Depth of Self-Disclosure for Social Media Influencers: Tensions and Tradeoffs by Woocheol Kim A dissertation accepted and approved in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Marketing Dissertation Committee: Conor Henderson, Chair Noelle Nelson, Core Member Jessica Gamlin, Core Member Julian Saint Clair, Core Member Edward Rubin, Institutional Representative University of Oregon Spring 2024 © 2024 Woocheol Kim This work is openly licensed via CC BY-NC-ND 4.0. 2 DISSERTATION ABSTRACT Woocheol Kim Doctor of Philosophy in Marketing Title: Depth of Self-Disclosure for Social Media Influencers: Tensions and Tradeoffs Social media influencers wrestle with the decision of how much to disclose about their personal lives. Must attention and financial opportunities come at the cost of privacy? Do risks from being personally vulnerable at least come with rewards? The present research investigates consumers’ responses to influencers’ depth of disclosure on social media. To do so, I adopt a multi-method research approach that incorporates data from controlled experiments, automated text analysis of social media posts, and qualitative theory-in- use interviews. With experiments, I find that greater disclosure is associated with being more relatable, but less aspirational, which combine for offsetting overall effects on influencer endorsement-ability. Consistent with these offsetting effects, text analysis of influencer social media posts reveals a nonlinear, curvilinear effect of depth of disclosure in a social media post on post evaluations (i.e., likes, retweets, comments). Interviews with social media influencers reveal a desire to present themselves as “whole” “well-rounded” people while also protecting their privacy. These influencers note that many brands push them to make posts very personal, but they question whether this is even necessary. My findings suggest these influencers’ intuition is correct, and the brands should not prescribe influencers to be overly personal and risk “over-sharing.” Theoretically, the research contributes to influencer marketing by highlighting the dual importance of relatable and 3 aspirational qualities while showing that the two are balanced against each other, at least in relation to depth of disclosure. 4 ACKNOWLEDGMENTS I would like to sincerely thank Professor Conor Henderson for the last five years of his guidance, encouragement, and positivity; I am incredibly privileged to be a student of yours. In addition, I wish to express appreciation to my committee members, Noelle Nelson, Jessica Gamlin, Julian Saint Clair, and Edward Rubin for their support and valuable input throughout this process. This research was supported in part by the Roger and Robin Best Fund for Doctoral Research. 5 To my family, thank you for always being by my side, supporting me through and through. This would not have been possible without you. 6 TABLE OF CONTENTS Chapter Page I. INTRODUCTION .................................................................................................... 11 II. THEORETICAL BACKGROUND ........................................................................ 13 The Tradeoff in Self-Disclosure ............................................................................ 13 The Roles of Being Relatable and Aspirational ..................................................... 15 III. OVERVIEW OF STUDIES ................................................................................... 18 Study 1: Empirical Analysis of Social Media Data ............................................... 19 Sponsored Post Moderation ............................................................................. 25 Study 2: Qualitative Interview on Being Personal ................................................ 27 Study 3: Survey of Two Qualities with Different Self-Disclosure Depth ............. 34 Study 4: Experiment Examining Dual Path Mediation.......................................... 38 Study 5: Temporal Variations in Self-Disclosure and Brand Positioning ............. 44 Study 5-1: Present versus Past Self-Disclosure on Dual Mechanism .............. 45 Study 5-2: Brand Positioning Moderated Mediation ....................................... 48 IV. GENERAL DISCUSSION .................................................................................... 50 Theoretical and Managerial Implications ............................................................. 51 Limitations and Future Research Directions.......................................................... 53 Conclusion ............................................................................................................. 55 APPENDICES ............................................................................................................. 56 A. STUDY 1 ROBESTNESS CHECKS ............................................................... 56 B. STUDY 4 STIMULI ......................................................................................... 60 C. STUDY 4 MEASUREMENT SCALES ........................................................... 62 7 D. STUDY 5 STIMULI ......................................................................................... 63 REFERENCES ............................................................................................................ 66 8 LIST OF FIGURES Figure Page 1. Figure 1: Conceptual framework ........................................................................... 17 2. Figure 2: Nonlinear relationship between depth of self-disclosure and likes ........ 25 3. Figure 3: Sponsored post moderation effect .......................................................... 26 4. Figure 4: Examples of low and high self-disclosure posts .................................... 31 5. Figure 5: Two different types of influencers and self-disclosure depth ................ 36 6. Figure 6: Contrasting characteristics of two types of influencers .......................... 37 7. Figure 7: Dual path mediation between self-disclosure depth and endorsement attitudes ................................................................................................... 43 8. Figure 8: Temporal variations in self-disclosure on dual mechanism ................... 47 9. Figure 9: Optimal self-disclosure depth ................................................................. 51 9 LIST OF TABLES Table Page 1. Table 1: The effects of self-disclosure depth on likes ........................................... 22 10 INTRODUCTION “Obviously, all relations which people have to one another are based on their knowing something about one another” Simmel (1964). Social media influencers are increasingly sharing their personal information, as well as feelings and emotions when advertising brands through a post. Whether this trend is voluntary or demanded by brands, prior research has shown that such disclosure enables them to better connect with audiences and capture their attention. Numerous studies provide evidence of the role of friendship-building in enhancing influencer marketing effectiveness (Chen 2016; Chung and Cho 2017; Kurtin et al. 2018; Labrecque 2014; Rihl and Wegener 2019). Unveiling oneself to others is thought to be an important step in the development of intimate relationships (Kaplan and Haenlein 2010). In line with this reasoning, marketing researchers recently started to investigate how influencers’ self-disclosure affects consumer attitudes and behaviors. In general, self-disclosure appears to have a positive impact on businesses. Prior research looking into the consequences of self-disclosure on social media demonstrates that it improves trust and product attitudes (Huang 2015), viewer attention (Baruh and Cemalcilar 2015), and consumer-brand connections (Leite and Baptista 2021). However, self-disclosure does not always lead to desirable results. Baruh and Cemalcilar (2015) indicate negative outcomes such as diminishing attraction. Lin and Utz (2017) find that the appropriateness of self-disclosure restricts the perception of social attraction and closeness. In the context of travel influencers, AlRabiah et al. (2022) imply only the negative effects of self-disclosure, which is mediated by perceived appropriateness. Source credibility also reduces as a result of their intimate self-disclosure (Leite et al. 2022). In summary, the 11 nascent but growing literature on social media influencers’ self-disclosure points to mixed findings. Most of the studies that uncover negative impacts focus on the appropriateness of self- disclosure as the mechanism. In other words, high disclosure is not considered adequate and thus weakens consumer attitudes. Low disclosure, however, enhances consumer attitudes due to its appropriateness. Such binary theorizing could be contributing to the conflicted findings. Leite et al. (2022) suggest a different mechanism: the need for relatedness. But their findings are predicated on the premise that the need for relatedness is predicted by how appropriate self- disclosure is. I conclude that although self-disclosure on social media is prevalent as a promotion tool by influencers and marketers alike, my knowledge of such disclosure’s impact and the mechanisms underlying it are limited. The current research describes linkages between self-disclosure and the underpinning as a continuum rather than a dichotomy and seeks to postulate a more comprehensive theoretical model that resolves mixed findings regarding self-disclosure in the marketing literature. Specifically, I focus on a tradeoff in the depth of self-disclosure: high disclosure may increase the relatable side of influencers at the expense of their aspirational side, and low disclosure may have the opposite effects. Building on this, I propose that high (vs. low) self-disclosure has a marginal impact on downstream consequences with the two paths canceling each other out as a result of the tradeoff. I identify the two conflicting paths—being relatable and being aspirational—as an underlying mechanism through which the depth of self-disclosure affects consumer attitudes. The findings of this research produce three major contributions. First, it advances the marketing literature on social media influencers by uncovering the novel mechanism of the self- 12 disclosure effects in businesses. The dual-path model serves to reconcile the contradictory findings to provide theoretical implications for marketing research. Second, I add to psychology research on self-disclosure by revealing that the relationship between self-disclosure and respondent reactions is nonlinear. Although this nonlinearity has been suggested in earlier studies, such investigations took place in a lab setting with face-to-face participant interactions (Cozby 1972; Worthy et al. 1969). Extending this research, I provide empirical evidence for the generalizability of the nonlinear relationship to online communications. Lastly, the findings will inform influencers and brands about how much personal information to share. Mega