Social computing: a sociological perspective

Computer-mediated social interactions are ubiquitous in today's world. Blogs, forums, wikis, social networking sites are all examples of how social technology facilitates collective participation and creates new forms of social expression. And the really great news for science is that most social computing systems leave a digital trace of the actions and interactions of thousands to millions of people in real time, as a byproduct of their normal operation. The growing availability of such digital trace data holds unprecedented opportunities to answer many questions that social scientists have entertained only theoretically for decades, largely due to the difficulty to directly observe social relations that are fundamental for social groups and networks. Among the first direct uses of the Internet by social scientists were online surveys and experiments (Johnson 2001; Salganik et al. 2006). Using the Web brings down the costs of conducting surveys with the conventionally chosen samples and panels, and makes it possible to work with larger samples and multiple experimental conditions. In many cases, researchers also recruit their subjects online, by selectively placing advertisements in portals and forums frequented by people from the target demographic groups, which may be hard to reach otherwise. This is an example of how social computing facilitates and extends traditional sociological research methodology. Yet social scientists are also beginning to embrace, albeit warily, the specific, novel opportunities presented by social computing systems, by focusing on the electronically registrable social interactions as opposed to self-reported data. Communities such as LiveJournal, Wikipedia, Advogato, and The World of Warcraft are being used as real life test labs to study the structure of large scale social networks (Backstrom et al. 2006), the dynamic processes of status attainment (Stewart 2005), technology adoption (Cosley et al. 2007) and even public health research (Balicer 2007). In particular, online communities are a treasure trove of data with which to study large, evolving social networks. A major problem for social network research has been the paucity of relational data. Up until recently, social scientists could do little more than speculate about the dynamics of large-scale self-organization in social networks. The problem has been finding data about human social interactions, particularly in the formative stages of communities when patterns first begin to emerge. Social interactions are fleeting and mostly private, making them hard to capture and arduous to hand-code and record. These problems are compounded by the need for repeated observations, in order to study how relations develop over time, and by the exponential increase in the number of relations as group size increases. As a result, the processes such as social influence, information diffusion, network evolution have not been well documented except in observational and ethnographic studies of small groups. All this is rapidly changing as human interactions move increasingly online. The growing tendency to interact online have created vast quantities of semi-structured data such as Web pages, e-mail list archives, blog postings, and forum logs. These computer-mediated interactions leave a digital record that opens up new opportunities to systematically capture relational data. There are, however, several technical difficulties in gathering and analyzing digital trace data that have deterred social scientists from embracing this approach. Researchers confront massive amounts of poorly structured data that must be organized into a usable form. For example, a crawl of the Web yields only a blurry snapshot of pages that may have been created over several weeks or months. Researchers trained in the use of survey data are not prepared for the methodological challenges posed by dynamic networks with millions of nodes and billions of edges measured at multiple time points. In many cases, one also must work out how to protect the privacy of research subjects (Backstrom et al. 2007) which is not trivial for relational data. These issues can be most effectively solved by cross-disciplinary collaboration.