if the dimensions are manifestations of the second-order factor or the construct leads to the dimensions (Law et al., 1998; Law & Wong, 1999; MacKenzie, Podsakoff, Shen & Podsakoff, 2006). An aggregate or formative second-order factor exists if the dimensions form the second-order construct (Law et al., 1998; Podsakoff et al., 2006). In our two alternative second-order factor models, we model the first-order dimensions as reflective (not formative) of the second-order factors because we expect the trust dimensions to co-vary with and even influence each other (Mackenzie et al., 2005; McKnight et al., 2002a: Petter, Straub & Rai, 2007). Also, we believe the first-order dimensions jointly reflect the overall trust concept and may be influenced by it (Diamantopouos, Piefler & Roth, 2008; Petter et al., 2007). Because we also model the first-order factors as reflective constructs (i.e., the measurement items for each first-order trust belief reflect that trust belief), we are examining reflective first-order and reflective second-order factors. This is consistent with other trust research that portrays trust dimensions as reflective first-order factors that reflect a second-order trust concept (McKnight et al., 2002a, Wang & Benbasat, 2005). Second-order factors are advantageous because they can explain the co-variation among the first-order factors in a more parsimonious manner (Law et al., 1998; Segars & Grover, 1998). However, researchers debate whether the general theories related to second-order factors have more utility than the more specific theories related to first-order factors, and whether second- 10 order factors have appropriate reliability and validity (see Edwards, 2001 for a detailed discussion of this debate). We propose that trust is a second-order factor based on trust theory (e.g., McKnight et al., 2002a). Further, we empirically analyze the appropriateness of alternative second-order factors based on prior research recommendations (Tanriverdi, 2006). Our first hypothesis predicts that the three technology trust beliefs will reflect a secondorder technology trust factor and the three interpersonal trust beliefs will reflect a separate second-order interpersonal trust factor (see Figure 1a). These second-order factors represent overall interpersonal and technology trust concepts. This hypothesis assumes that users probably view the technology trust beliefs as relating to the technology itself. By contrast, users probably view the interpersonal beliefs as relating to some kind of person-like characteristics of the technology (Wang & Benbasat, 2005). If this is true, it suggests that respondents are able to distinguish between interpersonal trust in Facebook and technology trust in Facebook. We believe Facebook respondents should be able to distinguish by these categories because they are able to perceive Facebook as either a technology (i.e., a website) or a quasi-person. H1: The three technology trust beliefs will reflect a second-order factor that is separate from the second-order factor reflected by the three interpersonal trust beliefs. [Insert Figure 1 Here] Our second hypothesis predicts that the pairs of conceptually related trust beliefs will reflect three second-order factors. We argued above that functionality is the technology analog of the interpersonal competence trust belief. That is, functionality is a specific perception that, for technology trust, is similar in nature to competence for interpersonal trust. If functionality is a technology trust instantiation of the interpersonal competence belief, then these beliefs will be significantly and highly correlated. The same is true of the integrity-reliability pair and the 11 benevolence-helpfulness pair because similar logic was used. Because we believe these three pairs are highly intracorrelated, we predict that another viable way to model these beliefs is to have the competence-functionality pair reflect one second-order trust factor, the integrityreliability pair reflect another second-order trust factor, and the benevolence-helpfulness pair reflect a third second-order factor (see Figure 1b). Literature evidence for this conceptual pairing comes from articles that try to synthesize the most important trusting beliefs from among the many trusting beliefs used. Mayer et al. (1995) distilled from the literature three beliefs—ability (similar to the competence-functionality pair), benevolence (like the benevolence-helpfulness pair), and integrity (like the integrityreliability pair). They showed in the literature that most of the oft-used types of trusting beliefs related to these three concepts. Similarly, McKnight et al.’s (2002a) Table 1 showed that most trusting beliefs could be clustered by meaning similarity into competence, integrity, and benevolence belief categories. H2: Competence-functionality, integrity-reliability, and benevolence-helpfulness will reflect three second-order factors that are distinct from each other (see Figure 1b). H1 and H2 may both be supported. Both ways of modeling these trust constructs may demonstrate good fit. But it is likely that one of them has better fit than the other. That is, the six trusting beliefs may be better modeled either with a technology-interpersonal split (as in Model 1) or with a conceptual split (as in Model 2). In effect, H1 and H2 offer alternative plausible ways of modeling how people perceive the six interpersonal and technology trusting beliefs. People tend to link like things and to separate unlike things (Levi-Strauss, 1968). H1 suggests trusting beliefs be modeled to separate interpersonal and technology categories regardless of conceptual type. H2 suggests they be modeled to distinguish among the three conceptual types 12 regardless of an interpersonal-technological distinction. We argued above for both ways of modeling these constructs. We now argue that Model 2 will be the better model. The more one gets to know another party, the better one is able to distinguish among their several characteristics (Lewicki et al. 1998). For example, one may trust one’s mother to fix dinner competently, but may not trust her to play a piano piece perfectly. On the other hand, we may not be able to distinguish such specific attributes in a stranger. Instead, we categorize the stranger into such broad categories as “competent” or “incompetent.” We come to know another by intensive interactive experience with them. The same is true with a technology; one comes to know it well through intense experience with it. We have found that the typical college-aged Facebook user has become very experienced with Facebook. In this study, subjects used Facebook every day on average, and they had used it for an average of 1.6 years. For this reason, they should be able to distinguish well between the competence, integrity, and benevolence attributes of Facebook. Early e-commerce studies have not always found this to be true (e.g., Bhattacherjee, 2002; Gefen et al. 2003). But Facebook users tend to have more experience with the website trustee than the respondents to early studies had with their e-commerce trustee. Further, we have defined the technology trust attributes to form matching pairs with the interpersonal trust attributes. That is, the functionality technology belief is very similar to the competence belief used in interpersonal trust. Reliability is a specific type of integrity, per McKnight et al. (2002a), who listed reliability as a component of their integrity cluster. Helpfulness is tied to benevolence, since it is a techno-form of the interpersonal benevolence concept. Because Model 2 reflects these close conceptual ties and because most Facebook users are experienced with it, we believe Model 2 will have a better fit than Model 1. 