Skip to content

Setting Up an AI Study Environment for Economics

This guide helps you set up an AI assistant that knows your course material. You will be able to ask questions, get explanations, and generate practice problems. The setup takes about 30 minutes.

What You Need

  • An account with an AI provider. I am using Claude for this guide because that's what I know.
  • Your lecture slides (PDF)
  • Other non-copyrighted course materials (problem sets, handouts, formula sheets)
  • Past exams (up to ~25 files)

Do not upload copyrighted textbooks.

Why Claude Projects?

Claude Projects let you upload files into a persistent knowledge base. Every conversation within the project can draw on those files. This means you upload your materials once and ask questions across many sessions.

Other tools exist (ChatGPT, Gemini, NotebookLM). The workflow below applies specifically to Claude, but the principles transfer. - NotebookLM has the benefit of zero setup and it provides links to files for every statement it makes. - But the user interface is meh.

Step 1: Create a Project

  1. Open claude.ai and log in.
  2. In the left sidebar, click Projects.
  3. Click Create Project.
  4. Name it something clear, e.g., Econ520 Fall 2026.

Step 2: Upload Your Materials

Inside the project, click Add content and upload your files. Organize them logically.

What to upload:

  • Lecture slides (all of them)
  • Problem sets and their solutions
  • Past exams and answer keys
  • Your own notes, if typed
  • Reading material from the slides' bibliographies.

File format tips:

  • PDF works best.
  • Keep file names descriptive: Lecture_04_ASAD_Model.pdf is better than slides4.pdf.

Limits: Claude Projects accept a large amount of text, but there is a ceiling. Twenty-five past exams plus a semester of lecture slides will typically fit. If you hit the limit, prioritize recent exams and core lecture slides.

Step 3: Write a System Prompt

Every project has a system prompt (also called project instructions). This tells the AI how to behave in every conversation within the project. A good system prompt makes the difference between vague answers and useful ones.

Click the Set project instructions field and paste something like the following. Adapt it to your course.

You are a tutor for an undergraduate macroeconomics course.
Your knowledge base contains the lecture slides, problem sets, and past exams for this course.

Rules:
1. Base your answers on the uploaded course materials. When you use a specific slide or exam, say which file and page/question you are referencing.
2. If the course materials do not cover a topic, say so clearly. Then offer a general explanation and flag that it may differ from what the instructor teaches.
3. Use the notation and terminology from the lecture slides. For example, if the slides write Y for output and P for the price level, use Y and P—not Q and π.
4. When explaining a model (e.g., AS/AD, IS/LM), walk through it step by step. State the assumptions first, then the mechanism, then the result.
5. When the student asks for practice problems, generate problems that match the style and difficulty of the uploaded past exams. Provide full solutions.
6. When you search the web for supplementary material, prefer university-hosted resources (.edu), central bank publications, and established economics encyclopedias (e.g., Econlib). Avoid anonymous blog posts and AI-generated summaries.
7. Keep explanations precise. Use short sentences. Avoid unnecessary jargon.

You can refine this prompt over time. If the AI gives answers that are too long, add "Keep answers concise." If it ignores your slides, strengthen rule 1.

Step 4: Start Asking Questions

Open a new conversation inside the project. The AI now has access to everything you uploaded. Here are examples of useful prompts.

Understanding a Concept

I don't understand how a negative supply shock shifts the AS curve in the short run. Walk me through it using the notation from the lecture slides.

Solving a Problem

Here is a question from the 2024 exam (Question 3a). I'm stuck on the second part. Can you show me how to derive the equilibrium price level after the fiscal expansion? Use the AS/AD framework from our slides.

Generating Practice Questions

Generate three exam-style questions on the IS/LM model. Make them similar in difficulty to the questions in the uploaded past exams. Give me the questions first, then the solutions separately so I can try them on my own.

Finding External Resources

I want to read more about the Phillips Curve and the expectations-augmented version. Search the web and find two or three reliable sources that explain it at an undergraduate level.

For this last type of query, Claude will search the web and return results. The system prompt instructs it to prefer reliable sources. Always check the links yourself—AI can sometimes return outdated or broken URLs.

Practical Tips

Ask one thing at a time. A focused question gets a better answer than a long paragraph with five sub-questions.

Correct the AI when it is wrong. If an answer contradicts your lecture slides, say so. For instance: "That's not how the instructor defined the natural rate. In Lecture 7, she defines it as..." The AI will adjust.

Use follow-up questions. Conversations have memory within a session. If an explanation is unclear, ask "Can you explain step 2 in more detail?" rather than starting a new chat.

Do not trust the AI blindly. AI assistants sometimes produce plausible but incorrect reasoning. This is especially true for mathematical derivations and graph interpretations. Always verify key steps against your lecture notes or a trusted source.

Separate study from assessment. Use the AI to learn and practice. Do not use it during graded work unless your instructor explicitly permits it.

Write detailed prompts. Think of the AI as a research assistant who has limited common sense. Spell out the details. State exactly what you want and why. You will get much better answers than from short, generic prompts.

Limitations

  • Mathematical notation. AI handles equations in text form reasonably well but can make errors in multi-step algebra. Double-check derivations.
  • Graphs. AI can describe what a graph should look like, but it cannot reliably read graphs embedded as images in your slides. If a question depends on a specific graph, describe it in words.
  • Instructor-specific content. Your instructor may use non-standard definitions or models. The AI follows the uploaded slides, but if those slides are ambiguous, the AI may guess wrong. When in doubt, ask your instructor.
  • Hallucination. AI can fabricate references, invent theorems, or misattribute ideas. The system prompt reduces this by anchoring answers in your uploaded files, but it does not eliminate it.

Summary

Step Action
1 Create a Claude Project for your course
2 Upload lecture slides, problem sets, and past exams
3 Write a system prompt that sets the rules
4 Ask focused questions, generate practice problems, and find reliable sources

The value of this setup grows as you use it. Refine your system prompt. Upload new materials as the semester progresses. Treat the AI as a study partner that has read all your slides—not as an authority that is always right.

Note: The first draft of this guide was written by Claude.