The International AAAI Conference on Weblogs and Social Media (ICWSM)

From the 8th to 11th of July 2013, Dong Nguyen, PhD-student at the University of Twente, and I, intern at the Meertens Institute for TINPOT (a Twitter research project on language, identities, conduits and rumors), attended the International AAAI (Association for the Advancement of Artificial Intelligence.) Conference on Weblogs and Social Media (ICWSM). The conference has been established in 2007 in order to bring together computer scientists and social scientists working on social media. This years ICWSM-13 was held at MIT in Boston, MA and included work drawing upon network science, machine learning, computational linguistics, sociology and communication. Accepted papers were not only presented in plenum presentations, but also in more interactive poster and demo sessions. Additionally, the conference was opened and closed with inspiring keynote talks: David Lazer, for example, outlined the strengths and weaknesses of social scientists and computer scientists, concluding that studying social media accurately requires approaches from both fields of study: while social scientists have expertise in social theory and topics involving human behavior, computer scientists are able to tackle large scale data and develop complex statistical methods. Combining these skills can lead to more meaningful and interesting findings. Certainly, this became clear in practice in the course of the conference.

For a poster that got an honorable mention, for instance, more than one billion tweets from different time zones and countries were collected and crawled for emoticons. Interestingly, the analysis shows that emoticons are subject to socio-cultural norms and that their meaning can vary depending on these norms and on the individual (Park et al. 2013). It also became clear that geography and language influence the emoticon style, since horizontal style emoticons were rather used in English speaking and Western countries, while vertical style emoticons were rather employed in Eastern countries. As a reason for this, it is mentioned that easterners pay more attention to the eyes when interpreting facial expressions, while westerners rather focus on the mouth. This might also be the case when interpreting emoticons, according to the authors. Surely, this research is a good example of a successful cooperation of computer scientists and social scientists.

The MIT campuskleiner

The paper we presented ourselves also deals with data from Twitter (Nguyen et al. 2013). In detail, it examines how well an automatic system can predict the age of a Twitter user. In general, the system is able to predict age categories and life stage categories with 86% correctness. For the exact age, the mean absolute error is less than 4 years. But the analyses also show that the age of younger users is easier to predict than the age of older users. As a reason for this, it might be hypothesized that the differences in language use are greater between, for example, a 15-year-old and a 20-year-old than they are between a 40-year-old and a 45-year-old, even though the age span between them is similar. Additionally, predictions of the automatic system and humans were compared – and it became clear that the automatic system does slightly better in predicting the age of Twitter users than humans do.

Many more of the presented papers and posters used data from Twitter, but there were also researchers that worked with data of families on Facebook (Burke et al. 2013), fashion blogs (Marwick 2013), online forums (Schoenebeck 2013), or Pinterest.com (Zarro 2013), etc. While some of the papers clearly bridged computational and social scientific approaches to their data, others rather leaned to one or the other side. This led to an interesting mix of topics and approaches. All in all, the conference offered a vast variety of perspectives on social media research by bringing together researchers with different fields of expertise. All accepted papers can be accessed online under www.aaai.org/.

References:
Burke, Moira, Lada Adamic, Karyn Marciniak. Families on Facebook. In Proceedings of ICWSM-13, July 8-11 2013, Boston, MA.
Marwick, Alice. “They’re Really Profound Women, They’re Entrepreneurs”: Conceptions of Authenticity in Fashion Blogging. In Proceedings of  ICWSM-13, July 8-11 2013, Boston, MA.
Nguyen, Dong, Rilana Gravel, Dolf Trieschnigg and Theo Meder. “How Old Do You Think I Am?”: A Study of Language and Age in Twitter. In Proceedings of ICWSM-13, July 8-11 2013, Boston, MA.
Park, Jaram, Vladimir Barash, Clay Fink and Meeyoung Cha. Emoticon Style: Interpreting Differences in Emoticons across Cultures. In Proceedings of ICWSM-13, July 8-11 2013, Boston, MA.
Schoenebeck, Sarita Yardi. The Secret Life of Online Moms: Anonymity and Disinhibition onYouBeMom.com. In Proceedings of ICWSM-13, July 8-11 2013, Boston, MA.
Zarro, Michael, Catherine Hall and Andrea Forte. Wedding Dresses and Wanted Criminals:Pinterest.com as an Infrastructure for Repository Building. In Proceedings of ICWSM-13, July 8-11 2013, Boston, MA.