What Elements of an Online Social Networking Profile Predict Target- Rater Agreement in Personality Impressions? David C. Evans Psychster LLC 16904 Juanita Dr. #145 Kenmore, WA 98028 david@psychster.com Samuel D. Gosling Department of Psychology University of Texas Austin, TX 78712 samg@mail.utexas.edu Anthony Carroll Psychster LLC 16904 Juanita Dr. #145 Kenmore, WA 98028 anthony@psychster.com Abstract Social-Networking Websites (SNWs) like Facebook and MySpace are playing an increasingly prevalent role in everyday social interactions. But very little is known about the effectiveness of the various profile elements in conveying information about the personality of the profile owner. Here we examine 5,303 impressions made on the basis of a specially designed social networking website (http://www.YouJustGetMe.com) and on the basis of Facebook profiles. Our findings suggest that profile owners are generally seen by others as they see themselves that when raters are judging unknown targets, rater-target agreement is stronger for female (vs. male) targets and for female (vs. male) raters and several specific elements of profiles are associated with increased or diminished levels of rater-target impression agreement. The findings are important because they are the first to show how impression agreement may be affected by specific elements in SNW profiles. Keywords Personality, impressions, self-other agreement, online identity, online profiles, social networking, social computing, Facebook. Introduction Social networking websites (SNWs) such as Facebook and MySpace are playing an increasingly prominent role in everyday social interactions. The particular role of SNWs varies across relationships���in some contexts SNWs supplement existing real-world social networks but in other contexts, interactions can be entirely mediated by SNWs [12]. People may even use them to gather information on others (e.g., prospective employers marketers) [1]. The result of all these changes is that SNWs have become a central medium of interpersonal perception. But how effectively do SNWs convey information about the profile owner? Are some elements more informative than others? Are there any elements that actually impede the flow of valid information? The present paper examines these questions in two SNW contexts���an SNW in which profile owners have their personalities rated by strangers and a Facebook application allowing visitors to rate the personalities of people on the basis of their Facebook profiles. The Realistic Accuracy Model (RAM), developed by personality psychologist David Funder [2,3], is the most fully developed system for understanding the parameters that can influence the quality of judgments. Funder proposes four parameters, which are associated with four questions that can be asked about judgments: (1) Are some traits associated with better judgments than others? (2) Are some judges (or raters) better than others? (3) Are some targets more easily judged than others? (4) Do some forms of information (e.g., movie preferences vs. photos of the target) elicit better judgments than others? There are several different metrics for evaluating the quality of a judgment [8]. One common metric is the accuracy of the judgment that is, do the judgments of a target reflect what that target is really like? However, in the present research we are interested in whether the target (i.e., profile owner) is able to successfully convey information about him or herself, regardless of whether the target���s self-view is accurate. If a target believes she is extraverted, even if in reality that is not true, the relevant question is whether the target can bring others to see her as she sees herself [10]. Therefore, the relevant metric for evaluating judgment quality is self-rater agreement, which we refer to as ���impression agreement.��� Research Questions The first of Funder���s questions (Are some traits easier to judge than others?) was addressed in previous research. Gosling, Gaddis, and Vazire [4] examined personality impressions based on Facebook profiles, and found that agreement was strongest for ratings of extraversion and openness to experience and weakest for ratings of emotional stability. The present study was designed to examine the remaining three questions: Research Question 1 (RQ1): Are some judges better than others? In particular we examine whether there are sex differences in judgmental ability in the context of SNWs. Research Question 2 (RQ2): Are some targets more easily judged than others? In particular we examine whether there are sex differences in judgability in the context of SNWs. Research Question 3 (RQ3): Do some forms of information elicit better judgments than others? Previous research has examined broad moderators of impression agreement, showing for example, that agreement is strongest for traits that are observable, non-evaluative, and associated with the personality domain of extraversion [7]. However, it is not known which specific elements of information promote successful judgments. We examine ICWSM���08, March 31 - April 2, 2008, Seattle, Washington, USA.

the degree to which 33 specific elements (e.g., the target���s nomination for a ���great book,��� or the ���proudest thing I ever did���) predict impression agreement. It has long been assumed that more information is better. But are there elements that actually distort impression agreement? It is possible that some forms of information lead judges astray, perhaps by activating false stereotypes [8]. Therefore, RQ3 is broken down into two narrower sub-questions: RQ3a: What specific elements of SNWs promote impression agreement? And RQ3b: What specific elements of SNWs hinder impression agreement? Method Materials To create a realistic environment in which to examine our research questions, a fully functional social networking website was built under the name http://www.YouJustGetMe.com. The site was equipped to allow participants to create and view profiles, send and receive messages, and browse for others. Additionally, to examine the generality of the effects, an application was built and launched on the Facebook Platform. Facebook is a social utility with 64 million active users. To register and use YouJustGetMe, participants were required to provide their gender, date of birth, and an anonymous display name. All participants also completed a self-assessment of personality using the Big Five Inventory (BFI-K Form S) [6] in which they rated their agreement on 5-point Likert scales with 21 personality statements that began with "I see myself as someone who���" Participants also answered 19 additional 5-point semantic-differential scales loosely designed to tap attitudes ("left of center" vs. "right of center") and preferences ("a wine person" vs. "a beer person") in a fun and engaging way. Participants Student participants were invited to use the website and Facebook application via three sources: (a) an introductory psychology class at a large university in the western United States, (b) invitations forwarded from members of a listserv for the Society of Personality and Social Psychology, and (c) invitations forwarded from other YouJustGetMe participants. At the time of analysis, 5,754 participants had joined the site (4,457 via Facebook, 1,297 via www.YouJustGetMe.com), but only the data from 5,216 participants who were 18 or older were analyzed. Procedure Participants were encouraged to upload a profile picture and to "give clues about your personality so people can guess what you're like" by completing any of 33 pre- defined fields. The fields, which were chosen to encourage self-disclosure on a diverse array of informational sources, are listed in Table 1. None of the fields nor the picture were Table 1. Profile elements as predictors of impression agreement. Elements with Sig. Positive Betas Beta p-value A link to funny video 0.096 0.000 What makes me glad to be alive? 0.089 0.007 Most embarrassing thing I ever did 0.074 0.018 Proudest thing I ever did 0.057 0.063 My spirituality 0.051 0.044 A great person 0.060 0.048 I believe this 0.040 0.080 Elements with Sig. Negative Betas Beta p-value Profile picture was a non-person -0.128 0.003 An awful website -0.079 0.003 An awful person -0.065 0.028 A great book -0.066 0.037 Elements with Non-sig. Betas Beta p-value Letters after my name 0.015 0.429 My relationship status -0.028 0.411 My relationship saga -0.016 0.489 My political leanings (5pt. scale) 0.007 0.725 My political views -0.023 0.312 A link to my other profile 0.003 0.884 A link to great art -0.005 0.845 My career path 0.034 0.268 A great song 0.019 0.557 An awful song -0.004 0.866 A great movie -0.033 0.437 An awful movie -0.039 0.156 An awful book 0.040 0.115 Delicious food -0.034 0.441 Terrible food 0.043 0.151 A great website -0.027 0.310 A great company 0.020 0.436 An awful company 0.028 0.340 Best time I ever had 0.017 0.563 Worst time I ever had -0.014 0.672 What have I been up to lately? -0.013 0.650 What drives me crazy? -0.041 0.207 Profile picture was a face/bust shot -0.016 0.595