User-Centered AI for Data Exploration -- Rethinking GenAI's Role in Visualization

2025-04-05

AI Summary

Existing GenAI-powered visualization systems overemphasize automation while neglecting the heterogeneity of users’ domain expertise and analytical needs, resulting in rigid interaction, shallow reasoning, and poor adaptability. This paper proposes a user-centered adaptive GenAI visualization paradigm that elevates GenAI from an “automated tool” to a “cognitive collaborator.” Methodologically, we introduce a user capability assessment framework integrating multi-granularity interaction understanding, interpretable reasoning generation, and dynamic visualization synthesis to enable real-time adaptation of interaction modalities and reasoning depth. Our key contributions are threefold: (1) the first systematic articulation of adaptive visualization design principles; (2) a principled pathway for GenAI-augmented cognitive collaboration that enhances human analytical reasoning; and (3) identification of critical challenges and open research questions, establishing a theoretical foundation and practical guidelines for next-generation human–AI collaborative visualization systems.

Technology Category

• Cognitive Modeling & Cognitive Systems: Adaptive Behavior
• Humans and AI: Intelligent User Interfaces
• Computer Vision: Visual Reasoning & Symbolic Representations

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

Recent advances in GenAI have enabled automation in data visualization, allowing users to generate visual representations using natural language. However, existing systems primarily focus on automation, overlooking users' varying expertise levels and analytical needs. In this position paper, we advocate for a shift toward adaptive GenAI-driven visualization tools that tailor interactions, reasoning, and visualizations to individual users. We first review existing automation-focused approaches and highlight their limitations. We then introduce methods for assessing user expertise, as well as key open challenges and research questions that must be addressed to allow for an adaptive approach. Finally, we present our vision for a user-centered system that leverages GenAI not only for automation but as an intelligent collaborator in visual data exploration. Our perspective contributes to the broader discussion on designing GenAI-based systems that enhance human cognition by dynamically adapting to the user, ultimately advancing toward systems that promote augmented cognition.

Problem

• Adapt GenAI tools to user expertise levels
• Enhance visualization beyond automation for analysis
• Design user-centered AI for data exploration

Innovation

• Adaptive GenAI-driven visualization tools
• Assessing user expertise dynamically
• User-centered AI as intelligent collaborator