Watching Me Watching You: How Observational Learning Affects Self-disclosure on Social Network Sites?

Watching Me Watching You: How Observational Learning Affects Self-disclosure on Social Network... Journal of Computer-Mediated Communication Watching Me Watching You: How Observational Learning Affects Self-disclosure on Social Network Sites? Tamar Ashuri Department of Communication, Tel Aviv University, Ramat Aviv, Israel Shira Dvir-Gvisman Department of Communication, Tel Aviv University, Ramat Aviv, Israel Ruth Halperin Oranim Academic College, Israel Many explanations have been proposed regarding people’s willingness to disclose information on social network sites (SNSs). Focussing on the reciprocal nature of such sites, this study explores the significant role observational learning (OL) plays in determining users’ willingness to self-disclose information on Facebook. It demonstrates how the ability to view other users’ actions—and the rewards and setbacks they encounter—impinge on their risk assessment and resulting disclosure behavior. Using an online survey of 742 Facebook users and an experiment conducted with 264 such participants, we demonstrated that users learn from others regarding self-disclosure behavior and resulting gains/ losses. We showed that the observation mechanism contributes to reward envy, that leads to a high level of self-disclosure behavior. By contrast, observation of risks has only a marginal effect on such undertakings. Keywords: Social Network Site (SNS), Facebook, Self-disclosure, Observational Learning (OL), Privacy. doi:10.1093/jcmc/zmx003 Introduction Every 60 seconds, 1.09 billion active Facebook users around the world post 510 comments, 293,000 statuses and 136,000 photos. On Twitter, 500 million tweets are submitted each day and billions of com- ments are shared on other social network sites (SNSs) (Zephoria digital marketing, 2017). This user- created information, that contains personal details, news, moods, and opinions, is critical to sustaining the Corresponding author: Tamar Ashuri; e-mail: tashuri@post.tsu.sc.il Editorial Record: First manuscript received on July 22, 2016. Revisions received on January 22, 2017, May 17, 2017, and September 13, 2017. Accepted by S. Shyam Sundar on October 27, 2017. Final manuscript received on November 08, 2017. First published online on 31 January 2018. 34 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You popularity and value of SNSs. Indeed, without massive production and consumption of personal informa- tion, SNSs would not be able to maintain the attention required to secure user loyalty (Chen, 2013). Several approaches have been offered to explain users’ motivation to disclose personal information on SNSs. Anticipation of benefits, such as enjoyment and social acceptance, have been suggested as incentives (e.g., Choi & Bazarova, 2015; Sledgianowski & Kulviwat, 2008), whereas perceived possibilities of loss, such as harassment, social rejection, browsing history tracking, third party use of personal data and identity theft, serve as deterrents (e.g., Acquisti, Brandimarte, & Loewenstein, 2015). Our overall motivation is improve- ment of our understanding of users’ willingness to disclose sensitive personal information on SNSs despite the substantial risks that stem from such behavior. More specifically, we aim to advance the current discus- sion on privacy concerns by tracing the early stages of decision-making processes and exploring how per- ceptions of gains and losses are learned and acquired.Tothisend,wepoint outthe significant role the reciprocal nature of SNSs plays in disclosure behavior. We argue that the ability to constantly monitor others and be monitored by others fosters social learning. Drawing on the seminal observational learning (OL) theory (e.g., Bandura, 1971b; Deutsch & Gerard, 1955; Simpson, Siguaw, & Cadogan, 2008), it is sug- gested that the visibility of other users’ self-disclosure behavior and the gains and setbacks they encounter in the process impinge on users’ risk assessment and by implication on their disclosure behavior. Theoretical framework and hypotheses: self-disclosure behavior in face-to face interactions and on SNSs Self-disclosure refers to the “process of making the self known to others” (Jourard & Lasakow, 1958, p. 91). SNSs offer many means of fostering and enhancing self-disclosure among users. While almost any action takenonan SNS (e.g., updating profile information, posting a status, photo, or video, clicking on commercial ads) directly or indirectly discloses information about the self (Tsay-Vogel, Shanahan, & Signorielli, 2016), in this study we relate to information SNS users intentionally reveal about themselves. Studies on self-disclosure—pre- and post-SNS—showed that such behavior encompasses various pur- poses, which dependent on the setting in which disclosure occurs (e.g., Antaki, Barnes, & Leudar, 2005; Jourard, 1971). Significantly, it has been noted that the decision to share personal information involves counterbalance between two conflicting elements: Anticipated gains and possibility of losses (e.g., Jourard, 1971). It may fulfill fundamental needs, such as social connectedness and belonging (e.g., Tamir & Mitchell, 2012), identity formation and enhancement of self-esteem (e.g., Hollenbaugh & Ferris, 2014), but it also bears potential losses: emotional distresses like jealousy in others’ supposedly “good life” life (Muise, Christofides, & Desmarais, 2009), and envy in the rewards they receive online (e.g., rewards like getting support in the form of receiving “Likes”), and other losses associated with undesired usage of the information disclosed that the discloser cannot fully control (Altman, 1975). To better understand disclosure decisions under conditions of risk and uncertainty, early information privacy scholars coined the term privacy calculus (Laufer & Wolfe, 1977). This metaphoric construct con- stitutes a cost-benefittradeoff analysis that accounts for the inhibitors and drivers that effect decisions regarding information disclosure. This construct was adopted by scholars who explored disclosure deci- sions in online interactions, primarily in ecommerce contexts (e.g., Dinev & Hart, 2004; Li, Sarathy, & Xu, 2010). More recently, it was used to explain users’ self-disclosure behavior on SNSs (chiefly Facebook). Here, scholars showed that perception of risk was negatively associated with self-disclosure on SNSs, while anticipated benefits were positively associated with it (e.g., Krasnova, Koroleva, Spiekermann, & Hildebrand, 2010; Dienlin & Metzger, 2016;see also McKnight, Lankton, & Tripp, 2011). While acknowledging the contribution of such works, we identify some weaknesses with regards to their conceptualization of cost-benefit analysis. Specifically, we highlight common tendency in the Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 35 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. literature to ask users to assess their general perceptions of SNS safety and possible gains (e.g., “overall, I see no real threat to my privacy due to my presence in an SNS” [Krasnova, Veltri, & Günther, 2012]; “Disclosing can improve my living and working efficiency” [Xu, Michael, & Chen, 2013]). Extensive research in risk assessment, however, indicates that people’s decision-making processes are far more com- plex (e.g., Sitkin & Pablo, 1992; Yates & Stone, 1992). When assessing both gains and losses, people consider two dimensions—the significance of a given loss/gain and its probability. For example, winning the lottery could result in a significant profit, but the low likelihood of winning deters people from participating: Uncertain consequences or outcomes that might follow one or more options should be evalu- ated and weighted in reaching an overall decision. Basically, the probability of each outcome is multiplied by the evaluation of that outcome and the score based on these products represents the overall utility of the option. This theory thus provides a model that serves as a rational or a normative basis for making decisions. (Van der Pligt, 2002, p. 258) This useful perspective was not incorporated in current studies on online self-disclosure. In recent studies, for example, Chang and Chen (2014) and Chang and Heo (2014) offered elaborate scales that mea- sured perceived benefits by using items such as “I feel that Facebook helps me interact with friends.” Nevertheless, the authors did not measure perceived importance of interactions with friends via Facebook. Instead, it appears that the study adopts a normative view according to which social ties are universally important. Overlooking the differences among users with regard to significance of specificgains (and losses), results in imprecise estimates of each user’s individual privacy calculus. The same shortcoming appears with regard to differences among scenarios. For example, although being hacked is clearly a worri- some situation, it is far less likely to occur than is receipt of comments from friends. By concentrating on magnitude alone, we risk misassessment of the perceived losses users ascribetothese twoscenarios. Having identified these weaknesses, we offer a new conceptualization of cost-benefit analysis in the context of self-disclosure (SD) on SNSs that includes both the likelihood and the magnitude of benefits and costs. We thus hypothesize: Hypothesis 1: Assessment of gains, the product of user-perceived magnitude and likelihood, will be associated with SD behavior: The more users expect gains by SD, the more they engage in self-disclosure behavior. Hypothesis 2: Assessment of losses, that is also the product of user-perceived magnitude and likelihood, will be associated with SD behavior: The more users anticipate losses by SD, the less they engage in self-disclosure behavior. The effects of social observation on technology perception and use As indicated, OL theory informs our predictions. In the 1960s and 1970s, Albert Bandura and his col- leagues demonstrated that social interactions constitute a significant component in learning processes that can accrue either intentionally or unintentionally. They showed that by observing the behavior of others and the benefits and losses they incur as a result of such actions, one forms an idea of the advis- ability of a particular behavior. Subsequently, this coded information serves as a guide for action (e.g., Bandura, 1971b). Learning accrues in cases in which the observed model is either actually present or only symbolically so (in mass media artifacts, for example). Significantly, media consumers can learn by viewing a specific role model or from their encounter with a “vast amount of information about human values, styles of thinking and behavior patterns (…) gained from the extensive modeling in the symbolic environment of the mass media” (Bandura, 2001, p. 271). Thus, in the context of mass 36 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You communication, consumers may also learn by watching behavior performed by numerous models, rather than by relying only on one specific model (e.g., Stefanone, Lackaff, & Rosen, 2010). OL scholarship highlights four dimensions that foster observation learning: Attention, retention, reproduction, and motivation (e.g., Bandura, 2009). We contend that this four-dimensional framework can be applied successfully to the online environment and especially to SNSs. We begin with the dimension of attention, positing that a particular behavior may be learned successfully only if one pays sufficient attention to relevant information regarding, norms, values, thinking patents, conducts, etc., gained from extensive modeling in the media’s environment (e.g., Ball-Rokeach & DeFleur, 1976; Bandura, 2001). It is well documented that the attention-drawing mechanism is a major component of popular SNSs (e.g., Hodas & Lerman, 2012). One outstanding example is SNSs algorithms, that attract users’ attention to others who share their values and attitudes, while making dissimilar users invisible (e.g., Ellison, Vitak, Gray, & Lampe, 2014; Pariser, 2011). Such a mechanism might reinforce learning, as OL scholars note repeatedly: When models are similar to the learner, the chances that the learner will be influenced by the model’s behavior increase (van de Bongardt, Reitz, Sandfort, & Deković, 2015). Another significant “attracting attention” feature is the first-in-first-out queue struc- ture of the platforms’ interfaces (e.g., “Time line” in Facebook and chorological queue in Twitter). The second dimension, retention, indicates that storing information in ways enabling subsequent access is crucial to the OL process. SNSs might promote retention because these sites constitute archives of social information (Margetts, John, Hale, & Yasseri, 2015) that can be retrieved at different points in time. For example, the individual user can easily utilize the research engine tool placed on top of a Facebook page. By typing a keyword in the designated “research window,” they can easily obtain requested information about network members’ past behaviors on the network, thereby learning their respective values and attendant benefits and costs. In reproduction, the learner gets to practice the observed behavior. SNSs are designed to enhance repetitive SD actions, because their sustainability depends on user disclosure of personal information. Several features enhance such activities. On illus- trative example is the ability to “comment on a comment” on one’s Facebook page, and clearly observe the chaining patterns. In Tweeter, it is the retweeting feature that facilitate and even encourage repro- duction. With regard to motivation, observed rewards (gains) and punishment (losses) either raise or lower the learner’s motivation to repeat the modeled behavior, as Bandura (1971b) argues: “Seeing others gain desired outcomes by their actions can create outcome expectancies that function as posi- tive incentives [referred to as perceived gains in this article]; observed punishing outcomes can create negative outcome expectancies that function as disincentives [referred to here as perceived losses]” (p. 276). SNS platforms allow users to “visually monitor” (Carroll & Bandura, 1982) their friends’ behav- ior and the outcomes of these actions. For example, a prime feature in Facebook is the “Like” function. Any Facebook users can easily view the number of “Likes” other people received for the posts they publish. In line with OL scholarship, SNS users will most likely be conform to the behavioral patterns they observe because they expect certain benefits or losses, such as social recognition or rejection, or an increase or decrease in social status. As SNS may constitute a conducive environment for OL, we aim to explore how the ability to observe other users on SNSs enables one to learn the advisability of the modeled behavior. We develop our hypotheses according to the three core elements users observe and from which they learn: Others’ behavior, and the gains—or alternatively, losses—they incur. Others’ SD behavior Following early theorization of OL, numerous studies documented the influence of observed behavior, stressing the significance of observing behavior performed by many (Bandura, 1971a). As noted by van de Bongardt et al. (2015): Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 37 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. The larger the number of peers who engage in a certain behavior, the more functional and cor- rect the behavior will be perceived to be and the more likely it is that adolescents will engage in the same behavior, based on the reasoning that if others are doing it, it is probably a good or wise thing to do the same. (p. 204) On the face of it, in SNSs such as Facebook, any user can, in principle, visually monitor the behav- ior of many network friends. Nevertheless, these platforms also embody a visibility constraint: Users who disclose more become more visible. Conversely, risk-averse users usually disclose less and hence are less visible. This may strengthen the perception of self-disclosure as the most common and there- fore most prudent course of action. The influence of others’ perceived behavior on users’ decisions has been documented in the con- text of online environments. For example, it was demonstrated repeatedly that in the domain of ecom- merce, consumers observe the purchase actions of previous individuals and tend to imitate them (e.g., Wang & Yu, 2017). Similarly, Burke, Marlow, and Lento (2009), who studied online distribution of news, demonstrated that Facebook users who see their friends contributing news content proceed to share more content of this kind themselves (for similar results see Cheung, Liu & Lee, 2015). Building on these insights, we posit: Hypothesis 3: Perceived SD of others will be positively associated with one’s own SD behavior. Others’ gains As indicated above, one significant element that impinges on the effectiveness of OL is the reward the observed model receives (Bandura, 1971a). In OL theory, if learners see that a model gains from a particular behavior, they develop the expectation that such behavior would afford them the same gains. In such instances, users might be motivatedtoperform theobservedbehavior. Although observed gain is described as a fundamental factor in OL, to the best of our knowledge, it has been overlooked in SD literature on SNSs. Recent studies acknowledged the importance of observing behavior of others, as well as the gain one receives directly from the learning process (Cheung et al., 2015), yet no research has addressed the link between the two components. Hence we propose the following: Hypothesis 4: Observed gains of others will be positively associated with one’s perception of gains to self. With regard to others’ losses, OL theory posits that punishments (or losses) incurred by others receive deter the learner from repeating the observed behavior (van de Bongardt et al., 2015). Recently, this idea was tested in a study on software piracy, in which scholars showed that people tend H4 Assessment of H1 Observed gains gains to self H3 Observed SD SD behavior behavior H5 Assessment of H2 Observed losses losses to self Figure 1 Suggested theoretical model. 