13 Another reason is that Model 1 makes a human-technology distinction that has often not been found in respondents’ minds in practice. For example, Wang and Benbasat (2005) found that subjects had no problem attributing interpersonal characteristics to a recommendation agent technology. Reeves and Nass (1996) generalized this finding across a number of technologies, finding that people comfortably attribute human-like attributes to technical artifacts. Hence, we think the Model 1 distinction between interpersonal trust and technology trust will be weaker than the Model 2 distinction between the three major conceptual types of trust. H3: Model 2 (distinguishing the three conceptual trusting belief types) will have a better fit than will Model 1 (distinguishing interpersonal from technology trusting beliefs). If H3 is true, then people relate less to Facebook’s human versus technology distinction than to Facebook’s competence, integrity, and benevolence distinction. This implies people are comfortable treating Facebook both as a technology and a quasi-person. METHODOLOGY We performed a survey in Fall 2006 to test the hypotheses. The survey used social networking websites as the target technology. The study participants were junior and senior business college students in a required introductory information systems course at a Midwestern U.S. public university. College students are an appropriate sample for investigating Facebook trusting beliefs because a sizeable percentage of Facebook users are college-aged. 40% of unique Facebook users were age 18-24 in 2006 and 29% were age 18-24 in 2007 (Lipsman, 2007). Procedure Of the 511 students enrolled in the course, 427 students (84%) completed the paper-based survey on a voluntary basis during class time. The survey measured the three technology trusting 14 beliefs, the three interpersonal trusting beliefs, reputation, privacy concern, ease of use, trusting intention, and continuance intention. The survey instructions asked subjects to indicate one social networking site in which they were currently a member or one site in which they might become a member. The survey then instructed subjects to answer all remaining questions referring to that social networking site, which the questions referred to as “MySNW.com.” Of the 427 responses, 362 both indicated Facebook as that social networking website and stated that they had previously used the site. These respondents were 54% male and on average 20 years old. We used this subsample to analyze the factor structure of Facebook users’ trusting beliefs. Testing the Hypothesized Second-Order Factor Models We tested the hypothesized second-order factor models in three ways. First, we assessed the convergent and discriminant validity of the first-order factors by performing a principal components analysis in SPSS, and analyzing the measurement model via a confirmatory factor analysis in EQS. Second, we ran a measurement model, also using EQS, for both hypothesized second-order factor structures. We assessed the appropriateness of the second-order factor structures following recommendations by Tanriverdi (2006). We also compared the measurement models by examining goodness-of-fit statistics and performing chi-squared difference tests. Third, we assessed the nomological validity of the hypothesized second-order factor structures by running structural equation models, using EQS. The structural model relationships we examine are based on trust theory (e.g., McKnight et al., 1998, 2002a). Specifically, we investigate reputation, privacy concern, and ease of use as antecedents to the second-order trust factors and trusting intention and usage continuance intention as direct and indirect consequents, respectively (see Figure 2). 15 [Insert Figure 2 Here] The first antecedent, reputation, means that one assigns attributes to another person based on second-hand information about the person (McKnight et al., 1998). For example, an individual may believe that another individual has a good reputation because their friends or coworkers have said good things about that person. If one has a good reputation another individual can develop trusting beliefs about that person even without first-hand knowledge (McKnight et al., 2002b). The second antecedent, privacy concern, means users believe the website will protect their personal information. Privacy concern can influence trusting beliefs because it makes the environment feel trustworthy. One important condition, especially in the Internet environment, is protection from the loss of privacy (McKnight et al., 2002a). Ease of use means that one perceives that using the website is free from effort. Gefen et al. (2003) posit that experiential factors about one’s use of a website can increase trusting beliefs. If one perceives the site is free from effort and easy to use, one will more likely ascribe positive attributes to the trustee (Gefen et al., 2003). Empirical research has confirmed the effects of these three antecedents on trusting beliefs (e.g., Al Abri, McGill & Dixon, 2009; Klein, 2006; Li, Rong & Thatcher, 2009; Gefen et al. 2003). We also examine how strongly the hypothesized trusting belief factor structures relate to trusting intention, which means a willingness to depend on the other person (McKnight et al., 2002a). Trusting beliefs relate positively to trusting intention because individuals with higher trusting beliefs will perceive the technology to have attributes that allow one to depend on it despite possible risks (McKnight et al., 2002a). Prior research has established this relationship in online contexts (e.g., Lowry et al., 2008; McKnight et al., 2002a, b). To complete our nomological network, we examine the association of trusting intention with usage continuance 16 intention or the intention to continue using the technology (i.e., Facebook.com) beyond an initial usage period. McKnight et al. (2002b) describes how being willing to depend on the trustor is a volitional preparedness to become vulnerable and is typically demonstrated by engaging in trusting behaviors. Because it is difficult to measure behaviors, many studies instead examine the relationship between trusting intention and other behavioral intentions. For example, ecommerce researchers have found that trusting intention influences intentions to make onsite purchases, share personal information with the site, and re-use web services (Jarvenpaa, Tractinsky & Vitale, 2000; Turel, Yuan & Connelly, 2008). Thus we predict that individuals with higher trusting intention will intend to continue using Facebook in the future. Measurement Scales The scales are shown in the Appendix. For the interpersonal trust beliefs, we adapted prior scales from McKnight et al. (2002a) and formatted them with headers like those of McKinney, Yoon & Zahedi (2002). We adapted the trusting intention items from McKnight et al. (2002a) and the ease of use items from Venkatesh and Davis (1996). For continuance intention we used items from Venkatesh et al. (2003) and made changes to reflect continuance of use. We also added one item that refers to continuing to use Facebook in the near future. We adapted the privacy concern items from the Smith et al. (1996) unauthorized secondary use and improper access privacy subscales. We also added an item pertaining to individual control over privacy. We measured functionality, helpfulness, reliability, and reputation using scales that we developed based on a pilot test using 233 students from the same course in a previous semester. We developed the pilot items to emphasize the constructs’ core meanings. For example, the helpfulness items relate to guidance and help, which is its core meaning. The reputation items relate to hearing favorable comments from others about using Facebook, which is its core 17 meaning. The pilot items had Cronbach’s alphas ranging from 0.88 to 0.96. We adopted all items directly from the pilot study to the current study except for two reliability items that we changed to refer more to the software’s consistency and accuracy. Because the pilot items referred to a different technology, we also re-worded all items to refer to “MySNW.com.” DATA ANALYSIS AND RESULTS First-Order Factor Measurement Model We first analyzed the fit and psychometric properties for the measurement model with the six trust beliefs, reputation, privacy concern, ease of use, trusting intention, and continuance intention to ensure that the first-order trusting beliefs are distinct. For goodness-of-fit a nonsignificant χ2 statistic can indicate the data fits the model well. However, as sample size increases the χ2 test has a tendency to be significant and as sample size decreases, it has a tendency to be non-significant (Schumaker & Lomax, 1996). Because our sample size is larger, we present both the χ2 value (and its significance level) and the χ2/df test. For the χ2/df test a value of 2 or less reflects good fit (Ullman, 2001), and 3 or less is acceptable (Kline, 1998). We also examine the non-normed fit index (NNFI), the comparative fit index (CFI), the root mean square error of approximation (RMSEA), and the Akaike information criteria (AIC). An NNFI and CFI of at least 0.95 is suggested for good fit (Hu & Bentler, 1999). Others suggest that this value is too stringent (Marsh, Hau & Wen, 2004), and many IS researchers consider 0.90 or above to be adequate fit (e.g., Gefen & Ridings, 2003; Rutner, Hardgrave & McKnight, 2008; Tanriverdi, 2006). Hu and Bentler (1999) also suggest that RMSEA should be 0.06 or lower to represent good fit, while others believe that values less than 0.05 indicate