New artificial intelligence tools are rapidly transforming how economists and policy analysts conduct research, dissect data, and communicate findings. Rather than replacing traditional research methods, ChatGPT, Claude, and Gemini are serving as force multipliers, allowing analysts to explore ideas more thoroughly and overcome common bottlenecks in the research process.
Here are some of the basics.
The Basics
As of writing, I use ChatGPT, Claude, and Gemini fairly interchangeably. Oftentimes, I will ask the same query of all three to compare the outputs; that being said, I tend to find ChatGPT is best at doing deep research, Claude is the best at coding, and Gemini is the best for writing and copyediting.
Here are other notable tools worth exploring:
- NotebookLM—Google’s tool for extracting information from large documents. Upload a paper and NotebookLM can make a podcast from it.
- Copilot—Excellent at programming in Visual Studio Code.
- Cursor—A code editor that many swear by.
- Perplexity—An all-purpose “AI-powered answer engine.”
- arXiv Xplorer—An AI-powered search engine for ArXiv papers.
- ResearchRabbit—This tool will find similar research from just three articles. It also monitors new literature and visualizes research landscapes.
- Consensus—Ask any question and get research papers as a response.
Prompts and Prompt Engineering
The key to getting useful outputs from these AI tools lies in crafting effective prompts, which are the specific instructions and questions you provide to guide the model’s responses.
Some of my most commonly used prompts include:
- “Write three variations on this sentence/paragraph ensuring each takes an active voice, varying the length, structure, and sentence order…”
- “Proofread the following. Fix grammar and spelling mistakes. Make suggestions that will improve the clarity of my writing…”
- “Summarize the text below and give me a list of bullet points with key insights and the most important facts…”
- “Draft three paragraphs from the outline below…”
Research from Google has found that simple commands like “Take a deep breath and work on this problem step-by-step” or “Break this down” are incredibly effective at garnering high quality results. Sometimes, the most powerful way to get positive results is through an all-purpose prompt, like the one below:
Help me craft a really good prompt.
First, ask me what I want to do. Pause and wait for my answer. Ask questions to clarify as needed.
Second, once you have the information suggest a prompt that includes context, examples, and chain of thought prompting.
Third, show what your response as ChatGPT would be to the prompt.
Fourth, ask if the user has any suggestions and help them revise the prompt.
For data extraction, multi-step workflows, and document processing, pure JSON commands might be your best bet. This would mean prompting the LLM with the following string:
{
“task”: “summarize this article”,
“audience”: “college students”,
“length”: “100 words”,
“tone”: “curious”
}
Prompting can become quite complex. For example, Chain of Density prompts generate increasingly concise, entity-dense summaries to ensure articles are summarized correctly. Another trick is to ask the LLM to adopt multiple personas and simulate a debate between them. For more advanced prompts, visit the Prompt Engineering Guide, check out the Prompt Engineer subreddit, and follow Ethan Mollick on X.
Interesting Use Cases
For economists and policy analysts, the ability to quickly create compelling visualizations and break down complex mathematical functions has traditionally required significant technical expertise and time investment. But by knowing how to use the tools at your disposal, AI can speed this up considerably.
As Ömer Özak explained on X, “A student of mine needed to draw a graph so I decided to check whether #ChatGPT would be able to create figures in #TiKZ #LaTeX and voilà. So useful!” The prompt was simple: “Write Tikz code to draw the supply and demand curve for a good. Supply curve is red and demand curve is blue.”
Economist Scott Cunningham went one step further, asking ChatGPT to create a directed acyclic graph or DAG, with the following prompt:
The results were exactly what he asked for:
LLMs can also be useful in understanding equations. From economist Simon Eskildsen comes this workflow:
- Find big scary equation that’s hard to parse
- Latex OCR it with Mathpix
- Ask ChatGPT to break it down into heavily commented Python
Eskildsen was able to use this workflow to transform the equation below…
Into this…
Economist James Brand has found ChatGPT useful for developing “example code for structural estimation from a description + latex equations. The goal was to get it to give me a function that could estimate an approximate mixed-logit model.”
As these AI tools continue to evolve, economists and policy analysts who master prompt engineering and integrate these technologies into their workflows will find themselves better equipped to tackle complex research questions, communicate findings more effectively, and contribute more meaningfully to their profession.
The post AI Tools for Economists and Policy Analysts appeared first on American Enterprise Institute – AEI.
