General chair: Prof. Markus Strohmaier (GESIS & U. of Koblenz)
The GESIS Computational Social Science Winter Symposium is a one day event held on Dec 1st 2014 in Cologne, Germany. The main objective of the event is to explore the manifold challenges and opportunities related to understanding social systems via computational approaches, and via new kinds of data (e.g. online social networks, physical contact sensors, communication traces, online- and offline behavioral data, etc).
The Computational Social Science Winter Symposium will provide an environment for networking and interdisciplinary discussions through three main activities: (i) an exciting program featuring a series of invited talks that will provide different perspectives on current advances and limitations of computational social science. (ii) an open call for contributed posters that will provide opportunities for computational social scientists to present and discuss their own work and (iii) an evening event at the famous Cologne Christmas markets that will provide plenty of opportunities for further discussions and informal networking.
The program will feature lectures from the following invited speakers (all confirmed):
When Webb et al. (1966) published their highly innovative work entitled „unobtrusive measures: nonreactive research in the social sciences“ they certainly were not aware that some decades later the internet would provide a huge mass of unobtrusive data. These data on numerous aspects of human behavior are highly useful for the investigation of many important research questions in the social sciences. The data have several advantages: They refer to the observation of objective behavior, the behavior is not distorted by the researcher’s instrument, and the samples are often large and contain hundreds of thousands observations or more. In my presentation, I will report on our study based on more than 300’000 eBay auctions. The aim of the analysis is to explore hypotheses on reputation effects, trust and free-rider problems, and the functioning of a system of social cooperation in general.
Our everyday usage of the Internet generates huge amounts of data on how humans collect and exchange information worldwide. In this talk, I will outline recent work in which we investigate whether data from sources such as Google, Wikipedia and Flickr can be used to measure and even predict real world human behaviour. I will provide case studies from the economic domain and beyond.
Over the last couple of years and from a range of theoretical perspectives, social science research has repeatedly pointed to new challenges and new opportunities for data collection, data analysis and sociological explanations (e.g. Latour et al. 2012; Lazer et al. 2009; Savage/Burrows 2007): given our ubiquitous use of digital media, we generate heaps of data that allow insights into, for instance, consumer and social behavior at large scale. Increasingly, social research does not use survey or interview data anymore, which relies on ex post or a priori evaluation, but where and when possible analyzes large amounts of relational data that is generated while social processes are taking, e.g. when communicating or consuming (e.g. Kossinets/Watts 2006; Pentland 2014).
These types of relational data are already available in digital form. However, there are also gigantic amounts of already existing textual data, which relate to each other and/or refer to the same things or actors: e.g. printed media reports, speeches, minutes of meetings, press statements, evaluative reviews, letters, contracts. My talk argues for the systematic use and analysis of texts as data using large-scale data sets. Based on different empirical studies, I show how large amounts of media reports can be first digitized and cleaned to then be formally analyzed using computational tools such as topic modeling. Originally developed in computer science, machine learning, and computational linguistics, the method of topic modeling sorts together terms that cooccur into semantic contexts of relational meaning and thus identifies the themes that structure a textual corpus. The themes can be understood as frames (Gamson 1992). It is thus a method that captures many elements of a cultural sociology interested in the measurement of meaning and taking relationality into account (e.g. Mohr 1998; Mohr/Bogdanov 2013). In sum, my presentation bridges between current discussions in digital humanities, where huge amounts of text and image data sources are digitized and related to each other, and in computational social science, which uses digitally generated data and algorithmic tools to explain social action – using a sociological perspective which searches for patterns in order to find explanations.
The global spread of infectious diseases, rumors, opinions, and innovations are complex, network-driven dynamic processes. The combined multiscale nature and intrinsic heterogeneity of the underlying networks make it difficult to develop an intuitive understanding of these processes, to distinguish relevant from peripheral factors, to predict their time course, and to locate their origin. I will show that complex spatiotemporal patterns can be reduced to surprisingly simple, homogeneous wave propagation patterns, if conventional geographic distance is replaced by a probabilistically motivated effective distance. I will explain how this approach works in scenarios such as the 2003 SARS epidemic, the H1N1 influenza pandemic of 2009, and the EHEC outbreak in Germany in 2011. I will also discuss how these insights can be used to understand the import risk in the context of the current Ebola outbreak in West Africa.
Wearable devices are opening up a new window on human social behavior at the fine resolution of individual face-to-face interactions, impacting diverse research domains that span the social sciences, public health, location-aware services, and more. In this talk we will focus on networks of human close-range proximity measured in real-world environments by means of wearable sensors. We will illustrate the structural and temporal heterogeneities exhibited by the empirical data and discuss the effect of these features on the dynamics of simple processes unfolding over the social network. In important social contexts - such as schools - where the interaction network is shaped by spatial constraints and by an externally-defined activity schedule, a rich meso-scale network structure may arise: we will show how techniques from machine learning can be successfully used to detect correlated activity patterns and assess their relevance. Finally, we will discuss the potential of high-resolution social network data to gain insight into specific issues such as gender homophily or the relation between mental health and network position.
Computational social science is only as good as the models used and the data analysed to describe and to understand social phenomena.
So, when is a model good enough? And how can the available data be used to calibrate a model?
In this talk, we will illustrate these problems by focusing on emotional influence. Different from opinions, emotions are short-lived psychological states that strongly bias individual behavior. Following Russell (1980), emotions can be classified along the dimensions of valence (the pleasure associated with emotions) and arousal (the degree of activity induced by emotions). We can quantify the emotions of individuals who, for example, participate in an online chat, by surveying their subjective response or by providing a sentiment analysis of the text they read and write. But how can this be linked to a model? We have developed an agent-based modeling framework where the dynamics of individual valence and arousal and the communication between agents is explicitely modeled. For the emotional response of agents, we can test different assumptions (a) by fitting these to the observed subjective response, and (b) by comparing the model output to the observed collective behavior. We will provide different examples (communication in online chats, product reviews, emotional cascades) to demonstrate that the agent-based models can remarkably well reproduce the real behavior of users in online social media.
Users of social media produce huge amounts of digital trace data. The wealth of these data has led scholars from diverse disciplinary backgrounds to study what might be concluded from online traces about processes in society at large. In particular, public opinion on political matters is one of the fields that has attracted much scholarly attention. Some analyses suggest that social media communication reflects current public opinion or future election outcomes, whereas others qualify or deny this claim. Despite its liveliness, the debate has not paid much attention to the mechanisms potentially linking digital trace data to processes in society at large, however. This lack of theoretical reasoning is an impediment to our understanding of the potentials and limitations of social media data in reflecting societal processes. In this talk, I will thus address the relationship between information gleaned from digital trace data and public opinion on political topics with an emphasis on the underlying mechanisms. Drawing on this perspective, hypotheses on social media communication in the run-up to the 2013 German federal election will be derived and tested.