good fit, and values as high as 0.08 represent adequate fit (Browne & Cudeck, 1993). There are no suggested minimum or maximum values for the AIC, because this goodness-of-fit statistic is mainly used to compare 18 models. Lower AIC values (sometimes the values can even be negative) indicate better fitting and more parsimonious models (Byrne, 2006). The measurement model for the model with the six first-order trust beliefs (no secondorder factor structures), reputation, privacy concern, ease of use, trusting intention, and continuance intention has adequate fit despite the significance of the χ2 statistic (χ2 = 1552.18, p < .001). The NNFI was .942, the CFI was .948, the RMSEA was .051, the χ2/df was 1.93 (χ2 = 1552.18, df = 805), and the AIC was -57.82. These fit statistics are all close to or at suggested levels. To assess convergent validity for this model, we assessed the item factor loadings, the internal consistency reliability (ICR), Cronbach’s alpha (CA), and the average variance extracted (AVE). We used a principal components factor analysis in SPSS with a direct oblimin rotation because the trusting beliefs should be correlated, as others have found. This analysis shows that all items load on their own factor at more than the 0.70 standard (Fornell & Larcker, 1981) except the first reliability item, which loads at 0.62 and the first competence and benevolence items that both load at 0.68 (see Table 2). The ICR for each construct was greater than 0.80 and the AVE for each construct was greater than 0.50 (see Table 3), which are the recommended minimums (Fornell & Larcker, 1981). The CAs are also above the required 0.70 minimum (Hair, Anderson, Tatham & Black, 1998). These tests indicate adequate convergent validity. [Insert Tables 2 and 3 Here] We assessed discriminant validity using three methods. First, we analyzed the Lagrange Multiplier (LM) statistics in EQS. Second, we examined the principal components cross loadings in SPSS. Third, we compared the square root of the AVEs to each construct’s correlations with other constructs. The LM test showed several significant but minor model misspecifications. The 19 incremental LM χ2 values were relatively small, ranging from 3.90 to 25.28, and the corresponding standardized parameter change values (i.e., the parameters’ estimated values if freely estimated in a subsequent test of the model) were also small, ranging from .04 to .33. This test demonstrates that cross-loadings are not a problem in our data (Byrne, 2006), supporting discriminant validity. The principal components analysis also supports discriminant validity as the SPSS cross-loadings are well below the .30 suggested maximum (Hair et al., 1998) (see Table 2). Finally, for each factor, the square root of the AVE is greater than the correlations in that construct’s row or column (Chin, 1998) (see Table 3). The AVE itself is also greater than the correlations in that construct’s row or column for all factors. This latter test constitutes stronger evidence than the test comparing correlations with AVE square roots (Fornell & Larcker, 1981). All the above tests support discriminant validity. We performed one additional test of the discriminant nature of the six trusting beliefs. Mean difference tests between the conceptually related trust beliefs show that mean reliability is significantly different from mean integrity (t = 3.077, p < .01), and mean functionality is significantly different from mean competence (t = 11.073, p < .001). This test found no difference between helpfulness and benevolence (t = .313). Thus, not only are the first two pairs of constructs distinct, but they are also perceived at different mean levels. We next tested for multicollinearity and common method variance at the first-order factor level. We assessed multicollinearity by examining variance inflation factors and condition indexes. Variance inflation factors range from 1.31 to 2.15, which are well below suggested cutoffs of 10.00 (Hair et al., 1998) and 4.00 (Fisher & Mason, 1981). Also, condition indexes are under the suggested maximum of 30 (Belsley, Kuh & Welsch, 1980) suggesting that multicollinearity is not a problem in this data. We assessed common method variance by adding 20 a factor with all measures as indicators to the theorized model (Widaman, 1985). This model shows that the non-normed fit index improves only minimally (.012), and the original factor loadings are still significant (Elangovan & Xie, 1999). Therefore, we conclude that common method variance is not a problem in this data. Second-Order Factor Measurement Models We next analyzed and compared the hypothesized second-order factor measurement models. These models are shown in Figure 1a-b. We evaluated the appropriateness of using the second-order structures using tests recommended by Tanriverdi (2006). The first test was to ensure the first-order factors for each second-order construct are significantly correlated and of moderate to high magnitude. We find the correlations among the six trusting beliefs (r = 0.28 to 0.69) are all statistically significant at p < .05 and of moderate to high magnitude (see Table 3). The second test was to ensure the factor loadings of the first-order factors on the second-order factors are significant. We find that the loadings range from 0.52 to 0.90 and are all significant at p<.001 (see Table 4). The third test was to ensure the second-order measurement models have similar fit with the first-order measurement model, remembering that the goodness-of-fit of a higher-order model can never be better than that of its first-order model (Marsh & Hocevar, 1985). Table 5 presents the fit indices for the second-order measurement models we tested. With the exception of the significant χ2 statistic all the fit statistics are within or close to suggested guidelines. Also, these fit statistics are similar to those of the first-order measurement model. We also calculated the target (t)-coefficient, which is calculated as the χ2 of the first-order factor model divided by the χ2 of the second-order factor model. It tells how much variance in the firstorder factors the second-order factor explains and is an alternative way to compare the fit of a second-order model with that of the first-order model (Marsh & Hocevar, 1985). We find that the 21 target coefficient value for Model 1 is .88 and for Model 2 is .95 (see Table 5). These values suggest that the second-order factors explain a sufficient amount of variance in the first-order factors. In all, these tests demonstrate the appropriateness of the second-order factor models, and show support for H1 and H2. [Insert Tables 4 and 5 Here] Given that the second-order factor models are appropriate, we then compare their measurement models. We find that the H2 model (Model 2) is the better fitting model with a lower χ2/df statistic, a higher NNFI and CFI, and a lower RMSEA than the H1 model (Model 1) (see Table 5). The chi-square difference test and the AIC statistic also show that Model 2 is better fitting than Model 1 (see Table 5). Therefore H3 is supported. Structural Equation Models: Nomological Validity We then tested how both second-order factor models behave in a nomological network (see Figure 2). If the second-order constructs predict and are predicted by the same kinds of variables used in other studies, this lends additional support for Models 1 and 2. To do so, we ran three structural models. The first structural model used only the first-order factors, and had reputation, privacy concern, and ease of use as antecedents to the six trusting beliefs and trusting intention and continuance intentions as direct and indirect consequents to these beliefs, respectively. The other two structural models used second-order factors and thus were used to test the nomological validity of Models 1 and 2. The results are presented in Table 6. In the firstorder factor model, the antecedents all have significant relationships with the trusting beliefs except for the reputation-helpfulness, ease of use-integrity, and ease of use-benevolence relationships. Also, all first-order trusting beliefs significantly influence trusting intention except integrity. For second-order factor models to be appropriate, not all the first-order factor 22 relationships have to be significant (Tanriverdi, 2006). In fact, one benefit of using second-order factors is to predict better than the first-order factors. [Insert Table 6 Here] We find that for the second-order structural models, the antecedents significantly influence the second-order trust factors. The only exception is the non-significant effect of ease of use on the benevolence-helpfulness second-order factor in Model 