38 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You to avoid software pirating when they see online that those who commit this offence suffer losses (e.g., Sun, Lim, & Fang, 2013). In accordance with these observations, we propose (Figure 1): Hypothesis 5: Observed losses of others will be positively associated with one’s assessment of losses. Study 1: survey design Method We conducted an online survey among Israeli Facebook users between 7 and 9 December 2015. Sample We used the service of an online survey company that holds an online panel of participants who receive gift vouchers in exchange for their participation in online surveys. For this study, a sample of 742 Israeli Facebook users was created. The sample is not representative of the population of Israel, but rather was designed to be representative of the Israeli population of Facebook. Marketing data show that this latter group is far younger than the population as a whole and comprises more men (53%) than women. Israeli Facebook users are also more secular than Israeli society as a whole. The sample constitutes a near- perfect match to data available with regards to age and gender composition of Israeli Facebook users (age ranging between 18 and 71 with M = 42.3 SD = 14.6; 54% male; 3% ultra-Orthodox; 45% holding an aca- demic degree; income: M = 3.0 SD = 1.2, on a 5-point scale). Measures For a full description of scale items and validation, see Appendix A. Dependent variable To measure SD behavior, we created two scales: Actions taken and content shared. For the first scale, we asked participants to rank how often they were involved in a list of activities (adapted from Tsay- Vogel et al., 2016; 9 items, Cronbach’salpha = .88), while the second inquired how often participants shared content concerning different aspects of their life (e.g., politics, consumption preferences, family, etc.; Cronbach’salpha = .88). In both cases, answers were expressed on a scale ranging from 1 (never) to 5 (more than once a day) (actions: M = 2.4, SD = .86; information sharing: M = 1.7, SD = .70). Independent variables Perceived SD of others. To measure participants’ perception of how much others disclose, we asked them to answer the same questions as above with regard to their Facebook friends (Actions: Cronbach’s alpha = .90; M = 4.0, SD = .94; information sharing: Cronbach’salpha = .93; M = 3.1, SD = 1.0). We measured significance and likelihood of both gains and losses (Yates & Stone, 1992). Perception of gains. To gauge perception of gains that could be attributed to SD, we created a list of pos- sible benefits (Xu et al., 2013). To assess their significance, we asked participants the following question: “To what extent do you consider the following to be advantages of Facebook use?” Answers were ranked on a scale from 1 (not an advantage at all) to 5 (a significant advantage) (6 items, Cronbach’salpha = .84; M = 3.3, SD = 1.0). To assess likelihood, we asked: “When you think about the list we mentioned earlier, to what extent do you believe you are able to make the most of these benefits?” Answers were ranked on a 1 to 5 scale, ranging between “notatall” to “very much” (6 items, Cronbach’salpha = .87; M = 3.2, SD = .96). Perception of losses. To measure losses possibly attributable to SD, we created a list of possible types of harm (Acquisti et al., 2015). As in the case of gains, we asked participants about magnitude, i.e., the extent to which they considered the listed events to be troubling: The answer scale ranged from 1 (not troubling at all) to 5 (very troubling) (7 items, Cronbach’s alpha = .87; M = 3.8, SD = .97). We also asked participants to evaluate the likelihood of each loss: “How likely is it that Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 39 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. such things will happen to you as a result of your Facebook use?” Answers were ranked on a 1 to 5 scale, ranging between “not at all” to “very” (7 items, Cronbach’s alpha = .91; M = 3.6, SD = .99). Observed losses and gains of others. To measure participants’ assessments of the gains others achieved by using Facebook, we asked: “In your opinion, how successful are your Facebook friends in making the most of these benefits?” using the same list as above. The response scale ranged from 1 (not at all successful) to 5 (very successful) (6 items, Cronbach’s alpha = .90; M = 4.0, SD = .83). To gauge participants’ evaluation of others’ losses, we also asked whether any of the above risks affected their Facebook friends. Responses reflected a 3-point scale: “Never,”“once” and “more than once” (ranging from 0 to 7, M = 3.3, SD = 2.8). Covariates Besides applying demographic criteria, the model used three questions to control for online behavior. Participants were asked: (a) if, and how often they browse the web, (b) if, and how often they access Facebook, and (c) if, and how often they access other SNSs, such as Instagram or Twitter. Answers ranged from 1 (never) to 5 (several times a day). Participants who reported they engage in any of these activities on a daily basis were then asked how many hours a day they spent doing so, with answers reflecting a 3-point scale: “Less than an hour,”“1–3 hours,” and “more than 3 hours.” In our sample, only 1% reported browsing and 11% reported using Facebook on a less than daily basis. Consequently, we only used the hours per day item to report web browsing behavior (for which the 1% that did not answer this question were coded as 0, M = 2.6, SD = .60). For Facebook and other SNS use, we created a combined 6-point scale based on the two measures (Facebook M = 4.8, SD = 1.0; other SNS M = 2.7, SD = 1.0). Past experience We also asked participants whether any of the above risks affected them, with responses ranging on a 3-point scale: “never,”“once,” and “more than once” (ranging from 0 to 7, self: M = 3.1, SD = 2.0). Statistical analysis As indicated above, this theoretical approach to cost-benefit analysis calls for a process of weighting the significance of each item in gain/loss scales according to its perceived likelihood ([Van der Pligt, 2002]; gains: M = 11.8, SD = 5.5; losses: M = 14.1, SD = 5.9, see Appendix B for distribution). The theoretical model described above requires mediation relations testing, for which we calculated Maximum Likelihood (ML) estimates, using IBM SPSS Amos 22.0 to assess the proposed model’s goodness of fit and test each of the hypotheses. To test mediation, we relied on bootstrapping (5,000 bootstrap samples) and following MacKinnon, Lockwood, Hoffman, West, and Sheets (2002) we opted for a bias-corrected confidence interval (Preacher & Hayes, 2008). Results Before testing our hypotheses, we note the correlations between the two dimensions mentioned— significance and likelihood of gains and losses—and SD behavior. In the case of gains, a strong pos- itive correlation was found between significance and likelihood of gains and SD (significance: acts: Pearson’s r = .49, P < .01, information: Pearson’s r = .35, P < .01; likelihood: acts: Pearson’s r = .55, P < .01, information: Pearson’s r = .41, P < .01). By contrast, in the case of losses, both diminutions were weakly correlated with SD behavior (significance: acts: Pearson’s r = .02, P = .75, information: Pearson’s r = -.07, P = .07). Importantly, likelihood was positively correlated with it (acts: Pearson’s r = .07, P = .08, information: Pearson’s r = .08, P = .07). We used Structural equation modeling (SEM) to test our hypotheses. The model was assessed with SD actions and sharing information creating a latent variable of SD behavior for both self and others (for full correlations table, see Appendix C). The theoretical model yielded satisfactory results 40 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Assessment of Observed losses Actions losses to self ** .46 SD behavior ** Assessment of Observed gains gains to self Information ** ** observed SD behavior Information Actions Covariates: Demographics and online usage, past experience Figure 2 SEM of SD behavior social learning. Notes: Parameters are standardized coefficients. ***p < .001, **p < .01, *p < .05; N = 742. in terms of goodness of fit indices. We obtained a chi-square to df ratio (CMIN/DF) of 1.58 and found that the model fit the data extremely well (χ = 55.4, df = 35, p = .02; RMSEA = .03, CFI = .987). Figure 2 shows that most of our hypotheses were confirmed. Insofar as gains are concerned, H1 and H4 were supported: observed gains were positively associated with gains assessment for self (H4), that were, in turn, positively associated with SD behavior. Observed gains had a positive and signifi- cant indirect effect on SD behavior (standardized estimates = .39; 95% C.I. [.21–.32]). As for losses, H5 was supported, with observed losses to others positively associated with assess- ment of loss to self. Contrary to H2, assessment of losses to self had no bearing on SD behavior. Finally, in keeping with H3, perceived SD of others behavior was positively associated with the partici- pants’ own activity in this respect. It is important to note that results of Study 1 are the product of a cross-sectional design. As such, they are open to alternative interpretations and do not establish a causal relation. Study 2: experimental design The primary advantage of our survey design is the ability to gather information on users’ perceptions of their social reality. The survey, however, cannot establish causality, but can only demonstrate the associa- tion among the variables. In the case at hand, there could be two alternative explanations to the observed correlations between perceptions of others’ behavior gains and losses and cost-benefit assessment for self. One option is that users learn from others (wherein the direction of causal relations is from others to self). Alternatively, it could be interpreted as a projection process in which one projects one’s own views on others. Consequently, we conducted an online experiment to demonstrate that OL does occur. We exposed participants to one of the most common gain and loss scenarios that Facebook users encounter: The sentiments expressed in responses to a given post (positive vs. negative). Method Participants Using the same recruitment platform, we sampled 512 participants for the experiment. Of these, 44 did not have a Facebook account and were therefore excluded. An additional 41 did not complete the Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 41 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Figure 3 Experimental stimuli. experiment, leaving 427 participants in the final sample (age ranged between 18 and 71 with M = 37 SD = 13.1; 59% female; 2% ultra-Orthodox; 42% with academic degrees; income: M = 2.8 SD = 1.2 on a 5-point scale). Manipulation Participants were divided randomly into four groups. Each was exposed to a segment of a Facebook page (see Figure 3) presenting a news feed in which a person shared a post and the reactions it received. In all cases, the post was identical, but the responses to it were manipulated. In the control case, there were no comments and 15 Likes. In the next three cases, we manipulated the type of responses, but held the amount of attention constant (that is, the number of comments and Likes were the same). In all cases, two comments were identical, one positive and one negative, while four additional comments varied in sentiment. The comments addressed four issues: Characteristics, post content, political issue described and public support regarding the issue (whether or not the post represents the opinion of many or few). The picture, user names of people commenting and length of comments were held constant across all cases. In the positive cases, the post received Likes and posi- tive emoticons, as well as supportive comments from other users. In the negative case, responses included negative emotions and comments. Finally, a mixed response case included both positive and negative emotions, along with three positive and three negative comments. Manipulation check. To ensure that participants noticed the manipulation, we included five state- ments at the end of the questionnaire and asked participants to indicate the extent of their agreement on a scale of 1 (disagree) to 5 (agree): (a) The comments to the post were unpleasant, (b) The com- ments to the post were supportive (excluding participants in the control, F 13.9, p < .01; (2,425) = F 32.0, p < .01), (c) It was important to post this content (F 5.5, p < .01), (d) The post (2,425) = (3,423) = was convincing (F 5.1, p < .01), and (e) The post was inappropriate (F 1.1, p = .40). In (3,423) = (3,423) = all but the last case, the differences among groups were significant and conformed to the anticipated directions (for means and std. errors, see Appendix D), suggesting that the manipulation worked. Measures. After exposure, participants were instructed to respond according to the scales described above for perceived losses (Cronbach’s alpha = .86; M = 3.8, SD = .90) and gains (Cronbach’s alpha = .86; M = 3.4, SD = .94) and were asked how likely they were to self-disclose, using the same items as earlier (Cronbach’s alpha = .86; M = 2.6, SD = .83). The order of scales was randomized across participants and ANOVA was conducted to test the hypothesis. Results Figure 4 presents the means and standard errors of all dependent variables, according to experimental case. As anticipated, participants in the positive case group reported a higher tendency to share when 42 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You 4.5 3.5 2.5 1.5 Self-disclosure Gains Risks Control Positive Mix Negative Figure 4 Means and standard errors: risk perception, gains perception and tendency towards SD behavior, by experimental case. Scale ranges from 1 (not at all) to 5 (a lot). using SNSs, while those in the negative case group reported a lower one (F = 5.5, p < .01; all dif- (3,424) ferences among groups are significant; for a full description of contrasts, see Appendix E). Perceptions of gains were also affected by the manipulation in the expected direction (F = 7.37, p < .01). (3,424) Participants in the positive case group perceived a greater number of advantages to self-disclosure behavior (in comparison to the negative case and the mixed case) while those in the negative case group reported the opposite (in comparison to the control group and positive case group). By contrast, there was no significant difference in perception of losses (F = .70, p = .55); con- (3,424) trary to expectations, participants in the negative case group did not demonstrate a higher perception of possible losses. We also observed correlations among the three dependent variables. In the current study, the correlation between SD tendency and perceived gains remains much stronger than the one between SD tendency and perceived losses (gains: r = .53, p < .001; losses: r = -.08, p = .20). We conjectured that OL shapes Facebook users’ perceptions of losses and gains, wherein users learn how much to share and the outcomes of various types of sharing behavior by observing others online. Most of these hypotheses were largely supported by empirical data: Surveys and experiments alike dem- onstrated that Facebook users learn from others regarding gains/losses and behavior. Moreover, per- ceived gains were positively correlated with SD behavior. Only partial confirmation was obtained concerning losses, however: While a significant association was found between perceived losses to self and to others, the applied manipulation did not alter perception and no effects on behavior were noted. Discussion Our study applied the privacy calculus perspective to assess the influence of OL on SD in SNS. We conceptualized a model that underscores three key mechanisms at work: (a) Facebook users’ associa- tion of losses and gains with SD on SNS, (b) their assumption that positive relations obtain in both cases, and (c) the powerful effect of OL on losses/gains perception and—by implication—on SD behavior. By focussing on OL’s role in determining SD perceptions and resulting behavior, this model adds a new dimension to existing studies emphasizing the effects of social influence on SD (e.g., Shibchurn & Yan, 2015). A core argument of these studies maintains that significant others are likely Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 43 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. to contribute to the formation of one’s intentions regarding SD behavior. For example, Shibchurn and Yan (2015) recently speculated that “[h]igh risk perceptions and low disclosure levels by significant others may lead individuals to disclose less” (p. 106). If significant others disclose information plenti- fully, users are likely to adopt such patterns and disclose more. To date, their hypothesis has not been tested empirically. We attempted to advance this discussion by focussing on aspects of social influence that ought to be accorded more attention in an age of networked media: rather than focussing on per- ceived social norms (what one believes others think), we addressed technological features that enable SNS users to retrieve information regarding others by viewing their conduct and the consequences thereof. We were thus able to shed new light on the complex effect of social influence of risk assess- ment on disclosure behavior. We discovered that when users perceive other Facebook users’ self-disclosure, together with the benefits they obtain thereby, they tend to believe that they too can leverage the platform to their own advantage. We showed that this conception—that can be summarized as “more self-disclosure = greater benefit”—encourages participants to intensify their SD behavior. The situation differs with regard to perception of losses. We found that assessment of losses has a marginal effect on users’ SD behavior. Although the first study (survey) demonstrated that users’ risk assessment was linked with their perceptions of other users’ losses, the second (experiment) did not replicate these results. In fact, manipulation influenced assessment of gains and not of losses. These findings are in line with existing scholarship indicating that although users ascribe high risk levels to SD (lack of safety, security, reli- ability and trustworthiness), they still disclose a considerable amount of apparently sensitive informa- tion (e.g., Acquisti & Gross, 2006; Taddicken, 2013). When predicting Facebook self-disclosure, contemporary scholars studying privacy calculus on SNSs noted that expected benefits have more pre- dictive power than perceived losses (Dienlin & Metzger, 2016). We suggest two reasons that might explain the difference between the predictive power of gains and losses on SD. First, as noted by Ellison et al. (2014) in their study of Facebook, this platform is designed to impel users to reward those who generate content, because the absence of response is a vague signal that might be interpreted as a lack of interest. In Ellison et al.’s words: SNS users must respond in a manner that leaves an observable marker of attention as a way of cultivating, or grooming, their connections. (…) On an SNS, explicitly responding to another user via activities that leave visible traces, such as commenting or clicking the “Like” button, is the most reliable way to indicate one has seen and attended to any individual piece of content on the site. (p. 858) Following this reasoning, gains may well become more apparent in users’ feeds, leading them to believe that other members of their network leverage the platform more than they do. We argue that this perception of the rewards other users are supposedly receiving is among the factors that motivate Facebook users to disclose information liberally. Another possible reason for the observed disparity concerns our approach to studying risk assess- ment in connection with SD on Facebook. Our findings suggest that in the case of perceived gains, the significance and likelihood of the gain have a similar effect on SD, but when it comes to losses, the pic- ture is more complex: The significance of a given loss is negatively correlated with SD, whereas the likelihood of a given loss is positively correlated with it. This may well reflect the complexity of risk- assessment as it applies to SD on Facebook: Those who attribute significant losses to SD on Facebook will be less likely to self-disclose, regardless of their perceived likelihood of these losses. By contrast, those who disclose more may realize that these activities increase their chances of being harmed, but they might not necessarily perceive the possible losses as significant. 