During the Fall 2024 semester, I took a psychology course built entirely around one major project. We were free to pursue any topic as long as it involved data analysis, so I chose to explore data visualization using artificial intelligence—a direction inspired by the rapidly evolving role of AI in the world.
The early stages of my project involved a lot of trial and error as I experimented with different datasets to understand what ChatGPT could reliably interpret and visualize. Pushing these boundaries on my own was eye-opening, especially when I discovered limitations such as the model’s cell cap for Excel files—going even one cell over often caused it to generate inaccurate values.
I met weekly with my professor to discuss progress, challenges, and potential next steps. These meetings strengthened my problem-solving skills and taught me how to clearly present what I had accomplished and what I planned to tackle next.
The first phase of the project focused on basic visualizations—box plots, dot plots, bar charts—where ChatGPT performed well with consistent accuracy. But my ultimate goal was to generate a heat map of the United States, and that proved far more difficult. None of the AI tools I tested could create a fully accurate map; while the general shape of the U.S. appeared correctly, the state labels were often wrong or replaced with nonsensical text. I’ve continued revisiting this challenge as AI capabilities improve, and recently was able to successfully map the data.