2. In both models, the second-order trust factor(s) significantly predict trusting intention with the exception of the integrity-reliability factor in Model 2. Also, in both models, trusting intention significantly predicts usage continuance intentions. These results show that in general both hypothesized second-order factor structures have nomological validity. As a supplementary analysis, we included the direct paths from the trusting beliefs to continuance intention in Models 1 and 2 in addition to the indirect paths through trusting intention (see Table 7). We then tested for mediation of trusting intention on the trusting beliefs—continuance intention relationship following the Sobel method (Sobel, 1982, 1986), which calculates the significance of the indirect path. For Model 1, we find that the indirect paths from both interpersonal trust (β = .09, p < .01) and technology trust (β = .17, p < .001) are significant (see Table 7). Because in Model 1 technology trust has a significant direct effect on continuance intention (β = .28, p < .001) and interpersonal trust does not, the mediation of trusting intention is partial for technology trust and full for interpersonal trust. For Model 2, we find that the indirect path to continuance intention through trusting intention is significant for both benevolence-helpfulness (β = .09, p < .01) and competence-functionality (β = .14, p < .001). Because in this model only the competence-functionality trusting belief has a significant direct effect on continuance intention (β = .36, p < .001), trusting intention has a partial mediation 23 effect for competence-functionality and a full mediation effect for benevolence-helpfulness. Adding the direct paths from the trusting beliefs second-order factors to continuance intentions does not significantly change the other relationships reported in Table 6. DISCUSSION AND LIMITATIONS Because trust can be an important factor in social networking use, this study explored Facebook users’ trust beliefs to determine what it means to say one trusts Facebook. No research to-date has contrasted the interpersonal versus technological attributes comprising Facebook trust. Our study’s main objective was to explore two alternative second-order factor structures composed of both interpersonal and technology trust beliefs. Other research examining technology trust has examined either interpersonal trust attributes or technology trust attributes, but has not compared them in one study. Our Model 1 depicts interpersonal trusting beliefs and technology trusting beliefs as separate second-order constructs (H1), Model 2 depicts each trusting belief conceptual pair as a second-order construct (H2). We first tested and compared the second-order factor measurement models. Then we assessed how the second-order factors behave in a nomological network including reputation, privacy concern, ease of use, trusting intention, and usage continuance intention. Our findings contribute to research on trust, social networking, and information systems continuance in general. We now discuss these findings and provide both research and practice implications. Measurement Model Comparison Implications By testing and comparing the two alternative second-order factor measurement models, we find that both models have adequate (or close to adequate) fit. However, the measurement model with conceptually-related trusting beliefs (competence-functionality, integrity-reliability, and benevolence-helpfulness) as separate second-order factors (Model 2) is the better fitting 24 measurement model. We found that each of the six trusting beliefs types is discriminant from the others. So to this extent respondents distinguish the technology-related trust characteristics of Facebook from its human-like trust characteristics. However, they distinguish Facebook trusting characteristics based on the characteristics’ conceptual nature (Model 2) more than they do the human-versus-technology categories (Model 1). For example, our respondents think the website’s reliability models better with integrity than with helpfulness. This confirms the theory that technology reliability really is an integrity-related construct. Similarly, we find helpfulness models well with benevolence, and functionality with competence. Also, the fact that respondents distinguish between human- and technology trust to a lesser degree suggests that they are relatively comfortable attributing both human and technology attributes to Facebook. In fact, our findings suggest that users blend human demonstrations of trust with technology demonstrations of trust. This could be because users think of the Facebook website both as a technology and a quasi-person, even though it is a technical artifact. Future research will benefit from our findings by reflecting individuals’ Facebook trust as conceptually-related trust belief pairs, rather than as trust beliefs that distinguish between the technology-like and human-like characteristics. Each second-order construct not only reflects two conceptually-related trust beliefs, but also the interrelationships between them. Future research should verify whether our results translate to other social networking websites. For example, Dwyer et al. (2007) find that there are some differences in trust between Facebook and MySpace users. Also, because social networking website interfaces, privacy policies, and functionality change over time, and sometimes in response to user feedback, researchers should explore whether these second-order trust beliefs and their structures change over time. Structural Model Nomological Validity Implications 25 The two structural models we examine test the nomological validity of the alternative second-order structures. We find that all paths depicted in Figure 2 are generally significant. This implies that the Models 1 and 2 second-order factors behave as predicted by trust theory, further confirming their validity. Research could explore how Models 1 and 2 constructs behave in other trust-related models. For example, research could include trust beliefs in the extended privacycalculus model (Dinev & Hart, 2006) and the TAM-Trust model (Gefen et al., 2003). An important research implication is that the antecedents predicted the trusting beliefs differently, revealing something about the antecedents’ nature. Table 6, Model 1 column shows that reputation predicts interpersonal trust ( =.29) slightly better than technology trust ( =.23). Similarly, privacy concern predicts interpersonal trust ( =.45) better than it does technology trust ( =.25). This likely indicates that privacy concern and reputation are interpersonal in nature. By contrast, ease of use predicts technology trust ( =.49) better than it does interpersonal trust ( =.28). This makes sense because ease of use is about the technology, not the person. Table 6, Model 2 column shows reputation relates about equally to competencefunctionality ( =.29) and integrity-reliability ( =.28), but relates slightly less to benevolencehelpfulness ( =.21). This suggests that individuals seem to be more influenced by second-hand information related to Facebook’s ability and honesty rather than whether it acts in their best interest. Privacy concern clearly relates better to benevolence-helpfulness ( =.55) and integrityreliability ( =.43) than to competence-functionality ( =.16). This finding suggests that privacy concern increases trust mostly because it gives users increased confidence that the site will act in their best interest and be ethical. By contrast, our findings show that individuals relate ease of use more to competence-functionality ( =.58) than to integrity-reliability ( =.19) or 26 benevolence-helpfulness ( =.08). This makes privacy concern and ease of use complementary predictors. Overall our findings can help future researchers clarify which trust antecedents are most effective predictors of the specific trusting beliefs and their second-order factors. Further, these findings might be able to explain non-significant findings regarding these antecedents (e.g., Li et al., 2008). This research also contributes by presenting a mechanism (trusting intention) by which trusting beliefs in social networking websites increase continuance intention. Specifically, the findings show that individuals’ trusting beliefs make them willing to depend or make themselves vulnerable to the social networking website, which in turn increases future use. We did find that the integrity-reliability conceptual pair has a non-significant effect on trusting intention. This shows that it was perhaps hard for subjects to identify honesty, truthfulness, and consistency with the Facebook