influencers such as celebrities who establish their value from their profession may seek to open up a little about their personal lives rather than keep everything in the veil in order to balance their relatable and aspirational sides. Nano or micro-influencers, on the other hand, may benefit from gradually revealing less about themselves as they move up the ladder of influencer status. I offer four multi-method studies that investigate the proposition that the relationship between self-disclosure and public reactions will not be linear and be mediated by relatable and aspirational aspects, which conflict with one another. In the following sections, I review the existing work on self-disclosure and relevant influencer attributes that facilitate consumer attitudes, from which I derive my propositions. THEORETICAL BACKGROUND The Tradeoff in Self-Disclosure 13 According to Altman and Taylor (1973), self-disclosure is the voluntary sharing of emotions, feelings, or other details that the discloser may feel vulnerable revealing since they are regarded to be private. Prior research studies the role of self-disclosure from a variety of angles and reveals its benefits. Self-disclosure has been shown to improve interpersonal relationships in verbal and nonverbal communications (Altman and Taylor 1973; Taylor et al. 1969), reciprocity (Levin and Gergen 1969; Worthy et al. 1969), likings (Fitzgerald 1963; Halverson and Shore 1969; Jourard and Lasakow 1958), and social approval (Taylor et al. 1969), all of which may be required for social media influencers. Self-disclosure plays an important role in influencer marketing as users rely on it to guide their evaluations (Djafarova and Trofimenko 2019). Since many followers want to learn about and feel a deep sense of connection to influencers, revealing oneself has become a strategy for gaining audiences and advertising brands (Bickart et al. 2015; Hoos 2019). However, the impacts of self-disclosure may not be straightforward. Dienlin and Metzger (2016) assert that self-disclosure requires a cost-benefit tradeoff, indicating that all the positive roles mentioned above may come at a price. The social penetration theory introduced by Altman and Taylor (1973) supports this argument. They initially compare the self-disclosure process to peeling back an onion’s layers: the outer layers of the onion correspond to surface characteristics of a person, such as appearance and voice, and the inner layers exhibit more private information, such as the person’s feelings, thoughts, and interpersonal interactions. As a person self-discloses to a friend or partner, she peels away the outer layers of herself to reveal her true nature. Importantly, Altman and Taylor (1973) propose a stage at which communicators may perceive the cost of self-disclosure outweighs its benefits, suggesting a turning point of self-disclosure. 14 Such change in signs from positive to negative is evidenced by other research. In an experiment using a hypothetical other, Cozby (1972) demonstrates that overly intimate information leads to anxiety, which makes the high discloser appear to be offering the subject negative results. According to Levin and Gergen (1969), medium levels of disclosure are optimal, while people conversing a lot about themselves may be viewed as indiscreet and unreliable. In the context of social media, there is indirect support for the nonlinear self-disclosure effect. Online interaction is more conducive than face-to-face to expressing one's true self, real aspects of the self that are not typically revealed to others (Bargh et al. 2002). Furthermore, people tend to self-disclose more when using their mobile phones as opposed to personal computers (Melumad and Meyer 2020). These facts lead me to suspect that self-disclosure on social media may be more excessive and uncontrolled than in person. The fact that self- disclosure in face-to-face interactions indicates diminishing returns of the subject reactions implies that the self-disclosure effect on social media could be highly nonlinear. Drawing on these research findings, I propose that self-disclosure and audience responses on social media have a nonlinear relationship, such that depth increases are associated with more likes at low levels of depth, but this positive association weakens as depth develops and eventually turns negative. The Roles of Being Relatable and Aspirational An important question has yet to be answered: What can explain the nonlinear relationship of self-disclosure depth and likes on social media? The extant research contradicts the idea of classifying self-disclosure as either positive (appropriate) or negative (inappropriate) 15 and of such binary categorization determining consumer attitudes. Ferchaud et al. (2018) uncover that the effect of self-disclosure does not depend on disclosure valence, such that negative, positive, or neutral disclosure increases perceived realism and authenticity as opposed to the absence of negative, positive, or neutral disclosure, respectively. Therefore, the mechanism accounting for the nonlinear self-disclosure effect may be more nuanced and multifaceted. Before exploring the mechanism, I first look at unique yet ambiguous characteristics of influencers, as these characteristics may alter how audiences perceive them and their disclosure. Influencers are a cross between peers and celebrities (Johnson et al. 2021). Their competitive advantage over traditional endorsers stems from this hybrid component. Although similarity to peers contributes in a special way to their credibility, they are highly admirable individuals that people look up to (Freberg et al. 2011; Moran and Muzellec 2017). All in all, because they are relatable (like peers) and aspirational (like celebrities), influencers can shift followers’ attitudes and behaviors. The importance of relatability in marketing is widely known. Research has demonstrated that a close identification with others increases their credibility, persuasiveness, and information flow (e.g., Brown and Reingen 1987; Djafarova and Rushworth 2017). In practice, using “someone like me” to promote brands is one of the most efficient tools to engage with Generation Z, those between the ages of 11 and 26 (Miachon 2018). Larson (2019) supports this argument by elucidating, “I like what I see in others – especially if what I see in others is ourselves.” 16 Figure 1. Conceptual framework Research on aspirational traits is relatively scarce due to its dynamic and complex nature. Ray (2006) proposes the “aspiration window,” indicating that aspirations emerge from uneven power relations and are motivated by the ideals and achievements of those who are present in one’s aspiration window. This explanation supports the idea that aspirational influencers could be effective in social media marketing since their followers rely on them to fulfill the intended function in relationships that appear to be unequal (Dahl 2021). Be Like Mike, a Gatorade advertisement in which Michael Jordan plays basketball with kids who emulate his moves, highlights the impacts of marketing that uses an idealized influencer (Rovell 2016). If an influencer falls within one’s aspiration window, an individual can accept the aspirational guidance to attain social acceptance, which also helps one’s self-concept improve. All in all, both relatable and aspirational aspects could improve attitudes toward influencers and businesses associated with them. In general, the two qualities can be compatible, with one characteristic predicting another (Hoffner and Cantor 1991). Based on this compatibility, Schouten et al. (2020) introduce 17 “identification,” which includes relatable and aspirational characteristics as they examine the relationship between endorser type and advertisement effectiveness. When self-disclosure comes into play, however, the two sides begin to contradict each other. In regard to self-disclosure, influencers on social media face inevitable tensions between sincerity on the one hand and impeccability on the other hand (Blum 2019). As noted, self-disclosure—even if it has negative information—enhances relatability (Ferchaud et al. 2018). However, it may come out as lacking in professionalism and competence, which would undermine an influencer’s aspirational quality (Orben and Dunbar 2017). Low self- disclosure or self-representation, a form of communication intended to project a desired self- image, may have the opposite effect by creating doubt about perceived similarity while boosting professionalism and expertise (Johnson 1981; Johnson et al. 2020). In addition, social influence’s aspirational side occurs when perceptions or beliefs are adopted to support one’s self-concept (Argo 2020). For audiences, a perception of an individual as relatable and aspirational may be subject to how much influencers reveal about themselves. It is apparent that there is a conflict between the positive effects of self-disclosure on being relatable and the negative effects of self- disclosure on being aspirational. To that end, I contend that when evaluating influencers’ self- disclosure, the relatable and aspirational aspects of identification should be disconnected. Considering these findings from previous studies, I hypothesize that self-disclosure impacts endorsement attitudes along the two separate paths, producing a limited overall effect where the advantages of one path are counterbalanced by the disadvantages of the other. OVERVIEW OF STUDIES 18 In five studies that employ a multi-method approach, I offer evidence for my conceptual model. Each study uses a different operationalizations of self-disclosure: self-disclosure by automated text analysis (Study 1), influencers’ insights on sharing personal information (Study 2), perceptions of self-selected celebrities (Study 3), randomly assigned self-disclosure depths (Study 4), and self-disclosure depth manipulated by temporal differences and positioning moderation (Study 5). The studies together demonstrate that greater self-disclosure depth does not always result in better engagement or more positive endorsement attitudes. Study 1 suggests a quadratic relationship between self-disclosure depth and user engagement, indicating an inverted U-shaped pattern. With real influencers sharing their thoughts and opinions on self- disclosure in social media, Study 2 indirectly confirms the findings of Study 1 and explains as to the mechanism of the nonlinear relationship. Study 3 demonstrates that the extent of their disclosure might result in the different attributes between relatable and aspirational influencers I examine, providing a foundation for my proposition. The underlying mechanism of the nonlinear relationship between disclosure depth and endorsement attitudes is shown in Study 4, when higher disclosure increases perceptions of one being relational, while simultaneously lowering perceptions of one being aspirational, together resulting in only a marginal overall effect. Study 5 that employed temporal variations in self-disclosure complements my previous findings and reaffirms the trade-off between the dual mechanism. STUDY 1: Empirical Analysis of Social Media Data My first study uses automated text analysis to examine the effect of disclosure depth on user engagement, as measured by number of likes. I predict that posts with deeper personal disclosure will not always gain more likes, showing a more nuanced pattern. 