Resources:
- Prompting Guide
- How to Use ChatGPT’s Advanced Data Analysis Feature
- Advanced Data Analysis is now an integrated part of ChatGPT Plus
- Coding helpers exist, including Git Guru
Demo:
- Session for generating a dataset
- The generated dataset ("never" share data on Github)
- Session for analyzing the dataset (live demo)
The first few prompts. Nothing fancy, just showing that one can indeed start with a very open-ended approach.
First:
Hey, I am getting ready to give a demo on how to use Advanced Data Analysis as a research tool. The first thing would be to create a prompt for a Chat-GPT-based research assistant. How should I go about prompting?
Second:
This is academic research, specifically. Let's consider CSCW as a potential outlet. The data set will be a synthetic sample of Reddit. We'll be initially focusing on the users of the platform. Specifically, identifying different types of users. Please recommend a prompt following the guidelines above.
ChatGPT as a research assistant can
- produce synthetic data
- write and edit code
- debug and fix code
- run the code
- analyze and interpret the output of the code
Concluding remarks
- Sharing the data with ChatGPT is a showstopper in many cases
- Creating a synthetic dataset enables using ChatGPT as a lead developer
- ... in which case you can run the code in your own environment
- Integrating code snippets definitely insists on development experience
- ... not to mention the need to evaluate and validate the approach that ChatGPT offers
- Bonus: try copying the description of your favorite analysis and using that as prompt