Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

InsightLens: Augmenting LLM-Powered Data Analysis with Interactive Insight Management and Navigation

Luoxuan Weng, Xingbo Wang, Junyu Lu, Yingchaojie Feng, Yihan Liu, Haozhe Feng, Danqing Huang, Wei Chen

TL;DR

InsightLens tackles the challenge of tracking and navigating LLM-generated insights during data analysis by automatically extracting, organizing, and visualizing insights and their evidence in real time. It uses an LLM-agent-based framework with an Insight Extraction (IE) and Insight Organization (IO) pipeline, plus five coordinated UI views including an Insight Minimap and Topic Canvas to support multi-level navigation. A formative study and a user study show significant reductions in manual effort and cognitive load while maintaining conversational workflow, and improved coverage of data attributes and analytic topics. The approach demonstrates a practical path toward scalable, explainable, and efficient LLM-powered data analysis across conversational interfaces and analytic workspaces.

Abstract

The proliferation of large language models (LLMs) has revolutionized the capabilities of natural language interfaces (NLIs) for data analysis. LLMs can perform multi-step and complex reasoning to generate data insights based on users' analytic intents. However, these insights often entangle with an abundance of contexts in analytic conversations such as code, visualizations, and natural language explanations. This hinders efficient recording, organization, and navigation of insights within the current chat-based LLM interfaces. In this paper, we first conduct a formative study with eight data analysts to understand their general workflow and pain points of insight management during LLM-powered data analysis. Accordingly, we introduce InsightLens, an interactive system to overcome such challenges. Built upon an LLM-agent-based framework that automates insight recording and organization along with the analysis process, InsightLens visualizes the complex conversational contexts from multiple aspects to facilitate insight navigation. A user study with twelve data analysts demonstrates the effectiveness of InsightLens, showing that it significantly reduces users' manual and cognitive effort without disrupting their conversational data analysis workflow, leading to a more efficient analysis experience.

InsightLens: Augmenting LLM-Powered Data Analysis with Interactive Insight Management and Navigation

TL;DR

InsightLens tackles the challenge of tracking and navigating LLM-generated insights during data analysis by automatically extracting, organizing, and visualizing insights and their evidence in real time. It uses an LLM-agent-based framework with an Insight Extraction (IE) and Insight Organization (IO) pipeline, plus five coordinated UI views including an Insight Minimap and Topic Canvas to support multi-level navigation. A formative study and a user study show significant reductions in manual effort and cognitive load while maintaining conversational workflow, and improved coverage of data attributes and analytic topics. The approach demonstrates a practical path toward scalable, explainable, and efficient LLM-powered data analysis across conversational interfaces and analytic workspaces.

Abstract

The proliferation of large language models (LLMs) has revolutionized the capabilities of natural language interfaces (NLIs) for data analysis. LLMs can perform multi-step and complex reasoning to generate data insights based on users' analytic intents. However, these insights often entangle with an abundance of contexts in analytic conversations such as code, visualizations, and natural language explanations. This hinders efficient recording, organization, and navigation of insights within the current chat-based LLM interfaces. In this paper, we first conduct a formative study with eight data analysts to understand their general workflow and pain points of insight management during LLM-powered data analysis. Accordingly, we introduce InsightLens, an interactive system to overcome such challenges. Built upon an LLM-agent-based framework that automates insight recording and organization along with the analysis process, InsightLens visualizes the complex conversational contexts from multiple aspects to facilitate insight navigation. A user study with twelve data analysts demonstrates the effectiveness of InsightLens, showing that it significantly reduces users' manual and cognitive effort without disrupting their conversational data analysis workflow, leading to a more efficient analysis experience.
Paper Structure (28 sections, 3 figures)

This paper contains 28 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. NLIs for Data Analysis
  4. Analytic Provenance in Data Analysis
  5. Exploration of LLM Responses
  6. Formative Study
  7. Participants and Procedure
  8. Findings
  9. Design Requirements
  10. InsightLens: Framework
  11. Insight Extraction
  12. Insight Organization
  13. InsightLens: User Interface
  14. User Interface Overview
  15. User Input & Insight Inspection
  16. ...and 13 more sections

Figures (3)

  • Figure 1: InsightLens consists of (A) a user interface and (B) an LLM-agent-based framework. While users are (A1) interacting with the analytical chatbot, the Insight Extraction (IE) Agent (B1) takes each round of Q&A for insight extraction and evidence association, as well as interestingness evaluation. Following this, the Insight Organization (IO) Agent (B2) organizes the insights by identifying their data context, analytic topics, and related insights. Users can then (A2) inspect the extracted insights and (A3) explore the structured topics with progressively-evolving visualizations.
  • Figure 2: The results of the measures and qualitative ratings regarding InsightLens's support for insight management and navigation.
  • Figure 3: The results of the questionnaire regarding InsightLens's effectiveness, usability, and impact on data analysis.