ASTK18274U Political Analysis of Social Media Data in Python: From Natural Language Processing, to Machine Learning, to the Ethics of Data Science

This course enables students to critically use a variety of advanced data analytical tools for the political analysis of social media and text data. While these tools are more broadly applicable, the course has a particular and hands-on focus on the Twitter data collected for the department’s ERC-funded project “Diplomatic Face-Work Between Confidential Negotiations and Public Display (DIPLOFACE)” (PI: Prof. Rebecca Adler-Nissen). The data covers more than 80 million Twitter status updates regarding “Brexit” since March 2019. The course participants will apply the learned data analytical techniques to samples of this dataset and thus investigate research questions of their choice in the final assignment.

The goal is to not just teach advanced methods in the abstract, but to give participants a hands-on understanding of the promises and challenges of applied data analytical research in the field of political science. More practically, participants of this course will learn:

• How to program and manage complex data in Python
• How to explore and understand basic patterns in the data via descriptive statistics and data visualization
• How to use natural language processing techniques (tokenizing, lemmatizing, stemming, part-of-speech tagging, etc.)
• How to apply unsupervised and supervised machine learning techniques to understand large amounts of text data

A key ambition of this course is not just to convey the technical know-how of how to apply these tools, but also to provide an understanding of their philosophical underpinnings, the practical pitfalls of data science, and the ethical and political challenges that emerge from these. Thus, the course encourages participants to critically reflect on what is meant by ‘artificial intelligence’ or ‘machine learning’, and on the extent to which the possibilities of these tools are ethically acceptable or politically desirable.

The course is structured into 14 weekly sessions of four hours each. Of these, two hours are dedicated to understanding and discussing the conceptual dimensions of the given data analytical techniques. Thereafter, the course will turn into a programming laboratory in which these techniques are applied in Python and to the Twitter data on Brexit. These lab sessions will introduce homework assignments that can be done in groups, and the solutions to which are provided and discussed in the following week. The instructor will provide for additional weekly ‘methods cafés’ as an opportunity to discuss individual questions and challenges beyond the classroom.

The weekly homework assignments prepare the participants for the writing of a final paper, in which they address a research question of their choice through the learned methods and using the provided Brexit Twitter data. The final assignment should involve the application of the learned techniques through Python, but can also address more abstract philosophical, political, or ethical issues.

N.B.: The course only provides a brief introduction to programming in Python. Participants without any prior knowledge of the language are encouraged to complete the ‘Introduction to Python’ course on DataCamp before the start of the course.