44 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Limitations and future research directions Several aspects of the research design may well have affected the findings of this study. Measures, for example, may constitute a significant source of bias. To reinforce design, we supplemented the survey with an experiment. Although the resulting combination provides powerful methodical strength, alter- native designs could have been considered as well, such as a study of organic interactions on Facebook, analyzing the interplay between visible social information and behavior patterns, made pos- sible by the advent of unobtrusive behavioral measurement. Its advantages notwithstanding, this method also entails certain constraints, particularly ethical concerns: when monitoring SNS personal profiles of consenting participants, for example, exposure of research to content posted by others who provided no such consent is virtually inevitable. Furthermore, this study used an opt-in sampling method, wherein participants selected themselves. Obviously, a non-random sample of paid partici- pants would engender various biases. Considering the extremely low response rates prevailing at pres- ent, however, the method appeared particularly useful despite its limitations. Another constraint worth mentioning involved the characteristics of the scales used. This study comprises some scales that contain “and” and “or” clauses. For example, we asked participants if they share personal information on Facebook, but did not specify whether this information contains a phone number, an email address, a password, etc. Clearly, while users might be willing to share phone numbers, they would not necessarily agree to sharing passwords. Future studies could focus on specific SD behavior and nuance our scales. Related to this, is a possible discrepancy between the use of the word “importance” of given benefit users report to, and the “gain” they attribute to it. For example, getting “Likes” might be a big benefit of SD on SNSs, but a respondent might not see it as “important” since the word might be associated with issues that carry social importance. Future research can address this limitation by “fine-tuning” the measurement. Several additional factors may have exerted a substantive effect on results. For example, the sample was limited in two respects, as it comprised Israeli users alone and was tailored to one platform only— Facebook. Although Facebook is the most popular SNS in Israel (Alexa, 2017), the study should be replicated in other cultural/national settings in which users may have a different predisposition regarding self-disclosure. Such replication could also be useful in countries whose privacy regulations and concerns differ from those of Israel. Considering the well-documented dissimilarities among countries (see, e.g., Krasnova & Veltri, 2010; Ribak & Turow, 2003), one would expect users in coun- tries with different approaches to privacy to ascribe differential gain/losses levels to SD on SNS. Studies of this kind might provide more intensive insight into the effects of culture on cost-benefit analysis and resulting SD behavior. Similarly, future efforts could consider the manner in which desig- nated visibility features inherent in certain online platforms such as WhatsApp, Instagram,or LinkedIn affect users’ perceptions of others and their behavior patterns. Participants’ ages may also contribute to research bias. Our sample comprises users of different ages (see discussion in the Methodology section), as we sought to provide a representative (nation- wide) sample of Facebook users (to whatever extent possible—demographic information concerning Facebook users is not available to the public). It may be valuable, however, to conduct a comparative study with cohorts representing different age groups (Generation Y and the older Generation X and Baby Boomers). As an increasing number of studies show that privacy concerns among young people differ from the anxieties manifested by adults, subsequent studies could shed considerable light on the manner in which participant age affects the value ascribed to SNS losses and gains. In this study, we did not distinguish between strong tie and weak tie friends. Realizing that social tie strength might influence OL, it could be valuable to conduct a study that explores the extent to which it does so. Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 45 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. The study might also contribute to practice-led research. One issue that received massive attention by both researchers and practitioners is means of fostering privacy literacy in the context of online networked technologies, particularly on SNSs. The findings of the current study, that underscored the importance of reciprocal features of SNS, could assist in the construction of interventions that enhance users’ capability to control the effects of their disclosure behavior. At present, Facebook responds to this challenge by sending its users personal reminders to review their privacy settings. One possible intervention that emerged from this study is an interface design that provides a visual display of friends’ privacy-related decisions, such as the relevant privacy settings. This study contributes to a growing body of research on SD behavior by pointing to the significant role OL plays in determining users’ willingness to self-disclose personal information on SNSs. It shows that the observation mechanism contributes to reward envy, that leads, in turn, to a high level of online SD. Subsequent studies could apply the model to examine the effects of OL on various manifes- tations of human conduct in online spaces. Acknowledgments The project was funded by the Israel Science Foundation. Grant number 1385/13. Notes 1 See http://www.fialkov.co.il/english/facebook-audience-insights, accessed 23 December 2017. 2 Ultra-Orthodox Jews, who comprise 8.8% of the population over age 20, often abstain from Internet use for ideological and economic reasons. See CBS, Statistical Abstract of Israel 2013— No. 64, Subject 7, Table 7.1, Persons aged 20 and over by selected characteristics., accessed 2 September 2014. Because of language barriers, the Israeli Arab population was not included in the sample. 3 The order of scales was randomized across participants. 4 We chose to present a certain degree of social response, given that typically any post receives some attention, thus a post with no response at all may appear unrealistic. 5 Our findings show that users discern positive relationships between levels of self-disclosure and of gains/losses. When observing others’ gains, users appear to perceive self-disclosure engagements on Facebook as a double-edged sword, exacerbating personal vulnerability as it increases platform usefulness. References Acquisti, A., Brandimarte, L., & Loewenstein, G. (2015). Privacy and human behavior in the age of information. Science, 347(6221), 509–514. Acquisti, A., & Gross, R. (2006). Imagined communities: Awareness, information sharing and privacy on the Facebook. In Privacy enhancing technologies (pp. 36–58). Berlin Heidelberg: Springer. Alexa (2017). Available at https://www.alexa.com/topsites/countries/IL Altman, I. (1975). The environment and social behavior: Privacy, personal space, territory, and crowding. Monterey, CA: Brooks/Cole Publishing. Antaki, C., Barnes, R. and Leudar, I., (2005). Self-disclosure as a situated interactional practice. British Journal of Social Psychology, 44, 181–200. Ball-Rokeach, S., & DeFleur, M. (1976). A dependency model of mass media effects. Communication Research,3,3–21. 46 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Bandura, A. (1971a). Vicarious and self-reinforcement processes. In R. Glaser (Ed.), The nature of reinforcement (pp. 228–278). New York: Academic Press. Bandura, A. (2001). Social cognitive theory of mass communication. Media Psychology, 3(3), 265–299. Bandura, A. (Ed.). (1971b). Psychological modeling: Conflicting theories. Chicago: Aldine-Atherton. Bazarova, N. N., & Choi, Y. H. (2014). Self‐disclosure in social media: Extending the functional approach to disclosure motivations and characteristics on social network sites. Journal of Communication, 64(4), 635–657. van de Bongardt, D., Reitz, E., Sandfort, T., & Deković, M. (2015). A meta-analysis of the relations between three types of peer norms and adolescent sexual behavior. Personality and Social Psychology Review, 19(3), 203–234. Burke, M., & Marlow, C., & Lento T. (2009). Feed me: Motivating newcomer contribution in social network sites, CHI 2009 Social Networking Sites, 7 April 2009. Boston, MA, USA. Carroll, W. R., & Bandura, A. (1982). The role of visual monitoring in observational learning of action patterns: Making the unobservable observable. Journal of Motor Behavior, 14(2), 153–167. Chang, C. W., & Chen, G. M. (2014). College students’ disclosure of location-related information on Facebook. Computers in Human Behavior, 35, 33–38. Chang, C. W., & Heo, J. (2014). Visiting theories that predict college students’ self-disclosure on Facebook. Computers in Human Behavior, 30, 79–86. Chen, R. (2013). Living a private life in public social networks: An exploration of member self- disclosure. Decision Support Systems, 55(3): 661–668. Cheung, C. M., Liu, I. L., & Lee, M. K. (2015). How online social interactions influence customer information contribution behavior in online social shopping communities: A social learning theory perspective. Journal of the Association for Information Science and Technology, 66(12), 2511–2521. Choi, Y. H., & Bazarova, N. N. (2015). Self-disclosure characteristics and motivations in social media: Extending the functional model to multiple social network sites. Human Communication Research, 41(4), 480–500. Deutsch, M., &. Gerard, H. B. (1955). A study of normative and informational social influences upon individual judgment. The Journal of Abnormal and Social Psychology, 51(3), 629–636. Dienlin, T., & Metzger, M. J. (2016). An extended privacy calculus model for SNSs: Analyzing self‐ disclosure and self‐withdrawal in a representative US sample. Journal of Computer‐Mediated Communication, 21(5), 368–383. Dinev, T., & Hart, P. (2004). Internet privacy concerns and their antecedents-measurement validity and a regression model. Behaviour & Information Technology, 23(6), 413–422. Ellison, N. B., Vitak, J., Gray, R., & Lampe, C. (2014). Cultivating social resources on social network sites: Facebook relationship maintenance behavior and their role in social capital processes. Journal of Computer‐Mediated Communication, 19(4), 855–870. Gross, R., & Acquisti, A. (2005). Information revelation and privacy in online social networks. Proceedings of the 2005 ACM Workshop on Privacy in the Electronic Society (pp. 71–80). ACM. Hodas, N. O., & Lerman, K. (2012). How visibility and divided attention constrain social contagion. Privacy, Security, Risk and Trust (PASSAT), 2012 International Conference on and 2012 International Conference on Social Computing (pp. 249–257). IEEE. Hollenbaugh, E. E., & Ferris, A. L. (2014). Facebook self-disclosure: Examining the role of traits, social cohesion, and motives. Computers in Human Behavior, 30, 50–58. doi:10.1016/j.chb.2013.07.055 Jourard, S. M. (1971). Self-disclosure: An experimental analysis of the transparent self. Journal of Experimental Social Psychology, 11(3), 233–260. Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 47 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Jourard, S. M., & Lasakow, P. (1958). Some factors in self-disclosure. The Journal of Abnormal and Social Psychology, 56(1), 91. Krasnova, H., Koroleva, K., Spiekermann, S., & Hildebrand, T. (2010). Online social networks: Why we disclose. Journal of Information Technology, 25(2), 109–125. Krasnova, H., & Veltri, N. F. (2010). Privacy calculus on social networking sites: Explorative evidence from Germany and USA. System Sciences (HICSS), 2010 43rd Hawaii international conference on (pp. 1–10). IEEE. Krasnova, H., Veltri, N. F., & Günther, O. (2012). Self-disclosure and privacy calculus on social networking sites: The role of culture. Business & Information Systems Engineering, 4(3), 127–135. Laufer, R. S., & Wolfe, M. (1977). Privacy as a concept and a social issue: A multidimensional developmental theory. Journal of social Issues, 33(3), 22–42. Li, H., Sarathy, R., & Xu, H. (2010). Understanding situational online information disclosure as a privacy calculus. Journal of Computer Information Systems, 51(1), 62–71. MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A comparison of methods to test mediation and other intervening variable effects. Psychological Methods, 7(1), 83–104. Margetts, H., John, P., Hale, S., & Yasseri, T. (2015). Political turbulence: How social media shape collective action. Princeton, NJ: Princeton University Press. McKnight, D. H., Lankton, N., & Tripp, J. (2011). Social networking information disclosure and continuance intention: A disconnect. Paper presented at the System Sciences (HICSS), 2011 44th Hawaii International Conference on. Muise, A., Christofides, E., & Desmarais, S. (2009). More information than you ever wanted: Does Facebook bring out the green-eyed monster of jealousy? CyberPsychology & Behavior, 12(4), 441–444. Pariser, E. (2011). The filter bubble: What the internet is hiding from you. London, UK: Penguin. Preacher, K. J., & Hayes, A. F. (2008). Contemporary approaches to assessing mediation in communication research. In A. F. Hayes, M. D. Slater & L. B. Snyder (Eds.), The Sage sourcebook of advanced data analysis methods for communication research (pp. 13–54). http://dx.doi.org/10. 4135/9781452272054.n2 Ribak, R., & Turow, J. (2003). Internet power and social context: A globalization approach to Web privacy concerns. Journal of Broadcasting & Electronic Media, 47(3), 328–349. Shibchurn, J., & Yan, X. (2015). Information disclosure on social networking sites: An intrinsic– extrinsic motivation perspective. Computers in Human Behavior, 44, 103–117. Simpson, P. M., Siguaw, J. A., & Cadogan, J. W. (2008). Understanding the consumer propensity to observe. European Journal of Marketing, 42(1/2), 196–221. Sitkin, S. B., & Pablo, A. L. (1992). Reconceptualizing the determinants of risk behavior. Academy of Management Review, 17(1), 9–38. Sledgianowski, D., & Kulviwat, S. (2008). Social network sites: Antecedents of user adoption and usage. AMCIS 2008 Proceedings, 83. Stefanone, M. A., Lackaff, D., & Rosen, D. (2010). The relationship between traditional mass media and “social media”: Reality television as a model for social network site behavior. Journal of Broadcasting & Electronic Media, 54(3), 508–525. Sun, Y., Lim, K. H., & Fang, Y. (2013). Do facts speak louder than words? Understanding the sources of punishment perceptions in software piracy behavior. Paper presented at the Pacific Asia Conference on information Systems, Jeju Island, South Korea (p. 167). Taddicken, M. (2013). The “privacy paradox” in the social web: The impact of privacy concerns, individual characteristics, and the perceived social relevance on different forms of self‐disclosure. Journal of Computer Mediated Communication, 19(2), 248–273. 48 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Tamir, D. I., & Mitchell, J. P. (2012). Disclosing information about the self is intrinsically rewarding. Proceedings of the National Academy of Sciences of the United States of America, 109(21), 8038–8043. Tsay-Vogel, M., Shanahan, J., & Signorielli, N. (2016). Social media cultivating perceptions of privacy: A 5-year analysis of privacy attitudes and self-disclosure behaviors among Facebook users. New Media & Society. doi:10.1177/1461444816660731. Van der Pligt, J. (2002). Cognition and affect in risk perception and risky decision-making. Psychology at the Turn of the Millennium: Social, Developmental, and Clinical Perspectives, 2, 247–270. Wang, Y., & Yu, C. (2017). Social interaction-based consumer decision-making model in social commerce: The role of word of mouth and observational learning. International Journal of Information Management, 37(3), 179–189. Williams, S., & Voon, Y. W. W. (1999). The effects of mood on managerial risk perceptions: Exploring affect and the dimensions of risk. The Journal of Social Psychology, 139(3), 268–287. Xu, F., Michael, K., & Chen, X. (2013). Factors affecting privacy disclosure on social network sites: An integrated model. Electronic Commerce Research, 13(2), 151–168. doi:10.1007/s10660-013-9111-6 Yardi, S., Golder, S. A., & Brzozowski, M. J. (2009). Blogging at work and the corporate attention economy. Paper presented at the Proceedings of the SIGCHI conference on human factors in computing systems. Yates, J. F., & Stone, E. R. (1992). The risk construct. Oxford, England: Wiley & Sons. Zephoria Digital Marketing. (2017). Available from https://zephoria.com/top-15-valuable-facebook- statistics/ Zhang, J., & Daugherty, T. (2009). Third-person effect and social networking: Implications for online marketing and word-of-mouth communication. American Journal of Business,24(2), 53–64. Appendix A: Measurements Confirmatory factor analysis SD acts—self act1 Write status updates act2 Comment on other people’s status updates act3 Share other people’s content act4 Upload photos act5 Tag yourself in photos act6 Tag other people in photos act7 Participate in groups act8 Indicate your geographical location, for example by mentioning a name of a restaurant, an event, or a landmark act9 Indicate you intend to go to an event, or check in Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 49 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. N Min. Max. Mean Std. D. Skewness Kurtosis act1 744 1 5 2.36 1.22 .64 −.60 act2 744 1 5 3.26 1.32 −.20 −1.16 act3 744 1 5 2.86 1.35 .15 −1.20 act4 744 1 5 2.32 1.10 .79 −.09 act5 744 1 5 1.91 1.07 1.20 .79 act6 744 1 5 1.96 1.13 1.13 .45 act7 744 1 5 3.04 1.45 −.01 −1.38 act8 744 1 5 1.75 .97 1.46 1.79 act9 744 1 5 1.94 1.11 1.17 .60 Fit of model: χ = 34, df = 11, p = .46; RMSEA = .00, CFI = .997 act4 act6 act9 act1 act2 act3 act5 act8 act7 .77 .80 .81 .77 .80 .85 Acts 50 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You SD information-self info1 Share content regarding politics and current events, such as your opinions, political content you support, or news stories you found interesting info2 Share content regarding your consumer preferences and shopping habits, such as companies and products you like or dislike info3 Share information regarding your family, such as your relationship status, the identity of your partner, your parental status, or activities in which your children took part info4 Share information regarding your work, such as your workplace, your profession, or the people you work with info5 Share information regarding your friends, such as the identity of the people you hang out with info6 Share information regarding your current location, such as restaurants or events you are currently attending info7 Share information regarding your location in retrospect, such as restaurants or events you previously attended info8 Indicate where you live on your profile info9 Share personal information, such as phone numbers, ID numbers, or passwords, or send it in messages N Min. Max. Mean Std. D. Skewness Std. Error Kurtosis Std. Error info1 744 1 5 2.04 1.13 .94 .09 .04 .18 info2 744 1 5 1.78 1.04 1.31 .09 .99 .18 info3 744 1 5 1.7 .98 1.54 .09 1.92 .18 info4 744 1 5 1.71 .98 1.53 .09 1.88 .18 info5 744 1 5 1.78 .98 1.31 .09 1.19 .18 info6 744 1 5 1.69 .92 1.37 .09 1.44 .18 info7 744 1 5 1.72 .91 1.39 .09 1.74 .18 info8 743 1 5 1.85 1.16 1.46 .09 1.27 .18 info9 744 1 5 1.3 .78 2.99 .09 8.78 .18 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 51 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Fit of model: χ = 15.9, df = 26, p = .87; RMSEA = .00, CFI = .995 inf5 inf6 inf7 inf8 inf9 inf1 inf2 inf3 inf4 .72 .76 .68 .63 .72 .60 information SD acts—Others act_fr1 Write status updates act_fr2 Comment on other people’s status updates act_fr3 Share other people’s content act_fr4 Upload photos act_fr5 Tag themselves in photos act_fr6 Tag other people in photos act_fr7 Indicate their geographical location, for example by mentioning a name of a restaurant, an event, or a landmark act_fr8 Indicate they intend to go to an event, or check in act_fr9 Participate in groups N Min. Max. Mean Std. D. Skewness Std. Error Kurtosis Std. Error act_fr1 698 1 5 3.97 1.04 −.80 .09 −.22 .19 act_fr2 697 1 5 4.21 .97 −1.21 .09 .91 .19 act_fr3 693 1 5 3.98 1.04 −.87 .09 .01 .19 act_fr4 710 1 5 3.94 1.04 −.80 .09 −.15 .18 act_fr5 668 1 5 3.71 1.14 −.56 .10 −.65 .19 act_fr6 668 1 5 3.62 1.16 −.51 .10 −.67 .19 act_fr7 677 1 5 3.46 1.20 −.37 .09 −.84 .19 act_fr8 650 1 5 3.44 1.18 −.34 .10 −.84 .19 act_fr9 633 1 5 3.93 1.07 −.83 .10 −.10 .19 52 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Fit of model: χ = 37.5, df = 15, p = .07; RMSEA = .03, CFI = .981 act6 act9 act1 act2 act3 act4 act5 act8 act7 .77 . 