DRAFT Manuscript: Submitted to CHI 2007 – September, 2006. Not for Circulation.
1
A Familiar Face(book): Profile Elements as Signals in an Online Social Network Cliff Lampe, Nicole Ellison, and Charles Steinfield Dept. of Telecommunication, Information Studies, and Media Michigan State University, East Lansing, MI ABSTRACT Using data from a popular online social network site, this paper explores the relationship between profile structure (namely, which fields are completed) and number of friends, giving designers insight into the importance of the profile and how it works to encourage connections and articulated relationships between users. We describe a theoretical framework that draws on aspects of signaling theory, common ground theory, and transaction costs theory to generate an understanding of why certain profile fields may be more predictive of friendship articulation on the site. Using a dataset consisting of 30,773 Facebook profiles, we determine which profile elements are most likely to predict friendship links and discuss the theoretical and design implications of our findings. Author Keywords social network sites, profile elements, signaling theory, ACM Classification Keywords H5.m. Information interfaces and presentation (e.g., HCI): Miscellaneous. INTRODUCTION User profiles are an integral part of social network sites and can include a vast array of user-contributed content. However, little is known about the specific effects user profiles have on interactions in online communities. Intuitively, we believe that profiles can help create a sense of presence and must garner positive outcomes for their users given the time commitment they require to complete and keep updated; yet we do not know what types of included information matter. This gap in understanding motivates our basic research question: how do elements in a profile influence the outcomes from using an online social network? Online communities have different goals, but a common
and important enterprise is forming connections between users. This is especially true for online communities that focus on articulating social networks, such as Facebook, MySpace, Friendster and Orkut, where the number of friends a user lists may act as a simple proxy for their connectedness in the network. Connections between users in an online community may be important for facilitating other tasks of the group [14][17], reducing misbehavior [7] [15], and building types of social capital [10,16], [28] among other potential benefits. To examine the role of profile elements in the formation of online connections, we focus on Facebook.com, an online social network site. Facebook, as with similar sites like MySpace and Friendster, allows users to create in-depth profiles describing themselves, and then to establish explicit links with other users, who are described as “friends” by the system. Facebook is a particularly appropriate site to study as it has profile creation and network articulation as primary community tasks, meaning that there is a consistency of action across the different users that allows for variance to be more clearly articulated and examined. Also, although Facebook is now open to those without academic affiliations, at the time data were collected, Facebook communities generally corresponded to existing offline network membership, typically related to academic environments like universities. This offline connection has several implications. First, it allows the establishment of a natural boundary around the network that assists when determining who is a member and who is not. Second, the connection to an offline network might increase the likelihood of profile use by offline contacts, as the chances that a relationship formed in the online environment will extend to an offline meeting. This means that profile information has more opportunities to be verified than in other online communities. Third, participation may be reinforced by offline connections, contributing to a take-up rate for a given population that may be higher than normal. As mentioned above, Facebook is divided into networks based on affiliation with a particular offline institution. Consequently, users from University A are not, by default, considered part of the network of University B. Our study focused on one network within Facebook, namely the network defined by membership in X State University (XSU). Membership in a university network is defined by Facebook as having a valid email address assigned by that university.

DRAFT Manuscript: Submitted to CHI 2007 – September, 2006. Not for Circulation.
2
In this paper, we report on an empirical study of the XSU Facebook community. Specifically, using data collected from all accessible XSU Facebook members via an automated script, we explore how profile elements relates to the number of friendship connections among users. The paper is organized as follows. We first review related literature, focusing on theoretical explanations for why profile information may influence online social network activity. The next section outlines the methodological approach for an empirical test of research questions generated by our review. We then present results, highlighting the explanatory power of selected profile elements. The paper closes with a discussion of the implications of our findings, as well as study limitations and overall conclusions. Literature Review In this section, we explore prior work that establishes the basis for our primary proposition: That the amount and type of information included in user profiles should affect the number of articulated relationships in the online community. Individuals form impressions of others in order to decide whether to pursue or continue a relationship [19]. In initial impression formation, individuals form impressions very quickly -- in as little as three minutes in face-to-face settings [20]. In order to achieve relational and other goals, individuals attempt to manage these impressions, strategically emphasizing some characteristics while de-emphasizing others [11]. These same self-presentational behaviors exist online, although online self-presentation is more malleable and subject to self-censorship than face-to-face self-presentation due to the asynchronous nature of computer-mediated communication (CMC) and the fact that CMC emphasizes verbal and linguistic cues over less controllable nonverbal communication cues [23]. These same processes of impression formation and management take place in online settings, albeit slightly differently due to the affordances and constraints of CMC. In online environments, traditional identity cues, such as accent and style of dress, are not available. Early research assumed this forced online interactants to operate in a vacuum of identity cues, with attendant negative consequences for interpersonal relationship and community formation [5] [17]. However, subsequent work developed a more optimistic assessment (for review, see [27]), noting that CMC groups just needed more communication time than face-to-face groups, in order to compensate for CMC’s slower rate of exchange [24]. Walther’s Social Information Processing theory posits that online users compensate for the lack of traditional cues in online environments by looking towards other kinds of cues, such as spelling ability [21] [26][25]. Evidence for SIP has been generated in other contexts such as MUDs [22] and online dating [10]. Online interactants seeking to form impressions of their communication partners must assess not only the content of
the identity claims made by others, but also the veracity of these claims. As Donath [8] writes, “In order for a signal to have its intended effect, the receiver must both understand and believe it.” Although deception in offline environments is common [6], the ability to selectively self-present online [23] means that some kinds of misrepresentation (e.g., “gender-bending”) are more easily accomplished via CMC. In some online environments such as online dating, misrepresentation is a significant concern [10]. Users in online environments rely on a variety of cues to make determinations about one another; however, all these cues are not deemed equally credible. For instance, Goffman [11] notes that identity cues can be intentionally given or unintentionally given off, and that we are more likely to privilege those cues that are perceived to be unintentional as opposed to strategically constructed. This ability to engage in deceptive self-presentation online is compounded when interactants do not share a social network and therefore have less access to “information triangles” such as mutual friends who might confirm or deny information [2]. Theoretical motivations We draw upon three theories to help explain how profile construction might affect participation in online communities. Signaling theory addresses the type of information that can be placed in profiles, suggesting that profile elements act as signals that may prove something about the identity of the user. These signals can be manipulated by senders to communicate personal qualities, or interpreted by receivers to make judgments about the characteristics of other users. We use common ground theory to explain the motivation of filling out profiles, which is to establish common frames of reference that enhance mutual understanding. Transaction cost theory bridges the two former theories and suggests that certain profile elements may facilitate the production of shared referents, which usually involves costly negotiations between participants, and makes it easier for interactants to engage in other forms of communication (such as email). Signaling and the verifiability of profile entries Signaling theory addresses a basic question: what keeps signals reliable in contexts where deception can be beneficial [8]? Based on work in biology [30, 31], Donath argues that a signaling system must evolve so that it is beneficial for participants to produce reliable signals, but costly to produce deceptive ones. Building on contemporary signaling theory, she distinguishes between many different kinds of signals, including those that reliably indicate possession of some quality simply through observation of the signal, which are termed assessment signals, and those that only indicate a quality through social convention, which are termed conventional signals. As she notes, lifting a heavy weight is an assessment signal that reliably indicates that a person is strong. Wearing a Golds® Gym T-shirt is a conventional signal that suggests that the person works out and therefore is likely to be in shape, but is easy enough to acquire that the wearer might actually be weak.