The rapid advancement of AI is transforming human-centered systems, with profound implications for human-AI interaction, human-data interaction, and visual analytics. In the AI era, data analysis increasingly involves large-scale, heterogeneous, and multimodal data that is predominantly unstructured, as well as foundation models such as LLMs and VLMs, which introduce additional uncertainty into analytical processes. These shifts expose persistent challenges for human-data interactive systems, including perceptually misaligned latency, scalability constraints, limitations of existing interaction and exploration paradigms, and growing uncertainty regarding the reliability and interpretability of AI-generated insights. Responding to these challenges requires moving beyond conventional efficiency and scalability metrics, redefining the roles of humans and machines in analytical workflows, and incorporating cognitive, perceptual, and design principles into every level of the human-data interaction stack. This paper investigates the challenges introduced by recent advances in AI and examines how these developments are reshaping the ways users engage with data, while outlining limitations and open research directions for building human-centered AI systems for interactive data analysis in the AI era.

We explore the design of visualizations for values spanning multiple orders of magnitude; we call them Orders of Magnitude Values (OMVs). Visualization researchers have shown that separating OMVs into two components, the mantissa and the exponent, and encoding them separately overcomes limitations of linear and logarithmic scales. However, only a small number of such visualizations have been tested, and the design guidelines for visualizing the mantissa and exponent separately remain under-explored. To initiate this exploration, better understand the factors influencing the effectiveness of these visualizations, and create guidelines, we adopt a multi-stage workflow. We introduce a design space for visualizing mantissa and exponent, systematically generating and qualitatively evaluating all possible visualizations within it. From this evaluation, we derive guidelines. We select two visualizations that align with our guidelines and test them using a crowdsourcing experiment, showing they facilitate quantitative comparisons and increase confidence in interpretation compared to the state-of-the-art.

We propose leveraging Large Language Models (LLMs) as an interaction layer for medical visualization systems. In domains like healthcare, where users must navigate high-dimensional, coded, and heterogeneous datasets, LLM-generated queries enable expert medical users to express complex analytical intents in natural language. These intents are then translated into editable and executable queries, replacing the dynamic query interfaces used by traditional visualization systems built around sliders, check boxes, and drop-downs. This interaction model reduces visual clutter and eliminates the need for users to memorize field names or system codes, supporting fluid exploration, with the drawback of not exposing all the filtering criteria. We also reintroduce dynamic queries on demand to better support interactive exploration. We posit that medical users are trained to know the possible filtering options but challenged to remember the details of the attribute names and code values. We demonstrate this paradigm in ParcoursVis, our scalable EventFlow-inspired patient care pathway visualization system powered by the French National Health Data System, one of the largest health data repositories in the world.

We present Bi-Scale density Plot (BSP), anew technique to enhance density plots by efficiently optimizing the local density variance in high- and mid-density regions while providing more details in low-density regions. When visualizing large and dense discrete point samples, scatterplots and thematic maps are often employed and we need density plots to further provide aggregated views. However, in the density plots, local patterns such as outliers can be filtered out and meaningful structures such as local density variations can be broken down. The key innovations in BSP include (i) the unified bin-summarize-decompose-combine framework for interactively bi-scale enhancing density plots through combining large- and small-scale density variations; and (ii) the variance-aware filter, which is reformulated based on the edge-preserving image filter, for maintaining the relative data density while reducing the excessive variability in the density plot. Further, BSP can be adopted with a 2D colormap, allowing simultaneous exploration of the enhanced structures and recovering the absolute aggregated densities to improve comparison and lookup tasks. We empirically evaluate our techniques in a controlled study and present two case studies to demonstrate their effectiveness in exploring large data.

While existing visualization libraries enable the reuse, extension, and combination of static visualizations, achieving the same for interactions remains nearly impossible. Therefore, we contribute an interaction model and its implementation to achieve this goal. Our model enables the creation of interactions that support direct manipulation, enforce software modularity by clearly separating visualizations from interactions, and ensure compatibility with existing visualization systems. Interaction management is achieved through an instrument that receives events from the view, dispatches these events to graphical layers containing objects, and then triggers actions. We present a JavaScript prototype implementation of our model called Libra.js, enabling the specification of interactions for visualizations created by different libraries. We demonstrate the effectiveness of Libra by describing and generating a wide range of existing interaction techniques. We evaluate Libra.js through diverse examples, a metric-based notation comparison, and a performance benchmark analysis.