80 .81 .77 .80 .85 Acts FR SD-information others info_fr1 Share their opinions regarding current events and political contents they support info_fr2 Share content regarding their consumer preferences and shopping habits, such as companies and products they like or dislike info_fr3 Share information regarding their family, such as their relationship status, the identity of their partner, their parental status, or activities in which their children took part info_fr4 Share information regarding their work, such as their workplace, their profession, or the people they work with (Continued) Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 53 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Continued info_fr5 Share information regarding their friends, such as the identity of the people they hang out with or how they hang out with them info_fr6 Share information regarding their current location, such as restaurants or events they are currently attending info_fr7 Share information regarding places they previously attended, such as restaurants and events info_fr8 Indicate where they live on their profile info_fr9 Share personal information, such as phone numbers, ID numbers, or passwords N Min. Max. Mean Std. D. Skewness Std. Error Kurtosis Std. Error info_fr1 688 1 5 3.37 1.17 −.28 .09 −.86 .19 info_fr2 669 1 5 3.06 1.26 .03 .09 −1.12 .19 info_fr3 681 1 5 3.03 1.25 .05 .09 −1.09 .19 info_fr4 676 1 5 2.91 1.28 .18 .09 −1.11 .19 info_fr5 690 1 5 3.3 1.19 −.16 .09 −1.01 .19 info_fr6 684 1 5 3.25 1.21 −.10 .09 −1.06 .19 info_fr7 692 1 5 3.21 1.17 −.10 .09 −.95 .19 info_fr8 644 1 5 2.89 1.37 .19 .10 −1.23 .19 info_fr9 611 1 5 1.96 1.34 1.12 .10 −.15 .20 54 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Fit of model: χ = 10.3, df = 16, p = .85; RMSEA = .00, CFI = .994 inf5 inf6 inf7 inf8 inf9 inf1 inf2 inf3 inf4 .85 .88 .80 .80 .88 .49 information SD-Gains self: adv1 Simply and rapidly obtaining information suiting my interests adv2 Obtaining likes and support from others adv3 Expressing myself—my opinions and feelings adv4 Maintaining and reinforcing social relations adv5 Getting personalized information, such as information about my close friends or about my fields of interest adv6 Effectively promoting myself professionally Significance N Min. Max. Mean Std. D Skewness Std. Error Kurtosis Std. Error adv1 744 1 5 3.64 1.23 −.63 .09 −.52 .18 adv2 744 1 5 3.13 1.35 −.18 .09 −1.13 .18 adv3 744 1 5 2.92 1.33 .01 .09 −1.15 .18 adv4 744 1 5 3.60 1.24 −.56 .09 −.63 .18 adv5 744 1 5 3.53 1.19 −.54 .09 −.48 .18 adv6 744 1 5 3.02 1.40 −.03 .09 −1.23 .18 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 55 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Fit of model: χ = 10.3, df = 5, p = .08; RMSEA = .03, CFI = .995 ad3 ad6 ad1 ad2 ad4 ad5 .70 .73 .71 .57 Gains SD-Gains self: Likelihood N Min. Max. Mean Std. D. Skewness Std. Error Kurtosis Std. Error adv1 743 1 5 3.49 1.17 −.37 .09 −.67 .18 adv2 743 1 5 3.24 1.27 −.20 .09 −.99 .18 adv3 743 1 5 2.96 1.31 .04 .09 −1.08 .18 adv4 743 1 5 3.38 1.23 −.36 .09 −.79 .18 adv5 743 1 5 3.40 1.20 −.31 .09 −.75 .18 adv6 743 1 5 2.47 1.29 .48 .09 −.84 .18 56 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Fit of model: χ = 5.39, df = 5, p = .14; RMSEA = .033, CFI = .998 ad3 ad6 ad1 ad2 ad4 ad5 .64 .80 .71 .79 .67 .71 Gains likelihood SD-Gains others: adv_fr1 Obtaining information suiting their interests adv_fr2 Obtaining likes and support from others adv_fr3 Publicly expressing themselves and their feelings adv_fr4 Reinforcing social relations adv_fr5 Getting personalized information, such as information about their close friends or about their fields of interest adv_fr6 Promoting themselves professionally N Min. Max. Mean Std. D. Skewness Std. Error Kurtosis Std. Error adv_fr1 579 1 5 4.07 .98 −.91 .10 .24 .20 adv_fr2 620 1 5 4.18 .95 −1.08 .10 .71 .20 adv_fr3 614 1 5 4.13 .99 −1.05 .10 .62 .20 adv_fr4 589 1 5 4.07 1.02 −1.02 .10 .53 .20 adv_fr5 575 1 5 3.98 1.00 −.77 .10 .04 .20 adv_fr6 560 1 5 3.63 1.19 −.51 .10 −.62 .21 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 57 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Fit of model: χ = 3.2, df = 5, p = .68; RMSEA = .00, CFI = .998 ad3 ad6 ad1 ad2 ad4 ad5 .78 .86 Gains others SD loss—self rsk1 Information I want to share with specific people will find its way to other people rsk2 People will share information about me without my consent, for example by tagging me in pictures rsk3 I will often be exposed to information I don’t care about rsk4 Commercial companies will use information I share to my disadvantage, for example by bombarding me with irrelevant advertisements rsk5 Commercial companies will sell my information to third parties without my knowledge or consent rsk6 The state will know more about me than I want it to rsk7 Hackers will steal my information Significance Min. Max. Mean Std. D. Skewness Std. Error Kurtosis Std. Error rsk1 1 5 3.61 1.35 −.62 .09 −.81 .18 rsk2 1 5 3.45 1.36 −.36 .09 −1.12 .18 rsk3 1 5 3.27 1.28 −.26 .09 −.94 .18 rsk4 1 5 3.94 1.20 −.96 .09 −.04 .18 rsk5 1 5 4.02 1.26 −1.10 .09 .02 .18 rsk6 1 5 3.43 1.37 −.39 .09 −1.05 .18 rsk7 1 5 4.04 1.32 −1.11 .09 −.09 .18 58 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Fit of model: χ = 8.0, df = 6, p = .24; RMSEA = .02, CFI = .997 rs7 rs1 rs2 rs3 rs4 rs5 rs6 1.0 .81 .95 .96 Losses significance Likelihood N Min. Max. Mean Std. D. Skewness Std. Error Kurtosis Std. Error rsk1 743 1 5 3.47 1.24 −.34 .09 −.91 .18 rsk2 743 1 5 3.56 1.25 −.46 .09 −.81 .18 rsk3 743 1 5 3.87 1.15 −.71 .09 −.47 .18 rsk4 743 1 5 3.9 1.19 −.87 .09 −.23 .18 rsk5 743 1 5 3.71 1.27 −.66 .09 −.66 .18 rsk6 743 1 5 3.46 1.30 −.41 .09 −.92 .18 rsk7 743 1 5 3.49 1.26 −.33 .09 −.98 .18 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 59 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. χ = 3.8, df = 6, p = .73; RMSEA = .00, CFI = .999 rs7 rs1 rs2 rs4 rs5 rs6 rs3 1.0 .93 .71 .97 .91 .92 Likelihood losses Losses—others rsk_fr1 Information they want to share with specific people will find its way to other people rsk_fr2 People will share information about them without their consent, for example by tagging them in pictures rsk_fr3 They will often be exposed to information they don’t care about rsk_fr4 Commercial companies will use information they share to their disadvantage, for example by bombarding them with irrelevant advertisements rsk_fr5 Commercial companies will sell their information to third parties without their knowledge or consent rsk_fr6 The state will know more about them than they want it to rsk_fr7 Hackers will steal their information N Min. Max. Mean Std. D. Skewness Std. Error Kurtosis Std. Error rsk_fr1 440 1 3 2.13 .72 −.20 .12 −1.06 .23 rsk_fr2 490 1 3 2.31 .69 −.50 .11 −.83 .22 rsk_fr3 509 1 3 2.51 .68 −1.07 .11 −.12 .22 rsk_fr4 454 1 3 2.34 .74 −.63 .12 −.92 .23 rsk_fr5 418 1 3 2.08 .79 −.14 .12 −1.37 .24 rsk_fr6 413 1 3 1.96 .77 .06 .12 −1.32 .24 rsk_fr7 433 1 3 1.85 .72 .23 .12 −1.04 .23 60 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You χ = 13.0, df = 9, p = .15; RMSEA = .02, CFI = .992 rs7 rs1 rs2 rs4 rs5 rs6 rs3 .89 .85 .80 .90 .81 .79 Losses others Construct validity To validate our measurements, we tested their correlations with related concepts on an additional sample, relying on previous research by Dienlin and Metzger (2016), who developed scales to measure Facebook benefits, privacy concerns, Facebook privacy self-efficacy, Facebook self-disclosure and Facebook self-withdrawal. Furthermore, to validate data about self-disclosure, we asked participants to provide us with their Facebook activity log and added an item measuring passive Facebook use (dem- onstrating distinction between self-disclosure and passive use). Sample We used an online sample of 261 participants, 124 of whom completed the survey, ranging in age between 19 and 64 with M = 40.5 SD = 12.7; 49% male; 3% ultra-Orthodox; 50% with academic degrees; income: M = 3.0 SD = 1.4 on a 5-point scale; participants were recruited using the same plat- form as the samples described in the article. Measures In addition, developing the above measures, we also asked participants about their Facebook benefits, privacy concerns, Facebook privacy self-efficacy, Facebook self-disclosure and Facebook self- withdrawal. All scales are taken from Dienlin and Metzger (2016) (full scales: https://osf.io/h4pef/? view_only=cf4c10222def4efdbecae20c1dca7dc6). Activity log data: Facebook allows users to download a log documenting all their activities. To pro- tect participants’ privacy, we did not ask them to disclose their log (containing a complete documenta- tion of photos, posts, etc.), but rather to report the size of the files downloaded (see Figure A1), based on the assumption that size is an indicator of amount of material shared. Nevertheless, it is important to note that 137 participants did not comply with this request, resulting in a high attrition rate. Furthermore, 22 participants failed to follow the instructions and thus supplied incorrect data. Seeking a uniform size indicator for all files (photos, posts, events, friends, etc.), we standardized each item (image files are much larger than text files, for example), ensuring that each participant was assigned an accurate relative score that could be compared with those of other participants. Passive use: To gauge passive use, we asked participants how often they read their Facebook friends’ posts (text and images) as part of the activity scale we developed to measure SD act frequency. Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 61 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Figure A1 Sample activity log data disclosed by participant. Results Table A1 presents the descriptive statistics of variables measured, while Table A2 displays the results of a factor analysis for all scales. As indicated, all items in our scales were loaded on one factor, except for the SD acts scale, for which a two-factor solution was identified: SD acts were loaded on one factor and another measuring passive use was loaded on another. Only one item displayed high loadings of both factors: Liking friends’ posts/images. The dual high score may be explained by the item’s rele- vance to self and others, as it measures the participant’s active use yet relies on observation by others as well. Table A3, that presents the correlation among the variables, provides additional validation for the measurements developed: Self-disclosure measurements were highly correlated with Dienlin and Matzear’s Facebook self-disclosure measurements, suggesting convergence validity. We also noted a positive correlation among perceived gains, Dienlin and Matzear’s Facebook benefits, self-disclosure and privacy concerns. Perceived losses were correlated with Facebook self-withdrawal and privacy concerns. 62 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Table A1 Descriptive Statistics of All Variables Mean Std. D. Skewness S.E. Kurtosis S.E. Cronbach’s α SD activity 2.43 .85 .66 .21 .11 .43 .89 SD information 1.78 .74 1.54 .23 3.10 .46 .88 Gains 3.50 1.00 −.34 .23 −.40 .46 .86 Risks 3.60 .95 −.48 .23 −.25 .46 .86 Privacy concerns (Dienlin and Matzear) 2.75 .79 −.35 .23 −.32 .46 .59 Facebook privacy self-efficacy (Dienlin 2.48 .96 .25 .23 −.63 .46 .88 & Matzear) Facebook benefits (Dienlin and 3.29 .86 −.37 .23 .63 .46 .90 Matzear) Facebook self-disclosure (Dienlin and 2.68 .94 .24 .23 −.26 .46 .86 Matzear) Facebook self-withdrawal (Dienlin and 1.69 .29 −1.07 .23 .54 .46 .61 Matzear) Activity log data .002 .66 1.04 .24 2.02 .47 .87 Type N Min. Max. Mean Std. D. Ads 101 2 62,919 9,547 10,773 Contact info 102 1.61 2,879,918 145,797 361,512 Events 102 2 20,187,833 329,373 2,032,835 Friends 102 2 191,275 13,994 28,121 Messages 101 1.01 43,798,254 3,173,460 6,615,034 Mobile device 102 1 1,538 858 681 Photos 102 1 65,703 6,687 11,083 Places 102 1 10,628 1,082 1,517 Synced photos 102 1 1,431 792 628 Timeline 102 3 8,214,581 569,150 1,050,921 Videos 102 1 42,768 4,305 7,629 Table A2 Exploratory Factor Analysis SD actions Item Component Write status updates .78 −.02 Comment on other people’s status updates .73 .38 Share other people’s content .73 .32 Upload photos .79 −.16 Tag yourself in photos .79 −.39 (Continued) Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 63 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Table A2 Continued SD actions Item Component Tag other people in photos .78 −.30 Participate in groups .63 .45 Indicate your geographical location, for example by mentioning a name of a restaurant, .76 −.36 an event, or a landmark Indicate you intend to go to an event, or check in .74 −.16 Read content uploaded by your friends .44 .62 SD information Item Component Share their opinions regarding current events and political contents they support .835 Share content regarding their consumer preferences and shopping habits, such as .814 companies and products they like or dislike Share information regarding their family, such as their relationship status, the identity of .778 their partner, their parental status, or activities in which their children took part Share information regarding their work, such as their workplace, their profession, or the .747 people they work with Share information regarding their friends, such as the identity of the people they hang .743 out with or how they hang out with them Share information regarding their current location, such as restaurants or events they .705 are currently attending Share information regarding places they previously attended, such as restaurants and .675 events Indicate where they live on their profile .62 Share personal information, such as phone numbers, ID numbers, or passwords .562 SD gains Significance Item Component Simply and rapidly obtaining information suiting my interests .687 Obtaining likes and support from others .845 Expressing myself—my opinions and feelings .818 Maintaining and reinforcing social relations .696 Getting personalized information, such as information about my close friends or about .794 my fields of interest Effectively promoting myself professionally .756 64 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Likelihood Item Component Simply and rapidly obtaining information suiting my interests .816 Obtaining likes and support from others .865 Expressing myself—my opinions and feelings .779 Maintaining and reinforcing social relations .733 Getting personalized information, such as information about my close friends or about .719 my fields of interest Effectively promoting myself professionally .631 SD losses Significance Item Component Information I want to share with specific people will find its way to other people .82 People will share information about me without my consent, for example by tagging me .714 in pictures I will often be exposed to information I don’t care about .523 Commercial companies will use information I share to my disadvantage, for example by .701 bombarding me with irrelevant advertisements Commercial companies will sell my information to third parties without my knowledge .823 or consent The state will know more about me than I want it to .779 Hackers will steal my information .781 Likelihood Item Component Information I want to share with specific people will find its way to other people .865 People will share information about me without my consent, for example by tagging me .815 in pictures I will often be exposed to information I don’t care about .734 Commercial companies will use information I share to my disadvantage, for example by .82 bombarding me with irrelevant advertisements Commercial companies will sell my information to third parties without my knowledge .816 or consent The state will know more about me than I want it to .717 Hackers will steal my information .775 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 65 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Table A3 Pearson’s Correlations for Measurements Gains- Gains- Loss- Loss- Activity Self- Acts Information sig lik sig sig log withdrawal Benefits SD Information .754** Gains-sig .487** .411** 121 121 121 Gains-lik .591** .516** .758** 121 121 121 Loss-sig .236** .239** .393** .433** 121 121 121 121 Loss-lik .301** .194* .419** .407** .586** 121 121 121 121 121 Activity log .274** .212* .349** .347** .219* .321** 102 102 102 102 102 102 Self-withdrawal .152 .159 .108 .165 .235* .221* .168 Dienlin and Matzear 119 119 119 119 119 119 100 Benefits .168 .212* .342** .404** −.011 −.007 −.022 .161 Dienlin and Matzear 121 121 121 121 121 121 102 119 SD .556** .573** .531** .622** .285** .210* .210* .145 .349** Dienlin and Matzear 121 121 121 121 121 121 102 119 121 Privacy con −.044 −.098 −.06 .008 .148 .341** .216* .198* −.275** −.118 Dienlin and Matzear 121 121 121 121 121 121 102 119 121 121 *p < .5; **p < .01 Appendix B: Distribution of Risk Assessment Measures 1 2 3 4 5 6 7 8 9 10 111213 141516 171819 2021 222324 25 Perceived Gains Perceived Losses Note: Percent refers to the percentage of participants for each scale. Perceived gains are measured as the likelihood of gain (on a scale of 1–5) weighted by its magnitude (on a 1–5 scale). Perceived losses measured as the likelihood of loss (on a scale of 1–5) weighted by its magnitude (on a scale of 1–5). 66 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Percent T. Ashuri et al. Watching Me Watching You Appendix C: Pearson’s Correlations Among Variables 12 345 678 9 10 11 2. Sharing information—self .665** 3. Sharing activity—others .278** .197** 774 724 4. Sharing information— .126** .321** .592** others 744 704 744 5. Browsing .231** .126** .058 .093* 732 742 732 704 6. Facebook use (hrs.) .473** .335** .103** .126** .425** 730 742 730 704 742 7.Other SNS use (hrs.) .344** .359** .254** .132** .218** .339** 240 742 240 704 742 742 8. Past damage—self .115** .222** .222** .119** .086* .06 .166** 742 742 742 704 742 742 742 9. Losses—others −.004 .166** .109** .130** .015 .048 .154** .447** 622 742 619 704 742 742 742 742 10. Losses—self .038 −.04 .063 .032 .009 −.029 .022 .281** .046 742 742 724 704 742 742 742 742 742 11. Gains—self .538** .391** .272** .158** .082* .369** .198** .036 .059 .076* 741 742 723 704 742 742 742 742 742 742 12. Perceived gains—others .201** .132** .400** .258** .046 .154** −.037 −.063 -.083* .095* .507** 616 632 616 623 632 632 632 632 632 632 632 Notes: *p < .5; **p < .01 Variables: (1) Sharing activity—self; (2) Sharing information—self; (3) Sharing activity—others; (4) Sharing infor- mation—others; (5) Browsing; (6) Facebook use in hours; (7) Other SNS use in hours; (8) Sustained damage in the past—self; (9) Losses—others; (10) Losses—self; (11) Gains –self; (12) Perceived gains—others. Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 67 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Appendix D: Manipulation Check 4.5 3.5 2.5 1.5 Comments unpleasant Comments supportive Important post Convincing post Post was in appropriate Control Positive Mix Negative Figure A Means and std. errors for post and comment evaluation, according to experimental condi- tion: 1 (disagree); 5 (agree). Appendix E Table 1 Results of Contrast between Experimental Conditions Contrast Value of Contrast Std. Error t Sig. Self-disclosure Control vs. Negative .34 .13 2.60 .01 Control vs. Positive −.31 .12 −2.57 .01 Mix vs. Negative .33 .13 2.48 .01 Mix vs. Positive −.32 .12 −2.55 .01 Negative vs. Positive −.32 .12 −2.55 .01 Gains Control vs. Negative .42 .17 2.42 .02 Control vs. Positive −.24 .16 −1.48 .14 Mix vs. Negative .25 .18 1.38 .17 Mix vs. Positive −.41 .17 −2.49 .01 Negative vs. Positive −.41 .17 −2.49 .01 68 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Computer-Mediated Communication Oxford University Press

Watching Me Watching You: How Observational Learning Affects Self-disclosure on Social Network Sites?

Free
35 pages

Loading next page...
 
/lp/ou_press/watching-me-watching-you-how-observational-learning-affects-self-5yIZQ9sHbR
Publisher
Oxford University Press
Copyright
© 2018 International Communication Association
eISSN
1083-6101
D.O.I.
