2023Activity reportProject-TeamAVIZ

2.1 Objectives

Aviz (Analysis and VIsualiZation) is a multidisciplinary project that seeks to improve data exploration methods, techniques, and tools based on Interactive Visualization. Visualization, in general, refers to the graphical representation of data or concepts to aid access, distribution or explanations of data. Card et al. give a general definition for visualization as

“the use of computer-supported, interactive, visual representations of data to amplify cognition.” 42

Taking this definition, visualization is a means of creating visual aids that lead to insight in the underlying data sets. It is not about producing nice pictures but about making data understandable and explorable so that visualizations help viewers gain knowledge about the data. It is about aiding the process of forming a mental model for the acquired data and so helping the viewer to understand underlying concepts, patterns, and connections within the data 67. In partiular, visualization has the goal to improve humans' sensemaking of complex data by taking advantage of the capabilities of their vision system: visual information can be processed in parallel and with a high bandwidth into the human cognitive centers 75. Ware defines five advantages of visualization 75:

1. Comprehension: Supports the comprehension of large amounts of data.
2. Pattern Perception: Previously unnoticed properties of data may emerge.
3. Problem Analysis: Problems within the data may become immediately apparent.
4. Adaptability: facilitates understanding of large- and small-scale features of data.
5. Interpretation: Hypothesis formulation is facilitated.

Three main areas of visualization have evolved in the computer science community: Scientific Visualization, Information Visualization, and Visual Analytics. Scientific visualization is primarily concerned with displaying real or simulated scientific data. Basic visualization techniques for this area include surface rendering, volume rendering, and animation. Typical examples include processing of satellite photographs, fluid flows, or medical data. Datasets in information visualization typically come from large information spaces or information systems and are both structured or unstructured. Examples include network data, multi-dimensional tables of abstract measurements, or unstructured data such as text. Visual analytics, finally, is concerned with augmenting human-led data exploration with automatic techniques such as machine learning. The Aviz team has expertise in all three areas of visualization.

The conceptual Data Analysis Pipeline related with the four themes of AVIZ.

As shown in Figure 1, visualization deals with the data analysis pipeline and research in visualization has been addressing all the stages with less emphasis on the two initial ones and the last one. In its initial incarnation, Aviz has been focusing on interaction in combination with visualization, physical presentation, and perception. We now want to expand our research to wider questions, both in the human-side and in the system side. For the human side, we want to better understand human perception and cognition to improve the visualization techniques, so as to better convey information to the human brain. We also want to better understand the human biases to overcome them when possible, or provide methods to avoid them otherwise.

On the system side, we want to expand the scope of visualization that is currently limited to relatively small datasets and relatively simple analytical methods. To achieve scalability in visualization, we will focus on a paradigm shift: progressive data analysis. Long-running computations currently hamper the exploration and visualization process because human's attention is limited by latency constraints. We want to design exploratory systems that provide continuous feedback and allow interactions at any time during computation. The new progressive data analysis paradigm offers these capabilities, but to be usable, it requires the whole analytical pipeline to be re-implemented, and visualization and interaction techniques to be adapted.

2.2 Research Themes

Aviz's research on Visualization and Visual Analytics is organized around four research themes, described in more detail in the next section. Instead of addressing point problems, each research theme will address several stages of the visualization pipeline in a holistic manner, as summarized in fig:diagram.

1. Progressive Data Analysis and Scalability will address visualization scalability problems. Existing data analysis systems (such as Tableau  68, R  69, or Python with its data analysis ecosystem  56) are not scalable for exploratory analysis because their latency is not controllable. This theme will lay out the foundations of progressive data analysis systems, which generate estimates of the results and updates the analyst continuously at a bounded pace. It will focus on all the stages of the data analysis pipeline: data management mechanisms, data analysis modules, as well as visualizations, perception, understanding, and decision making.

2. Physicality in Input and Output will seek to better understand the benefits of physicality for information. Although beyond-desktop environments for visualization are generating more and more interest, theories and empirical data are lacking. This theme will consolidate the nascent areas of data physicalization, situated visualization, and immersive visualization.

3. Perception, Cognition, and Decision Making will study how we perceive and understand visualizations in order to develop generalized guidelines for optimizing effectiveness. It will generalize results obtained with simple charts to more complex visualizations of large datasets, establish theories on the use of abstraction in visualization, and contribute new empirical knowledge on decision making with visualizations.

4. Methodologies for Visualization Research will develop new methods to ground the study of the above scientific questions, and to benefit visualization more generally. This theme will develop evidence-based strategies for communicating quantitative empirical findings, and will promote methodological discussions and open research practices within the field.