We present a novel visualization-driven illumination model for density plots, a new technique to enhance density plots by effectively revealing the detailed structures in high- and medium-density regions and outliers in low-density regions, while avoiding artifacts in the density field's colors. When visualizing large and dense discrete point samples, scatterplots and dot density maps often suffer from overplotting, and density plots are commonly employed to provide aggregated views while revealing underlying structures. Yet, in such density plots, existing illumination models may produce color distortion and hide details in low-density regions, making it challenging to look up density values, compare them, and find outliers. The key novelty in this work includes (i) a visualization-driven illumination model that inherently supports density-plot-specific analysis tasks and (ii) a new image composition technique to reduce the interference between the image shading and the color-encoded density values. To demonstrate the effectiveness of our technique, we conducted a quantitative study, an empirical evaluation of our technique in a controlled study, and two case studies, exploring twelve datasets with up to two million data point samples.

Numeracy—the ability to understand and reason with numbers—is essential for interpreting and acting on quantitative information, yet it remains under-addressed in visualization research. In this position paper, we argue that numeracy should be a guiding principle in the design and evaluation of public-facing visualizations. Drawing on cognitive psychology and decision science, we propose a numeracy-informed design process: (1) measuring individual differences in numeracy, (2) designing visualizations that support gist and verbatim understanding, reduce cognitive effort, and provide evaluative meaning, and (3) evaluating quantitative comprehension across numeracy levels. Our goal is to spark a broader conversation about inclusive visualization practices that reflect numerical diversity and support public understanding of quantitative information.

We present ParcoursVis, an open-source Progressive Visual Analytics tool designed to explore aggregated electronic health record sequences of patients at scale. Existing tools are limited to about 20k patients that they can process fast enough to remain interactive, under human latency limits. They need to process the whole dataset before showing the visualization, taking a time proportional to the data size. Yet, managing large datasets allows for discovering rare medical conditions and unexpected patient pathways, contributing to improving treatments. To overcome this limitation, ParcoursVis relies on a progressive aggregation algorithm that quickly computes an approximate initial result, visualized as an Icicle tree, and improves it iteratively, until the whole computation is done. With its architecture, ParcoursVis remains interactive while visualizing the sequences of millions of patients – three orders of magnitude more than similar tools. We describe our PVA architecture, which achieves scalability with fast convergence and visual stability.

Communicating climate change remains challenging, as climate reports, though rich in data and visualizations, often feel too abstract or technical for the public. Although personalization can enhance communication, most tools still lack the narrative and visualization tailoring needed to connect with individual experiences. We present CLAImate, an AI-enabled prototype that personalizes conversation narratives and localizes visualizations based on users' climate knowledge and geographic location. We evaluated CLAImate through internal verification of factual correctness, a formative study with experts, and a pilot with UK residents. CLAImate achieved 66% SNLI accuracy and 70% FACTSCORE. Visualization experts appreciated its clarity and personalization, and seven out of ten UK participants reported better understanding and local relevance of climate risks with CLAImate. We also discuss design challenges in personalization, accuracy, and scalability, and outline future directions for integrating visualizations in personalized conversational interfaces.

We introduce our ongoing work toward an insight-based evaluation methodology aimed at understanding practitioners' mental models when exploring medical data. It is based on ParcoursVis, a Progressive Visual Analytics system designed to visualize event sequences derived from Electronic Health Records at scale (millions of patients, billions of events), developed in collaboration with the Emergency Departments of 16 Parisian hospitals and with the French Social Security. Building on prior usability validation, our current evaluation focuses on the insights generated by expert users and aims to better understand the exploration strategies they employ when engaging with exploration visualization tools. We describe our system and outline our evaluation protocol, analysis strategy, and preliminary findings. Building on this approach and our pilot results, we contribute a design protocol for conducting insight-based studies under real-world constraints, including the availability of health practitioners whom we were fortunate to interview. Our findings highlight a loop, where the use of the system helps refine data variables identification and the system itself. We aim to shed light on generated insights, to highlight the utility of exploratory tools in health data analysis contexts.

Scalability is an issue in visualization and visual analytics. The dataset sizes we can handle are lagging behind by several levels of magnitude compared to domains such as database, artificial intelligence, and simulation. The standard method for addressing scalability consists of adding more resources: more processors, more GPUs, more memory, and faster networks. Unfortunately, this method will not solve the visualization scalability problem alone. It does not solve the crucial issues of maintaining latency under critical limits to allow exploration and taming human attention during long-lasting computations. Progressive Data Analysis (PDA) emerged about a decade ago to address this scalability problem, showing promising but challenging solutions. I will show a few examples of applications. However, PDA is still lagging behind, mainly because of domain boundaries coming from academic research.

