Using LLMs: Text Data and Children's Aspirations

I am working with a unique dataset of 10,000 essays written by children in 1969 describing how they imagined their future. These essays are linked to more than 50 years of follow-up data on the same individuals, now in their 60s. The goal of the project is to explore the relationship between early-life aspirations and later-life outcomes.

The RA will work with local large language models (LLMs) running on the Yale Cluster to analyze and categorize the content of these essays. Responsibilities will include:

• iIteratively developing and refining prompts for LLMs to improve classification and extraction accuracy.
• Evaluating and cleaning model outputs for consistency and reliability.
• Assisting in the organization, documentation, and preliminary analysis of the categorized data.
• Collaborating with the research team to troubleshoot model performance and suggest enhancements.