Keywords: Sensemaking, Large Language Model (LLM), Visual Analytics
Overview
Fig 1.An example of human visual workspace for sensemaking.
Fig 2. Our Human-AI Sensemaking framework
We investigate the parallels between human cognition and AI for sensemaking tasks. We focus on the commonality of the concept of Space to Think, the use of external intermediate representations, as important to both human and AI sensemaking processes. Empirical results demonstrate the value of Space to Think for both, and how human-generated Spaces can improve LLM performance. We show how Space to Think can be used as a form of common ground in visual analytics systems to enable interactive human-AI sensemaking that accelerates the overall sensemaking process.
papers:
Steering LLM Summarization with Visual Workspaces for Sensemaking