Data Science for Social Challenges

Abstract

Many of today's social challenges cannot be adequately grasped simply by observing human behavior. To make these challenges visible and address their causes, we can use advanced statistics to disentangle complex interdependencies between the driving factors.In this course, we build up methodological skills and places a strong focus on interpretation and reflection of results.

Objective

A successful participant of this course will be able to - interpret the results of data analysis with regard to the methodological choices and the operationalization of theoretical concepts - assess potential flaws in research designs that can lead to flawed interpretations of results - apply a wide variety of statistical models (e.g., regressions, difference-in-difference, network models) to different data sources - and name the difference between statistical models and the advantages (or drawbacks) they hold for different data types - name the limitations of observational data analysis, especially with regard to causality - explain the importance of sensitivity and robustness checks for statistical analyses In summary, a successful participant is able to assess quantitative social science research with regard to its research design, the model choice as well as the interpretation drawn from the estimates and make suggestions for improvements.

Content

Data Science for Social Challenges offers a practical approach to the quantitative analysis of human behavior and social interactions. While the course `Social Data Science' focuses on data retrieval and processing, this course focuses on data analysis and interpretation of results. The course is organized in three blocks of increasing data complexity. The first block tackles linear data analyses, where a dependent variable is modeled based on a set of independent and control variables. The second block tackles causal inference, where experimental settings are approximated with observational data to allow for causal interpretation of results. The third block tackles data sources where observations are not independent of each other and therefore defy most statistical models. Here, we examine how people interact with each other and how these interactions affect the people involved in turn. The course covers various application of quantitative social sciences: - measuring biases in societies - analyzing behavior changes (due to internal or external events) - studying deviant behavior and peer effects - exploring coordination between people The course makes the link to sociological theories and shows how they can be used to derive testable hypotheses. A strong focus is laid upon the operationalization of different concepts, such as finding an appropriate measure of deviant behavior or the level of animosity that exists between people at a given time. These measures are tested using appropriate statistical models. Here, the focus is put upon the interpretation (e.g., coefficient sizes and power) as well as the presentation of results (e.g., through marginal effects). Lastly, the course fosters critical thinking by discussing sensitivity and robustness tests. As such, the course offers insights into quantitative research design by following a hands-on approach to the study of societal challenges through social data science. The course includes a lecture, presentations by external researchers, and an accompanying exercise class. In the exercise class students get the opportunity to run through the whole data analysis process. Starting with data inspection, students operationalize theoretical concepts and test them on various statistical models. Strong focus is put on sensitivity checks, where the effect of changes to the model (i.e., adding another control variable) is assessed. Students complete graded R exercises after the coding tutorial.

Resources