19 Method First, I collected social media data from fourteen renowned celebrities’ Twitter and Instagram accounts. Those celebrities are Taylor Swift, Selena Gomez, Ellen DeGeneres, LeBron James, Kylie Jenner, Kim Kardashian, Justin Bieber, Kevin Hart, Miley Cyrus, Katy Perry, Jennifer Lopez, Drake, Demi Lovato, and Rihanna. I selected celebrities with follower counts on both platforms among the top 100 worldwide as of early April 2022 with the idea that they would be active in posting and engaging with users. The majority of them are musicians, but there are also actors, media personalities, an athlete, and a TV host. At the beginning of August 2022, I scraped 16,493 posts from their accounts, spanning from January 1st, 2021, to June 30th, 2022. To ensure steady like counts following data scraping, I had approximately a month gap between the last posting date in the dataset and the date of data collection because the effects of posts may continue for several days (Instagram 2021). Second, I measured the depth of self-disclosure on each post in the dataset. I drew on prior research (Melumad and Meyer 2020) that indicates six linguistic markers or writing styles of greater self-disclosure in text. According to the research, texts that are deeply personal frequently employ the following terms: 1) first-person pronoun (such as I, me), 2) family, 3) friends, 4) words that describe emotions—particularly negative emotions, 5) a more authentic writing style, and 6) a less analytical writing style. These widespread language indicators served as the basis for algorithms created to automatically capture the extent of disclosure in online communications. Building on the existing literature, I examined the 16,493 posts using Linguistic Inquiry and Word Count (LIWC; Boyd et al. 2022), which includes dictionaries for each of the six indicators (first-person pronouns, references to family and/or friends, negative 20 emotions, authentic writing style, and analytical writing style). The dataset was first cleaned up by removing the URL linkages from each post. After reversing the analytical style variable, I standardized each variable by subtracting the mean and dividing that by the standard deviation. Then, I took the average of all, which I named a depth of disclosure. As a final step, I centered the minimum to make zero the base disclosure depth before excluding posts without text (i.e., word count = 0) from the analysis. To test whether greater disclosure depth leads to more likes, I employed a fixed effect negative binomial regression. Like count in the dataset is an over-dispersed count variable, where its variance (SD = 1,376,719) is substantially higher than its mean (491,863). Additionally, I constructed a squared term of disclosure depth and examined the possibility that the disclosure depth effect on likes would not be linear. A more nuanced effect may be revealed if the coefficient of the squared term of disclosure depth is significant. For the squared variable, I estimated the following equation: Log (Likesiym) = 𝛽1𝐷𝑖𝑠𝑐𝑙𝑜𝑠𝑢𝑟𝑒 𝐷𝑒𝑝𝑡ℎ𝑖𝑦𝑚 + 𝛽2𝐷𝑖𝑠𝑐𝑙𝑜𝑠𝑢𝑟𝑒 𝐷𝑒𝑝𝑡ℎ 2 𝑖𝑦𝑚 + 𝛼𝑖 + 𝛾𝑦 + 𝛿𝑚 + 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖𝑦𝑚 + 𝜀𝑖𝑦𝑚, where i represents an influencer, y indicates a year, and m illustrates a month. 𝛼𝑖 is an influencer effect, 𝛾𝑦 is a year effect, and 𝛿𝑚 is a month effect. 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖𝑦𝑚 include a log of word count, log of follower count, platform (Twitter vs. Instagram), text-only post (1 if no images are included, 0 otherwise), hashtag count (#), mention count (@), and weekend post (1 if posted on Saturday or Sunday, 0 otherwise). I excluded several factors that are available only on 21 either Twitter or Instagram. The linear effect of disclosure depth is captured by 𝛽1, and the nonlinear effect of disclosure depth is captured by 𝛽2, the main object of interest in Study 1. Table 1. The effects of self-disclosure depth on likes Results 22 As predicted, posts with more disclosure depth did not necessarily receive more likes (b1 = 0.331, se = .136, p < .05; b2 = -0.098, se = .045, p < .05, see Table 1 and Figure 2). Increases in disclosure depth were associated with greater likes at low levels of depth, but this positive association became weaker as depth increased and eventually turned negative. Even as I included more controls, the effects were stable and consistent. One could wonder if negative emotions were driving the effects. Moderate level of negative emotions might be ideal because a low level of negative emotions might not be engaging, and a high level of negative emotions might be inappropriate. To address this concern, I constructed a depth of self-disclosure without negative emotions and one with positive emotions and negative emotions. Results were robust to different methods of measuring disclosure depth. In other words, the effects I uncovered were not influenced by valence alone. I also addressed another concern that the self-disclosure effects might vary depending on influencer status. In other words, users are likely to receive mega influencers’ self-disclosure differently than micro influencers’ self-disclosure. Self-disclosure by a higher status person undermines influence and relationships while self-disclosure by a peer does not show those negative outcomes. (Gibson et al. 2018). Given that micro influencers are more akin to peers than mega influencers, one could wonder if the self-disclosure effects would only be linear and positive for micro influencers. To deal with such possibility, I gathered more than 10,260 social media posts from fourteen micro influencers, and examined whether the interaction term of influencer status and depth of self-disclosure were statistically significant. The regression results demonstrate that although the inverted U-shape relationships still emerge, the interaction term is not significant, indicating that influencer status does not shift how users react to depth of self- disclosure. 23 Likes as a dependent variable could also raise further questions in the sense that users may choose to like posts even without reading them. I focused on like count, a widely used metric that enables viewers to express their approval of the content without making a comment (Li and Xie 2020). One could raise a concern that likes could be misleading because anyone may click “like” without actually being interested in what influencers are saying. Users invest more time and effort in commenting or sharing than they do in liking a post, making one of those actions a more accurate measure of user involvement (Dwoskin 2022). I addressed this concern by using other variables such as comments and sharing in lieu of likes. I examined the self- disclosure effects on comments with Instagram posts. For sharing, I regressed the retweet count on the depth of self-disclosure and its quadratic term. The results were robust to different dependent variables and remained quite stable. Lastly, one could also question if women influencers, who make up the vast majority of the dataset, were responsible for the outcomes. Prior studies that discovered the nonlinear relationship between self-disclosure and liking used female subjects in their experiments (Cozby 1972; Worthy et al. 1969). A study with male participants shows that there is no difference in liking between low and high self-disclosure (Ehrlich and Graeven 1971). To test the potential gender effect, I exclusively analyzed four male celebrity influencers. Even though my sample size substantially shrank, the results remained reliable and stable. I also carried out a different robustness check regarding gender. Instead of running an analysis with only males, I controlled for gender. Again, the results were consistent. This shows that, at least for social media communications, the nonlinear self-disclosure effects are generalizable to both genders (see Appendix A). 24 Figure 2. Nonlinear relationship between depth of self-disclosure and likes Sponsored Post Moderation To further explore the disclosure depth effect on user engagement, I additionally took post type into consideration. Specifically, I explored how the non-linear relationship between depth of self-disclosure and likes is affected by sponsored posts as opposed to non-sponsored posts. This might be a more relevant question for marketers and influencers who are attempting to promote products or services via social media. I had three research assistants classify each of the over 16,000 social media posts into either sponsored or non-sponsored posts. When there was a disagreement, I used the majority rule to determine whether a post is sponsored. I ran a regression with the two interaction terms: disclosure depth and sponsored posts and the square of 25 disclosure depth and sponsored posts. The outcome variable was likes. My focus is whether the disclosure depth effect varies with the type of post. The results demonstrated that the inverted U-shape relationship still existed for both sponsored and non-sponsored posts, replicating the outcome I previously displayed. However, as I included more covariates, the relationship was substantially attenuated for sponsored posts, which were flattening the curve (b3 = -0.495, se = .118, p < .001; b4 = 0.093, se = .048, p < .10; see Figure 3). I assume that consumers activate their persuasion knowledge when they notice sponsorships or partnerships in postings, making them perceive that influencers and posts are not as sincere. Importantly, disclosure depth leading to the highest level of engagement was much lower for sponsored posts, Sponsored posts reached the highest point at about 1.5 disclosure depth while non-sponsored posts peaked at 2.5 disclosure depth. Figure 3. Sponsored post moderation effect 26 Discussion An automated text analysis of more than sixteen thousand social media posts provides preliminary evidence on how disclosure depth affects user engagement in online communications. The results support the nonlinear pattern hypothesis, implying diminishing returns of like count. Initially, like count increases as disclosure depth increases but begins to decline after a certain point. This relationship continues to hold even as I include other variables that might explain the variations in like count. Robustness checks ruled out a few alternative explanations that valence might be the primary cause of the effects, and that influencer status might be responsible for the results. I also demonstrated that the nonlinear self-disclosure effects are generalizable to other outcome variables and all genders. Additionally, I