The rapid growth of data has made accessing, integrating, and analyzing information increasingly challenging. While large-scale systems support processing and querying, interactive visualizations are essential for exploring complex datasets. Understanding how users gain insights from these visualizations requires capturing their interactions. Provenance data offers a natural solution, but current methods often fail to capture interaction-level provenance effectively. This paper presents an approach to capture and record user interaction provenance and integrate it with both prospective and retrospective provenance. We implement this approach in the Curio framework, which builds urban data visualization pipelines. Results demonstrate its effectiveness in capturing user behavior during visual exploration.
Oliveira et al. (Mon,) studied this question.