10.1093/jcmc/zmx003
Publisher site
See Article on Publisher Site

Abstract

Journal of Computer-Mediated Communication Watching Me Watching You: How Observational Learning Affects Self-disclosure on Social Network Sites? Tamar Ashuri Department of Communication, Tel Aviv University, Ramat Aviv, Israel Shira Dvir-Gvisman Department of Communication, Tel Aviv University, Ramat Aviv, Israel Ruth Halperin Oranim Academic College, Israel Many explanations have been proposed regarding people’s willingness to disclose information on social network sites (SNSs). Focussing on the reciprocal nature of such sites, this study explores the significant role observational learning (OL) plays in determining users’ willingness to self-disclose information on Facebook. It demonstrates how the ability to view other users’ actions—and the rewards and setbacks they encounter—impinge on their risk assessment and resulting disclosure behavior. Using an online survey of 742 Facebook users and an experiment conducted with 264 such participants, we demonstrated that users learn from others regarding self-disclosure behavior and resulting gains/ losses. We showed that the observation mechanism contributes to reward envy, that leads to a high level of self-disclosure behavior. By contrast, observation of risks has only a marginal effect on such undertakings. Keywords: Social Network Site (SNS), Facebook, Self-disclosure, Observational Learning (OL), Privacy. doi:10.1093/jcmc/zmx003 Introduction Every 60 seconds, 1.09 billion active Facebook users around the world post 510 comments, 293,000 statuses and 136,000 photos. On Twitter, 500 million tweets are submitted each day and billions of com- ments are shared on other social network sites (SNSs) (Zephoria digital marketing, 2017). This user- created information, that contains personal details, news, moods, and opinions, is critical to sustaining the Corresponding author: Tamar Ashuri; e-mail: tashuri@post.tsu.sc.il Editorial Record: First manuscript received on July 22, 2016. Revisions received on January 22, 2017, May 17, 2017, and September 13, 2017. Accepted by S. Shyam Sundar on October 27, 2017. Final manuscript received on November 08, 2017. First published online on 31 January 2018. 34 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You popularity and value of SNSs. Indeed, without massive production and consumption of personal informa- tion, SNSs would not be able to maintain the attention required to secure user loyalty (Chen, 2013). Several approaches have been offered to explain users’ motivation to disclose personal information on SNSs. Anticipation of benefits, such as enjoyment and social acceptance, have been suggested as incentives (e.g., Choi & Bazarova, 2015; Sledgianowski & Kulviwat, 2008), whereas perceived possibilities of loss, such as harassment, social rejection, browsing history tracking, third party use of personal data and identity theft, serve as deterrents (e.g., Acquisti, Brandimarte, & Loewenstein, 2015). Our overall motivation is improve- ment of our understanding of users’ willingness to disclose sensitive personal information on SNSs despite the substantial risks that stem from such behavior. More specifically, we aim to advance the current discus- sion on privacy concerns by tracing the early stages of decision-making processes and exploring how per- ceptions of gains and losses are learned and acquired.Tothisend,wepoint outthe significant role the reciprocal nature of SNSs plays in disclosure behavior. We argue that the ability to constantly monitor others and be monitored by others fosters social learning. Drawing on the seminal observational learning (OL) theory (e.g., Bandura, 1971b; Deutsch & Gerard, 1955; Simpson, Siguaw, & Cadogan, 2008), it is sug- gested that the visibility of other users’ self-disclosure behavior and the gains and setbacks they encounter in the process impinge on users’ risk assessment and by implication on their disclosure behavior. Theoretical framework and hypotheses: self-disclosure behavior in face-to face interactions and on SNSs Self-disclosure refers to the “process of making the self known to others” (Jourard & Lasakow, 1958, p. 91). SNSs offer many means of fostering and enhancing self-disclosure among users. While almost any action takenonan SNS (e.g., updating profile information, posting a status, photo, or video, clicking on commercial ads) directly or indirectly discloses information about the self (Tsay-Vogel, Shanahan, & Signorielli, 2016), in this study we relate to information SNS users intentionally reveal about themselves. Studies on self-disclosure—pre- and post-SNS—showed that such behavior encompasses various pur- poses, which dependent on the setting in which disclosure occurs (e.g., Antaki, Barnes, & Leudar, 2005; Jourard, 1971). Significantly, it has been noted that the decision to share personal information involves counterbalance between two conflicting elements: Anticipated gains and possibility of losses (e.g., Jourard, 1971). It may fulfill fundamental needs, such as social connectedness and belonging (e.g., Tamir & Mitchell, 2012), identity formation and enhancement of self-esteem (e.g., Hollenbaugh & Ferris, 2014), but it also bears potential losses: emotional distresses like jealousy in others’ supposedly “good life” life (Muise, Christofides, & Desmarais, 2009), and envy in the rewards they receive online (e.g., rewards like getting support in the form of receiving “Likes”), and other losses associated with undesired usage of the information disclosed that the discloser cannot fully control (Altman, 1975). To better understand disclosure decisions under conditions of risk and uncertainty, early information privacy scholars coined the term privacy calculus (Laufer & Wolfe, 1977). This metaphoric construct con- stitutes a cost-benefittradeoff analysis that accounts for the inhibitors and drivers that effect decisions regarding information disclosure. This construct was adopted by scholars who explored disclosure deci- sions in online interactions, primarily in ecommerce contexts (e.g., Dinev & Hart, 2004; Li, Sarathy, & Xu, 2010). More recently, it was used to explain users’ self-disclosure behavior on SNSs (chiefly Facebook). Here, scholars showed that perception of risk was negatively associated with self-disclosure on SNSs, while anticipated benefits were positively associated with it (e.g., Krasnova, Koroleva, Spiekermann, & Hildebrand, 2010; Dienlin & Metzger, 2016;see also McKnight, Lankton, & Tripp, 2011). While acknowledging the contribution of such works, we identify some weaknesses with regards to their conceptualization of cost-benefit analysis. Specifically, we highlight common tendency in the Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 35 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. literature to ask users to assess their general perceptions of SNS safety and possible gains (e.g., “overall, I see no real threat to my privacy due to my presence in an SNS” [Krasnova, Veltri, & Günther, 2012]; “Disclosing can improve my living and working efficiency” [Xu, Michael, & Chen, 2013]). Extensive research in risk assessment, however, indicates that people’s decision-making processes are far more com- plex (e.g., Sitkin & Pablo, 1992; Yates & Stone, 1992). When assessing both gains and losses, people consider two dimensions—the significance of a given loss/gain and its probability. For example, winning the lottery could result in a significant profit, but the low likelihood of winning deters people from participating: Uncertain consequences or outcomes that might follow one or more options should be evalu- ated and weighted in reaching an overall decision. Basically, the probability of each outcome is multiplied by the evaluation of that outcome and the score based on these products represents the overall utility of the option. This theory thus provides a model that serves as a rational or a normative basis for making decisions. (Van der Pligt, 2002, p. 258) This useful perspective was not incorporated in current studies on online self-disclosure. In recent studies, for example, Chang and Chen (2014) and Chang and Heo (2014) offered elaborate scales that mea- sured perceived benefits by using items such as “I feel that Facebook helps me interact with friends.” Nevertheless, the authors did not measure perceived importance of interactions with friends via Facebook. Instead, it appears that the study adopts a normative view according to which social ties are universally important. Overlooking the differences among users with regard to significance of specificgains (and losses), results in imprecise estimates of each user’s individual privacy calculus. The same shortcoming appears with regard to differences among scenarios. For example, although being hacked is clearly a worri- some situation, it is far less likely to occur than is receipt of comments from friends. By concentrating on magnitude alone, we risk misassessment of the perceived losses users ascribetothese twoscenarios. Having identified these weaknesses, we offer a new conceptualization of cost-benefit analysis in the context of self-disclosure (SD) on SNSs that includes both the likelihood and the magnitude of benefits and costs. We thus hypothesize: Hypothesis 1: Assessment of gains, the product of user-perceived magnitude and likelihood, will be associated with SD behavior: The more users expect gains by SD, the more they engage in self-disclosure behavior. Hypothesis 2: Assessment of losses, that is also the product of user-perceived magnitude and likelihood, will be associated with SD behavior: The more users anticipate losses by SD, the less they engage in self-disclosure behavior. The effects of social observation on technology perception and use As indicated, OL theory informs our predictions. In the 1960s and 1970s, Albert Bandura and his col- leagues demonstrated that social interactions constitute a significant component in learning processes that can accrue either intentionally or unintentionally. They showed that by observing the behavior of others and the benefits and losses they incur as a result of such actions, one forms an idea of the advis- ability of a particular behavior. Subsequently, this coded information serves as a guide for action (e.g., Bandura, 1971b). Learning accrues in cases in which the observed model is either actually present or only symbolically so (in mass media artifacts, for example). Significantly, media consumers can learn by viewing a specific role model or from their encounter with a “vast amount of information about human values, styles of thinking and behavior patterns (…) gained from the extensive modeling in the symbolic environment of the mass media” (Bandura, 2001, p. 271). Thus, in the context of mass 36 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You communication, consumers may also learn by watching behavior performed by numerous models, rather than by relying only on one specific model (e.g., Stefanone, Lackaff, & Rosen, 2010). OL scholarship highlights four dimensions that foster observation learning: Attention, retention, reproduction, and motivation (e.g., Bandura, 2009). We contend that this four-dimensional framework can be applied successfully to the online environment and especially to SNSs. We begin with the dimension of attention, positing that a particular behavior may be learned successfully only if one pays sufficient attention to relevant information regarding, norms, values, thinking patents, conducts, etc., gained from extensive modeling in the media’s environment (e.g., Ball-Rokeach & DeFleur, 1976; Bandura, 2001). It is well documented that the attention-drawing mechanism is a major component of popular SNSs (e.g., Hodas & Lerman, 2012). One outstanding example is SNSs algorithms, that attract users’ attention to others who share their values and attitudes, while making dissimilar users invisible (e.g., Ellison, Vitak, Gray, & Lampe, 2014; Pariser, 2011). Such a mechanism might reinforce learning, as OL scholars note repeatedly: When models are similar to the learner, the chances that the learner will be influenced by the model’s behavior increase (van de Bongardt, Reitz, Sandfort, & Deković, 2015). Another significant “attracting attention” feature is the first-in-first-out queue struc- ture of the platforms’ interfaces (e.g., “Time line” in Facebook and chorological queue in Twitter). The second dimension, retention, indicates that storing information in ways enabling subsequent access is crucial to the OL process. SNSs might promote retention because these sites constitute archives of social information (Margetts, John, Hale, & Yasseri, 2015) that can be retrieved at different points in time. For example, the individual user can easily utilize the research engine tool placed on top of a Facebook page. By typing a keyword in the designated “research window,” they can easily obtain requested information about network members’ past behaviors on the network, thereby learning their respective values and attendant benefits and costs. In reproduction, the learner gets to practice the observed behavior. SNSs are designed to enhance repetitive SD actions, because their sustainability depends on user disclosure of personal information. Several features enhance such activities. On illus- trative example is the ability to “comment on a comment” on one’s Facebook page, and clearly observe the chaining patterns. In Tweeter, it is the retweeting feature that facilitate and even encourage repro- duction. With regard to motivation, observed rewards (gains) and punishment (losses) either raise or lower the learner’s motivation to repeat the modeled behavior, as Bandura (1971b) argues: “Seeing others gain desired outcomes by their actions can create outcome expectancies that function as posi- tive incentives [referred to as perceived gains in this article]; observed punishing outcomes can create negative outcome expectancies that function as disincentives [referred to here as perceived losses]” (p. 276). SNS platforms allow users to “visually monitor” (Carroll & Bandura, 1982) their friends’ behav- ior and the outcomes of these actions. For example, a prime feature in Facebook is the “Like” function. Any Facebook users can easily view the number of “Likes” other people received for the posts they publish. In line with OL scholarship, SNS users will most likely be conform to the behavioral patterns they observe because they expect certain benefits or losses, such as social recognition or rejection, or an increase or decrease in social status. As SNS may constitute a conducive environment for OL, we aim to explore how the ability to observe other users on SNSs enables one to learn the advisability of the modeled behavior. We develop our hypotheses according to the three core elements users observe and from which they learn: Others’ behavior, and the gains—or alternatively, losses—they incur. Others’ SD behavior Following early theorization of OL, numerous studies documented the influence of observed behavior, stressing the significance of observing behavior performed by many (Bandura, 1971a). As noted by van de Bongardt et al. (2015): Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 37 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. The larger the number of peers who engage in a certain behavior, the more functional and cor- rect the behavior will be perceived to be and the more likely it is that adolescents will engage in the same behavior, based on the reasoning that if others are doing it, it is probably a good or wise thing to do the same. (p. 204) On the face of it, in SNSs such as Facebook, any user can, in principle, visually monitor the behav- ior of many network friends. Nevertheless, these platforms also embody a visibility constraint: Users who disclose more become more visible. Conversely, risk-averse users usually disclose less and hence are less visible. This may strengthen the perception of self-disclosure as the most common and there- fore most prudent course of action. The influence of others’ perceived behavior on users’ decisions has been documented in the con- text of online environments. For example, it was demonstrated repeatedly that in the domain of ecom- merce, consumers observe the purchase actions of previous individuals and tend to imitate them (e.g., Wang & Yu, 2017). Similarly, Burke, Marlow, and Lento (2009), who studied online distribution of news, demonstrated that Facebook users who see their friends contributing news content proceed to share more content of this kind themselves (for similar results see Cheung, Liu & Lee, 2015). Building on these insights, we posit: Hypothesis 3: Perceived SD of others will be positively associated with one’s own SD behavior. Others’ gains As indicated above, one significant element that impinges on the effectiveness of OL is the reward the observed model receives (Bandura, 1971a). In OL theory, if learners see that a model gains from a particular behavior, they develop the expectation that such behavior would afford them the same gains. In such instances, users might be motivatedtoperform theobservedbehavior. Although observed gain is described as a fundamental factor in OL, to the best of our knowledge, it has been overlooked in SD literature on SNSs. Recent studies acknowledged the importance of observing behavior of others, as well as the gain one receives directly from the learning process (Cheung et al., 2015), yet no research has addressed the link between the two components. Hence we propose the following: Hypothesis 4: Observed gains of others will be positively associated with one’s perception of gains to self. With regard to others’ losses, OL theory posits that punishments (or losses) incurred by others receive deter the learner from repeating the observed behavior (van de Bongardt et al., 2015). Recently, this idea was tested in a study on software piracy, in which scholars showed that people tend H4 Assessment of H1 Observed gains gains to self H3 Observed SD SD behavior behavior H5 Assessment of H2 Observed losses losses to self Figure 1 Suggested theoretical model. 