tested the moderation effect of sponsored posts, specifically how those posts impact the nonlinear relationship between depth of self-disclosure and user engagement. I revealed that sponsored (vs. non-sponsored) posts weaken the impact of self-disclosure depth on likes, indicating that the non-linear effect is particularly the case for posts purely about influencers themselves. In sponsored posts, the self-disclosure effect turned in negative direction much sooner, so that influencers should exercise greater caution in the amount of information they disclose when endorsing brands. STUDY 2: Qualitative Interview on Being Personal The findings of Study 1 are consistent with my postulation, but they provide only preliminary evidence for a nonlinear relationship between disclosure depth and user engagement. Furthermore, the data analyzed are from social media accounts of worldwide celebrities whose 27 main revenue sources are not social media activities. To reaffirm if Study 1’s results align with the thoughts of influencers and to acquire deeper insights into personal sharing on social media, I conducted a qualitative study using in-depth interviews. Method My interviewees were three student influencers and one general influencer. Among those student influencers, two were attending the University of Oregon (UO), and one was a third-year student at the University of California, Riverside (UCR). Cami Wilson was a UO college athlete on the acrobatics team promoting various brands, including skincare products, clothing, and dating apps on Instagram. As of July 1st, 2021, college athletes can take advantage of name, image, and likeness (NIL) opportunities and receive compensation for their NIL (NCAA 2021), and brands approached her to involve and support female student-athletes in their marketing. She had 31,000 followers on Instagram. The other UO student, Tia Bagha, was a general non-athlete student who had chosen to specialize in health and beauty. She gained 330,000 followers on TikTok for workout, fitness, and vlog style videos. Calvin Chang, a UCR non-athlete influencer with about 35,000 followers, had produced daily outfit vides on TikTok with an emphasis on fashion-oriented content. Katie Prentiss, who introduced her as actor, filmmaker, and photographer, was also interviewed. She had 14,000 Instagram followers, the majority of which she said were females in their 30s or 40s. Each of the four was perceived as micro-to-macro influencers. During semi- structured interviews via video calls that lasted between 30 and 40 minutes in 2023 and 2024, I listened to their thoughts on personal disclosure on social media. With their consent, formal interviews were recorded and then transcribed. 28 Results The themes that arose in the interviewees’ responses supported my proposition and were consistent with my findings in Study 1: social media endorsement may not always benefit from self-disclosure. First, all of them concurred on the importance of showing the personal side of oneself on social media as a content creation strategy: I really try and think of my overall brand as three different buckets of my life. I have my Oregon acrobatic bucket. I have just my lifestyle, my friends, and my family posting together. And then I also have my cheerleading background because that will always be a part of my life. The way I go about it is, just looking at it, I never want one to be too heavy than the other, but obviously when I'm in acrobatic season, there will be more acrobatic, but I always will make sure ‘okay, I've had a bunch of acrobatic stuff. Maybe it's time to share a photo of my friends and family, see what we're up to there.’ And by doing so, it kind of lets everybody get a full picture of my life. Rather than just being known as just a cheerleader, just an acrobatic athlete, and just being able to share those tidbits of my life, it creates a rounder picture of who I am. (Cami Wilson) I am very careful about making sure that nothing gives me a bad image, especially because people can take things out of context, and you can get canceled very quickly these days. They can get at you for all sorts of things. So just posting content that you know is not going to trigger anyone. It really depends on your preferences of what you like to post, but it also depends on what makes you you and what really shows the people 29 who you are. That's like the kind of content that aligns with everything you stand for and your niche. (Tia Bagha) I honestly don’t show too much personality on my account. I still feel like people don’t really know who I am. But I think that showcasing yourself more definitely has a potential of attracting more brands. If I was to put more of my personality out there, it would open up a new area for brands, and a new different kind of brands would reach out to me, rather than just fashion. I think it would open a lot more doors. (Calvin Chang) Although Tia and Calvin seemed more questioning of revealing personal information than Cami, I confirmed that expressing oneself is a basic strategy that helps the influencers gain likes and followers. Calvin mentioned another important point. Sharing personal details and feelings might allow influencers to diversify their partnership portfolio because showing more of oneself can appeal to different companies. Katie explained that “the more humanity, the better”. Second, all suggested two-sided consequences of being personal on social media in line with my findings and hypotheses. According to them, almost all the brands they have worked with wanted them to be personal and share their feelings in promoting their products or services, but they were unsure if that might be good or effective. All the interviewees were aware of the pros and cons of sharing personal information on social media. I found that despite firms’ desires for influencers to be more personal in promoting their brands, they were cautious of crossing the fine line between “enough” and “too much.” They agreed that it is critical to create content that relates to viewers without irritating them. 30 There are a lot of brands that value being real and raw. You think about campaigns for women who show authentic bodies and embrace aging. I care about positivity and sharing hopeful and optimistic perspective with people, but I also feel the responsibility to have people know that I fail, struggle, and am insecure, too. It is such a fine line. I don’t know how much to share, what part is my responsibility, and what part is the viewers’ responsibility, how we perceive each other, and how we compare and judge all those things. (Katie Prentiss) If you are more personal and revealing, it could lead to people following you more and really liking you for who you are. And that's the case with a lot of social media influencers. But it could also hinder you in ways because if they know more about you, they have more leverage on you to be negative. I think that's one of those things where you're more vulnerable if you really show who you are. But again, if you're going full into TikTok, this is your life. (Tia Bagha) I protect myself in certain ways, but I do at times get extremely personal in terms of just sharing my background, sharing my story, and I think that it really depends on every creator's comfortability with it because at the end of the day, audiences can be brutal. (Cami Wilson) Figure 4. Examples of low and high self-disclosure posts 31 Illustration of low self-disclosure Illustration of high self-disclosure 32 Third, they implied a possible mechanism that underlies the inverted U-shape pattern I observed in Study 1. They brought up the concept of being relatable and aspirational, which might explain a positive and negative association between disclosure depth and user attitudes, respectively. Tia argued, “Getting emotions out could work in people relating. You can relate to somebody and see, ‘Oh, they are going through the same thing.’” Cami indicated, “Revealing creates a deeper connection with people and that happens when you share those not-so-perfect moments in your life.” Revealing not-so-perfect moments might in turn reduce a perception of one being aspirational. Katie brought up how she changed from being relatable to being aspirational. “As my follower numbers have grown on social media, I’ve gotten more protective of some of my personal stuff. I took a lot of my family stuff down off of my account. I share some little tidbit that might feel encouraging to other people or might make people think their perspective.” Calvin summarized that “I don’t think there is too much benefit in posting too much of your life. I just show off products and don’t think any brands see me in a negative way. There has got to be a balance.” Discussion Study 2 demonstrates the uncertainty and tradeoffs influencers face when deciding how much to share. Both interviewees address opportunities and risks of personal disclosure on social media, in line with my Study 1 findings. They suggest revealing oneself will increase one’s popularity and grow one’s network of followers, but it will also expose one to potential criticism. That is, one’s choice to disclose personal information means compromising one's self-image in exchange for favorable attention. 33 In addition, I had a conversation with Tyler Fairchild, the chief executive officer of Courser, an Italian handmade sneaker about personal sharing. It allowed me to understand that brands see influencers’ personal sharing from a different perspective. “Influencers are creating the whole persona. We're hypervigilant that they're staying true to the image that they're portraying, which is very important for a new luxury brand like us. I really want to make sure that they eat, sleep, and breathe what they're representing. They have to be in the right places with right friends. We don't want them kind of veering into a space that is counterintuitive to our message. It is trying to walk that fine line”, He said. Marketers are cautious of whether the lifestyles and messages of influencers seem to be genuine, and in line with their brand image. Put another way, influencers' ability to convey their true selves—that is, the consistency between their online persona and real life—is what counts most to marketers. STUDY 3: Survey of Two Qualities with Different Self-Disclosure Depth Study 3 uses a survey to look at how relatable and aspirational influencers are perceived on a range of attributes. Drawing on Study 2 and extant literature, I categorize influencers into two categories: relatable influencers who share a lot about themselves and connect with followers and aspirational influencers who share less about themselves and manage impressions. Two types of influencers driven by the difference in the extent of their disclosure might lead consumers to evaluate them differently. Study 3 aims to confirm my prediction that relatable (vs. aspirational) influencers tend to reveal more about themselves than aspirational influencers, leading