The article discusses a study conducted on the challenges faced by data scientists when interacting with Large Language Models (LLMs) powered chatbots. The study, conducted via interviews and a survey, identified key issues including difficulty in retrieving contextual data, formulating prompts for complex tasks, adapting generated code, and refining prompts iteratively. The authors propose design recommendations, such as data brushing and inquisitive feedback loops, to improve communication with AI-based assistants in data science tools.
Publication date: 24 Oct 2023
Project Page: https://chat.openai.com/
Paper: https://arxiv.org/pdf/2310.16164