3.1 Progressive Data Analysis and Scalability

Permanent involved: Jean-Daniel Fekete

While data analysis has made tremendous progress in scalability in the last decade, this progress has only benefited “confirmatory” analysis or model-based computation; progress in data exploration has lagged behind. Existing data analysis systems do not support data exploration at scale because, for large amounts of data or for expensive computations, their latency is not controllable: computations can take minutes, hours, even days and months. Cognitive psychologists have shown that humans' cognitive capabilities degrade when latency increases  62, 66. Miller  62 points out that the feedback of a system should remain below 10 seconds to maintain the user's attention. Therefore, to try to limit the latency, analysts currently resort to complex, inefficient, and unsatisfactory strategies, such as sampling with its issues related to representativity.

To address the scalability problem under controlled latency, instead of performing each computation in one long step that forces the analyst to wait for an unbounded amount of time, a progressive system generates estimates of the results and updates the analyst continuously at a bounded pace. The process continues until the computation is complete, or it stops early if the analyst considers that the quality of the estimates is sufficient to make a decision. During the process, a progressive system allows users to monitor the computation with visualizations and steer it with interactions.

While the topic of progressive data analysis has started to emerge in the late 90's, it has remained marginal practically because it touches three fields of computer science that are traditionally separate: data management, data analysis, and visualization. Research on progressive data analysis remains fragmented; the solutions proposed are partial and the different solutions cannot always be combined. We have organized a Dagstuhl seminar on Progressive Data Analysis and Visualization  46, 70 that acknowledged the harm of this topical separation and devised a research agenda. Aviz will participate in this agenda with specific assets.

Aviz is actively working on designing and implementing the ProgressiVis language that is natively progressive  47. The language relies on a Python interpreter but its execution semantics is different in the sense that all the operations that would take time to execute are performed progressively. The ProgressiVis system touches all the stages of the conceptual data analysis pipeline of fig:diagram; it integrates data management mechanisms, data analysis modules, as well as visualizations, perception, understanding, and decision making. Aviz will strengthen its work on the implementation of a natively progressive data science system. Such a system will lead to the following research topics:

1. Progressive language kernel and data management mechanisms
2. Progressive algorithms and computation strategies
3. Progressive visualizations
4. Management of uncertainties, computed from the algorithms and conveyed to the analysts.

Visualizations:

Like algorithms need to be adapted or transformed, visualization techniques and user interfaces also need to be adapted to be used in a progressive setting.

For the visualization pipeline, including the rendering phase, there has been previous work related to progressive layout computation and progressive rendering. For example, in network visualization, there is an old tradition of using iterative force-directed solvers and showing their results progressively when computing complex graph layouts. We generalize this approach to all the visualizations. Progressive rendering is popular in the real-time computer graphics and gaming domains, but new in visualization. We are designing visualization techniques based on GPU and accelerated algorithms to better render aggregated visualizations, such as heatmaps and sampled visualizations.

We have also started to explore the problem of requirements for progressive user interfaces  38, leading the community of researchers to understand that progressive systems need to provide effective assessments of the quality of their results to allow analysts to make early decisions. This justifies the next research topic on the management of uncertainties. We have also started to propose new data models for managing aggregated visualizations  55. We now continue our research to adapt existing visualization techniques and interactive environments to deal with progressive results and parameter steering.

(1) simple bar char. (2) with confidence intervals. (3) with recent history, and (4) with recent history and confidence intervals.

3.2 Physicality in Input and Output

Permanents involved: Petra Isenberg, Tobias Isenberg, Jean-Daniel Fekete

 

 

 

This is a compilation of images showing work the Aviz team did in the area of "physicality in input and output". The first image shows the Zooids project about miniature robots that can produce data physicalizations. The second shows a visualization for a smartwatch and fitness band about sleep data. The third shows situated visualizations in a video game. The third shows a visualization in augmented reality and a selection technique for choosing certain points.

This is a compilation of images showing work the Aviz team did in the area of "physicality in input and output". The first image shows the Zooids project about miniature robots that can produce data physicalizations. The second shows a visualization for a smartwatch and fitness band about sleep data. The third shows situated visualizations in a video game. The third shows a visualization in augmented reality and a selection technique for choosing certain points.

During the last five years, we expanded our work on beyond-desktop environments for visualization. Our team has made contributions in the areas of data physicalization, visualization for wearable devices, situated and embedded visualization, and visualization in augmented reality.

