I’ve been happy to follow the recent emergence of computational data analysis in the domain of social science. Finally there are possibilities to handle large data sets or apply new methods to gain understanding of the surrounding society. Levy’s and Franklin’s (2013) work used topic models to demonstrate that different stakeholders discuss different aspects when commenting regulations, analyzing about 3500 documents from lobbyists, employees and employers. Shor et al. (2014) examined how male and female names were presented in newspaper stories, applying computational search for these names. As a final example, Jurek and Scime (2013) examined patterns that explain democratic leadership applying rather interesting unsupervised method to explore the data.
I’m eager to see this adaptation, especially how it differs from the early ideas on computational social scientists who focused on collection and exploration of large data sets on society, or more familiarly big data (Cioffi-Revilla 2010;Lazer at al 2009). I’m hoping these examples indicate that computational methods, like computation and machine learning, are taking taken as serious methods that can be applied. Naturally, there’s work that needs to be done, not only in research using these methods and understanding the scope and validity questions related to computational methods (such as Grimmer and Stewart 2013), but also in teaching these tools to (under)graduate students and making sure they can work with these.
Recently the term digitalization has become significant term in Finland. It refers to the rising role of ICT in functions of the civic society, and especially how information technology takes over the world changes traditional institutions, their roles and interactions between them. Jansson and Erlingsson (2014) present how computer systems take a decision-making role in social security processes. Wright and Street (2007) discuss the influence of design on online participation platforms. These examples link to notation on algorithms’ power (e.g. Musiani 2013) and how digitalization influences political outcomes. Looking at social scientists: are we again one step behind the rest of the world?
For maximal societal impact, shouldn’t we instead of analyzing the past using computational tools build the future with them? I mean interactive systems that face the users and that are used outside the research chambers. This might be an interesting area for quasi-experimental studies (e.g.Manosevitch et al. 2014), but even interesting are systems that do some fancy data-tricks based on users involvement. This is not just my dreaming: Kriplean, et al. (2012a; 2012b) and Faridani et al. (2010) present systems where users input is handled through computation and given back to the users. The possibilities to test different hypotheses – like the famous Facebook emotion experiment – are huge, if we want to move into there.
(Cross-posted from Matti’s personal blog, Science & Industry)