38 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You to avoid software pirating when they see online that those who commit this offence suffer losses (e.g., Sun, Lim, & Fang, 2013). In accordance with these observations, we propose (Figure 1): Hypothesis 5: Observed losses of others will be positively associated with one’s assessment of losses. Study 1: survey design Method We conducted an online survey among Israeli Facebook users between 7 and 9 December 2015. Sample We used the service of an online survey company that holds an online panel of participants who receive gift vouchers in exchange for their participation in online surveys. For this study, a sample of 742 Israeli Facebook users was created. The sample is not representative of the population of Israel, but rather was designed to be representative of the Israeli population of Facebook. Marketing data show that this latter group is far younger than the population as a whole and comprises more men (53%) than women. Israeli Facebook users are also more secular than Israeli society as a whole. The sample constitutes a near- perfect match to data available with regards to age and gender composition of Israeli Facebook users (age ranging between 18 and 71 with M = 42.3 SD = 14.6; 54% male; 3% ultra-Orthodox; 45% holding an aca- demic degree; income: M = 3.0 SD = 1.2, on a 5-point scale). Measures For a full description of scale items and validation, see Appendix A. Dependent variable To measure SD behavior, we created two scales: Actions taken and content shared. For the first scale, we asked participants to rank how often they were involved in a list of activities (adapted from Tsay- Vogel et al., 2016; 9 items, Cronbach’salpha = .88), while the second inquired how often participants shared content concerning different aspects of their life (e.g., politics, consumption preferences, family, etc.; Cronbach’salpha = .88). In both cases, answers were expressed on a scale ranging from 1 (never) to 5 (more than once a day) (actions: M = 2.4, SD = .86; information sharing: M = 1.7, SD = .70). Independent variables Perceived SD of others. To measure participants’ perception of how much others disclose, we asked them to answer the same questions as above with regard to their Facebook friends (Actions: Cronbach’s alpha = .90; M = 4.0, SD = .94; information sharing: Cronbach’salpha = .93; M = 3.1, SD = 1.0). We measured significance and likelihood of both gains and losses (Yates & Stone, 1992). Perception of gains. To gauge perception of gains that could be attributed to SD, we created a list of pos- sible benefits (Xu et al., 2013). To assess their significance, we asked participants the following question: “To what extent do you consider the following to be advantages of Facebook use?” Answers were ranked on a scale from 1 (not an advantage at all) to 5 (a significant advantage) (6 items, Cronbach’salpha = .84; M = 3.3, SD = 1.0). To assess likelihood, we asked: “When you think about the list we mentioned earlier, to what extent do you believe you are able to make the most of these benefits?” Answers were ranked on a 1 to 5 scale, ranging between “notatall” to “very much” (6 items, Cronbach’salpha = .87; M = 3.2, SD = .96). Perception of losses. To measure losses possibly attributable to SD, we created a list of possible types of harm (Acquisti et al., 2015). As in the case of gains, we asked participants about magnitude, i.e., the extent to which they considered the listed events to be troubling: The answer scale ranged from 1 (not troubling at all) to 5 (very troubling) (7 items, Cronbach’s alpha = .87; M = 3.8, SD = .97). We also asked participants to evaluate the likelihood of each loss: “How likely is it that Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 39 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. such things will happen to you as a result of your Facebook use?” Answers were ranked on a 1 to 5 scale, ranging between “not at all” to “very” (7 items, Cronbach’s alpha = .91; M = 3.6, SD = .99). Observed losses and gains of others. To measure participants’ assessments of the gains others achieved by using Facebook, we asked: “In your opinion, how successful are your Facebook friends in making the most of these benefits?” using the same list as above. The response scale ranged from 1 (not at all successful) to 5 (very successful) (6 items, Cronbach’s alpha = .90; M = 4.0, SD = .83). To gauge participants’ evaluation of others’ losses, we also asked whether any of the above risks affected their Facebook friends. Responses reflected a 3-point scale: “Never,”“once” and “more than once” (ranging from 0 to 7, M = 3.3, SD = 2.8). Covariates Besides applying demographic criteria, the model used three questions to control for online behavior. Participants were asked: (a) if, and how often they browse the web, (b) if, and how often they access Facebook, and (c) if, and how often they access other SNSs, such as Instagram or Twitter. Answers ranged from 1 (never) to 5 (several times a day). Participants who reported they engage in any of these activities on a daily basis were then asked how many hours a day they spent doing so, with answers reflecting a 3-point scale: “Less than an hour,”“1–3 hours,” and “more than 3 hours.” In our sample, only 1% reported browsing and 11% reported using Facebook on a less than daily basis. Consequently, we only used the hours per day item to report web browsing behavior (for which the 1% that did not answer this question were coded as 0, M = 2.6, SD = .60). For Facebook and other SNS use, we created a combined 6-point scale based on the two measures (Facebook M = 4.8, SD = 1.0; other SNS M = 2.7, SD = 1.0). Past experience We also asked participants whether any of the above risks affected them, with responses ranging on a 3-point scale: “never,”“once,” and “more than once” (ranging from 0 to 7, self: M = 3.1, SD = 2.0). Statistical analysis As indicated above, this theoretical approach to cost-benefit analysis calls for a process of weighting the significance of each item in gain/loss scales according to its perceived likelihood ([Van der Pligt, 2002]; gains: M = 11.8, SD = 5.5; losses: M = 14.1, SD = 5.9, see Appendix B for distribution). The theoretical model described above requires mediation relations testing, for which we calculated Maximum Likelihood (ML) estimates, using IBM SPSS Amos 22.0 to assess the proposed model’s goodness of fit and test each of the hypotheses. To test mediation, we relied on bootstrapping (5,000 bootstrap samples) and following MacKinnon, Lockwood, Hoffman, West, and Sheets (2002) we opted for a bias-corrected confidence interval (Preacher & Hayes, 2008). Results Before testing our hypotheses, we note the correlations between the two dimensions mentioned— significance and likelihood of gains and losses—and SD behavior. In the case of gains, a strong pos- itive correlation was found between significance and likelihood of gains and SD (significance: acts: Pearson’s r = .49, P < .01, information: Pearson’s r = .35, P < .01; likelihood: acts: Pearson’s r = .55, P < .01, information: Pearson’s r = .41, P < .01). By contrast, in the case of losses, both diminutions were weakly correlated with SD behavior (significance: acts: Pearson’s r = .02, P = .75, information: Pearson’s r = -.07, P = .07). Importantly, likelihood was positively correlated with it (acts: Pearson’s r = .07, P = .08, information: Pearson’s r = .08, P = .07). We used Structural equation modeling (SEM) to test our hypotheses. The model was assessed with SD actions and sharing information creating a latent variable of SD behavior for both self and others (for full correlations table, see Appendix C). The theoretical model yielded satisfactory results 40 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Assessment of Observed losses Actions losses to self ** .46 SD behavior ** Assessment of Observed gains gains to self Information ** ** observed SD behavior Information Actions Covariates: Demographics and online usage, past experience Figure 2 SEM of SD behavior social learning. Notes: Parameters are standardized coefficients. ***p < .001, **p < .01, *p < .05; N = 742. in terms of goodness of fit indices. We obtained a chi-square to df ratio (CMIN/DF) of 1.58 and found that the model fit the data extremely well (χ = 55.4, df = 35, p = .02; RMSEA = .03, CFI = .987). Figure 2 shows that most of our hypotheses were confirmed. Insofar as gains are concerned, H1 and H4 were supported: observed gains were positively associated with gains assessment for self (H4), that were, in turn, positively associated with SD behavior. Observed gains had a positive and signifi- cant indirect effect on SD behavior (standardized estimates = .39; 95% C.I. [.21–.32]). As for losses, H5 was supported, with observed losses to others positively associated with assess- ment of loss to self. Contrary to H2, assessment of losses to self had no bearing on SD behavior. Finally, in keeping with H3, perceived SD of others behavior was positively associated with the partici- pants’ own activity in this respect. It is important to note that results of Study 1 are the product of a cross-sectional design. As such, they are open to alternative interpretations and do not establish a causal relation. Study 2: experimental design The primary advantage of our survey design is the ability to gather information on users’ perceptions of their social reality. The survey, however, cannot establish causality, but can only demonstrate the associa- tion among the variables. In the case at hand, there could be two alternative explanations to the observed correlations between perceptions of others’ behavior gains and losses and cost-benefit assessment for self. One option is that users learn from others (wherein the direction of causal relations is from others to self). Alternatively, it could be interpreted as a projection process in which one projects one’s own views on others. Consequently, we conducted an online experiment to demonstrate that OL does occur. We exposed participants to one of the most common gain and loss scenarios that Facebook users encounter: The sentiments expressed in responses to a given post (positive vs. negative). Method Participants Using the same recruitment platform, we sampled 512 participants for the experiment. Of these, 44 did not have a Facebook account and were therefore excluded. An additional 41 did not complete the Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 41 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Figure 3 Experimental stimuli. experiment, leaving 427 participants in the final sample (age ranged between 18 and 71 with M = 37 SD = 13.1; 59% female; 2% ultra-Orthodox; 42% with academic degrees; income: M = 2.8 SD = 1.2 on a 5-point scale). Manipulation Participants were divided randomly into four groups. Each was exposed to a segment of a Facebook page (see Figure 3) presenting a news feed in which a person shared a post and the reactions it received. In all cases, the post was identical, but the responses to it were manipulated. In the control case, there were no comments and 15 Likes. In the next three cases, we manipulated the type of responses, but held the amount of attention constant (that is, the number of comments and Likes were the same). In all cases, two comments were identical, one positive and one negative, while four additional comments varied in sentiment. The comments addressed four issues: Characteristics, post content, political issue described and public support regarding the issue (whether or not the post represents the opinion of many or few). The picture, user names of people commenting and length of comments were held constant across all cases. In the positive cases, the post received Likes and posi- tive emoticons, as well as supportive comments from other users. In the negative case, responses included negative emotions and comments. Finally, a mixed response case included both positive and negative emotions, along with three positive and three negative comments. Manipulation check. To ensure that participants noticed the manipulation, we included five state- ments at the end of the questionnaire and asked participants to indicate the extent of their agreement on a scale of 1 (disagree) to 5 (agree): (a) The comments to the post were unpleasant, (b) The com- ments to the post were supportive (excluding participants in the control, F 13.9, p < .01; (2,425) = F 32.0, p < .01), (c) It was important to post this content (F 5.5, p < .01), (d) The post (2,425) = (3,423) = was convincing (F 5.1, p < .01), and (e) The post was inappropriate (F 1.1, p = .40). In (3,423) = (3,423) = all but the last case, the differences among groups were significant and conformed to the anticipated directions (for means and std. errors, see Appendix D), suggesting that the manipulation worked. Measures. After exposure, participants were instructed to respond according to the scales described above for perceived losses (Cronbach’s alpha = .86; M = 3.8, SD = .90) and gains (Cronbach’s alpha = .86; M = 3.4, SD = .94) and were asked how likely they were to self-disclose, using the same items as earlier (Cronbach’s alpha = .86; M = 2.6, SD = .83). The order of scales was randomized across participants and ANOVA was conducted to test the hypothesis. Results Figure 4 presents the means and standard errors of all dependent variables, according to experimental case. As anticipated, participants in the positive case group reported a higher tendency to share when 42 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You 4.5 3.5 2.5 1.5 Self-disclosure Gains Risks Control Positive Mix Negative Figure 4 Means and standard errors: risk perception, gains perception and tendency towards SD behavior, by experimental case. Scale ranges from 1 (not at all) to 5 (a lot). using SNSs, while those in the negative case group reported a lower one (F = 5.5, p < .01; all dif- (3,424) ferences among groups are significant; for a full description of contrasts, see Appendix E). Perceptions of gains were also affected by the manipulation in the expected direction (F = 7.37, p < .01). (3,424) Participants in the positive case group perceived a greater number of advantages to self-disclosure behavior (in comparison to the negative case and the mixed case) while those in the negative case group reported the opposite (in comparison to the control group and positive case group). By contrast, there was no significant difference in perception of losses (F = .70, p = .55); con- (3,424) trary to expectations, participants in the negative case group did not demonstrate a higher perception of possible losses. We also observed correlations among the three dependent variables. In the current study, the correlation between SD tendency and perceived gains remains much stronger than the one between SD tendency and perceived losses (gains: r = .53, p < .001; losses: r = -.08, p = .20). We conjectured that OL shapes Facebook users’ perceptions of losses and gains, wherein users learn how much to share and the outcomes of various types of sharing behavior by observing others online. Most of these hypotheses were largely supported by empirical data: Surveys and experiments alike dem- onstrated that Facebook users learn from others regarding gains/losses and behavior. Moreover, per- ceived gains were positively correlated with SD behavior. Only partial confirmation was obtained concerning losses, however: While a significant association was found between perceived losses to self and to others, the applied manipulation did not alter perception and no effects on behavior were noted. Discussion Our study applied the privacy calculus perspective to assess the influence of OL on SD in SNS. We conceptualized a model that underscores three key mechanisms at work: (a) Facebook users’ associa- tion of losses and gains with SD on SNS, (b) their assumption that positive relations obtain in both cases, and (c) the powerful effect of OL on losses/gains perception and—by implication—on SD behavior. By focussing on OL’s role in determining SD perceptions and resulting behavior, this model adds a new dimension to existing studies emphasizing the effects of social influence on SD (e.g., Shibchurn & Yan, 2015). A core argument of these studies maintains that significant others are likely Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 43 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. to contribute to the formation of one’s intentions regarding SD behavior. For example, Shibchurn and Yan (2015) recently speculated that “[h]igh risk perceptions and low disclosure levels by significant others may lead individuals to disclose less” (p. 106). If significant others disclose information plenti- fully, users are likely to adopt such patterns and disclose more. To date, their hypothesis has not been tested empirically. We attempted to advance this discussion by focussing on aspects of social influence that ought to be accorded more attention in an age of networked media: rather than focussing on per- ceived social norms (what one believes others think), we addressed technological features that enable SNS users to retrieve information regarding others by viewing their conduct and the consequences thereof. We were thus able to shed new light on the complex effect of social influence of risk assess- ment on disclosure behavior. We discovered that when users perceive other Facebook users’ self-disclosure, together with the benefits they obtain thereby, they tend to believe that they too can leverage the platform to their own advantage. We showed that this conception—that can be summarized as “more self-disclosure = greater benefit”—encourages participants to intensify their SD behavior. The situation differs with regard to perception of losses. We found that assessment of losses has a marginal effect on users’ SD behavior. Although the first study (survey) demonstrated that users’ risk assessment was linked with their perceptions of other users’ losses, the second (experiment) did not replicate these results. In fact, manipulation influenced assessment of gains and not of losses. These findings are in line with existing scholarship indicating that although users ascribe high risk levels to SD (lack of safety, security, reli- ability and trustworthiness), they still disclose a considerable amount of apparently sensitive informa- tion (e.g., Acquisti & Gross, 2006; Taddicken, 2013). When predicting Facebook self-disclosure, contemporary scholars studying privacy calculus on SNSs noted that expected benefits have more pre- dictive power than perceived losses (Dienlin & Metzger, 2016). We suggest two reasons that might explain the difference between the predictive power of gains and losses on SD. First, as noted by Ellison et al. (2014) in their study of Facebook, this platform is designed to impel users to reward those who generate content, because the absence of response is a vague signal that might be interpreted as a lack of interest. In Ellison et al.’s words: SNS users must respond in a manner that leaves an observable marker of attention as a way of cultivating, or grooming, their connections. (…) On an SNS, explicitly responding to another user via activities that leave visible traces, such as commenting or clicking the “Like” button, is the most reliable way to indicate one has seen and attended to any individual piece of content on the site. (p. 858) Following this reasoning, gains may well become more apparent in users’ feeds, leading them to believe that other members of their network leverage the platform more than they do. We argue that this perception of the rewards other users are supposedly receiving is among the factors that motivate Facebook users to disclose information liberally. Another possible reason for the observed disparity concerns our approach to studying risk assess- ment in connection with SD on Facebook. Our findings suggest that in the case of perceived gains, the significance and likelihood of the gain have a similar effect on SD, but when it comes to losses, the pic- ture is more complex: The significance of a given loss is negatively correlated with SD, whereas the likelihood of a given loss is positively correlated with it. This may well reflect the complexity of risk- assessment as it applies to SD on Facebook: Those who attribute significant losses to SD on Facebook will be less likely to self-disclose, regardless of their perceived likelihood of these losses. By contrast, those who disclose more may realize that these activities increase their chances of being harmed, but they might not necessarily perceive the possible losses as significant. 