ICWSM 2014 took place at the University of Michigan, in sunny Ann Arbor, with chairs Eytan Adar @eytanadar and Paul Resnick @presnick. For me and quite a few others, this was a first time at this conference, so I would like to share some general impressions (for a more complete overview see the conference website and preface).
Keynote speakers were Eric Horvitz from Microsoft Research and Keith Hampton from Rutgers University. Program chairs were Munmun De Choudhury @munmun10 (Georgia Tech), Bernie Hogan @blurky (Oxford Internet Institute), and Alice Oh @aliceoh (KAIST), who did an excellent job of selecting 64 full papers and 18 posters, with 23% acceptance rate. The conference had no parallel sessions. Instead, some papers were presented in full-length talks, whereas others in few-minute talks, followed by posters. Interesting format. I definitely appreciated being able to hear all talks and not having to manage parallel sessions. There were discussions of whether and how to change the current format and I am curious to see what next year’s ICWSM will be like.
The research presented at ICWSM is interdisciplinary, mainly computer science and social sciences, which is a challenging mix, but was handled excellently. As a psychologist, I found all talks easy to follow and not too technical. I realize “not too technical” is not necessarily a good thing. However, at ICWSM there was plenty of time to discuss methods and procedures in detail during the poster sessions and well-timed coffee breaks. Or evening drinks, of course.
Overall, my impression was very positive. Inspiring research, great crowd, culture of openness and sharing (much of the data is made publicly available). ICWSM 2015 will be held in Oxford, keep an eye on http://www.icwsm.org/ and @icwsm for details and deadlines.
Following are brief summaries of some papers presented at ICWSM 2014 that relate to what we do at ReDefTie. I strongly recommend checking out the full list of papers here: Conference Proceedings of ICWSM 14
How Community Feedback Shapes User Behavior
Justin Cheng, Cristian Danescu-Niculescu-Mizil, Jure Leskovec
The authors investigated how receiving feedback in online news communities (twiends.com, breitbart.com, ign.com, allkpop.com) influences subsequent behavior. They found that, following negative feedback, such as down-votes, bad ratings etc., users begin to post more and their posts are of poorer quality. No effects of positive effects were observed. Check out the full paper at: http://www.aaai.org/ocs/index.php/ICWSM/ICWSM14/paper/view/8066/8104
VADER: A Parsimonious Rule-Based Model for Sentiment Analysis of Social Media
C. J. Hutto, Eric Gilbert
VADER is a human-validated sentiment analysis model, specifically tailored towards microblogging and social media. It outperforms other established algorithms, most notably LIWC (pronounced Luke; pun intended and hilarious). For anyone who does or considers doing sentiment analysis, Hutto & Gilbert’s paper is an absolute must: http://www.aaai.org/ocs/index.php/ICWSM/ICWSM14/paper/view/8109/8122
Emoticons and Phrases: Status Symbols in Social Media
Simo Editha Tchokni, Diarmuid Ó Séaghdha, Daniele Quercia
Linguistic features, such as word choice, sentiment, and emoticon use, predicted the status of Twitter users (number of followers, Klout score) with an impressive accuracy of up to 82%. Some of the results are rather surprising. For example, smilies : ) and abbreviations a la “thx” and “yal” seem to be associated with high status. The complete picture, of course, is far more nuanced and very much worth looking into:
http://www.aaai.org/ocs/index.php/ICWSM/ICWSM14/paper/view/8103/8149
Quantifying Information Overload in Social Media and Its Impact on Social Contagions
Manuel Gomez Rodriguez, Krishna Gummadi, Bernhard Schoelkopf
The paper describes ingenious ways of quantifying information overload on Twitter. Findings include that following more than 100 users and having an influx of over 30 tweets per hour interferes with people’s ability to discover information. The paper offers other interesting insights and future directions:
http://www.aaai.org/ocs/index.php/ICWSM/ICWSM14/paper/view/8108/8117
Understanding Loneliness in Social Awareness Streams: Expressions and Responses
Funda Kivran-Swaine, Jeremy Ting, Jed Richards Brubaker, Rannie Teodoro, Mor Naaman
After managing to identify tweets expressing loneliness (and distinguish them from song lyrics!), authors classified those tweets according to duration of the experience, context, and interactivity. In a “rich get richer” fashion, men expressed enduring loneliness less frequently than women, but when they did, they were more like to receive responses and support. Independent of gender effects, tweets expressing enduring loneliness, as opposed to transient, were less likely to receive replies. Read the full paper here:
http://www.aaai.org/ocs/index.php/ICWSM/ICWSM14/paper/view/8038/8126
More than Liking and Bookmarking? Towards Understanding Twitter Favouriting Behaviour
Florian Meier, David Craig Elsweiler, Max L. Wilson
This paper offers a qualitative analysis of the motives for using Twitter’s Favorite function, which has been largely understudied. Content analysis of open-ended responses to the question why participants marked a tweet as favorite, resulted in a taxonomy of 25 different motives. Complete list of motives and their frequency is available in the paper:
http://www.aaai.org/ocs/index.php/ICWSM/ICWSM14/paper/view/8094/8133
Counting on Friends: Cues to Perceived Trustworthiness in Facebook Profiles
Catalina Laura Toma
Participants reviewed snippets of real Facebook profiles and judged the trustworthiness of the profile owners. The impact of self-generated and other-generated cues on trustworthiness ratings was compared. In line with prior research, other-generated were the stronger predictor. For details and effects of additional factors, such as a smiling profile photo and gender, see the full paper.
http://www.aaai.org/ocs/index.php/ICWSM/ICWSM14/paper/view/8044/8150
The Tweets They Are a-Changin’: Evolution of Twitter Users and Behavior
Yabing Liu, Chloe Kliman-Silver, Alan Mislove
Under the entertaining (thanks for the earworm!) title is a comprehensive overview of changes in Twitter users and behavior, including user characteristics and demographics, Twitter-specific behaviors (retweet, RT, reply) and Twitter content (URLs, #hashtags, @mentions). Trends between 2006-2009 and 2014 are presented visually. I found it curious to read that the sharing of URLs has decreased, whereas the number of mentions has increased, suggesting that Twitter is becoming more conversational.
http://www.aaai.org/ocs/index.php/ICWSM/ICWSM14/paper/view/8043