Data Physicalization: Data physicalization is a rich and vast research area that studies the use of physical artifacts to convey data. It overlaps with a number of research areas including information/scientific visualization, visual analytics, tangible user interfaces, shape-changing interfaces, fabrication, as well as graphic design, architecture, and art. Physical data visualizations tap into our lifelong experience of perceiving and manipulating the physical world, either alone or with other people. Among the earliest man-made artifacts are physical representations of semantic concepts that provide physical metaphors that allow us to reason, remember, and communicate. With the advent of computers, we have substituted physical representations with pixels on a computer screen. The resurgence of physicalization as a research area, following our early definitions 53, asks what we have lost in this transformation. Certainly, a computer-based visualization allows us to zoom an image, transform variables in real time, and to zoom through virtual computer-based world. However, these representations can sever the relationship to the natural world, depriving us of the touch, feel, and emotion that comes from interacting with real objects. We studied several aspects of data physicalization including technical challenges of constructing physicalizations, potential benefits, and how historical examples could transfer to a modern world.

Visualization for Wearable and Mobile Devices: In the area of wearable and mobile-devices we engaged in device-driven research where we considered how small device form factors may influence how we need to design visualizations and how we can use them. Portable and wearable personal devices such as fitness tracking armbands, hand-held GPS trackers, smart watches, or mobile phones are very small displays that are capable of producing data themselves (through sensors), downloading it from other sources (through Wifi or Bluetooth), and displaying it immediately 4. Often the data is shown in the form of visualizations that have to be adapted to the small display size. We consider very small visualizations that are often used on such devices under the term “micro visualization” and have been working towards a better understanding of the complexities involved in designing and using micro visualizations but also studied the influence of the unique context of use of mobile devices on visualization use and design.

Situated and Embedded Data Representations: We study how embedding data visualizations in the context of the data sources can empower people to make effective use of their data in a variety of application contexts. Our goal in this work is to go beyond the traditional platforms of data analytics by using situated data visualizations on various types of non-traditional displays. In a situated data visualization, the data is directly visualized near the physical space, object, or person it originates from 76. For example, a person may attach small e-ink displays embedded with sensors at various locations of their house or their workplace, to better understand their use of space, of equipment, or of energy resources. Or a person who wishes to exercise more may use an augmented reality device to visualize their past running performance in-place. New situated data visualizations like these can surface information in the environment—allowing viewers to interpret data in-context and take action in response to it 8.

Visualization using Augmented Reality Devices: Many datasets are 3D-spatial in nature and researchers and practitioners could benefit from seeing them in true 3D space. This is where immersive technologies shine, and the recent advances in VR and AR headset technologies have made such displays accessible to the general public—the lack of large dedicated VR installations such as a CAVE is not preventing the use of immersive rendering anymore. Nonetheless, the investigation of 3D datasets also frequently requires researchers to use tools such as scripted analysis and statistical evaluation, and such direction of investigation will continue to be a cornerstone of scientific work. In our investigations we are thus interested in looking at, in particular, hybrid setups that allow researchers to use the best of both worlds: traditional workstations combined with an AR overlay for stereoscopic rendering of 3D data  74.

3.3 Research Axis 3: Perception, Cognition and Decision-Making

Permanents involved: Petra Isenberg, Tobias Isenberg, Jean-Daniel Fekete

As we collect increasingly large amounts of data in fields such as climate science, finance, and medicine, the need to understand and communicate that data becomes more important. Data visualizations are often used to give an overview of information, however it can be challenging to predict whether these visualizations will be effective before spending resources to develop them. Consequently, researchers make use of experimental methods from visual perception and cognition to study how we perceive and understand visualizations in order to develop generalized guidelines for optimizing effectiveness. On this research axis we have three focus areas:

Perception of Visualizations in Novel Contexts. Novel technology and usage contexts have several characteristics for data visualizations that warrant a (re-)evaluation of how well people can perceive visualizations. Characteristics we were particularly interested in include:

• Physical factors: Many now common displays have characteristics that warrant (re-)evaluation of what we know about visualizations to be displayed on them. Small form factors of displays such as smartwatches and fitness bands are an obvious characteristic. Some screens deviate from the standard rectangular form we are familiar with and use a circular geometry, which is another interesting design constraint for visualization.
• Data display mobility: Data display mobility captures the movement of the display(s) containing visual representations of data. Fixed, movable, carryable, wearable, and independently moving displays can be differentiated along this dimension. We conducted some of our research on carryable devices, including mobile devices such as smartphones and tablets but also wearable displays such as smartwatches and head-mounted displays.
• Context-of-use: Many novel displays are used in contexts that are much unlike traditional office settings. Visualizations here may be subjected to different lighting conditions and viewers may only afford very quick glances at the displays themselves. For example, when in a car the driver can only afford very quick glances at the GPS before returning to the primary task of arriving safely.

Within this focus area we worked on some of these aspects; such as the need for quick glances, the display size of visualizations, the movement of viewers or the data, and the understanding of 3D augmented reality spaces. We relied mostly on mixed-methods user studies where a quantitative analysis methodology was coupled with interviews or questionnaires.

