Computational
social science
I wanted a degree in Computational Methods after doing my senior honors thesis, but then I realized it would put me in debt, so I designed this open self-study curriculum. No tuition. No prerequisites beyond curiosity. Traditional degrees have their place and there are many benefits to learning from faculty experts, but I'm not sold on a life in research yet, and wanted to upskill somehow. This is very preliminary and will be updated as I read academic literature and design assignments.
The phase materials below are still being designed and are password-protected for now. Getting started is open to everyone.
Required software
- R
- RStudio
- Python
- VS Code (GUI)
For a more modern GUI you may use R extensions in VS Code rather than RStudio. I use R and R Plot Pro.
However, if you are new to R, I recommend learning the most basic skills through a few swirl lessons, where you can learn R directly in RStudio. This is how I started with zero statistics experience. AI can do most coding these days, but having a grasp on the fundamentals enables you to challenge AI and critique its output. When I first started learning R, I had access to AI models, but I coded in the terminal for weeks to enhance my understanding. Once you have a baseline, AI will exponentially speed up your workflows — and once you're more advanced, it's helpful to work in R within VS Code for machine learning applications and practical builds.
Start with an ethics debrief
- Ethics of computational sociology by Christopher Andrew Bail youtube.com/watch?v=vwleR-9-788
- Can generative AI improve social science? by Christopher Andrew Bail pnas.org/doi/10.1073/pnas.2314021121
optional · not a priority
Functions, data structures, classes, control flow. Don't move on until you can write something small from scratch without looking things up.
Start a GitHub repository to show your interest and engagement with the project.
Distributions, hypothesis testing, p-values.
Pandas, seaborn, matplotlib. These tools appear in every many data projects.
Linear to logistic to ordinal logistic regression. McElreath's lectures are free on YouTube. Ordinal models are my personal favorite and relevant to user survey feedback modeling.
Many organizations run on SQL at some layer. Fluency here is a practical baseline. Reddit has a great community for SQL questions and answers.
Chris Bail is one of the leading figures in computational social science and designed this bootcamp to be fully open. Covers APIs, text analysis, machine learning, and network analysis in R — directly applied to social science questions. The closest thing to a real CSS course without the tuition.
Webscrape reviews for any product you love. Clean, analyze, visualize, and write it up. Think of ways to turn text into actionable insights.
Likert scales, question wording effects, sampling bias. The foundation under any ordinal model is a well-designed survey.
Sentiment analysis, topic modeling, embeddings. Directly applicable to customer feedback analysis.
For when the relationship between variables isn't a straight line. Useful complement to the regression work in Phase 2.
Design a small survey, collect real responses, and model the results with ordinal logistic regression. Write up what the model says and what it doesn't.
Design an experiment around a product decision. Apply ordinal logistic regression to the results. Write the findings clearly enough that a non-technical reader can follow the logic. This is the portfolio.
The ability to explain what the numbers mean to people who don't think in numbers is a crucial skill.
Curriculum complete.
No debt. A GitHub portfolio and the ability to think clearly about data.
Now build something that matters.