Tracking the Yak: An Empirical Study of Yik Yak

To investigate the effects of anonymity on user behavior, we conduct an empirical study of the new and controversial social app, Yik Yak. First, we examine how users use the platform, analyzing patterns in posting, popularity of yaks, and vocabulary. As a comparison, we look at posting patterns on Twitter, which has similar limitations on lengths of posts, but is public and global rather than anonymous and local. Upon a sample of 2.9M posts (1.9M yaks and 1M geotagged tweets) from 20 locations across the USA, we find that interactions on Yik Yak are specific to its location limitations and reflect the schedules of its targeted demographic, college students. Second, we test two hypotheses related to anonymity and communication: (i) whether vulgarity usage is more likely to be acceptable, and (ii) whether unique topics emerge in conversations on Yik Yak. We find that posts on Yik Yak are only slightly more likely to contain vulgarities, and we do not find any significant bias in topic distributions on Yik Yak versus on Twitter; however, differences in vocabulary and most discriminative words used suggest the need for further analysis. Introduction While secrecy has always had its allure, in recent years, anonymous social networks have been on the rise, expanding from online forums to mobile applications. Several platforms for smartphone users, such as Secret, Whisper, and the Insider, have emerged, and work like “The Many Shades of Anonymity”—which studies content traces from Whisper— highlight some of the differences in posting behavior when compared to public networks (Correa et al. 2015). In the past year, Yik Yak, a social media platform founded in 2013, has gained considerable attention in the press as the subject of several controversies concerning privacy and free speech (Mahler 2015). Like Whisper, Yik Yak focuses on creating local, anonymized communities. However, it targets college campuses specifically. Other past research on anonymity on the internet—such as the work by Bernstein et al. on 4chan and /b/ focuses on web forums that are accessible to the public and cover random and broad topics, often ephemeral and/or vulgar in nature (Bernstein et al. 2011). Yik Yak differs in that its users are highly targeted, Copyright c © 2016, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. through marketing and a limited posting radius, to a specific demographic. Moreover, the platform is designed with a 200-character limit and other restrictions that guide its usage to specific use cases. It is both anonymous, and less open than freeform online forums. In this study, we seek to summarize basic user interactions on Yik Yak in addition to examining changes in language and behavior due to anonymity. Anonymity is linked to the “Online Disinhibition Effect,” which predicts increased aggressiveness and other negative social behaviors (as in the case of online “trolls”), but conversely, can be argued to lead to more openness and willingness to discuss taboo subjects (Suler 2004). In this study, we set out to answer questions about user behaviors on Yik Yak, including: What are the temporal patterns of posting? What affects the popularity or deletion (by vote) of a post? What sort of vocabulary is characteristic of posts? We also test the following hypotheses concerning anonymity and online disinhibition effects. H1: Vulgarity usage is more likely to be acceptable on an anonymous platform (Yik Yak) vs. a public one (Twitter). H2: Unique topics (potentially taboo ones) emerge on an anonymous platform that are undiscussed on a public one. For a baseline, we compare our analyses of Yik Yak with geotagged, public data from Twitter. Yik Yak Yik Yak, the focus of our study, is a Twitter-like anonymous social smartphone application. It functions similarly to an online community bulletin board. Within a 10 mile radius, users can leave posts (yaks) of up to 200 characters to the community, and interact with other community member’s posts as well. Outside of the radius, yaks are read-only with the Peek feature. The community determines the popularity and persistence of yaks with upvotes and downvotes; if a post receives -5 votes, it is deleted, creating a self-selecting mechanism for filtering and censoring content. The app is designed to encourage users to enforce this mechanism: If users engage with content (upvote/downvote posts) or post yaks that get upvoted they win Yakarma points. At the time of the study, Yik Yak was completely anonymized—with no persistent identities. (Currently, icons on a single comment thread link users who post more than once.) Additionally, it is localized, operating on a 10-mile 0 50 100 150 200 250 Dartmouth College Lehigh university West Virginia University Brown University University of Pennsylvania Cornell University Georgia Institute of Technology University of Iowa UC Berkeley Syracuse University Purdue University West Lafayette University of Wisconsin Madison Carnegie Mellon University University of Illinois at Urbana C. US Santa Barbara University of Michigan Ann Arbor University of Texas Austin University of Florida MIT / Harvard / Simmons College Penn State University Number of posts (Thousands) Yaks Tweets Figure 1: Locations selected for the study and the volume of yaks and tweets in each. radius, mostly surrounding college campuses. Users can upvote, downvote, and reply to yaks. The target demographic is young, with marketing efforts heavily directed towards college campuses (Shontell 2015). Conversely, Twitter enables users to send short 140character limited tweets. Users on Twitter have a persistant username and identity, which can be a pseudonym or a real id, or an organization. A portion of users are “verified”, or tied to their real ids. Specific to Twitter are various conventions such as @mentions, #hashtags, and the sharing of urls, all of which are not available on Yik Yak. The demographic of Twitter is less specific than the targeted audience of Yik Yak. Also, Twitter does not provide a self-selecting mechanism for filtering and censoring content, but allows users to favorite tweets and report posts that they find inappropriate. Although, Yik Yak and Twitter are different in many aspects, using Twitter as a baseline allows us: (i) to put the results in perspective, and (ii) help future researchers understand how their existing work on Twitter—which is commonly studied—relates to this new platform. Data Collection We collected data from 30 hotspots over a duration of 19 days: Monday, April 4th to Friday, April 24th, 2015. For variety, we sampled from Newsweeks lists: top 10 engineering schools, top 10 women’s colleges, top 10 party schools, and the Ivy League. After data collection, we narrowed the list down to 20 colleges due to sparse data in some locations, and an overlap of radiuses in others. Figure 1 shows the final set of locations and the volume of yaks and tweets in each. For Yik Yak, we set up scrapers to run in 5-minute intervals, with a 12-hour lookback to collect replies. For Twitter, we approximated Yik Yak hotspots by looking at geotagged tweets in a 10-mile radius around the same locations. We used the GNIP Historical API to retrieve all public tweets posted in these locations. Finally, we ended up with 1.9M yaks (569K posts and 1.4M replies) and 1M tweets. 9 AM 3 AM 6 PM Time of Day 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 F ra ct io n of P os ts Midnight A YakYak Twitter Noon Mon Tue Wed Thu Fri Sat Sun Day 0 2 4 6 8 10 12 14 16 18 V ol um e (% ) B Twitter YikYak Figure 2: Daily (A) and weekly (B) activity patterns on Yak Yak (green) and Twitter (blue). 10 10 10 Number of replies (log) 10 10 10 10 F re qu en cy (l og ) Slope: -2.71 A 0 10 15 5 Rating 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07