to divergent views about them. Specifically, I predict that relatable influencers will be 34 perceived as more open and vulnerable, whereas aspirational influencers will be seen as more idolized and manufactured. Method Ninety-six participants were recruited from Prolific (Mage = 35.8, 50.0% female). Initially, I had one hundred participants, but four of them who said they do not follow influencers on social media were removed from the study at the very beginning. The rest of the participants were first asked to “Think of two influencers that are similar in terms of how much you like them but differ in why you like them. One is easy to understand, like, and have sympathy for because of their similarities to you or your own experiences. The other is associated with appealing qualities, success, and status that make you desire to become like them or to enjoy imagining being like them.” The first was referred to as a relatable influencer, and the second as an aspirational influencer. Next, participants rated the degree to which each of the two influencers listed reveals personal information, frailties, faults, vulnerabilities, flaws, and is transparent in their media posts and appearances on an 8-point scale ranging from “Not at all” to “Very much so.” They also indicated the extent to which each of the two maintains a level of mysteriousness, their media posts and appearances look to be curated displays, and those influencers appear to be intended to impress on the same 8-point scale. Last, they indicated whether each of the two influencers’ media posts is more like a contribution to an autobiography or a mythology and how much participants idolize each of them. I gathered the participants’ age and gender. Results 35 A one-way ANOVA indicated that relatable influencers and aspirational influencers differed in a variety of attributes. As predicted, relatable influencers were thought to reveal more personal information or feelings than aspirational influencers (Mrel = 4.56, SDrel = 2.41 vs. Masp = 4.03, SDasp = 2.26; F(1.95) = 5.21, p < .05). Due to the variations in the disclosure levels across the selected influencers, I controlled for participants. I found that participants perceived relatable influencers as more frail (Mrel = 5.03, SDrel = 2.20 vs. Masp = 4.43, SDasp = 2.03; F(1.95) = 8.65, p < .01), more faulty (Mrel = 5.00, SDrel = 1.97 vs. Masp = 4.05, SDasp = 2.01; F(1.95) = 16.11, p < .001), more vulnerable (Mrel = 5.44, SDrel = 2.36 vs. Masp = 4.47, SDasp = 2.08; F(1.95) = 16.82, p < .001), more flawed (Mrel = 5.00, SDrel = 2.08 vs. Masp = 3.85, SDasp = 1.91; F(1.95) = 31.35, p < .001), and more transparent (Mrel = 6.83, SDrel = 1.23 vs. Masp = 5.95, SDasp = 1.85; F(1.95) = 17.54, p < .001). The depth of self-disclosure was positive and significant for those five attributes, too. Figure 5. Two different types of influencers and self-disclosure depth 36 Furthermore, I learned that participants perceived aspirational influencers as more mysterious (Mrel = 3.56, SDrel = 1.83 vs. Masp = 4.52, SDasp = 2.03; F(1.95) = 16.77, p < .001), more curated (Mrel = 3.91, SDrel = 2.13 vs. Masp = 5.17, SDasp = 2.37; F(1.95) = 26.32, p < .001), more intended to impress (Mrel = 4.00, SDrel = 2.02 vs. Masp = 5.41, SDasp = 2.15; F(1.95) = 40.70, p < .001), and more mythical (Mrel = 2.59, SDrel = 1.71 vs. Masp = 3.78, SDasp = 1.94; F(1.95) = 24.42, p < .001). Aspirational influencers were also marginally more idolized than relatable influencers (Mrel = 4.71, SDrel = 2.17 vs. Masp = 5.02, SDasp = 2.23; F(1.95) = 2.94, p < .10). Three (mysterious, curated, and mythical) of the five attributes where aspirational influencers outperform relational influencers were accounted for by disclosure depth. Figure 6. Contrasting characteristics of two types of influencers Discussion 37 In Study 3, I look at and compare two influencers with distinctive disclosure levels on various characteristics in an attempt to delve into the consequences of self-disclosure. Based on the findings, I may infer that relatable influencers and aspirational influencers differ in certain ways due to their self-disclosure levels. The findings indicate that relatable influencers are considered more forthcoming about their susceptibility whereas aspirational influencers are considered less so, presumably preserving their aura and semblance. Studies 2 and 3 together provide convergent evidence for the nonlinear self-disclosure effects and my hypothesis on the mechanism. In Study 4, I seek to display such nuanced effects of self-disclosure in a more controlled setting. STUDY 4: Experiment Examining Dual Path Mediation Study 4 tests the hypothesized mediating role of being relatable and being aspirational between self-disclosure and endorsement attitudes to shed light on the mechanism that underlies the nonlinear self-disclosure effects. Study 3 falls short of supporting my theorizing with concrete evidence as participants self-selected influencers they are following, resulting in a varying degree of disclosure across participants. To address this concern and directly test the causal effect of self-disclosure, I manipulate the depth of self-disclosure in this study, holding an influencer constant. I speculate that disclosure depth affects endorsement attitudes along the two paths of being relatable and being aspirational, but the effects of the two paths cancel out one another, leaving a marginal total effect. Method 38 Following preregistering my study plan, two hundred forty Prolific workers (Mage = 35.2, 51.3% female) completed it in exchange for small cash compensation. For stimuli where I adjusted the disclosure depth, I used and modified tweets from American musician Taylor Swift. Participants were randomly assigned to one of the four conditions: two with low disclosure and two with high disclosure. The two low-disclosure stimuli described her thoughts about either her birthday party or her brother, while the two high-disclosure stimuli shared her personal struggles with mental health issues or eating disorders. I used two stimuli for each condition to mitigate the possible topic effect. To align an image with a message, I kept a photo constant within a condition but utilized different photos across conditions (see Appendix B). Participants were forced to evaluate a tweet for at least 10 seconds before responding to questions. Then, I measured participants’ endorsement attitudes, a main dependent variable, using three items. Participants were asked to indicate how much Taylor Swift’s brand recommendations influence them, how much they like to use the brands that she endorses, and how much they trust her brand suggestions on a 9-point scale (1=Not at all, 9=Very much so). Next, participants indicated the extent to which they find Taylor Swift to be relatable and aspirational, each with three items. To measure the relatable mediator, I asked how much Taylor Swift seems relatable, like a normal person, and a lot like them. For the aspirational mediator, I asked how much Taylor Swift seems aspirational, like an ideal person, and how much they want to be like her on the same 9-point scale (1=Not at all, 9=Very much so, see Appendix C). To help participants understand what those terms (i.e., relatable and aspirational) mean, they read the description of a relatable and aspirational person before answering the questions. Then, I tested a manipulation check. On a 9-point scale ranging from “Not at all” to “Very much so” as earlier, participants rated the extent to which they feel Taylor Swift seems 39 transparent, candid, forthright, vulnerable, at risk of disapproval, exposed to judgment, controversial, offensive, impulsive, flawless, perfect, and impeccable. Because they explain perceptions associated with the depth of her disclosure, I used those items to determine whether participants were sensitive to the levels of disclosure of personal information in social media posts. To investigate potential common method variance, I assessed attitude toward the color blue using 7 items on a 7-point scale as a marker variable. Participants also indicated whether they follow Taylor Swift on social media before reporting their age and gender. Results I excluded two participants who did not pass an attention check question, leaving a final sample of two hundred thirty-eight participants. Before analyzing the data, I inspected whether the two low disclosure posts within the condition or two high disclosure posts within the condition differed on the variables investigated. I did not find significant differences between the two high-disclosure posts or the two low-disclosure posts on them. Thus, I collapsed them into a two-cell design for analysis: self-disclosure depth (low vs. high). Manipulation Check. To check whether my disclosure depth manipulation worked as intended, I compared the averages of those twelve items mentioned above that are relevant to the depth of disclosure on Taylor Swift’s tweets. First, I reversed code flawless, perfect, and impeccable as I expected that, in contrast to the others, each of these items would have a negative association with disclosure depth. Second, I ran a one-way ANOVA after averaging all the items. The result showed that participants in the high disclosure condition had a greater 40 average score than those in the low disclosure condition (Mlow = 4.83, SDlow = 0.97 vs. Mhigh = 5.15, SDasp = 0.93; F(1.236) = 6.70, p < .05). Measurement Model. I proposed that two independent factors, namely being relatable and being aspirational, would explain the covariance in the measures. One could wonder if being relatable and aspirational might be a single factor. To ensure that the two factors were distinctive enough to be used separately, I ran a confirmatory factor analysis. The results supported my proposition. Model fit of a two-factor model (CFI = 0.970, RMSEA = 0.128, SRMR = 0.045; 𝜒2(8) = 39.14) outperformed that of a single-factor model (CFI = 0.771, RMSEA = 0.333, SRMR = 0.102; 𝜒2(9) = 246.02). This was supported by an ANOVA test, which suggested that it was worthwhile to give up one degree of freedom for a better fit (p < 0.001). Due to my conceptual model and the two-factor model’s better fit, being relatable and being aspirational were treated as different constructs in my analyses. Direct Effect of Disclosure Depth. Before conducting a mediation analysis, I studied the main effect of disclosure depth. A one-way ANOVA presented that participants in both conditions did not differ in endorsement attitudes (Mlow = 4.11, SDlow = 2.07 vs. Mhigh = 4.09, SDhigh = 2.12; F(1.236) = .007, p = 0.94), indicating disclosure depth itself was insignificant for consumer attitudes. This result aligned with the outcomes I found in Study 1—the nonlinear pattern of declining returns for disclosure depth. That is, high disclosure did not necessarily enhance consumer attitudes toward endorsements as opposed to low disclosure. 