.png){kind=icon}
.png){kind=icon}
Return
InsightLab: A Lower-Barrier EDA Workflow Using LLMs for Insight Discovery
.png){kind=icon}
PRE
AI
DOI：10.1007/978-3-032-06634-3_2.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Exploratory Data Analysis (EDA) is a critical, yet time-consuming, and labor-intensive process, especially for novice analysts who face a steep learning curve due to the high barrier to coding. Large Language Models (LLMs) have shown promise in reducing these challenges by enabling automatic chart generation and insight discovery through natural language interfaces. However, automating too much of the EDA process risks removing the analyst from the loop, leading to misaligned insights and diminished analyst independent thinking. This paper introduces InsightLab, an interactive system that streamlines the EDA process by integrating its key stages into a cohesive workflow. InsightLab uses a modular approach that balances automation and human control, offering three workflows with varying levels of AI assistance. A betweensubjects study with 27 participants revealed that each workflow creates different effectiveness patterns based on user expertise. Automation-focused tools efficiently handle routine tasks but provide limited assistance for insight generation, prompt-driven interfaces enhance analytical thinking among experienced users while overwhelming beginners, and more generative systems serve both novices and experts effectively, but risk reducing analytical engagement. These results highlight the importance of adaptive design in AI-assisted analytical tools.
Keywords:
Data analysis
Visual communication
Large language model
Journal
N
IF:
0
Papers:
90
Citations:
0