We introduce a conceptual model for scalability designed for visualization research. With this model, we systematically analyze over 120 visualization publications from 1990 to 2020 to characterize the different notions of scalability in these works. While many article have addressed scalability issues, our survey identifies a lack of consistency in the use of the term in the visualization research community. We address this issue by introducing a consistent terminology meant to help visualization researchers better characterize the scalability aspects in their research. It also helps in providing multiple methods for supporting the claim that a work is "scalable." Our model is centered around an effort function with inputs and outputs. The inputs are the problem size and resources, whereas the outputs are the actual efforts, for instance, in terms of computational run time or visual clutter. We select representative examples to illustrate different approaches and facets of what scalability can mean in visualization literature. Finally, targeting the diverse crowd of visualization researchers without a scalability tradition, we provide a set of recommendations for how scalability can be presented in a clear and consistent way to improve fair comparison between visualization techniques and systems and foster reproducibility.

Currently, growing data sources and long-running algorithms impede user attention and interaction with visual analytics applications. Progressive visualization (PV) and visual analytics (PVA) alleviate this problem by allowing immediate feedback and interaction with large datasets and complex computations, avoiding waiting for complete results by using partial results improving with time. Yet, creating a progressive visualization requires more effort than a regular visualization but also opens up new possibilities, such as steering the computations towards more relevant parts of the data, thus saving computational resources. However, there is currently no comprehensive overview of the design space for progressive visualization systems. We surveyed the related work of PV and derived a new taxonomy for progressive visualizations by systematically categorizing all PV publications that included visualizations with progressive features. Progressive visualizations can be categorized by well-known visualization taxonomies, but we also found that progressive visualizations can be distinguished by the way they manage their data processing, data domain, and visual update. Furthermore, we identified key properties such as uncertainty, steering, visual stability, and real-time processing that are significantly different with progressive applications. We also collected evaluation methodologies reported by the publications and conclude with statistical findings, research gaps, and open challenges.

Inria Raweb dataset is built with data extracted from Inria Raweb platform.

We present ComBiNet, a visualization, query, and comparison system for exploring bipartite multivariate dynamic social networks. Historians and sociologists study social networks constructed from textual sources mentioning events related to people, such as marriage acts, birth certificates and contracts. We model this type of data using bipartite multivariate dynamic networks to maintain a representation faithful to the original sources while not too complex. Relying on this data model, ComBiNet allows exploring networks using both visual and textual queries using the Cypher language, the two being synchronized to specify queries using the most suitable modality; simple queries are easy to express visually and can be refined textually when they become complex. These queries are used for applying topological and attribute-based selection on the network. Query results are visualized in the context of the whole network and over a geographical map for geolocalized entities. We also present the design of our interaction techniques for querying social networks to visually compare the selections in terms of topology, measures and attribute distributions. We validate the query and comparison systems by showing how they have been used to answer historical questions and by explaining how they have been improved through a usability study conducted with historians.

Cartolabe LISN 2023.11.17.1 dataset contains the data extracted from the HAL open archive which stores scholarly documents from all academic fields. We've filtered this dataset to include all articles published by authors from LISN (Interdisciplinary Laboratory of Digital Sciences), LRI (Laboratoire de Recherche en Informatique) and LIMSI (Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur) between 2000 and 2023.

This article reports on an in-depth study that investigates barriers to network exploration with visualizations. Network visualization tools are becoming increasingly popular, but little is known about how analysts plan and engage in the visual exploration of network data—which exploration strategies they employ, and how they prepare their data, define questions, and decide on visual mappings. Our study involved a series of workshops, interaction logging, and observations from a 6-week network exploration course. Our findings shed light on the stages that define analysts' approaches to network visualization and barriers experienced by some analysts during their network visualization processes. These barriers mainly appear before using a specific tool and include defining exploration goals, identifying relevant network structures and abstractions, or creating appropriate visual mappings for their network data. Our findings inform future work in visualization education and analyst-centered network visualization tool design.

Designing big data visualization applications is challenging due to their complex yet isolated development. One of the most common issues is an increase in latency that can be experienced while interacting with the system. There exists a variety of optimization techniques to handle this issue in specific scenarios, but we lack models for integrating them in a holistic way, hindering the integration of complementary functionality and hampering consistent evaluation across systems. In response, we present a framework for modeling the big data visualization pipeline which builds a bridge between the Visualization, Human-Computer Interaction, and Database communities by integrating their individual contributions within a single, easily interpretable pipeline. With this framework, visualization applications can become aware of the full end-to-end context, making it easier to determine which subset of optimizations best suits the current context.