41 Dual Path Mediation. I employed a structural equation model (SEM) to estimate the effect of self-disclosure on consumer attitudes via two constructs: relatable and aspirational. There were several reasons for using SEM over other traditional methods. First, I needed to take measurement error into account because items I used for latent variables did not gain empirical evidence to confirm construct validity (MacKenzie 2001). Second, my focus was on latent variables and the relationships between them. Bagozzi and Yi (1999) demonstrate that when theoretical constructs underpin dependent variables, structural equation models can be helpful. Third, measured items across different constructs may be correlated with one another, and, unlike other methods that assume orthogonality between them, SEM adequately accounts for these covariances between measured items in terms of latent variables (Jöreskog and Sörbom 1982). I conducted two steps of analysis to test the dual path mediation. I first constructed three latent variables: endorsement attitudes, relatable, and aspirational. Next, I analyzed the effect of disclosure depth on being relational and aspirational, establishing the front two mediation paths. Then, I regressed endorsement attitudes on two constructs and disclosure depth. The models controlled for the mediators’ influence on each other when assessing the disclosure depth effect on each of the two mediators as one can predict another (Hoffner and Cantor 1991). That way, I could isolate the unique impacts of disclosure depth on the two mediators. I also controlled for age, gender, and blue, the marker variable. 42 Figure 7. Dual path mediation between self-disclosure depth and endorsement attitudes Notes: p-value: * < 0.05, ** < 0.01, *** < 0.001 As shown in Figure 7, I theorized that perceptions of being relatable and being aspirational emerge from disclosure depth, and the two paths are in opposition to one another in terms of their effects. These two perceptions improved endorsement attitudes. SEM analysis of the conceptual model exhibited a decent fit to the data (CFI = 0.950, RMSEA = 0.088, SRMR = 0.133). The results showed a full mediation along the two paths—one through relatable and another through aspirational. High disclosure led to a greater perception of Taylor Swift being relatable compared to low disclosure (b = 0.376, se = 0.151, p < .05), while high disclosure decreased the perception of her being aspirational relative to low disclosure (b = -0.352, se = 0.143, p < .05). Furthermore, both being relatable (b = 0.497, se = 0.133, p < .001) and being aspirational (b = 0.821, se = 0.154, p < .001) increased endorsement attitudes. The effect of disclosure depth on endorsement attitudes again was not significant (b = 0.064, se = 0.188, p = 43 .733). In summary, the results from the SEM analysis indicated that the positive effects of the relatable path on endorsement attitudes were offset by the aspirational path’s negative effects, yielding only the minimal overall effect of disclosure depth in support of my hypothesis. Discussion The results of Study 4 offer insight into the mechanism underlying the nonlinear self- disclosure effects. Consistent with my prior studies, high disclosure (vs. low disclosure) does not lead to greater attitudes toward endorsement. Specifically, a relatable-aspirational tradeoff emerging from disclosure depth impedes greater endorsement attitudes, even though both being relatable and being aspirational serve to increase attitudes. To test a possible concern that this effect might be confounded by participants who have extensive interests and knowledge about Taylor Swift, I conduct SEM analysis without her social media followers in the dataset or controlling for them. The results are stable and consistent (w/o followers: CFI = 0.949, RMSEA = 0.085, SRMR = 0.160; control for followers: CFI = 0.941, RMSEA = 0.085, SRMR = 0.141). These findings demonstrate that endorsement attitudes are not just contingent on pre-existing attitudes toward an influencer but also on the depth of one’s personal information or feelings expressed in communications. STUDY 5: Temporal Variations in Self-Disclosure and Brand Positioning Thus far, I have tested the self-disclosure effect within a post. In study 5, I aim to explore the self-disclosure effect between posts while delving into self-disclosure and offering practical implications. I attempt to examine two propositions: temporal variations in disclosure on the dual 44 mechanism and moderated mediation of brand positioning. I carried out two sub-studies to provide some insight into suggestions for brands in relation with depth of disclosure. First, I investigated the hypothesis that past disclosure, as opposed to present disclosure, strikes a better balance between relatable and aspirational. Present disclosure leads to higher disclosure, which leads to a perception of being more relatable and less aspirational, while past disclosure decreases depth of disclosure, which results in a perception of being more aspirational and somewhat relatable. Second, drawing on the results from study 5-1, I test whether positioning moderates the dual mechanism effect that exists between temporal variations in disclosure and attitudes toward brands. I posit that luxurious brand positioning enhances brand attitudes when influencers use past disclosure, thus being perceived as more aspirational whereas affordable positioning leads to greater brand attitudes with influencers using present disclosure and being perceived as more relatable. STUDY 5-1: Present versus Past Self-Disclosure on Dual Mechanism Method One hundred eighty Prolific workers (Mage = 40.0, 98.9% female) completed a single- factor study. One of the three conditions—one with past disclosure, one with present disclosure, and one with minimal disclosure—was randomly assigned to them. I made an image of a woman jogging beside a river using a generative artificial intelligence program, and I utilized it for all three conditions. Captions differ across conditions. In the past disclosure condition, the influencer explained that she felt good and had healthy habits, but that is no longer the case, 45 adding that she is stress-eating. We used the same content in the present disclosure condition but flipped the time sequence. She said that she felt bad and had unhealthy habits, but that is not anymore. In the no disclosure condition, she just encouraged users to get healthy habits and be at their prime, not revealing her feelings and details (see Appendix D). After evaluating one of those posts, on a 7-point scale (1=Not at all, 7=Very much), participants indicated the extent to which they perceived the influencer as relatable and aspirational, each with two items. I also measured depth of self-disclosure by asking them to rate the extent to which she reveals personal information, feelings, or thoughts on social media on a 7-point scale (1=To a very small extent, 7=To a very large extent). Results I averaged two relatable and two aspirational items, respectively to make a single relatable (𝛼 = .90) and aspirational score (𝛼 = .86). A one-way ANOVA demonstrated that present disclosure increases a perception of being relatable as opposed to past disclosure and minimal disclosure (Mpresent = 4.77, SDpresent = 1.52 vs. Mpast = 3.93, SDpast = 1.72 vs. Mminimal = 3.34, SDminimal = 1.60; F(2.177) = 11.92, p = 0.001). Although there is no significant difference between conditions, past disclosure leads to a greater perception of being aspirational than present disclosure and minimal disclosure (Mpresent = 3.71, SDpresent = 1.70 vs. Mpast = 4.21, SDpast = 1.82 vs. Mminimal = 4.16, SDminimal = 1.63; F(2.177) = 1.52, p > 0.10). When looking into the difference between relatable and aspirational, past disclosure showed a better balance between the two qualities—that is, a difference that was closer to zero—than present disclosure and minimal disclosure and the difference was significant (Mpresent = 1.06, SDpresent = 1.57 vs. Mpast = 46 -0.275, SDpast = 1.54 vs. Mminimal = -0.828, SDminimal = 1.45; F(2.177) = 24.41, p < 0.001, see Figure 8). Figure 8. Temporal variations in self-disclosure on dual mechanism I also looked into depth of self-disclosure across three conditions. Participants gave the present disclosure condition the highest rating for self-disclosure depth with the past disclosure condition coming in second (Mpresent = 3.97, SDpresent = 1.50 vs. Mpast = 2.57, SDpast = 1.48 vs. Mminimal = 1.87, SDminimal = 1.23; F(2.177) = 34.52, p < 0.001). The present disclosure leads to a perception of more relatable, while the minimal disclosure condition is viewed as more aspirational. The past disclosure is in the middle, not skewed in one direction or the other. This outcome was consistent with the findings from the prior studies—the impact of disclosure depth on the trade-off between 47 relatable and aspirational qualities. Building on study 5-1, I went on to further study downstream consequences of positioning, how luxurious versus affordable framing changes the effects of dual mechanism on brand attitudes. STUDY 5-2: Brand Positioning Moderated Mediation Method I recruited six hundred participants from Prolific (Mage = 41.5, 99.3% female) for a 3 (temporal disclosure: present vs. past vs. minimal) by 2 (positioning: affordable vs. luxurious) between-subjects study. Participants first evaluated one of the three conditions used in study 5-1, and subsequently assessed one of the two promotion posts, in which the woman influencer was endorsing sunglasses. One post was framed as affordable sunglasses, while another post was highlighted as luxurious sunglasses. Then, all participants indicated their attitudes toward the sunglasses with the pairs of adjectives (Negative vs. Positive, Dislike vs. Like, Unfavorable vs. Favorable) on 7-point bipolar scales (Grohmann 2009). They also indicated the degree to which she appears relatable and aspirational as in study 5-1. For screening purposes, I asked as to whether they had previously finished similar versions of the study. Results After dropping participants who said to have previously completed comparable versions of this study before, I conducted a moderated mediation analysis with temporal disclosure as the independent variable, relatable and aspirational qualities as dual mediators, sunglasses attitudes as the outcome variable, and positioning as a moderator that impacts the relationships between 48 the dual mechanisms and sunglasses attitudes. I did not find significant results. First of all, past disclosure, directly or through relatable or aspirational quality, did not lead to greater sunglasses attitudes compared to present disclosure and minimal disclosure (direct: b = -0.092, SE = 0.120, 95% CI = [-0.322, 0.139]; through relatable quality: b = 0.027, SE = 0.064, 95% CI = [-0.093, 0.161]; through relatable quality: b = -0.026, SE = 0.034, 95% CI = [-0.121, 0.021]). Second, although directionally consistent with my predictions, positioning moderations with either a relatable or aspirational quality