Machine learning and big data

6-7 October 2022

Dre Elisa Volpi, CUSO

Dr. Aleksandra Urman, Postdoctoral researcher, University of Zurich

Email: [email protected] 

State of the art in the use of (big) social media data and machine learning for political science research

The main aim of the module is to familiarize the students with the state-of-the-art computational approaches to the analysis of social media data for political science research. Within the module, the students will learn about the following:

● What is the range of research questions that social media data is often used to address with concrete examples

● Pros and cons of using social media data for research

● The challenges and opportunities in acquiring such data

● What ethical issues arise when using social media data and complex black-box models for its processing, and the foundations of handling these issues ethically

● Which computational techniques are commonly used to process big data from social media and other sources, with a focus on natural language processing and machine learning

● What are common misconceptions and pitfalls of using complex ML models for political science research

To sum up, the module will provide a general overview of what's possible and not possible in addressing political science questions with the current computational techniques and social media data. This is not a programming course - it's neither possible nor desirable to attempt to teach the practical programming implementations of the techniques covered in the module within 2 days. So no programming experience is required to attend. However, the module will include pointers to online tutorials and courses in Switzerland and neighbouring countries that participants interested in acquiring relevant programming skills can attend in the future.

Below is the preliminary schedule of the module, note that it is subject to minor changes prior to the start of the course. Each session will include a short Q&A part, some sessions include group work on case studies, and the last session will include an extensive structured discussion of potential applications of different techniques and social media data for students' own research.

Day 1 - October 6th 2022

9.30-11.00 - Introduction to using social media data for political science research: what's different about social media data and how is it useful for political scientists?

11.00-11.30 - Coffee break

11.30-13.00 - Collecting social media data: what is possible and how; how data collection differs across platforms - and why you should keep this in mind when coming up with research questions

13.00-14.00 - Lunch

14.00-15.30 - Foundations of ethical collection, use and sharing of social media data for political science research

15.30-16.00 - Coffee break

16.00-17.30 - Intro to Machine Learning and Deep Learning and their use for political science research

Day 2 - October 7th 2022

9.30-11.00 - Computational text analysis for political science research in the last 10 years: from "vanilla" topic modeling to black-box ML models and transformers

11.00-11.30 - Coffee break

11.30-13.00 - Computational analysis of image and video data

13.00-14.00 - Lunch

14.00-15.30 - Network analysis on large social media data

15.30-16.00 - Coffee break

16.00-17.30 - Wrap-up: summary of the state-of-the-art and extensive discussion of students' interests in potential application of computational techniques and social media data for their own research

Genève

The event takes place at the University of Geneva (Uni-Mail). More details about the exact room will follow.

18