website and their willingness to become vulnerable to it. Our findings imply that users may be more influenced by certain trust-related conceptual pairs than others when taking risks with technologies. Future research should explore this finding further. Also, the finding that trusting intention only partially mediates the effects of technology trust is important. In the past, most research has used interpersonal trust beliefs, which are fully mediated in their effects on continuance intention. We find that technology trust was a more powerful predictor than interpersonal trust, and had a direct effect on Facebook continuance intentions, suggesting that researchers should consider its use in such studies. Similarly, the competence-functionality construct had a direct effect on continuance intention, while the other two conceptual second-order factors did not predict as well. These mediation findings imply that future research should include trusting intention when examining social networking trust beliefs 27 and continuance intentions. Future research could examine other possible belief-intention mediators such as attitudes. Another contribution our study makes is the development of three technology trust beliefs—functionality, reliability, and helpfulness that are derived from the three most common interpersonal trust beliefs—competence, integrity, and benevolence, respectively. While other researchers have developed technology trust beliefs (e.g., Lippert, 2001; Muir & Moray 1996), our study makes the conceptual linkage between trust beliefs based on how humans demonstrate the trust characteristics versus how technology demonstrates the trust characteristics. This process is important for ensuring that we do not ascribe human traits to a technology inappropriately. Future research should explore whether there are more technology trust beliefs than the three we examined. For example, interpersonal trust includes other trust beliefs, such as predictability (Rempel et al., 1985) and carefulness (Gabarro, 1978). Likewise, researchers may find that certain technologies demonstrate additional trust characteristics that users consider important when deciding to use a technology. Several limitations to the study exist. First, this study uses only one data set. Just because a model fits one sample or one technology does not imply that it fits all (Doll, Xia & Torkzadeh, 1994). There may be other technologies in which Model 2 is not the best fitting or most parsimonious model. For example, research involving technology that tends to have strong human-like characteristics (e.g., recommendation agents) may find Model 1 is the best model. Second, because we only examined usage intentions, our research does not specifically test the extent to which technology trust increases use of social networking websites. While technology trust has been shown to increase use of other technologies, this area needs more research. Third, we examine subjects from just one university. While these subjects appear to represent typical 28 Facebook users in terms of age at the time of the study, students at other colleges and universities, and older individuals may have different perceptions relating to the Facebook website. Also, more recent statistics show that the largest percent of Facebook users are people aged 35-54, and the highest growth age group is people age 55 plus (Corbett 2010). Future research should examine more diverse samples to enhance generalizability. Implications for Practice Our results have practical implications that are important for managers and developers who are trying to help social networking users deal with technology risk. While many factors exist that organizations should consider when developing software, we show that they should also try to make the software more trustworthy. In considering ways to do this, practitioners should keep in mind that Facebook users perceive the website as a technology and as a quasiperson. Hence, on the technology side, practitioners should consciously work on making the website use experience have more desired functionality, and be more reliable and more helpful to users. Website simplicity and following usability guidelines like those of Nielsen and Loranger (2006) will help. On the interpersonal side, practitioners should maintain a good reputation for integrity, competence and benevolence towards users. This can be done both by making the website use experience positive and by nurturing a positive image through press releases that emphasize these three qualities. Also, by respecting personal privacy issues and by putting user needs over profit motive, website firms can maintain a positive image in the public eye. Users also distinguish among the three conceptually similar trusting characteristics more than they distinguish the human-like trusting characteristics from the technology-like trusting characteristics. That is, each conceptual pair (e.g., competence-functionality) forms a tight duo. This implies that practitioners should focus on the conceptual types of trust regardless of whether 29 they reflect the technology’s human or technology characteristics. For example, the benevolencehelpfulness duo is about the principle of putting user needs first. This principle should be kept in mind in every change to the way business is done (organizationally) and in every change to the web site operation itself (technically). Every helpful website change may influence user beliefs about both helpfulness and benevolence. Every functional website improvement may influence both user functionality belief and competence belief. Practitioners should also be aware that certain trusting beliefs have more influence on trusting intentions than others. We find competence, functionality, and benevolence to be the most highly correlated to both trusting intention and continuance intention. When placed in second-order models, we find technology trust predicts trusting intention better than does interpersonal trust. We find competence–functionality to predict trusting intention better than benevolence-helpfulness, which predicts better than integrity-reliability. This implies that competence and functionality are the two most important trusting beliefs for Facebook users. Social network providers should try to emphasize these attributes in their design decisions by providing excellence and functional richness in how the website operates. If the social networking website provides the functionality that users want, these users will be more willing to depend on it to network socially online. Next in predictive power was benevolence, suggesting social network providers should show they have the interests of the user in mind. For example, social networking companies should show they wish to address user needs. They should emphasize the security aspects of their websites that can keep users safe from unwanted solicitation and/or identity theft. While there might be other factors that influence intentions to continue using a social networking website, our study finds that the more a user becomes willing to depend on a social 30 networking website, the more they are likely to continue using it (= 0.50***). Because some trust beliefs have direct effects on continuance intention (technology trust, functionalitycompetence) that are only partially mediated by trusting intention, these are also important for ensuring users continue to use the website. Managers and developers should also recognize the factors that are most likely to increase trust in social networking websites. For example, we find that privacy and ease of use have moderate to large influences on trust beliefs. However, these perceptions drive trust in different ways. Therefore, not only should practitioners focus on ensuring confidentiality of personal information, they should also ensure the website is usable. Competition among social networking websites can make these factors even more critical to maintaining memberships and sustaining growth. For example, users may be more willing to switch to another social networking website because they perceive lower risk with the new site due to its greater ease of use and its better privacy policies. Another practical implication is that because our study shows that a variety of trusting beliefs are important to Facebook users’ trusting intention and continuance intentions, it will be important for web developers and managers to monitor and track user trusting beliefs. They could collect this information by using online surveys or creating focus groups, or researching related blogs. The questions shown in the Appendix can be used. Also, managers could monitor whether or