were not significant (interaction of luxurious and relatable: b = - 0.103, SE = 0.093, 95% CI = [-0.279, 0.081]; interaction of luxurious and aspirational: b = 0.121, SE = 0.092, 95% CI = [-0.054, 0.296]. I did find, however, that a relatable quality mediates the effect of present disclosure (vs. minimal disclosure) on attitudes toward sunglasses (b = 0.222, SE = 0.078, 95% CI = [0.092, 0.397]. Additionally, it was discovered that relatable and aspirational qualities improved sunglasses attitudes (relatable: b = 0.367, SE = 0.069, 95% CI = [0.237, 0.506]; aspirational: b = 0.177, SE = 0.071, 95% CI = [0.035, 0.313], which validated the results of study 4, in which I tested the dual mechanism using Taylor Swift’s social media posts. Discussion Study 5 offers incomplete results. I revealed that temporal differences in disclosure lead to varied degree of relatable and aspirational qualities, and past disclosure was perceived as more balanced between the two than present and minimal disclosure. In other words, past disclosure might be viewed as somewhat relatable and somewhat aspirational, which perhaps corresponds to the maximum point of the inverted U-shape relationship. Past disclosure, however, did not result in greater brand attitudes, as opposed to present and minimal disclosure. Brand positioning did not moderate the effects of the dual mechanism on brand attitudes, either. I regressed 49 sunglasses attitudes on temporal variations in disclosure, positioning, and the interaction term of those two variables. Once again, the results were not significant. Study 1 illustrates that in promotional posts, the effects of self-disclosure are substantially muted. This study adds to that by implying that there is limited carryover of the self-disclosure effects to the subsequent promotional post. GENERAL DISCUSSION Across five studies adopting different methodologies, including quantitative and qualitative approaches, I find consistent evidence for the nonlinear effects of self-disclosure. Increases in disclosure depth are linked to greater user engagement or consumer attitudes at low levels of depth, but this positive relationship gets weaker as depth rises. After a certain point, it turns negative, suggesting the inverted U-shaped pattern of diminishing returns. The theme arising from interviews with four micro influencers supports this pattern, suggesting a tension in regard to sharing personal thoughts and follower reactions. Facing a decision on how much to reveal themselves, influencers might have to make a tradeoff: risking negative attitudes toward them for gaining more likings or followers. What can justify this tradeoff? Drawing on the extant literature and the survey exploring distinctive attributes of relatable and aspirational influencers, I find that consumer attitudes are mediated by perceptions of being relatable and being aspirational. Because more disclosure raises relatability and weakens an aspirational characteristic, both paths contradict each other in terms of effects, contributing to the overall null effect. It is worth noting that the non-linear relationship between self-disclosure depth and user 50 engagement was significantly reduced in sponsored posts. Additionally, I discovered that the impacts of self-disclosure do not extend to adjacent posts—sponsored ones in particular. Theoretical and Managerial Implications The current research makes several theoretical and managerial contributions. Extending existing literature on how self-disclosure affects audience reactions, I shed light on the generalizability of the nonlinear self-disclosure effects. I show that in the context of social media, self-disclosure increases liking from low to medium disclosure but decreases it from medium to high disclosure. Second, this research identifies theoretically and managerially relevant mediators (relatable and aspirational aspects) that have been understudied in marketing literature. I disentangle two characteristics rather than tying them together and demonstrate that the effects of self-disclosure on consumer attitudes are a function of being relatable and being aspirational, which have opposing effects as a result of varied levels of self-disclosure. I also uncovered that time variations in self-disclosure leads to different perceptions of being relatable and aspirational. Past disclosure achieves a better balance between the two qualities relative to present disclosure, which raises relatability but drops aspirational perception. Thus, influencers might want to bring their past struggles and hardships up rather than current ones in order to be viewed as approachable and inspiring figures. This research also helps resolve the mixed findings that marketing research on self-disclosure has exhibited. Figure 9. Optimal self-disclosure depth 51 Third, I provide influencers and brands with practical guidelines on the level of self- disclosure in creating content on social media. I find that brands appear to believe that more disclosure is beneficial for promotions compared to less disclosure. My results, however, indicate a tradeoff between more and less disclosure. Brand managers may first want to assess influencers’ standing to determine whether they are perceived more as celebrities (aspirational) or peers (relatable). Based on that, managers can communicate with influencers on how much to share to balance out ordinary and extraordinary aspects of themselves. For example, nano influencers having relatively few followers can benefit from sharing more about themselves at first and shifting toward less revealing as they gain broader reach. Overall, a moderate level of disclosure positions influencers well, increasing the likelihood that consumers will regard their 52 message to be more persuasive and trustworthy (see Figure 9). By incorporating the qualities of being relatable and aspirational, influencers can elevate the efficacy of their content. Limitations and Future Research Directions Although I uncover the nonlinear and nuanced effects of disclosure depth on consumer attitudes, further work has to be done. First, I need to ensure the generalizability of those effects in terms of influencer status. Influencers I have used in my research so far to test my hypotheses are mostly renowned celebrities who have their own jobs or activities. Thus, one could question whether consumer reactions to their personal revealing might be confounded by the popularity and image they have cultivated through offline performance. One could also wonder if the effects of disclosure depth on perceptions of being relatable and aspirational would vary across message senders, which might shift their downstream consequences. Schouten et al. (2020) suggest such a possibility, indicating that people feel more similar to influencers than celebrities and also identify more with influencers. To address this issue, I want to examine if my findings obtained from celebrities are applicable to different types of influencers such as micro social media influencers whose main jobs are promoting brands on social media. Investigating such influencers will help me solidify the evidence of my hypotheses. Even if I do not get to replicate the results of my studies, distinctive findings may also provide me with a potential moderator or interesting implications for self-disclosure. For empirical analysis, I will scrape full-time influencers’ social media posts to determine whether the results of Study 1 hold true. To experimentally test my theorizing in a controlled setting, I may use fictitious influencers generated by artificial intelligence. By varying the number of followers, I can also examine how influencer status moderates the self-disclosure depth effects. 53 Second, it is worth investigating the role of product type as a potential moderator. The transfer model suggests that the meaning that begins in the roles of a celebrity passes from celebrity to product and from product to the consumer (McCracken 1989). Alternately, consumers may form their brand attitudes based on what those influencers represent to them, which might be shifted by the influencers’ level of disclosure. Having said that, promoting up- scale hotels or luxury goods with low personal disclosure may amplify the impacts of aspirational qualities on consumer attitudes while attenuating the effects of relatable. Ward and Dhal (2014) support this reasoning, demonstrating that consumers choose products from brands associated with aspirational reference groups in order to signal their ideal selves. High disclosure in a dating app advertisement, on the other hand, could increase the relatable path’s positive effect on consumer attitudes. I examined the moderating role of brand positioning on user attitudes in association with depth of self-disclosure, but the results were not noteworthy. Perhaps, the self-disclosure effects in the prior post do not transfer to the subsequent sponsored post. Incorporating brand positioning with self-disclosure in a single post might open up doors for further examination. Lastly, self-disclosure on social media may not always take the form of writing. Influencers often express their emotions and opinions through images or videos, and Instagram especially is a medium with a strong visual component. To my best knowledge, no study quantifies the depth of self-disclosure from visual information. I believe that people in general post pictures to supplement and complement their written material. That said, the fact that an automated measure of disclosure depth from text accounts for outcomes, particularly in Study 1, implies that the impact of self-disclosure depth may be even greater when taking images or videos into consideration. 54 Conclusion Social media influencers are sharing not only their personal information, but also genuine feelings and emotions when advertising brands through a post. Since high self- disclosure is regarded to increase intimacy, and followers look to connect with influencers on a deeper level, influencers and brands are increasingly adopting such personal sharing as part of their content strategy to impact consumer decisions from clicking like on posts to searching for or purchasing the advertised brands. Against these common ideas, this research identifies an important divergence. People’s attitudes toward an influencer and associated brands do not always improve in proportion to the disclosure depth of the influencer. A tradeoff between perceptions of being relatable and being aspirational drives the nonlinear relationship, with an increase in one offset by a decrease in another. Influencers may go deeper and fare worse. Motivated by uncertainties around how much to share as an influencer, I shed light on the underpinning mechanism of how self-disclosure depth impacts consumer attitudes. 