44 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Limitations and future research directions Several aspects of the research design may well have affected the findings of this study. Measures, for example, may constitute a significant source of bias. To reinforce design, we supplemented the survey with an experiment. Although the resulting combination provides powerful methodical strength, alter- native designs could have been considered as well, such as a study of organic interactions on Facebook, analyzing the interplay between visible social information and behavior patterns, made pos- sible by the advent of unobtrusive behavioral measurement. Its advantages notwithstanding, this method also entails certain constraints, particularly ethical concerns: when monitoring SNS personal profiles of consenting participants, for example, exposure of research to content posted by others who provided no such consent is virtually inevitable. Furthermore, this study used an opt-in sampling method, wherein participants selected themselves. Obviously, a non-random sample of paid partici- pants would engender various biases. Considering the extremely low response rates prevailing at pres- ent, however, the method appeared particularly useful despite its limitations. Another constraint worth mentioning involved the characteristics of the scales used. This study comprises some scales that contain “and” and “or” clauses. For example, we asked participants if they share personal information on Facebook, but did not specify whether this information contains a phone number, an email address, a password, etc. Clearly, while users might be willing to share phone numbers, they would not necessarily agree to sharing passwords. Future studies could focus on specific SD behavior and nuance our scales. Related to this, is a possible discrepancy between the use of the word “importance” of given benefit users report to, and the “gain” they attribute to it. For example, getting “Likes” might be a big benefit of SD on SNSs, but a respondent might not see it as “important” since the word might be associated with issues that carry social importance. Future research can address this limitation by “fine-tuning” the measurement. Several additional factors may have exerted a substantive effect on results. For example, the sample was limited in two respects, as it comprised Israeli users alone and was tailored to one platform only— Facebook. Although Facebook is the most popular SNS in Israel (Alexa, 2017), the study should be replicated in other cultural/national settings in which users may have a different predisposition regarding self-disclosure. Such replication could also be useful in countries whose privacy regulations and concerns differ from those of Israel. Considering the well-documented dissimilarities among countries (see, e.g., Krasnova & Veltri, 2010; Ribak & Turow, 2003), one would expect users in coun- tries with different approaches to privacy to ascribe differential gain/losses levels to SD on SNS. Studies of this kind might provide more intensive insight into the effects of culture on cost-benefit analysis and resulting SD behavior. Similarly, future efforts could consider the manner in which desig- nated visibility features inherent in certain online platforms such as WhatsApp, Instagram,or LinkedIn affect users’ perceptions of others and their behavior patterns. Participants’ ages may also contribute to research bias. Our sample comprises users of different ages (see discussion in the Methodology section), as we sought to provide a representative (nation- wide) sample of Facebook users (to whatever extent possible—demographic information concerning Facebook users is not available to the public). It may be valuable, however, to conduct a comparative study with cohorts representing different age groups (Generation Y and the older Generation X and Baby Boomers). As an increasing number of studies show that privacy concerns among young people differ from the anxieties manifested by adults, subsequent studies could shed considerable light on the manner in which participant age affects the value ascribed to SNS losses and gains. In this study, we did not distinguish between strong tie and weak tie friends. Realizing that social tie strength might influence OL, it could be valuable to conduct a study that explores the extent to which it does so. Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 45 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. The study might also contribute to practice-led research. One issue that received massive attention by both researchers and practitioners is means of fostering privacy literacy in the context of online networked technologies, particularly on SNSs. The findings of the current study, that underscored the importance of reciprocal features of SNS, could assist in the construction of interventions that enhance users’ capability to control the effects of their disclosure behavior. At present, Facebook responds to this challenge by sending its users personal reminders to review their privacy settings. One possible intervention that emerged from this study is an interface design that provides a visual display of friends’ privacy-related decisions, such as the relevant privacy settings. This study contributes to a growing body of research on SD behavior by pointing to the significant role OL plays in determining users’ willingness to self-disclose personal information on SNSs. It shows that the observation mechanism contributes to reward envy, that leads, in turn, to a high level of online SD. Subsequent studies could apply the model to examine the effects of OL on various manifes- tations of human conduct in online spaces. Acknowledgments The project was funded by the Israel Science Foundation. Grant number 1385/13. Notes 1 See http://www.fialkov.co.il/english/facebook-audience-insights, accessed 23 December 2017. 2 Ultra-Orthodox Jews, who comprise 8.8% of the population over age 20, often abstain from Internet use for ideological and economic reasons. See CBS, Statistical Abstract of Israel 2013— No. 64, Subject 7, Table 7.1, Persons aged 20 and over by selected characteristics., accessed 2 September 2014. Because of language barriers, the Israeli Arab population was not included in the sample. 3 The order of scales was randomized across participants. 4 We chose to present a certain degree of social response, given that typically any post receives some attention, thus a post with no response at all may appear unrealistic. 5 Our findings show that users discern positive relationships between levels of self-disclosure and of gains/losses. When observing others’ gains, users appear to perceive self-disclosure engagements on Facebook as a double-edged sword, exacerbating personal vulnerability as it increases platform usefulness. References Acquisti, A., Brandimarte, L., & Loewenstein, G. (2015). Privacy and human behavior in the age of information. Science, 347(6221), 509–514. Acquisti, A., & Gross, R. (2006). Imagined communities: Awareness, information sharing and privacy on the Facebook. In Privacy enhancing technologies (pp. 36–58). Berlin Heidelberg: Springer. Alexa (2017). Available at https://www.alexa.com/topsites/countries/IL Altman, I. (1975). The environment and social behavior: Privacy, personal space, territory, and crowding. Monterey, CA: Brooks/Cole Publishing. Antaki, C., Barnes, R. and Leudar, I., (2005). Self-disclosure as a situated interactional practice. British Journal of Social Psychology, 44, 181–200. Ball-Rokeach, S., & DeFleur, M. (1976). A dependency model of mass media effects. Communication Research,3,3–21. 46 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Bandura, A. (1971a). Vicarious and self-reinforcement processes. In R. Glaser (Ed.), The nature of reinforcement (pp. 228–278). New York: Academic Press. Bandura, A. (2001). Social cognitive theory of mass communication. Media Psychology, 3(3), 265–299. Bandura, A. (Ed.). (1971b). Psychological modeling: Conflicting theories. Chicago: Aldine-Atherton. Bazarova, N. N., & Choi, Y. H. (2014). Self‐disclosure in social media: Extending the functional approach to disclosure motivations and characteristics on social network sites. Journal of Communication, 64(4), 635–657. van de Bongardt, D., Reitz, E., Sandfort, T., & Deković, M. (2015). A meta-analysis of the relations between three types of peer norms and adolescent sexual behavior. Personality and Social Psychology Review, 19(3), 203–234. Burke, M., & Marlow, C., & Lento T. (2009). Feed me: Motivating newcomer contribution in social network sites, CHI 2009 Social Networking Sites, 7 April 2009. Boston, MA, USA. Carroll, W. R., & Bandura, A. (1982). The role of visual monitoring in observational learning of action patterns: Making the unobservable observable. Journal of Motor Behavior, 14(2), 153–167. Chang, C. W., & Chen, G. M. (2014). College students’ disclosure of location-related information on Facebook. Computers in Human Behavior, 35, 33–38. Chang, C. W., & Heo, J. (2014). Visiting theories that predict college students’ self-disclosure on Facebook. Computers in Human Behavior, 30, 79–86. Chen, R. (2013). Living a private life in public social networks: An exploration of member self- disclosure. Decision Support Systems, 55(3): 661–668. Cheung, C. M., Liu, I. L., & Lee, M. K. (2015). How online social interactions influence customer information contribution behavior in online social shopping communities: A social learning theory perspective. Journal of the Association for Information Science and Technology, 66(12), 2511–2521. Choi, Y. H., & Bazarova, N. N. (2015). Self-disclosure characteristics and motivations in social media: Extending the functional model to multiple social network sites. Human Communication Research, 41(4), 480–500. Deutsch, M., &. Gerard, H. B. (1955). A study of normative and informational social influences upon individual judgment. The Journal of Abnormal and Social Psychology, 51(3), 629–636. Dienlin, T., & Metzger, M. J. (2016). An extended privacy calculus model for SNSs: Analyzing self‐ disclosure and self‐withdrawal in a representative US sample. Journal of Computer‐Mediated Communication, 21(5), 368–383. Dinev, T., & Hart, P. (2004). Internet privacy concerns and their antecedents-measurement validity and a regression model. Behaviour & Information Technology, 23(6), 413–422. Ellison, N. B., Vitak, J., Gray, R., & Lampe, C. (2014). Cultivating social resources on social network sites: Facebook relationship maintenance behavior and their role in social capital processes. Journal of Computer‐Mediated Communication, 19(4), 855–870. Gross, R., & Acquisti, A. (2005). Information revelation and privacy in online social networks. Proceedings of the 2005 ACM Workshop on Privacy in the Electronic Society (pp. 71–80). ACM. Hodas, N. O., & Lerman, K. (2012). How visibility and divided attention constrain social contagion. Privacy, Security, Risk and Trust (PASSAT), 2012 International Conference on and 2012 International Conference on Social Computing (pp. 249–257). IEEE. Hollenbaugh, E. E., & Ferris, A. L. (2014). Facebook self-disclosure: Examining the role of traits, social cohesion, and motives. Computers in Human Behavior, 30, 50–58. doi:10.1016/j.chb.2013.07.055 Jourard, S. M. (1971). Self-disclosure: An experimental analysis of the transparent self. Journal of Experimental Social Psychology, 11(3), 233–260. Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 47 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Jourard, S. M., & Lasakow, P. (1958). Some factors in self-disclosure. The Journal of Abnormal and Social Psychology, 56(1), 91. Krasnova, H., Koroleva, K., Spiekermann, S., & Hildebrand, T. (2010). Online social networks: Why we disclose. Journal of Information Technology, 25(2), 109–125. Krasnova, H., & Veltri, N. F. (2010). Privacy calculus on social networking sites: Explorative evidence from Germany and USA. System Sciences (HICSS), 2010 43rd Hawaii international conference on (pp. 1–10). IEEE. Krasnova, H., Veltri, N. F., & Günther, O. (2012). Self-disclosure and privacy calculus on social networking sites: The role of culture. Business & Information Systems Engineering, 4(3), 127–135. Laufer, R. S., & Wolfe, M. (1977). Privacy as a concept and a social issue: A multidimensional developmental theory. Journal of social Issues, 33(3), 22–42. Li, H., Sarathy, R., & Xu, H. (2010). Understanding situational online information disclosure as a privacy calculus. Journal of Computer Information Systems, 51(1), 62–71. MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A comparison of methods to test mediation and other intervening variable effects. Psychological Methods, 7(1), 83–104. Margetts, H., John, P., Hale, S., & Yasseri, T. (2015). Political turbulence: How social media shape collective action. Princeton, NJ: Princeton University Press. McKnight, D. H., Lankton, N., & Tripp, J. (2011). Social networking information disclosure and continuance intention: A disconnect. Paper presented at the System Sciences (HICSS), 2011 44th Hawaii International Conference on. Muise, A., Christofides, E., & Desmarais, S. (2009). More information than you ever wanted: Does Facebook bring out the green-eyed monster of jealousy? CyberPsychology & Behavior, 12(4), 441–444. Pariser, E. (2011). The filter bubble: What the internet is hiding from you. London, UK: Penguin. Preacher, K. J., & Hayes, A. F. (2008). Contemporary approaches to assessing mediation in communication research. In A. F. Hayes, M. D. Slater & L. B. Snyder (Eds.), The Sage sourcebook of advanced data analysis methods for communication research (pp. 13–54). http://dx.doi.org/10. 4135/9781452272054.n2 Ribak, R., & Turow, J. (2003). Internet power and social context: A globalization approach to Web privacy concerns. Journal of Broadcasting & Electronic Media, 47(3), 328–349. Shibchurn, J., & Yan, X. (2015). Information disclosure on social networking sites: An intrinsic– extrinsic motivation perspective. Computers in Human Behavior, 44, 103–117. Simpson, P. M., Siguaw, J. A., & Cadogan, J. W. (2008). Understanding the consumer propensity to observe. European Journal of Marketing, 42(1/2), 196–221. Sitkin, S. B., & Pablo, A. L. (1992). Reconceptualizing the determinants of risk behavior. Academy of Management Review, 17(1), 9–38. Sledgianowski, D., & Kulviwat, S. (2008). Social network sites: Antecedents of user adoption and usage. AMCIS 2008 Proceedings, 83. Stefanone, M. A., Lackaff, D., & Rosen, D. (2010). The relationship between traditional mass media and “social media”: Reality television as a model for social network site behavior. Journal of Broadcasting & Electronic Media, 54(3), 508–525. Sun, Y., Lim, K. H., & Fang, Y. (2013). Do facts speak louder than words? Understanding the sources of punishment perceptions in software piracy behavior. Paper presented at the Pacific Asia Conference on information Systems, Jeju Island, South Korea (p. 167). Taddicken, M. (2013). The “privacy paradox” in the social web: The impact of privacy concerns, individual characteristics, and the perceived social relevance on different forms of self‐disclosure. Journal of Computer Mediated Communication, 19(2), 248–273. 48 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Tamir, D. I., & Mitchell, J. P. (2012). Disclosing information about the self is intrinsically rewarding. Proceedings of the National Academy of Sciences of the United States of America, 109(21), 8038–8043. Tsay-Vogel, M., Shanahan, J., & Signorielli, N. (2016). Social media cultivating perceptions of privacy: A 5-year analysis of privacy attitudes and self-disclosure behaviors among Facebook users. New Media & Society. doi:10.1177/1461444816660731. Van der Pligt, J. (2002). Cognition and affect in risk perception and risky decision-making. Psychology at the Turn of the Millennium: Social, Developmental, and Clinical Perspectives, 2, 247–270. Wang, Y., & Yu, C. (2017). Social interaction-based consumer decision-making model in social commerce: The role of word of mouth and observational learning. International Journal of Information Management, 37(3), 179–189. Williams, S., & Voon, Y. W. W. (1999). The effects of mood on managerial risk perceptions: Exploring affect and the dimensions of risk. The Journal of Social Psychology, 139(3), 268–287. Xu, F., Michael, K., & Chen, X. (2013). Factors affecting privacy disclosure on social network sites: An integrated model. Electronic Commerce Research, 13(2), 151–168. doi:10.1007/s10660-013-9111-6 Yardi, S., Golder, S. A., & Brzozowski, M. J. (2009). Blogging at work and the corporate attention economy. Paper presented at the Proceedings of the SIGCHI conference on human factors in computing systems. Yates, J. F., & Stone, E. R. (1992). The risk construct. Oxford, England: Wiley & Sons. Zephoria Digital Marketing. (2017). Available from https://zephoria.com/top-15-valuable-facebook- statistics/ Zhang, J., & Daugherty, T. (2009). Third-person effect and social networking: Implications for online marketing and word-of-mouth communication. American Journal of Business,24(2), 53–64. Appendix A: Measurements Confirmatory factor analysis SD acts—self act1 Write status updates act2 Comment on other people’s status updates act3 Share other people’s content act4 Upload photos act5 Tag yourself in photos act6 Tag other people in photos act7 Participate in groups act8 Indicate your geographical location, for example by mentioning a name of a restaurant, an event, or a landmark act9 Indicate you intend to go to an event, or check in Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 49 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. N Min. Max. Mean Std. D. Skewness Kurtosis act1 744 1 5 2.36 1.22 .64 −.60 act2 744 1 5 3.26 1.32 −.20 −1.16 act3 744 1 5 2.86 1.35 .15 −1.20 act4 744 1 5 2.32 1.10 .79 −.09 act5 744 1 5 1.91 1.07 1.20 .79 act6 744 1 5 1.96 1.13 1.13 .45 act7 744 1 5 3.04 1.45 −.01 −1.38 act8 744 1 5 1.75 .97 1.46 1.79 act9 744 1 5 1.94 1.11 1.17 .60 Fit of model: χ = 34, df = 11, p = .46; RMSEA = .00, CFI = .997 act4 act6 act9 act1 act2 act3 act5 act8 act7 .77 .80 .81 .77 .80 .85 Acts 50 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You SD information-self info1 Share content regarding politics and current events, such as your opinions, political content you support, or news stories you found interesting info2 Share content regarding your consumer preferences and shopping habits, such as companies and products you like or dislike info3 Share information regarding your family, such as your relationship status, the identity of your partner, your parental status, or activities in which your children took part info4 Share information regarding your work, such as your workplace, your profession, or the people you work with info5 Share information regarding your friends, such as the identity of the people you hang out with info6 Share information regarding your current location, such as restaurants or events you are currently attending info7 Share information regarding your location in retrospect, such as restaurants or events you previously attended info8 Indicate where you live on your profile info9 Share personal information, such as phone numbers, ID numbers, or passwords, or send it in messages N Min. Max. Mean Std. D. Skewness Std. Error Kurtosis Std. Error info1 744 1 5 2.04 1.13 .94 .09 .04 .18 info2 744 1 5 1.78 1.04 1.31 .09 .99 .18 info3 744 1 5 1.7 .98 1.54 .09 1.92 .18 info4 744 1 5 1.71 .98 1.53 .09 1.88 .18 info5 744 1 5 1.78 .98 1.31 .09 1.19 .18 info6 744 1 5 1.69 .92 1.37 .09 1.44 .18 info7 744 1 5 1.72 .91 1.39 .09 1.74 .18 info8 743 1 5 1.85 1.16 1.46 .09 1.27 .18 info9 744 1 5 1.3 .78 2.99 .09 8.78 .18 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 51 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Fit of model: χ = 15.9, df = 26, p = .87; RMSEA = .00, CFI = .995 inf5 inf6 inf7 inf8 inf9 inf1 inf2 inf3 inf4 .72 .76 .68 .63 .72 .60 information SD acts—Others act_fr1 Write status updates act_fr2 Comment on other people’s status updates act_fr3 Share other people’s content act_fr4 Upload photos act_fr5 Tag themselves in photos act_fr6 Tag other people in photos act_fr7 Indicate their geographical location, for example by mentioning a name of a restaurant, an event, or a landmark act_fr8 Indicate they intend to go to an event, or check in act_fr9 Participate in groups N Min. Max. Mean Std. D. Skewness Std. Error Kurtosis Std. Error act_fr1 698 1 5 3.97 1.04 −.80 .09 −.22 .19 act_fr2 697 1 5 4.21 .97 −1.21 .09 .91 .19 act_fr3 693 1 5 3.98 1.04 −.87 .09 .01 .19 act_fr4 710 1 5 3.94 1.04 −.80 .09 −.15 .18 act_fr5 668 1 5 3.71 1.14 −.56 .10 −.65 .19 act_fr6 668 1 5 3.62 1.16 −.51 .10 −.67 .19 act_fr7 677 1 5 3.46 1.20 −.37 .09 −.84 .19 act_fr8 650 1 5 3.44 1.18 −.34 .10 −.84 .19 act_fr9 633 1 5 3.93 1.07 −.83 .10 −.10 .19 52 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Fit of model: χ = 37.5, df = 15, p = .07; RMSEA = .03, CFI = .981 act6 act9 act1 act2 act3 act4 act5 act8 act7 .77 . 