Like Quantitative Social Science, Computational Social Science (CSS) is often concerned with the problem of explaining correlations in observational data. But beyond that, CSS is also concerned with predicting the numerical properties of observations or what categories they belong to. While explanations are also done in CSS with conventional statistical models (like the Generalized Linear Model), predictions are the turf of machine learning (ML). In the workshop, we will provide a basic understanding of ML, how predictions are made, and to what extent explanations are possible. We will touch upon the basics of supervised and unsupervised ML. Within supervised ML, regression is about predicting numbers, and classification is about predicting categories. Within unsupervised learning, clustering is about grouping observations, and dimensionality reduction is about grouping variables (in ML called features). In many cases, ML is performed on tables of observations (in rows) and features (in columns). We will be using such a non-social toy dataset to demonstrate the methods and a social dataset to learn about the practice of CSS. The workshop will alternate between live-coding demonstrations and periods in which participants apply that knowledge in context, both using Jupyter Notebooks. The software we will be using is scikit-learn, a standard Python library that is simple to understand, provides a breadth of options, and has a large user community. At the end of the workshop, participants will have an intuition about what ML can and cannot do. We will close with an outlook on how ML relates to Artificial Intelligence.

Undergraduate, master students, doctoral candidates, postdoctoral researchers, and experienced researchers who want to get introduced to the practice of Computational Social Science.

Participants are expected to know the basics of Python and have at least some experience using it.

For the workshops, participants should bring a running system on which they can execute Jupyter Notebooks. We will be using Python 3.9 and several standard libraries that are part of the Anaconda 2022.10 distribution or can be installed on top of that. A list of libraries and versions of these libraries that participants should import will be circulated before the workshops.

We recommend that participants install Anaconda 2022.10. Feel free to also work in a cloud-like Google Colab. Consult this link for more detailed instructions on how to set up your computing environment.