55 APPENDIX A Study 1 robustness checks Self-disclosure effects without negative emotions in an independent variable Self-disclosure effects with negative and positive emotions in an independent variable 56 Self-disclosure effects with influencer status (mega vs. micro) moderation 57 Self-disclosure effects on comments as an outcome variable Self-disclosure effects on retweets as an outcome variable 58 Self-disclosure effects controlling for gender Self-disclosure effects with male influencers only 59 APPENDIX B Study 4 Stimuli Low self-disclosure 60 High self-disclosure 61 APPENDIX C Study 4 measurement scales Constructs Items Endorsement attitudes • influential: To what extent would Taylor Swift's product, service, and brand recommendations influence you? • intent to use: To what extent would you like to use the products, services, and brands that Taylor Swift endorses? • trustworthy: To what extent would do you trust Taylor Swift to make good product, service, and brand suggestions? Aspirational • aspirational: To what extent does Taylor Swift seem aspirational? • be like: To what extent do you want to be like Taylor Swift? • ideal: To what extent does Taylor Swift seem like an ideal person? Relatable • relatable: To what extent does Taylor Swift seem relatable? • like me: To what extent does Taylor Swift seem a lot like you? • normal: To what extent does Taylor Swift seem like a normal person? 62 APPENDIX D Study 5 stimuli Minimal self-disclosure 63 Present self-disclosure 64 Past self-disclosure 65 References AlRabiah, S., Marder, B., Marshall, D., & Angell, R. (2022), “Too much information: An examination of the effects of social self-disclosure embedded within influencer eWOM campaigns,” Journal of Business Research, 152, 93-105. Altman, I., and Taylor, D. A. (1973), Social penetration: The development of interpersonal Relationships. Holt, Rinehart & Winston. Argo, J. J. (2020), “A contemporary review of three types of social influence in consumer psychology”, Consumer Psychology Review, 3(1), 126-140. Bagozzi, R. P., and Yi, Y. (1989), “On the use of structural equation models in experimental designs”, Journal of Marketing Research, 26(3), 271-284. Bargh, J. A., McKenna, K. Y., and Fitzsimons, G. M. (2002), “Can you see the real me? Activation and expression of the “true self” on the Internet”, Journal of social issues, 58(1), 33-48. Baruh, L., and Cemalcılar, Z. (2015), Rubbernecking effect of intimate information on Twitter: When getting attention works against interpersonal attraction, Cyberpsychology, Behavior, and Social Networking, 18(9), 506-513. Bickart, B., Kim, S., Pai, S., and Brunel, F. (2015), “How social media influencers build a brand following by sharing secrets,” In Strong brands, strong relationships (pp. 172-184). Routledge. Boyd, R. L., Ashokkumar, A., Seraj, S., & Pennebaker, J. W. (2022), The development and psychometric properties of LIWC-22. Austin, TX: University of Texas at Austin, 1-47. Brown, J. J., and Reingen, P. H. (1987), “Social ties and word-of-mouth referral behavior,” Journal of Consumer research, 14(3), 350-362. Chen, C. P. (2016), “Forming digital self and parasocial relationships on YouTube”, Journal of Consumer culture, 16(1), 232-254. Chung, S., and Cho, H. (2017), “Fostering parasocial relationships with celebrities on social media: Implications for celebrity endorsement”, Psychology & Marketing, 34(4), 481- 495. 66 Cozby, P. C. (1972), “Self-disclosure, reciprocity and liking,” Sociometry, 151-160. Dahl, S. (2021), “What is the role of aspirations in social media influencer marketing?” Dienlin, T., and Metzger, M. J. (2016), “An extended privacy calculus model for SNSs: Analyzing self-disclosure and self-withdrawal in a representative ME sample”, Journal of Computer-Mediated Communication, 21(5), 368-383. Djafarova, E., and Rushworth, C. (2017), “Exploring the credibility of online celebrities' Instagram profiles in influencing the purchase decisions of young female users”, Computers in human behavior, 68, 1-7. Djafarova, E., and Trofimenko, O. (2019), “‘Instafamous’–credibility and self-presentation of micro-celebrities on social media,” Information, communication & society, 22(10), 1432- 1446. Dwoskin, Jeff (2022), There's A Big Difference Between Likes and Comments, Retrieved May 16, 2023, https://www.stampede.social/there-s-a-big-difference-between-likes-and- comments#:~:text=Likes%20are%20quick%2C%20easy%20to,for%20their%20opinions %20as%20well!. Ehrlich, H. J., and Graeven, D. B. (1971), “Reciprocal self-disclosure in a dyad”, Journal of Experimental Social Psychology, 7(4), 389-400. Ferchaud, A., Grzeslo, J., Orme, S., and LaGroue, J. (2018), “Parasocial attributes and YouTube personalities: Exploring content trends across the most subscribed YouTube channels”, Computers in Human Behavior, 80, 88-96. Fitzgerald, M. P. (1963), “Self-disclosure and expressed self-esteem, social distance and areas of the self revealed”, The Journal of Psychology, 56(2), 405-412. Gibson, K. R., Harari, D., & Marr, J. C. (2018), “When sharing hurts: How and why self- disclosing weakness undermines the task-oriented relationships of higher status disclosers”, Organizational behavior and human decision processes, 144, 25-43. Grohmann, B. (2009). “Gender dimensions of brand personality”, Journal of Marketing Research, 46(1), 105–119. Halverson Jr, C. F., and Shore, R. E. (1969), “Self-disclosure and interpersonal functioning”, Journal of Consulting and Clinical Psychology, 33(2), 213. 67 Hoffner, C., and Cantor, J. (1991), “Perceiving and responding to mass media characters”, Responding to the screen: Reception and reaction processes, 1, 63-101. Hoos, B. (2019), The Psychology of Influencer Marketing, Forbes, Retrieved May 23, 2023, https://www.forbes.com/sites/forbesagencycouncil/2019/08/22/the-psychology-of- influencer-marketing/?sh=4f55f807e1be Huang, L. S. (2015), “Trust in product review blogs: The influence of self-disclosure and popularity”, Behaviour & Information Technology, 34(1), 33-44. Instagram (2021), How Instagram feed works, Retrieved May 15, 2023, https://help.instagram.com/1986234648360433. Johnson, J. A. (1981), “The" self-disclosure" and" self-presentation" views of item response dynamics and personality scale validity”, Journal of Personality and Social Psychology, 40(4), 761. Johnson, B. K., Bradshaw, A. S., Davis, J., Diegue, V., Frost, L., Hinds, J., ... and Wang, R. (2021), “Credible influencers: Sponsored YouTube personalities and effects of warranting cues”, Journal of Media Psychology: Theories, Methods, and Applications. Jöreskog, K. G., and Sörbom, D. (1982), “Recent developments in structural equation modeling”, Journal of Marketing Research, 19(4), 404-416. Jourard, Sydney M. and P. Lasakow (1958), “Some Factors in Self- Disclosure,” Journal of Abnormal and Social Psychology, 56 (1), 91–98. Kaplan, A. M., and Haenlein, M. (2010), “Users of the world, unite! The challenges and opportunities of Social Media”, Business horizons, 53(1), 59-68. Kurtin, K. S., O'Brien, N., Roy, D., and Dam, L. (2018), “The development of parasocial interaction relationships on YouTube”, The Journal of Social Media in Society, 7(1), 233- 252. Labrecque, L. I. (2014), “Fostering consumer–brand relationships in social media environments: The role of parasocial interaction”, Journal of interactive marketing, 28(2), 134-148. 68 Larson, J. D. (2019), Why do I obsessed over what’s ‘relatable’?, New York Times, Retrieved May 22, 2023, https://www.nytimes.com/2019/01/08/magazine/the-scourge-of-relatable- in-art-and-politics.html Leite, F. P., and Baptista, P. D. P. (2022), Influencers' intimate self-disclosure and its impact on consumers' self-brand connections: scale development, validation, and application, Journal of Research in Interactive Marketing, 16(3), 420-437. Leite, F. P., Pontes, N., and de Paula Baptista, P. (2022), “Oops, I've overshared! When social media influencers' self-disclosure damage perceptions of source credibility,” Computers in Human Behavior, 133, 107274. Levin, F. M., and Gergen, K. J. (1969), “Revealingness, ingratiation, and the disclosure of self”, In Proceedings of the Annual Convention of the American Psychological Association. American Psychological Association. Li, Y., and Xie, Y. (2020), “Is a picture worth a thousand words? An empirical study of image content and social media engagement”, Journal of Marketing Research, 57(1), 1-19. Lin, R., & Utz, S. (2017), “Self-disclosure on SNS: Do disclosure intimacy and narrativity influence interpersonal closeness and social attraction?”, Computers in Human Behavior, 70, 426-436. MacKenzie, S. B. (2001), “Opportunities for improving consumer research through latent variable structural equation modeling”, Journal of Consumer Research, 28(1), 159-166. McCracken, G. (1989), “Who is the celebrity endorser? Cultural foundations of the endorsement process”, Journal of consumer research, 16(3), 310-321. Miachon N. (2018), Reaching Gen z with Influencer Marketing: Four Trends for Success. Forbes, Retrieved May 22, 2023, https://www.forbes.com/sites/forbescommunicationscouncil/2018/10/09/reaching- gen-z-with-influencer-marketing-four-trends-for-success/?sh=21e74cab5050 Melumad, S., and Meyer, R. (2020), “Full disclosure: How smartphones enhance consumer self- disclosure”, Journal of Marketing, 84(3), 28-45. NCAA (2021), NCAA adopts interim name, image and likeness policy, Retrieved May 16, 2023, https://www.ncaa.org/news/2021/6/30/ncaa-adopts-interim-name-image-and-likeness- policy.aspx 69 Orben, A. C., and Dunbar, R. I. (2017), “Social media and relationship development: The effect of valence and intimacy of posts,” Computers in Human Behavior, 73, 489-498. Ray, D. (2006), “Aspirations, poverty, and economic change”, Understanding poverty, 1, 409- 421. Rihl, A., and Wegener, C. (2019), YouTube celebrities and parasocial interaction: Using feedback channels in mediatized relationships, Convergence, 25(3), 554-566. Rovell. D (2016), Famed 'Be Like Mike' Gatorade Ad Debuted 25 Years Ago, ESPN, Retrieved May 23, 2023, https://www.espn.com/nba/story/_/id/17246999/michael-jordan-famous- mike-gatorade-commercial-debuted-25-years-ago-monday Schouten, A. P., Janssen, L., and Verspaget, M. (2020), “Celebrity vs. Influencer endorsements in advertising: the role of identification, credibility, and Product-Endorser fit”, International journal of advertising, 39(2), 258-281. Shulman, S., Laursen, B., Kalman, Z., and Karpovsky, S. (1997), “Adolescent intimacy revisited”, Journal of Youth and adolescence, 26, 597-617. Simmel, G. (1964), “The secret society/Simmel G”, The sociology of Georg Simmel. Taylor, D. A., Altman, I., and Sorrentino, R. (1969), “Interpersonal exchange as a function of rewards and costs and situational factors: Expectancy confirmation- disconfirmation,” Journal of Experimental Social Psychology, 5(3), 324-339. Ward, M. K., and Dahl, D. W. (2014), “Should the devil sell Prada? Retail rejection increases aspiring consumers' desire for the brand,” Journal of Consumer Research, 41(3), 590- 609. Worthy, M., Gary, A. L., and Kahn, G. M. (1969), “Self-disclosure as an exchange process,” Journal of personality and social psychology, 13(1), 59. 70