This discussion of the practical aspects of teaching computational social science comes with two caveats. First, a computational social science project should start with a theoretically driven research question rather than the latest computational technique. Collecting the largest possible volume of social media data or applying the most powerful machine learning technique does not directly add value to political science research. These tools become relevant when they provide a new understanding of key theoretical questions in the relevant literature. In addition, linking computational methods to key political science topics has pedagogical value. It enables students to discover how they could apply these methods to explore their research interests (Williams et al. 2021).
Second, emphasizing the value of data-intensive social science research does not undermine the importance of careful human investigation. Big data is not the same as good data. A researcher still needs to assess whether the data can answer their research question regarding measurement and internal and external validity (Meng 2018). Because of this, computational methods are better applied when researchers are well-trained in research design and methods. A strong foundation in descriptive analysis, hypothesis testing, statistical modeling, and causal inference remains critical to answering empirical research questions.
However, having a strong motivation or a deeper understanding of a substantive prob- lem does not automatically turn a research idea into a research project. A successful student-designed research project involving computational methods requires an adequate understanding of programming and its implementation. Nevertheless, programming is chal- lenging for political science students because most have little prior experience. This challenge is especially acute among students from disadvantaged backgrounds because they have had relatively poor access to high-quality quantitative and computational education (for a review, see Xie, Fang, and Shauman 2015). This article tackles this problem by presenting principles and specific steps for lowering the barriers to learning these exciting new tools and techniques.
Designing small exercises can be problematic. If the exercises are not challenging, the students will not learn new information. If they are too complicated, students can lose confidence. To hit the sweet spot, some principles need to be observed. First, the exercises should be based on the content covered in the lecture. Second, instructors should encourage students to apply their skills beyond the contexts addressed in class. Third, the exercises should have an accompanying step-by-step guide and a template (e.g., R Markdown or Jupyter notebook). In essence, students need to be provided with a framework that they can use, opportunities that they can explore, and constraints within which they can engage in exploration without excessive risk (Wickham 2015).
- On Explanations, Fairness, and Appropriate Reliance in Human-AI Decision-Making - September 27, 2022
- Adverse effects of information personalization on human learning - September 26, 2022
- eXtended [Reality] Artificial Intelligence: New Prospects of Human-AI Interaction Research - September 19, 2022