not the trusting beliefs and their influences change over time, and if they differ based on demographics like age or gender. CONCLUSION What does it mean to trust Facebook? This study contributes by showing that Facebook users trust the website as both a technology and a quasi-person. They appear to relate well to 31 Facebook both as a technology (in terms of functionality, reliability, and helpfulness beliefs) and as a “person” (in terms of competence, integrity, and benevolence beliefs). Model 2 performs better than Model 1. Thus, people distinguish more clearly between Facebook’s conceptual attributes (Model 2) than they do between Facebook’s personal versus technology nature (Model 1). This result suggests people conceive of Facebook as both a technology and a quasi-person. We also find certain trusting beliefs like competence and functionality predict trusting intention towards Facebook better than other beliefs. In addition, we find that technology trust beliefs predict Facebook use continuance intention better than do interpersonal beliefs. This research provides many opportunities for future researchers to extend these findings and further our understanding of what it means to trust a technology artifact like the Facebook. REFERENCES Al Abri, D., McGill, T., and Dixon, M. (2009). Examining the Impact of E-Privacy Risk Concerns on Citizens’ Intentions to Use E-Government Services: An Oman Perspective. Journal of Information Privacy & Security, 5(2), 3-26. Bausch, S., and Han, L. (2006). Podcasting Gains an Important Foothold among U.S. Adult Online Population. Nielsen/NetRatings Retrieved 9/2/2007, from http://www.nielsennetratings.com/pr/pr_060712.pdf. Belsley, D. A., Kuh, E., and Welsch, R. E. (1980). Regression Diagnostics: Identifying Influential Data and Sources of Collinearity, New York: John Wiley & Sons. Bhattacherjee, A. (2002). Individual Trust in Online Firms: Scale Development and Initial Trust, Journal of Management Information Systems, 19(1), 213-243. Browne, M. W., and Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen and J. S. Long (Eds.). Testing Structural Equation Models, 136-161. Newbury Park, CA: Sage. Byrne, B. M. (2006). Structural Equation Modeling with EQS, Mahwah, NJ: Lawrence Erlbaum Associates, Inc. Chin, W. (1998). The partial least squares approach to structural equation modeling. In G. A. Marcoulides (Ed.). Modern Methods for Business Research, 295-336. Mahwah, NJ: Lawrence Erlbaum Associates. 32 Corbett, P. (2010) Facebook demographics and statistics report 2010 – 145% growth in 1 year. Retrieved 05/20/10, from http://www.istrategylabs.com/2010/01/facebook-demographicsand-statistics-report-2010-145-growth-in-1-year/. Diamantopoulos, A., Riefler, P., and Roth, K. P. (2008). Advancing Formative Measurement Models. Journal of Business Research, 61, 1203-1218 Dinev, T., and Hart, P. (2006). An Extended Privacy Calculus Model for E-Commerce Transactions. Information Systems Research, 17(1), 61-80. Doll, W. J., Xia, W., and Torkzadeh, G. (1994). A Confirmatory Analysis of the End-User Computing Satisfaction Instrument. MIS Quarterly, 18(4), 453-461. Dwyer, C., Hiltz, S. R., and Passerini, K. (2007). Trust and Privacy Concern within Social Networking Sites: A Comparison of Facebook and MySpace. Proceedings of the Thirteenth Americas Conference on Information Systems, Keystone Colorado, August 9 -12. Edwards, J. R. (2001). Multidimensional Constructs in Organizational Behavior Research: An Integrative Analytical Framework. Organizational Research Methods, 4(2), 144-182. Elangovan, A. R., and Xie, J. L. (1999). Effects of Perceived Power of Supervisor on Subordinate Stress and Motivation: The Moderating Role of Subordinate Characteristics. Journal of Organizational Behavior, 20, 359-373. Fisher, J., and Mason, R. (1981). The analysis of multicollinear data in criminology. In J. Fox (Ed.). Quantitative Criminology. New York: Academic Press. Fogel, J., and Nehmad, E. (2009). Internet Social Network Communities: Risk Taking, Trust, and Privacy Concerns. Computers in Human Behavior, 25, 153-160. Fornell, C., and Larcker, D. F. (1981). Evaluating Structural Equations with Unobservable Variables and Measurement Error. Journal of Marketing Research, 18, 39-50. Friedman, B., Kahn, Jr., P. H., and Howe, D. C. (2000). Trust Online. Communications of the ACM, 43(12), 34-40. Gabarro, J. J. (1978). The development of trust, influence, and expectations.” In A. G. Athos and J. J. Gabarro (Eds.). Interpersonal Behavior: Communication and Understanding in Relationships, 290-303. Englewood Cliffs, NJ: Prentice-Hall. Gefen, D., Karahanna, E., and Straub, D. W. (2003). Trust and TAM in Online Shopping: An Integrated Model. MIS Quarterly, 27(1), 51-90. Gefen, D., and Ridings, C. M. (2003). IT Acceptance: Managing Use—IT Group Boundaries. Database for Advances in Information Systems, 34(3), 25-30. 33 Hair, J. F., Anderson, R. E., Tatham, R. L., and Black W. C. (1998). Multivariate Data Analysis, Upper Saddle River: Prentice Hall. Hu, L., and Bentler, P. M. (1999). Cutoff Criteria for Fit Indexes in Covariance Structure Analysis: Conventional Criteria versus New Alternatives. Structural Equation Modeling, 6(1), 1-55. Jarvenpaa, S. L., Tractinsky, N., and Vitale, M. (2000). Consumer Trust in an Internet Store. Information Technology and Management, 1(1-2), 45-71. Kennedy, M., and Sakaguchi, T. (2009). Trust in social networking: Definitions from a global, cultural viewpoint. In C. Romm-Livermore and K. Setzekorn (Eds.) Social Networking Communities and E-Dating Services: Concepts and Implications, 225-238. Hershey, PA: Information Science Reference. Kim, D. (2008). Self-Perception Based versus Transference-Based Trust Determinants in Computer-Mediated Transactions: A Cross-Cultural Comparison Study. Journal of Management Information Systems, 24(4), 13-45. Klein, R. (1996). Internet-Based Patient-Physician Electronic Communication Applications: Patient Acceptance and Trust. e-Service Journal, 5(2), 27-51. Kline, R. B. (1998). Principles and Practice of Structural Equation Modeling, New York: The Guildford Press. Komiak, S. Y. X., and Benbasat, I. (2006). The Effects of Personalization and Familiarity on Trust and Adoption of Recommendation Agents. MIS Quarterly, 30(4), 941-960. Law, K. S., Wong, C. S., and Mobley, W. H. (1998). Towards a Taxonomy of Multidimensional Constructs. The Academy of Management Review, 23(4), 741-755. Law, K. S., and Wong, C. S. (1999). Multidimensional Constructs in Structural Equation Analysis: An Illustration Using the Job Perception and Job Satisfaction Constructs. Journal of Management, 25(2), 143-160. Lee, J. D., and See, K. A. (2004). Trust in Automation: Designing for Appropriate Reliance. Human Factors, 46(1), 50-81. Lee, M. K. O., and Turban, E. (2001). A Trust Model for Consumer Internet Shopping. International Journal of Electronic Commerce, 6(1), 75-91. Levi-Strauss, L. (1968). The Savage Mind. Chicago: University Of Chicago Press, 1968. Lewicki, R. J., McAllister, D. J., and Bies, R. J. (1998) Trust and distrust: New relationships and realities. Academy of Management Review, 23(3), 438-458. 34 Li, X., Rong, G., and Thatcher, J. (2009). Swift Trust in Web Vendors: The Role of Appearance and Functionality. Journal of Organizational and End User Computing, 2191), 88-108. Lippert, S. K.(2001). An Exploratory Study into the Relevance of Trust in the Context of Information Systems Technology. PhD Dissertation, The George Washington University. Lippert, S. K. (2007). Investigating Post-adoption Utilization: An Examination into the Role of Interorganizational and Technology Trust. IEEE Transactions on Engineering Management, 54(3), 468-483. Lippert, S. K., and Swiercz, P. (2005). Human Resources Information Systems (HRIS) and Technology Trust. Journal of Information Science, 31(5), 340-353. Lipsman, A. (2007). Facebook Sees Flood of New Traffic from Teenagers and Adults, Comscore. Retrieved 8/14/2007, from http://www.comsocire.com/press/relaease.asp?press=1519. Lowry, P. B., Vance, A., Moody, G., Beckman, B., and Read, A. (2008). Explaining and Predicting the Impact of Branding Alliances and Web Site Quality on Initial Consumer Trust of e-Commerce Web Sites. Journal of Management Information Systems, 24(4), 199-224. MacKenzie, S. B., Podsakoff, P. N., and Jarvis, C. B. (2005). The Problem of Measurement Model Misspecification in Behavioral and Organizational Research and Some Recommended Solutions. Journal of Applied Psychology, 90(4), 710-730. Marsh, H. W., Hau, K. T., and Wen, Z. (2004). In Search of Golden Rules: Comment on Hypothesis Testing Approaches to Setting Cutoff Values for Fit Indexes and Dangers in Overgeneralizing Hu and Bentler’s (1999) Findings. Structural Equation Modeling, 11(3), 320-341. Marsh, H. W., and Hocevar, D. (1985). Application of Confirmatory Factor Analysis to the Study of Self-Concept: First and Higher Order Factor Models and Their Invariance across Groups. Psychological Bulletin, 97, 562-582. Mayer, R. C., Davis, J. H., and Schoorman, F. D. (1995). An Integrative Model of Organizational Trust. The Academy of Management Review, 20(3), 709-734. McKinney, V., Yoon, K. and Zahedi, F. M. (2002). The Measurement of Web Customer Satisfaction: An Expectation and Disconfirmation Approach. Information Systems Research, 13(3), 296-315. McKnight, D. H. (2005). Trust in Information Technology. In G.B. Davis (Ed.). The Blackwell Encyclopedia of Management, Management Information Systems, 7, 329-331. Malden, MA: Blackwell,. 