Social Data Science
The Internet has changed our lives pretty drastically over the last few decades. With the average user now spending over two hours on online social media every day, it’s changing our social lives as well. The huge amounts of produced online data offer interesting possibilities for research too. One interesting fact is that in 2018 Uppsala University was the Swedish higher education institution mentioned most frequently abroad online.
One of the researchers at Uppsala University dealing extensively with social media data is Matteo Magnani. To capture more of the complexity of online social interactions he has contributed to the development of the field of multilayer network analysis. This field studies a recent extension of the more classical methods for analysing social networks. Multilayer network theory starts from the consideration that social networks are often very complicated and multi-faceted, for example including multiple types of relationships between individuals and organisations. Being able to mathematically model all the facets of our social lives allows for better understanding of processes such as the spread of information. With the advent of the Internet, social networks have expanded to a new domain: online. Previously developed methods can now be applied to world-spanning networks such as those existing on social media. Using mathematical models together with online generated social data, it is now possible like never before to learn about people’s behavior.
Magnani’s research takes full advantage of these possibilities. Having a background in Computer Science, he saw the potential of applying modern computational methods to social science research. One of his current research projects is about using algorithms to identify behaviours that may be associated to malicious users and bots. Social data science can also be used to automatically summarise and monitor online information and discussions. Advancements are being made through the inclusion of new types of information into traditional network models, more notably text, pictures, time and uncertainty about the existence of relationships. Another recent project performed at the AI4Research center involves the large-scale analysis of visual political communication using social and deep neural networks.
Despite the promising prospects, there are limitations to this approach as well. Online data is notoriously unstructured and can be of low quality, making the development of analysis methods a complicated task. A lot of manual qualitative assessment of data and results must be applied in order to deal with low quality data. In many cases this means that data turns out to be insufficient to answer all our questions. Another issue is the variety of human behaviours, making it difficult to capture models of "typical" users.
Whether we can ever depend on algorithms to detect and remove malicious content from our online lives remains debatable. There is also the question of whether we would want an artificial intelligence enforcing some form of online censorship. If an algorithm could with confidence classify malicious bots in our online social networks, it would probably be up to governmental institutions and social websites themselves to put that information to some positive use, which is also a scenario not free from concerns. For this purpose, Magnani is coordinating the Nordic Network on Online Disinformation (NORDIS). NORDIS hosts multi-disciplinary gatherings attracting researchers from several disciplines to cooperate on this topic of common interest.
The analysis of social data will continue to be an important part of data science activities at the IT department. Furthermore, general methods for network analysis are also taught as part of the curriculum for the International Masters's Programme in Data Science. You can find more information about Matteo Magnani and his research on the Uppsala University InfoLab website, or read his book Multilayer Social Networks.
Article and interviews by Erik Jan Bootsma, MSc Computational Science