Arcano: Generating Data Questions to Support Non-Expert Analysts in Hypothesis Formulation
摘要
Analysts and designers often face challenges in formulating hypotheses and asking meaningful questions when exploring unfamiliar datasets. Although general guidelines may offer some support, the early stages of exploratory data analysis remain challenging, particularly for those without specialized knowledge or prior experience producing data visualizations. Recent research indicates that recommending data questions can support non-expert analysts in raising or refining hypotheses during exploratory analysis. In this context, we introduce Arcano, a visualization recommendation system that generates data questions and recommends effective visual representations to assist in the discovery of data answers and insights. Using an uploaded data set and its data properties, Arcano guides users through a mixed-initiative approach via system-generated questions that can assist users in raising or refining their analytical inquiries. We describe the rationale underlying Arcano and present its user interface. As future work, we aim to evaluate how system-generated data questions and Generative AI models can influence hypothesis generation by non-expert analysts when interacting with unfamiliar datasets.