35 McKnight, D. H., Choudhury, V., and Kacmar, C. J. (2002a). Developing and Validating Trust Measures for e-Commerce: An Integrative Typology. Information Systems Research, 13(3), 334-359. McKnight, D. H., Choudhury, V. and Kacmar, C. J. (2002b). The Impact of Initial Consumer Trust on Intentions to Transact with a Web Site: A Trust Building Model. Journal of Strategic Information Systems, 11(3), 297-323. McKnight, D. H., Cummings, L. L., and Chervany, N. L. (1998). Initial Trust Formation in New Organizational Relationships. Academy of Management Review, 23(3), 473-490. Muir, B. M., and Moray, N. (1996). Trust in Automation. Part II. Experimental Studies of Trust and Human Intervention in a Process Control Simulation. Ergonomics, 39(3), 429-460. Nielsen, J. and Loranger, H. (2006). Prioritizing Web Usability. Berkeley, CA: New Riders. Petter, S., Straub, D. W., and Rai, A. (2007). Specifying Formative Constructs in Information Systems Research. MIS Quarterly, 31(4), 623-656 Podsakoff, N., Shen, W., Podsakoff, P. M. (2006). The role of formative measurement models in strategic management research: Review, critique and implications for future research. In D. En, D. Ketcheny and D. Bergh (Eds.) Research Methodology in Strategic Management, 197252. Greenwich, CT: JAI Press. Reeves, B., and Nass, C. (1996). The Media Equation: How People Treat Computers, Television, and New Media like Real People and Places. Cambridge: Cambridge University Press. Rempel, J. K., Holmes, J. G., and Zanna, M. P. (1985). Trust in Close Relationships. Journal of Personality and Social Psychology, 49, 95-112. Rutner, P. S., Hardgrave, B. C., and McKnight, D. H. (2008). Emotional Dissonance and the Information Technology Professional. MIS Quarterly, 32(3), 635-652. Schumaker, R. E., and Lomax, R. G. (1996). A Beginner’s Guide to Structural Equation Modeling, Mahwah, New Jersey: Lawrence Erlbaum Associations. Segars, A. H., and Grover, V. (1998). Strategic Information Systems Planning Success: An Investigation of the Construct and its Measurement. MIS Quarterly, 22(2), 139-163. Sledgianowski, D., and Kulviwat, S. (2009). Using Social Network Sites: The Effects of Playfulness, Critical Mass and Trust in an Hedonic Context. The Journal of Computer Information Systems, 48(4), 74-83. Smith, H. J., Milberg, S. J., and Burke, S. J. (1996). Information Privacy: Measuring Individuals’ Concerns About Organizational Practices. MIS Quarterly, 20(2), 167-196. 36 Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structural equation models. In S. Leinhart (Ed.) Sociological Methodology, 290-312, San Francisco: JosseyBass. Sobel, M. E. (1986). Some new results on indirect effects and their standard errors in covariance structure models. In N. Tuma (Ed.) Sociological Methodology, 159-186, Washington D.C.: American Sociological Association. Straub, D., Boudreau, M. C., and Gefen, D. (2004). Valuation Guidelines for IS Positivist Research. Communications of the Association for Information Systems, 13, 380-427. Tanriverdi, H. (2006). Performance Effects of Information Technology Synergies in Multibusiness Firms. MIS Quarterly, 30(1), 57-77. Turel, O., Yuan, Y., and Connelly, C. E. (2008). In Justice we Trust: Predicting User Acceptance of e-Customer Services. Journal of Management Information Systems, 24(4), 123-151. Ullman, J. G. (2001). Structural equation modeling. In B. G. Tabachnick and L. S. Fidell (Eds.). Understanding Multivariate Statistics, 4th edition, 653-771. Nedham Heights, MA: Allyn and Bacon. Venkatesh, V., Morris, M. G., Davis, G. B., and Davis, F. D. (2003). User acceptance of information technology: Toward a unified view. MIS Quarterly, 27(3), 425-478. Venkatesh, V., and Davis, F. D. (1996). A Model of the Antecedents of Perceived Ease of Use: Development and Test. Decision Sciences, 27(3), 451-481. Wang, W., and Benbasat, I. (2005). Trust in and Adoption of Online Recommendation Agents. Journal of the Association for Information Systems, 6(3), 72-101. Widaman, K. (1985), Hierarchically Nested Covariance Structure Models for MultitraitMultimethod Data. Applied Psychological Measurement, 9(1), 1-23. 37 Figure 1. Hypothesized Second-Order Factor Models a) Model 1: H1 b) Model 2: H2 38 Figure 2. Structural Model: Nomological Validity Test 39 Table 1. Conceptual Origins of Technology Trust Beliefs Conceptual Origins Technology Trust Beliefs Trust in Information Systems Functionality The degree to which one anticipates the technology will have the functions or features needed to accomplish one’s task(s) Reliability The degree to which an individual anticipates the technology will continually operate properly, or will operate in a consistent flawless manner. Helpfulness The degree to which an individual anticipates the technology will provide adequate and responsive help. Reliability The technology is fully functioning and not experiencing system downtime when completing job related tasks (Lippert, 2001). Trust in Online Environments and Online Agents Trust in Automation Interpersonal Trust Competence The trustee has the ability, skills, and expertise to perform effectively in specific domains (Wang and Benbasat, 2005, p. 76). Competence The extent the technology performs its functions properly (Muir and Moray, 1996, p. 434) Ability (Competence) The group of skills, competencies, and characteristics that enable a party to have influence with some specific domain (Mayer et al., 1995, p. 717). Integrity An individual believes that the trustee adheres to a set of principles (Wang and Benbasat, 2005, p. 76). Reliability The extent the technology responds similarly to similar circumstances at different points in time (Muir and Moray, 1996, p. 434). Integrity The trustor's perception that the trustee adheres to a set of principles that the trustor finds acceptable (Mayer et al., 1995, p. 719). Benevolence The trustee cares about her and acts in her interests (Wang and Benbasat, 2005, p. 76). 40 Benevolence The extent to which the trustee is believed to want to do good to the trustor, aside from an egocentric motive Mayer et al., 1995, p. 718). Table 2. Factor Loadings* Item 1 2 3 4 5 6 7 8 9 10 11 Reliability 1 .62 .08 .09 .15 .11 .09 .14 .04 .12 .07 .03 Reliability 2 .82 .02 .01 .07 .05 .05 .04 .02 .13 .01 .01 Reliability 3 .83 .04 .02 .01 .05 .07 .02 .02 .07 .08 .07 Reliability 4 .89 .03 .05 .05 .07 .02 .02 .03 .08 .04 .03 Reliability 5 .86 .01 .05 .02 .09 .09 .01 .01 .09 .02 .02 Functionality 1 .12 .74 .01 .03 .00 .07 .12 ;07 .04 .02 .13 Functionality 2 .08 .80 .01 .09 .01 .03 .06 .03 .03 .10 .01 Functionality 3 .05 .79 .06 .01 .10 .01 .11 .11 .00 .03 .02 Functionality 4 .06 .75 .10 .03 .15 .01 .01 .00 .05 .01 .00 Helpfulness 1 .00 .02 .85 .01 .03 .01 .03 .08 .01 .06 .02 Helpfulness 2 .02 .02 .90 .03 .01 .02 .02 .03 .01 .04 .01 Helpfulness 3 .01 .05 .91 .02 .01 .01 .02 .03 .01 .01 .01 Helpfulness 4 .01 .02 .85 .08 .08 .05 .00 .07 .04 .05 .00 Integrity 1 .00 .06 .06 .89 .05 .01 .03 .04 .02 .01 .01 Integrity 2 .01 .02 .02 .94 .00 .03 .04 .01 .04 .02 .01 Integrity 3 .02 .00 .01 .86 .04 .07 .02 .01 .03 .00 .02 Integrity 4 .05 .06 .01 .81 .09 .10 .01 .01 .05 .03 .01 Competence 1 .02 .18 .05 .09 .68 .03 .10 .00 .12 .02 .01 Competence 2 .02 .01 .00 .10 .83 .01 .06 .01 .09 .03 .05 Competence 3 .03 .04 .01 .01 .83 .02 .07 .02 .06 .04 .03 Competence 4 .07 .09 .02 .02 .72 .14 .01 .01 .03 .08 .03 Benevolence 1 .04 .01 .05 .17 .11 .68 .04 .05 .10 .02 .11 Benevolence 2 .01 .00 .14 .00 .01 .85 .02 .01 .02 .04 .03 Benevolence 3 .01 .01 .02 .06 .00 .86 .00 .07 .02 .02 .00 Reputation 1 .01 .08 .00 .02 .09 .01 .83 .01 .08 .03 .04 Reputation 2 .02 .06 .02 .00 .05 .04 .85 .03 .03 .03 .02 Reputation 3 .03 .04 .07 .03 .09 .02 .86 .06 .06 .01 .00 Reputation 4 .08 .07 .02 .01 .06 .03 .85 .04 .02 .05 .02 Privacy concern 1 .01 .11 .05 .01 .01 .12 .10 .72 .01 .05 .01 Privacy concern 2 .02 .12 .07 .02 .12 .08 .12 .72 .05 .00 .06 