Tech Setup

Why it matters: GitHub is where coursework lives this term: module assignments, your final project, and, if you build one, your personal portfolio site. It is also a standard professional workflow in data science and policy analysis.

How to set it up:

1. Create a free account at github.com if you do not already have one.
2. Sign up for the GitHub Student Developer Pack using your ETH email address. Apply early because student verification may take some time.
3. Install Git on your machine.
4. Configure Git with your name and email:
   git config --global user.name "Your Name"
git config --global user.email "you@example.com"
5. Join the course organization on GitHub using the link provided in class. Module templates and your project repository will live there.

Course recommendation: We will use RStudio for live coding, so we recommend it if you are new to programming or want to follow the demonstrations closely. You are welcome to use whichever IDE you prefer, including VS Code.

Choose where you want to work:

• RStudio (recommended): download RStudio Desktop after installing R. RStudio integrates scripts, the console, plots, files, packages, and help in one interface.
• VS Code: download Visual Studio Code, then install the R extension if you plan to use R.
• GitHub Codespaces: open a course repository on GitHub, select Code, then Codespaces, then Create codespace on main. This browser-based VS Code environment is a good option if you cannot install software, are using a shared computer, or encounter problems installing spatial-data libraries.
• Another IDE: you may use another editor or IDE you already know. Course support and live-coding demonstrations will focus on RStudio, so menus and keyboard shortcuts may differ.

Free GitHub accounts include 120 core-hours of Codespaces use per month. Verified students receive additional benefits through the GitHub Student Developer Pack. Stop your codespace when you finish working so it does not continue using your allowance.

Why it matters: These are the languages used for analysis: cleaning data, mapping it, modeling it, and evaluating policy. R is the primary course language, with strong ecosystems for spatial data and causal inference. Python alternatives are welcome for your own exploration.

How to set it up:

• If you chose GitHub Codespaces: R and Python are provided in the prepared course environment, so you do not need to install them on your computer.
• If you will use R locally: install R before installing RStudio or configuring your preferred IDE.
• If you will use Python locally: install Python through Anaconda or Miniconda, then configure it in your preferred IDE.

Course recommendation: We will focus on using a local model through Ollama. This gives us a shared way to examine what an AI model can and cannot do while keeping prompts, code, and data on your own computer. Follow the Ollama setup in the next section if your computer can run it.

Compare your options:

• Ollama (recommended for class): runs models locally, can work without sending prompts to a hosted AI provider, and gives you control over the model you use. Its speed and capability depend on your computer, local models can require substantial memory and disk space, and smaller models may produce weaker answers.
• GitHub Copilot: is easy to integrate with your IDE and often provides faster, more capable assistance for larger coding tasks. It requires an account and internet connection, and your requests are processed by a hosted service. It is free for verified students through the GitHub Student Developer Pack.
• Claude: can be strong at explanation, writing, and working through larger problems. It also requires an account and internet connection, sends requests to a hosted service, and may have usage limits or costs.
• Using Codespaces: hosted assistants work normally, but Ollama running on your laptop is not directly available at localhost inside a Codespace. For the Ollama exercises, use a local IDE where possible; contact the instructors if your computer cannot run a local model.

Engage critically: AI tools, models, prices, privacy terms, and capabilities change quickly. Do not treat any recommendation as permanent. Check what service and model you are using, what data it receives, and whether its output is supported by evidence. You remain responsible for the analysis and code you submit.

Why it matters: Local Ollama models are the main AI tools we will explore in class. Running a model yourself makes privacy, model choice, hardware constraints, and differences in model quality visible rather than hiding them behind a hosted service.

How to set it up:

1. Download and install Ollama for macOS, Windows, or Linux.
2. Open a terminal and confirm that Ollama is installed:
   ollama --version
3. Download and start a small coding model:
   ollama run qwen2.5-coder:3b
4. Enter a prompt to test the model. Type /bye when you are ready to exit.
5. To use the model in VS Code, install the Continue extension and select Ollama as the model provider. Ollama runs at http://localhost:11434 by default.

Local models run best on computers with at least 8 GB of memory; more memory generally allows larger and more capable models. If Ollama is too slow or cannot run on your machine, use GitHub Copilot or Claude and contact the instructors so you can still participate in the local-model exercises.