Hellenic Mediterranean University & Archimedes/Athena RC · University of Pittsburgh · Tsinghua University · Athena RC · Xi’an Jiaotong-Liverpool University
Classic challenges remain important, while AI introduces new concerns around LLMs, trust, explanations, and broader user access.
This paper summarizes a BigVis 2024 community survey on challenges, emerging topics, and opportunities in human-data interaction and visual analytics in the AI era. The survey involved experts from databases, information visualization, HCI, industry, and academia, and revisited the challenge agenda from the 2020 community report.
The central message is that many classic challenges remain relevant, especially scalability, efficiency, user understanding, guidance, and evaluation. At the same time, AI has reshaped the agenda: LLM-based visualization and analytics, fairness, trustworthiness, explainability, non-expert users, progressive analysis, and immersive visualization now stand out as prominent research directions.
All ten challenges from the 2020 report still receive at least 50% agreement or strong agreement on their current importance.
Scalability and efficiency are still the strongest carry-over challenge, with nearly universal agreement among survey participants.
LLMs, fairness, trustworthiness, and explainability appear as new agenda-setting themes that were absent from the older challenge list.
Human-in-the-loop processing and interactive, human-centered ML are among the most frequently selected emerging topics.
Visualization for non-expert users is identified as a key current challenge, connecting usability, guidance, explanation, and trust.
Progressive analytics, high-dimensional/stream data, and immersive visualization appear as newer directions in the AI-era agenda.
The most frequently mentioned current challenge is how to use LLMs effectively, safely, and meaningfully in visual analytics workflows.
Keep users actively involved in analysis, correction, interpretation, and decision-making rather than treating AI as a fully autonomous analyst.
AI-era analytics must address bias, uncertainty, provenance, and reliability across data, models, interfaces, and explanations.
Design ML systems around user goals, feedback, inspection, and iterative exploration.
Broader access requires systems that guide analysis, clarify uncertainty, and explain results in understandable ways.
Incremental, responsive analysis can reduce latency and support sensemaking over large, high-dimensional, or streaming data.
Large, dynamic, and complex data remain hard to analyze interactively, especially when AI models add cost and opacity.
AR/VR and spatial interfaces offer new ways to explore data, but require careful human-centered design and evaluation.
The recent 2026 community report Human-Data Interaction, Exploration, and Visualization in the AI Era discusses key challenges and opportunities for Human-AI Collaboration and Human-Data Interaction systems. It frames AI-era human-centered systems as a tightly coupled ecosystem of interaction, AI models, and multimodal data systems, highlighting the need for human-centered, trustworthy, and scalable analytics.
@article{bikakis2025visualAnalyticsChallengesAI,
title = {Visual Analytics Challenges and Trends in the Age of AI: The BigVis Community Perspective},
author = {Bikakis, Nikos and Chrysanthis, Panos K. and Li, Guoliang and Papastefanatos, George and Yu, Lingyun},
journal = {ACM SIGMOD Record},
year = {2025}
}