Privacy concern 3 .01 .06 .02 .04 .06 .01 .08 .84 .10 .00 .09 Privacy concern 4 .01 .01 .02 .06 .05 .08 .03 .83 .09 .01 .01 Ease of Use 1 .02 .02 .02 .09 .01 .12 .04 .01 .83 .08 .05 Ease of Use 2 .03 .04 .06 .02 .17 .02 .02 .07 .81 .03 .09 Ease of Use 3 .07 .10 .06 .08 .08 .17 .05 .02 .71 .02 .01 Trusting Intention 1 .01 .10 .05 .03 .11 .03 .00 .00 .03 .84 .06 Trusting Intention 2 .01 .01 .01 .06 .06 .02 .02 .00 .02 .89 .01 Trusting Intention 3 .01 .03 .04 .05 .04 .01 .00 .02 .01 .97 .03 41 Trusting Intention 4 .04 .01 .03 .02 .04 .02 .02 .04 .02 .89 .03 Continuance Intention 1 .01 .02 .02 .01 .03 .01 .01 .00 .00 .00 .98 Continuance Intention 2 .00 .01 .02 .02 .02 .01 .01 .01 .04 .02 .95 Continuance Intention 3 .02 .00 .02 .00 .01 .00 .00 .01 .00 .02 .97 Continuance Intention 4 .03 .01 .03 .04 .00 .01 .01 .01 .01 .00 .97 * SPSS Principal Components Analysis with Direct Oblimin rotation 42 Table 3. Means, SDs, ICRs, CAs, AVEs, and Correlations among Latent Constructs Means SD ICR CA AVE 1 1. Reliability 4.62 1.11 .91 .91 .67 .82 2. Functionality 5.17 0.98 .90 .90 .69 .53 .83 3. Helpfulness 3.86 1.08 .93 .93 .76 .37 .41 .87 4. Integrity 4.42 1.22 .95 .95 .82 .51 .46 .38 .90 5. Competence 5.68 1.01 .93 .93 .78 .36 .65 .28 .39 .88 6. Benevolence 3.83 1.24 .88 .88 .71 .44 .40 .41 .69 .35 .84 7. Reputation 5.31 0.99 .90 .90 .70 .33 .45 .29 .41 .54 .35 .84 8. Privacy 4.21 1.46 .84 .84 .56 .27 .39 .27 .46 .35 .52 .39 .75 9. Ease of Use 5.68 .89 .86 .85 .68 .34 .53 .30 .30 .65 .20 .46 .33 .82 10. Trusting Intention 4.71 1.16 .94 .94 .81 .42 .53 .38 .43 .50 .46 .40 .37 .48 .90 11. Continuance Intention 5.47 1.31 .98 .99 .94 .18 .40 .18 .21 .36 .24 .31 .23 SD = standard deviation, ICR = internal consistency reliability, CA = Cronbach’s alpha, AVE = average variance extracted *Diagonal elements are the square roots of the AVE; off-diagonal elements are correlations between latent constructs. **All correlations are significant at p < .05 .46 .50 First-Order Factor 2 3 4 Table 4. First-Order Factor Loadings on Second-Order Factors* Model 1 (H1) Model 2 (H2) Loading Second-Order Factor Loading Second-Order Factor Reliability .70 Technology Trust .64 Integrity-Reliability Functionality .73 Technology Trust .90 Competence-Functionality Helpfulness .56 Technology Trust .52 Benevolence-Helpfulness Integrity .82 Interpersonal Trust .79 Reliability-Integrity Competence .55 Interpersonal Trust .72 Competence-Functionality Benevolence .77 Interpersonal Trust .78 Helpfulness-Benevolence * All factor loadings are significant at p < .001. 43 5 6 7 8 9 10 11 .97 Table 5. Model Fit Statistics Model Fit Statistics χ2 df χ2 / df NNFI CFI RMSEA First-Order Factor Model 650.42, p < .001 237 2.74 .935 .945 Second-Order Factor Model 1 (H1) 740.77, p < .001 245 3.02 .925 Second-Order Factor Model 2 (H2) 683.83, p < .001 243 2.81 .933 Model # .070 Target Coefficient 176.42 na χ2 Difference Test na .934 .075 250.77 .88 1 & 2, p < .001 .941 .071 197.83 .95 AIC Note: These are the fit statistics for measurement models with the trusting beliefs only. Their relative nature and the result of the χ2 difference test is similar to that for the measurement models with all the study variables (i.e., trusting beliefs, reputation, privacy concern, ease of use, trusting intention, and continuance intention). 44 Table 6. Nomological Validity Test First-Order Factor Model Second-Order Factor Model 1 (H1) Standardized Path Coefficients Rep Reliability .14* Rep Functionality .16** Rep Helpfulness .11 Rep Integrity .20*** Rep Competence .26*** Rep Benevolence .15* Rep Technology Trust Rep Interpersonal Trust Rep Integrity-Reliability Rep Competence-Functionality Rep Benevolence-Helpfulness Priv Reliability .23*** Priv Functionality .24*** Priv Helpfulness .23*** Priv Integrity .44*** Priv Competence .11* Priv Benevolence .55*** Priv Technology Trust Priv Interpersonal Trust Priv Integrity-Reliability Priv Competence-Functionality Priv Benevolence-Helpfulness EOU Reliability .25*** EOU Functionality .44*** EOU Helpfulness .21** EOU Integrity .11 EOU Competence .68*** EOU Benevolence .01 EOU Technology Trust EOU Interpersonal Trust EOU Integrity-Reliability EOU Competence-Functionality EOU Benevolence-Helpfulness Reliability TI .10* Functionality TI .22*** Helpfulness TI .12* Integrity TI .04 Competence TI .23*** Benevolence TI .19*** Technology Trust TI Interpersonal Trust TI Integrity-Reliability TI Competence-Functionality TI Benevolence-Helpfulness TI TI Continuance Intention .49*** Model Goodness-of-Fit Statistics and Variance Explained NNFI .923 CFI .929 RMSEA .059 Second OrderFactor Model 2 (H2) .23*** .29*** .28*** .29*** .21** .25*** .45*** .43*** .16** .55*** .47*** .22*** .19** .58*** .08 .49*** .28*** 45 .50*** .10 .43*** .28*** .50*** .924 .929 .058 .928 .933 .056 χ2/df AIC R2 Reliability R2 Functionality R2 Helpfulness R2 Integrity R2 Competence R2 Benevolence R2 Technology Trust R2 Interpersonal Trust R2 Integrity-Reliability R2 Competence-Functionality R2 Benevolence-Helpfulness R2 Trusting Intention R2 Continuance Intention * = p < .05, ** = p < .01, *** = p < .001 1864.70/ 832 = 2.24 200.70 23.4% 45.1% 18.2% 36.8% 53.9% 39.1% 1863.00/ 842 = 2.21 179.00 1801.00/ 838 = 2.15 125.00 56.2% 56.5% 38.2% 24.4% 48.7% 69.3% 49.2% 45.0% 25.2% 46.0% 25.0% Table 7. Supplementary Analysis: Effects of Trusting Beliefs on Continuance Intentions Direct Indirect (through Total Mediation (%) Trusting Intention) Model 1 Technology Trust .28*** .17*** .45*** Partial (27%) Interpersonal Trust -.04 .09** .05 Full Integrity-Reliability -.14 .04 -.10 None Competence-Functionality .36*** .14*** .50*** Partial (22%) Benevolence-Helpfulness .02 .11 Full Model 2 .09** * = p < .05, ** = p < .01, *** = p < .001 46 APPENDIX Technology Trusting Beliefs Functionality I believe MySNW.com is functional. It: 1. has the functionality I need. 2. has the features required for my online social activities. 3. has the ability to do what I want it to do. 4. has the overall capabilities I need. Reliability I believe MySNW.com is reliable. It: 1. is a very reliable website. 2. does not fail me. 3. is extremely dependable. 4. does not malfunction for me. 5. provides error-free results. Helpfulness I believe MySNW.com is Helpful. It: 1. supplies my need for help through a help function. 2. provides competent guidance (as needed) through a help function. 3. provides whatever help I need. 4. provides very sensible and effective advice, if needed. Interpersonal Trusting Beliefs Competence I believe MySNW.com is competent. It: 1. is competent and effective in providing online social networking. 2. performs its role of facilitating online social networking very well. 3. is a capable and proficient online social networking provider. 4. is very knowledgeable about online social networking. Integrity I believe MySNW.com has Integrity. It: 1. is truthful in its dealings with me. 2. is honest. 3. keeps its commitments. 4. is sincere and genuine. Benevolence I believe MySNW.com is Benevolent. It: 1. acts in my best interest. 2. does its best to help me if I need help. 3. is interested in my well-being, not just its own. 47 Reputation 1. Others have mentioned good things about using MySNW.com. 2. I have heard others speak favorably about using MySNW.com. 3. Other people have told me they are satisfied with using MySNW.com. 4. I have heard that most others are pleased with using MySNW.com Privacy concern 1. MySNW.com strives to keep my personal information confidential. 2. MySNW.com will never allow unauthorized access to my MySNW.com page. 3. No one except those I designate will ever be allowed to see what I am doing on my MySNW.com page 4. MySNW.com does a good job of letting me control my privacy . Ease of Use 1. Interacting with MySNW.com does not require a lot of my mental effort. 2. I find MySNW.com easy to use. 3. I find it easy to get MySNW.com to do what I want it to do. Trusting Intention (Measured on a 7-point Likert scale from (1) Not true at all to (7) Very true) 1. When I network socially online, I feel I can depend on MySNW.com. 2. I can always rely on MySNW.com for online social networking. 3. MySNW.com is a website on which I feel I can fully rely on when networking on the web. 4. I feel I can count on MySNW.com when networking online. Usage Continuance Intention (Measured on a 7-point Likert scale from (1) Not true at all to (7) Very true) 1. In the near future, I intend to continue using MySNW.com. 2. I intend to continue using MySNW.com. 3. I predict that I would continue using MySNW.com. 4. I plan to continue using MySNW.com. Note: Unless otherwise indicated all items measured on a 7-Point Likert scale from (1) Strongly disagree to (2) Strongly agree 48