Second Symposium "Computational Social Science"

Computational Social Science is becoming mature as a scientific discipline, thanks to substantial contributions from different disciplines.
In particular, social and computing sciences have recently engaged in interdisciplinary research, to demonstrate how social phenomena can be understood through digital traces, and how theories from the social sciences can be informed through computational and digital methods.

Our second symposium on Computational Social Science features three showcases of successful research in this area.
We discuss the validity of digital methods for social science research, in particular political science, and demonstrate how these methods can be used to understand collective phenomena in natural scenarios. Specifically, we focus on the emotional and social responses of societies to external shocks. How these phenomena can be observed through digital traces and what are their implications to our understanding of human behavior will be addressed from a wider theoretical perspective.

Program

9:00 - 9:15 Opening (Frank Schweitzer)

9:15 - 10:00 Pablo Barberá, School of International Relations, University of Southern California Less is more? How demographic sample weights can improve public opinion estimates based on Twitter data.

Twitter data is widely acknowledged to hold great promise for the study of political behavior and public opinion. However, a key limitation in previous studies is the lack of information about the sociodemographic characteristics of individual users, which raises concerns about the validity of inferences based on this source of data. This paper addresses this challenge by employing supervised machine learning methods to estimate the age, gender, race, party affiliation, propensity to vote, and income of any Twitter user in the U.S. The training dataset for these classifiers was obtained by matching a large dataset of 1 billion geolocated Twitter messages with voting registration records and estimates of home values across 15 different states, resulting in a sample of nearly 250,000 Twitter users whose sociodemographic traits are known. To illustrate the value of this approach, I offer three applications that use information about the predicted demographic composition of a random sample of 500,000 U.S. Twitter users. First, I explore how attention to politics varies across demographics groups. Then, I apply multilevel regression and postratification methods to recover valid estimate of presidential and candidate approval that can serve as early indicators of public opinion changes and thus complement traditional surveys. Finally, I demonstrate the value of Twitter data to study questions that may suffer from social desirability bias.

10:00 - 10:45 David Garcia, Chair of Systems Design, ETH Zurich Understanding Collective Emotions through Digital Traces of Human Behavior

Collective emotions are emotional states temporarily shared by large amounts of individuals. From riots to sport events, from viral content to online quarrels, collective emotions are subject to appear in various situations and have long-lasting consequences. Understanding collective emotions has been challenging due to their fast evolution, large scale, and complex dynamics, limiting their tractability in natural scenarios and their controllability in experiments. Recent developments in computational social science, in particular agent-based modeling and sentiment analysis of digital traces, allow us to quantify and model emotions at unprecedented scales and resolutions, offering new opportunities to study collective emotions.
I will present an overview of modeling and analysis of collective emotions at scale, with an emphasis on how these can be observed through social media and how they can be reproduced in computational models. I will illustrate the application of digital traces of collective emotions to understand the collective responses to the Paris terrorist attacks of November, 2015. Using a large-scale dataset of Twitter public messages, we measured affect as expressed through text and social synchronization as captured by Twitter metadata. We find the traces of the simultaneous negative reaction to the attacks, which are later replaced by positive affect, shared values, and group identity. Our analysis supports the hypothesis that synchronization leads to positive affect after a collective trauma, illustrating how resilience mechanisms foster adaptation through collective emotions.

10:45 -11:15  Coffee Break

11:15 - 12:00      Daniel Romero, School of Information, University of Michigan     Examining the effects of exogenous shocks on social networks and     the collaborative dynamics in organizations and crowds

Computational Social Science has begun to take advantage of rich communication and behavioral data regarding coordination, decision making, and knowledge sharing among groups of people. Most studies, however, have not generally analyzed how exegenous events are associated with a group's social network structure, communicative properties, and collaborative dynamics. In this talk, I will address these issues in two different settings. First, we analyze the complete dataset of millions of instant messages among the decision-makers in a large hedge fund and their network of outside contacts. We investigate the links between price shocks, network structure, and changes in the affect and cognition of decision-makers embedded in the network. When price shocks occur, the networks display a propensity for higher clustering, strong tie interaction, and an intensification of insider vs. outsider communication. Second, we examine changes in the collaborative behaviour of editors of Chinese Wikipedia that arise due to the 2005 government censorship in mainland China. Using the exogenous variation in the fraction of editors blocked across different articles due to the censorship, we examine the impact of the shock on overall activity, centralization, and conflict among editors. Overall, activity and conflict drop with the fraction of editors blocked, whereas centralization increases. The findings in both settings provide support for threat rigidity theory - orignally introduced in the organizational theory literature - in the context of decision makers in an organization and large-scale collaborative crowds.