80 .81 .77 .80 .85 Acts FR SD-information others info_fr1 Share their opinions regarding current events and political contents they support info_fr2 Share content regarding their consumer preferences and shopping habits, such as companies and products they like or dislike info_fr3 Share information regarding their family, such as their relationship status, the identity of their partner, their parental status, or activities in which their children took part info_fr4 Share information regarding their work, such as their workplace, their profession, or the people they work with (Continued) Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 53 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Continued info_fr5 Share information regarding their friends, such as the identity of the people they hang out with or how they hang out with them info_fr6 Share information regarding their current location, such as restaurants or events they are currently attending info_fr7 Share information regarding places they previously attended, such as restaurants and events info_fr8 Indicate where they live on their profile info_fr9 Share personal information, such as phone numbers, ID numbers, or passwords N Min. Max. Mean Std. D. Skewness Std. Error Kurtosis Std. Error info_fr1 688 1 5 3.37 1.17 −.28 .09 −.86 .19 info_fr2 669 1 5 3.06 1.26 .03 .09 −1.12 .19 info_fr3 681 1 5 3.03 1.25 .05 .09 −1.09 .19 info_fr4 676 1 5 2.91 1.28 .18 .09 −1.11 .19 info_fr5 690 1 5 3.3 1.19 −.16 .09 −1.01 .19 info_fr6 684 1 5 3.25 1.21 −.10 .09 −1.06 .19 info_fr7 692 1 5 3.21 1.17 −.10 .09 −.95 .19 info_fr8 644 1 5 2.89 1.37 .19 .10 −1.23 .19 info_fr9 611 1 5 1.96 1.34 1.12 .10 −.15 .20 54 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Fit of model: χ = 10.3, df = 16, p = .85; RMSEA = .00, CFI = .994 inf5 inf6 inf7 inf8 inf9 inf1 inf2 inf3 inf4 .85 .88 .80 .80 .88 .49 information SD-Gains self: adv1 Simply and rapidly obtaining information suiting my interests adv2 Obtaining likes and support from others adv3 Expressing myself—my opinions and feelings adv4 Maintaining and reinforcing social relations adv5 Getting personalized information, such as information about my close friends or about my fields of interest adv6 Effectively promoting myself professionally Significance N Min. Max. Mean Std. D Skewness Std. Error Kurtosis Std. Error adv1 744 1 5 3.64 1.23 −.63 .09 −.52 .18 adv2 744 1 5 3.13 1.35 −.18 .09 −1.13 .18 adv3 744 1 5 2.92 1.33 .01 .09 −1.15 .18 adv4 744 1 5 3.60 1.24 −.56 .09 −.63 .18 adv5 744 1 5 3.53 1.19 −.54 .09 −.48 .18 adv6 744 1 5 3.02 1.40 −.03 .09 −1.23 .18 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 55 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Fit of model: χ = 10.3, df = 5, p = .08; RMSEA = .03, CFI = .995 ad3 ad6 ad1 ad2 ad4 ad5 .70 .73 .71 .57 Gains SD-Gains self: Likelihood N Min. Max. Mean Std. D. Skewness Std. Error Kurtosis Std. Error adv1 743 1 5 3.49 1.17 −.37 .09 −.67 .18 adv2 743 1 5 3.24 1.27 −.20 .09 −.99 .18 adv3 743 1 5 2.96 1.31 .04 .09 −1.08 .18 adv4 743 1 5 3.38 1.23 −.36 .09 −.79 .18 adv5 743 1 5 3.40 1.20 −.31 .09 −.75 .18 adv6 743 1 5 2.47 1.29 .48 .09 −.84 .18 56 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Fit of model: χ = 5.39, df = 5, p = .14; RMSEA = .033, CFI = .998 ad3 ad6 ad1 ad2 ad4 ad5 .64 .80 .71 .79 .67 .71 Gains likelihood SD-Gains others: adv_fr1 Obtaining information suiting their interests adv_fr2 Obtaining likes and support from others adv_fr3 Publicly expressing themselves and their feelings adv_fr4 Reinforcing social relations adv_fr5 Getting personalized information, such as information about their close friends or about their fields of interest adv_fr6 Promoting themselves professionally N Min. Max. Mean Std. D. Skewness Std. Error Kurtosis Std. Error adv_fr1 579 1 5 4.07 .98 −.91 .10 .24 .20 adv_fr2 620 1 5 4.18 .95 −1.08 .10 .71 .20 adv_fr3 614 1 5 4.13 .99 −1.05 .10 .62 .20 adv_fr4 589 1 5 4.07 1.02 −1.02 .10 .53 .20 adv_fr5 575 1 5 3.98 1.00 −.77 .10 .04 .20 adv_fr6 560 1 5 3.63 1.19 −.51 .10 −.62 .21 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 57 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Fit of model: χ = 3.2, df = 5, p = .68; RMSEA = .00, CFI = .998 ad3 ad6 ad1 ad2 ad4 ad5 .78 .86 Gains others SD loss—self rsk1 Information I want to share with specific people will find its way to other people rsk2 People will share information about me without my consent, for example by tagging me in pictures rsk3 I will often be exposed to information I don’t care about rsk4 Commercial companies will use information I share to my disadvantage, for example by bombarding me with irrelevant advertisements rsk5 Commercial companies will sell my information to third parties without my knowledge or consent rsk6 The state will know more about me than I want it to rsk7 Hackers will steal my information Significance Min. Max. Mean Std. D. Skewness Std. Error Kurtosis Std. Error rsk1 1 5 3.61 1.35 −.62 .09 −.81 .18 rsk2 1 5 3.45 1.36 −.36 .09 −1.12 .18 rsk3 1 5 3.27 1.28 −.26 .09 −.94 .18 rsk4 1 5 3.94 1.20 −.96 .09 −.04 .18 rsk5 1 5 4.02 1.26 −1.10 .09 .02 .18 rsk6 1 5 3.43 1.37 −.39 .09 −1.05 .18 rsk7 1 5 4.04 1.32 −1.11 .09 −.09 .18 58 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Fit of model: χ = 8.0, df = 6, p = .24; RMSEA = .02, CFI = .997 rs7 rs1 rs2 rs3 rs4 rs5 rs6 1.0 .81 .95 .96 Losses significance Likelihood N Min. Max. Mean Std. D. Skewness Std. Error Kurtosis Std. Error rsk1 743 1 5 3.47 1.24 −.34 .09 −.91 .18 rsk2 743 1 5 3.56 1.25 −.46 .09 −.81 .18 rsk3 743 1 5 3.87 1.15 −.71 .09 −.47 .18 rsk4 743 1 5 3.9 1.19 −.87 .09 −.23 .18 rsk5 743 1 5 3.71 1.27 −.66 .09 −.66 .18 rsk6 743 1 5 3.46 1.30 −.41 .09 −.92 .18 rsk7 743 1 5 3.49 1.26 −.33 .09 −.98 .18 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 59 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. χ = 3.8, df = 6, p = .73; RMSEA = .00, CFI = .999 rs7 rs1 rs2 rs4 rs5 rs6 rs3 1.0 .93 .71 .97 .91 .92 Likelihood losses Losses—others rsk_fr1 Information they want to share with specific people will find its way to other people rsk_fr2 People will share information about them without their consent, for example by tagging them in pictures rsk_fr3 They will often be exposed to information they don’t care about rsk_fr4 Commercial companies will use information they share to their disadvantage, for example by bombarding them with irrelevant advertisements rsk_fr5 Commercial companies will sell their information to third parties without their knowledge or consent rsk_fr6 The state will know more about them than they want it to rsk_fr7 Hackers will steal their information N Min. Max. Mean Std. D. Skewness Std. Error Kurtosis Std. Error rsk_fr1 440 1 3 2.13 .72 −.20 .12 −1.06 .23 rsk_fr2 490 1 3 2.31 .69 −.50 .11 −.83 .22 rsk_fr3 509 1 3 2.51 .68 −1.07 .11 −.12 .22 rsk_fr4 454 1 3 2.34 .74 −.63 .12 −.92 .23 rsk_fr5 418 1 3 2.08 .79 −.14 .12 −1.37 .24 rsk_fr6 413 1 3 1.96 .77 .06 .12 −1.32 .24 rsk_fr7 433 1 3 1.85 .72 .23 .12 −1.04 .23 60 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You χ = 13.0, df = 9, p = .15; RMSEA = .02, CFI = .992 rs7 rs1 rs2 rs4 rs5 rs6 rs3 .89 .85 .80 .90 .81 .79 Losses others Construct validity To validate our measurements, we tested their correlations with related concepts on an additional sample, relying on previous research by Dienlin and Metzger (2016), who developed scales to measure Facebook benefits, privacy concerns, Facebook privacy self-efficacy, Facebook self-disclosure and Facebook self-withdrawal. Furthermore, to validate data about self-disclosure, we asked participants to provide us with their Facebook activity log and added an item measuring passive Facebook use (dem- onstrating distinction between self-disclosure and passive use). Sample We used an online sample of 261 participants, 124 of whom completed the survey, ranging in age between 19 and 64 with M = 40.5 SD = 12.7; 49% male; 3% ultra-Orthodox; 50% with academic degrees; income: M = 3.0 SD = 1.4 on a 5-point scale; participants were recruited using the same plat- form as the samples described in the article. Measures In addition, developing the above measures, we also asked participants about their Facebook benefits, privacy concerns, Facebook privacy self-efficacy, Facebook self-disclosure and Facebook self- withdrawal. All scales are taken from Dienlin and Metzger (2016) (full scales: https://osf.io/h4pef/? view_only=cf4c10222def4efdbecae20c1dca7dc6). Activity log data: Facebook allows users to download a log documenting all their activities. To pro- tect participants’ privacy, we did not ask them to disclose their log (containing a complete documenta- tion of photos, posts, etc.), but rather to report the size of the files downloaded (see Figure A1), based on the assumption that size is an indicator of amount of material shared. Nevertheless, it is important to note that 137 participants did not comply with this request, resulting in a high attrition rate. Furthermore, 22 participants failed to follow the instructions and thus supplied incorrect data. Seeking a uniform size indicator for all files (photos, posts, events, friends, etc.), we standardized each item (image files are much larger than text files, for example), ensuring that each participant was assigned an accurate relative score that could be compared with those of other participants. Passive use: To gauge passive use, we asked participants how often they read their Facebook friends’ posts (text and images) as part of the activity scale we developed to measure SD act frequency. Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 61 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Figure A1 Sample activity log data disclosed by participant. Results Table A1 presents the descriptive statistics of variables measured, while Table A2 displays the results of a factor analysis for all scales. As indicated, all items in our scales were loaded on one factor, except for the SD acts scale, for which a two-factor solution was identified: SD acts were loaded on one factor and another measuring passive use was loaded on another. Only one item displayed high loadings of both factors: Liking friends’ posts/images. The dual high score may be explained by the item’s rele- vance to self and others, as it measures the participant’s active use yet relies on observation by others as well. Table A3, that presents the correlation among the variables, provides additional validation for the measurements developed: Self-disclosure measurements were highly correlated with Dienlin and Matzear’s Facebook self-disclosure measurements, suggesting convergence validity. We also noted a positive correlation among perceived gains, Dienlin and Matzear’s Facebook benefits, self-disclosure and privacy concerns. Perceived losses were correlated with Facebook self-withdrawal and privacy concerns. 62 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Table A1 Descriptive Statistics of All Variables Mean Std. D. Skewness S.E. Kurtosis S.E. Cronbach’s α SD activity 2.43 .85 .66 .21 .11 .43 .89 SD information 1.78 .74 1.54 .23 3.10 .46 .88 Gains 3.50 1.00 −.34 .23 −.40 .46 .86 Risks 3.60 .95 −.48 .23 −.25 .46 .86 Privacy concerns (Dienlin and Matzear) 2.75 .79 −.35 .23 −.32 .46 .59 Facebook privacy self-efficacy (Dienlin 2.48 .96 .25 .23 −.63 .46 .88 & Matzear) Facebook benefits (Dienlin and 3.29 .86 −.37 .23 .63 .46 .90 Matzear) Facebook self-disclosure (Dienlin and 2.68 .94 .24 .23 −.26 .46 .86 Matzear) Facebook self-withdrawal (Dienlin and 1.69 .29 −1.07 .23 .54 .46 .61 Matzear) Activity log data .002 .66 1.04 .24 2.02 .47 .87 Type N Min. Max. Mean Std. D. Ads 101 2 62,919 9,547 10,773 Contact info 102 1.61 2,879,918 145,797 361,512 Events 102 2 20,187,833 329,373 2,032,835 Friends 102 2 191,275 13,994 28,121 Messages 101 1.01 43,798,254 3,173,460 6,615,034 Mobile device 102 1 1,538 858 681 Photos 102 1 65,703 6,687 11,083 Places 102 1 10,628 1,082 1,517 Synced photos 102 1 1,431 792 628 Timeline 102 3 8,214,581 569,150 1,050,921 Videos 102 1 42,768 4,305 7,629 Table A2 Exploratory Factor Analysis SD actions Item Component Write status updates .78 −.02 Comment on other people’s status updates .73 .38 Share other people’s content .73 .32 Upload photos .79 −.16 Tag yourself in photos .79 −.39 (Continued) Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 63 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Table A2 Continued SD actions Item Component Tag other people in photos .78 −.30 Participate in groups .63 .45 Indicate your geographical location, for example by mentioning a name of a restaurant, .76 −.36 an event, or a landmark Indicate you intend to go to an event, or check in .74 −.16 Read content uploaded by your friends .44 .62 SD information Item Component Share their opinions regarding current events and political contents they support .835 Share content regarding their consumer preferences and shopping habits, such as .814 companies and products they like or dislike Share information regarding their family, such as their relationship status, the identity of .778 their partner, their parental status, or activities in which their children took part Share information regarding their work, such as their workplace, their profession, or the .747 people they work with Share information regarding their friends, such as the identity of the people they hang .743 out with or how they hang out with them Share information regarding their current location, such as restaurants or events they .705 are currently attending Share information regarding places they previously attended, such as restaurants and .675 events Indicate where they live on their profile .62 Share personal information, such as phone numbers, ID numbers, or passwords .562 SD gains Significance Item Component Simply and rapidly obtaining information suiting my interests .687 Obtaining likes and support from others .845 Expressing myself—my opinions and feelings .818 Maintaining and reinforcing social relations .696 Getting personalized information, such as information about my close friends or about .794 my fields of interest Effectively promoting myself professionally .756 64 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 T. Ashuri et al. Watching Me Watching You Likelihood Item Component Simply and rapidly obtaining information suiting my interests .816 Obtaining likes and support from others .865 Expressing myself—my opinions and feelings .779 Maintaining and reinforcing social relations .733 Getting personalized information, such as information about my close friends or about .719 my fields of interest Effectively promoting myself professionally .631 SD losses Significance Item Component Information I want to share with specific people will find its way to other people .82 People will share information about me without my consent, for example by tagging me .714 in pictures I will often be exposed to information I don’t care about .523 Commercial companies will use information I share to my disadvantage, for example by .701 bombarding me with irrelevant advertisements Commercial companies will sell my information to third parties without my knowledge .823 or consent The state will know more about me than I want it to .779 Hackers will steal my information .781 Likelihood Item Component Information I want to share with specific people will find its way to other people .865 People will share information about me without my consent, for example by tagging me .815 in pictures I will often be exposed to information I don’t care about .734 Commercial companies will use information I share to my disadvantage, for example by .82 bombarding me with irrelevant advertisements Commercial companies will sell my information to third parties without my knowledge .816 or consent The state will know more about me than I want it to .717 Hackers will steal my information .775 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 65 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Table A3 Pearson’s Correlations for Measurements Gains- Gains- Loss- Loss- Activity Self- Acts Information sig lik sig sig log withdrawal Benefits SD Information .754** Gains-sig .487** .411** 121 121 121 Gains-lik .591** .516** .758** 121 121 121 Loss-sig .236** .239** .393** .433** 121 121 121 121 Loss-lik .301** .194* .419** .407** .586** 121 121 121 121 121 Activity log .274** .212* .349** .347** .219* .321** 102 102 102 102 102 102 Self-withdrawal .152 .159 .108 .165 .235* .221* .168 Dienlin and Matzear 119 119 119 119 119 119 100 Benefits .168 .212* .342** .404** −.011 −.007 −.022 .161 Dienlin and Matzear 121 121 121 121 121 121 102 119 SD .556** .573** .531** .622** .285** .210* .210* .145 .349** Dienlin and Matzear 121 121 121 121 121 121 102 119 121 Privacy con −.044 −.098 −.06 .008 .148 .341** .216* .198* −.275** −.118 Dienlin and Matzear 121 121 121 121 121 121 102 119 121 121 *p < .5; **p < .01 Appendix B: Distribution of Risk Assessment Measures 1 2 3 4 5 6 7 8 9 10 111213 141516 171819 2021 222324 25 Perceived Gains Perceived Losses Note: Percent refers to the percentage of participants for each scale. Perceived gains are measured as the likelihood of gain (on a scale of 1–5) weighted by its magnitude (on a 1–5 scale). Perceived losses measured as the likelihood of loss (on a scale of 1–5) weighted by its magnitude (on a scale of 1–5). 66 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Percent T. Ashuri et al. Watching Me Watching You Appendix C: Pearson’s Correlations Among Variables 12 345 678 9 10 11 2. Sharing information—self .665** 3. Sharing activity—others .278** .197** 774 724 4. Sharing information— .126** .321** .592** others 744 704 744 5. Browsing .231** .126** .058 .093* 732 742 732 704 6. Facebook use (hrs.) .473** .335** .103** .126** .425** 730 742 730 704 742 7.Other SNS use (hrs.) .344** .359** .254** .132** .218** .339** 240 742 240 704 742 742 8. Past damage—self .115** .222** .222** .119** .086* .06 .166** 742 742 742 704 742 742 742 9. Losses—others −.004 .166** .109** .130** .015 .048 .154** .447** 622 742 619 704 742 742 742 742 10. Losses—self .038 −.04 .063 .032 .009 −.029 .022 .281** .046 742 742 724 704 742 742 742 742 742 11. Gains—self .538** .391** .272** .158** .082* .369** .198** .036 .059 .076* 741 742 723 704 742 742 742 742 742 742 12. Perceived gains—others .201** .132** .400** .258** .046 .154** −.037 −.063 -.083* .095* .507** 616 632 616 623 632 632 632 632 632 632 632 Notes: *p < .5; **p < .01 Variables: (1) Sharing activity—self; (2) Sharing information—self; (3) Sharing activity—others; (4) Sharing infor- mation—others; (5) Browsing; (6) Facebook use in hours; (7) Other SNS use in hours; (8) Sustained damage in the past—self; (9) Losses—others; (10) Losses—self; (11) Gains –self; (12) Perceived gains—others. Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association 67 Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018 Watching Me Watching You T. Ashuri et al. Appendix D: Manipulation Check 4.5 3.5 2.5 1.5 Comments unpleasant Comments supportive Important post Convincing post Post was in appropriate Control Positive Mix Negative Figure A Means and std. errors for post and comment evaluation, according to experimental condi- tion: 1 (disagree); 5 (agree). Appendix E Table 1 Results of Contrast between Experimental Conditions Contrast Value of Contrast Std. Error t Sig. Self-disclosure Control vs. Negative .34 .13 2.60 .01 Control vs. Positive −.31 .12 −2.57 .01 Mix vs. Negative .33 .13 2.48 .01 Mix vs. Positive −.32 .12 −2.55 .01 Negative vs. Positive −.32 .12 −2.55 .01 Gains Control vs. Negative .42 .17 2.42 .02 Control vs. Positive −.24 .16 −1.48 .14 Mix vs. Negative .25 .18 1.38 .17 Mix vs. Positive −.41 .17 −2.49 .01 Negative vs. Positive −.41 .17 −2.49 .01 68 Journal of Computer-Mediated Communication 23 (2018) 34–68 © 2018 International Communication Association Downloaded from https://academic.oup.com/jcmc/article-abstract/23/1/34/4832997 by Ed 'DeepDyve' Gillespie user on 16 March 2018

Journal

Journal of Computer-Mediated CommunicationOxford University Press

Published: Jan 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off