Visualization and Automation in Data Science: Exploring the Paradox of Humans-in-the-Loop

This position paper explores the interplay between automation and human involvement in data science. It synthesizes perspectives from Automated Data Science (AutoDS) and Interactive Data Visualization (VIS), which traditionally represent opposing ends of the human-machine spectrum. While AutoDS aims to enhance efficiency by reducing human tasks, VIS emphasizes the importance of nuanced understanding, innovation, and context provided by human involvement. This paper examines these dichotomies through an online survey and advocates for a balanced approach that harmonizes the efficiency of automation with the irreplaceable insights of human expertise. Ultimately, we address the essential question of not just what we can automate, but what we should automate, seeking strategies that prioritize technological advancement alongside the fundamental need for human oversight.