Incorporating Citizen-Generated Data into Large Language Models

This study investigates the use of citizen-generated data to optimize a large language model (LLM) chatbot that gives nutrition advice. By actively participating in the data collection and annotation process from FDA-approved websites, citizens provided insightful information that was essential for improving the model and addressing biases. The study highlights the difficulties in gathering and annotating data, especially in situations where nuances matter, such as pregnancy nutrition. The results show that the use of citizen-generated data improves the efficacy and efficiency of data collection procedures, providing a practical viewpoint and encouraging community involvement. In addition to guaranteeing data quality, the iterative process raises stakeholders’ awareness of and proficiency with data. Thus, citizen-generated data becomes an essential tool for creating information systems that are more reliable and inclusive.