Presenter: Tad Falk
Mentor: Sanjay Srivastava
Oral Presentation
Major: Psychology
The creation and increasing popularity of social media websites have changed the way people around the world interact with one another. Having our information available to the world allows others, such as potential employers, business contacts, new friends and even potential romantic partners to get to know us without ever needing to meet in person. Thus, it is important to understand which types of cues people use when forming judgments and whether these judgments are consistent across various observers. This study examined the extent to which participants agreed on their impressions of Twitter users based only on information presented in the users’ Twitter profile. 131 Amazon Mechanical Turk Workers and 89 University of Oregon Human Subject Pool participants evaluated a random selection of Twitter users’ profile pages. Multilevel modeling techniques were used to separate observer and target effects in order to estimate trait level agreement. This multilevel model with random effects for profile and observer allowed us to obtain estimates of variance attributable to profiles and observers. Following Shrout and Fleiss (1979) and Shrout (1993) those variance components were then used to calculate intraclass correlations (ICC), which were used as the measure of consensus. On average, participants agreed most about the degree of users’ thoroughness, intelligence, social economic status, and arrogance. Participants agreed least about the degree of users’ assertiveness, artistic interest, nervousness, and sense of humor. The findings of this study support existing literature regarding the ability of third party observers to make consistent judgments of strangers based on limited amounts of information present on social media websites. This study also extends previous literature to Twitter, a microblogging social media platform and one of the most popular social media websites in the world.