Illustrative Visualization. This focus area takes inspiration from illustrators' decades to centuries of experience on perception and cognition to better portray scientific subject matter. Another input arises from the field of non-photorealistic rendering which has developed numerous techniques of stylizing images and other input data. Traditionally, illustrative visualization has thus been applied primarily to data with a concrete spatial mapping in 2D and, more frequently, in 3D space.

Another main direction of research in this context is what the role of abstraction is in illustrative visualization  73, 72 as well as visualization in general, and specifically how we can provide dedicated means to control the abstraction being applied to visual representations of data. This means that we need to go beyond seeing abstraction only as a side-product of stylization as it has traditionally been viewed in many approaches in non-photorealistic rendering as well as illustrative visualization to date, and investigate how we can interactively adjust it to provide practitioners with a means to find the best visual representations for a given task. For example, we have investigated this question in the context of structural biology 79 and DNA nanostructures  59, 60, 61, but also want to expand this work to other application domains in the future.

Decision-Making with Visualizations. Human decision-making and cognitive biases are important research topics in the fields of psychology, economics and marketing. Visualization systems are increasingly used to support decision-making: large companies switch to visualization solutions to improve their decisions in a range of areas, where large sums of money or human lives are at stake. More and more, the ultimate goal of visualization is not to understand patterns in the data and get insights as was traditionally assumed, but to make good decisions. In order to fully understand how information visualizations can support decision-making, it is important to go beyond traditional evaluations based on data understanding, and study how visualizations interact with human judgment, human heuristics, and cognitive biases.

We pursue this important stream of research by investigating decision-making in the presence of uncertainty and incomplete information (in connection with the topics discussed in sec:progressive and the use of visualizations to support social choice and group decisions in the presence of conflicts of interest. How cognitive biases interact with visual perception is also an important and difficult question that has remained largely unexplored.

3.4 Research Axis 4: Methodology for Visualization Research

Permanents involved: Petra Isenberg, Tobias Isenberg, Jean-Daniel Fekete

An important aspect of any scientific research is to establish and follow rigorous and effective methodologies for acquiring new knowledge. In the field of Visualization in particular, scientific discourse on the validity, use, and establishment of methodologies is important as the field is highly interdisciplinary, with diverse influences and opinions. It is important to establish, for example, what level of rigor the field should require of its methods, how to choose among established methods and methodologies, and how to best communicate the results of our empirical research. We focus our efforts on three main topics related to visualization research methodologies.

Promoting and Following Open Research Practices

For issues with open research practices to be addressed, educational materials and guidelines need to be written, so researchers have clarity about how to make their research more credible. Aviz members are working with the organizing bodies of the visualization research community to establish incentives for making research artifacts and potentially establish minimal requirements for openness in published articles. Meanwhile, it is important to continue measuring and cataloging openness in the field to monitor progress. The goal is to improve the credibility and applicability of the field’s research.

Shaping the Scientific Visualization Community Aviz researchers are heavily involved in the organization structure of IEEE visualization conferences, the most prestigious conference in our field, by proposing workshops, tutorials, serving on various organizing committees, steering committees, editorial boards. We, in particular, aid the process by providing data collection and analysis services through the vispubdata.org dataset that we are collecting, cleaning, and making available to the community. The dataset has already been used in research (e. g., 58) but also to shape the scientific community by proposing program committee members, new processes, and was used by the Visualization Restructuring Committee (ReVISe). We are also involved in the EuroVis community and participate at multiple levels to its organization and management.

4.1 Natural Sciences

As part of a CORDI PhD project, we collaborated with researchers at CERN on interactive data visualization using augmented reality, with the goal to better understand this new visualization environment and to support the physicists in analysing their 3D particle collision data. As part of another CORDI PhD project, we collaborated with researchers at the German Center for Climate Computation (DKRZ) and the Helmholtz-Zentrum Hereon (formerly Helmholtz-Zentrum Geesthacht, HZG), to better understand collaborative data exploration and interaction in immersive analytics contexts. Finally, as part of the Inria IPL “Naviscope,” we collaborate with researchers at INRAE (as well as other Inria teams) on interactive visualization tools for the exploration of plant embryo development. We also work with experts from the natural sciences in general to create illustrative visualizations of scientific data, such as a continuous zooming technique from the nucleus of a cell all the way down to the atom configuration of the DNA, for example for the application in education.

4.2 Social Sciences

We collaborate with social science researchers from EHESS Paris on the visualization of dynamic networks; they use our systems (GeneaQuilts 39, Vistorian 65, PAOHVis 71, PK-Clustering 63) and teach them to students and researchers. Our tools are used daily by ethnographers and historians to study the evolution of social relations over time. In the social sciences, many datasets are gathered by individual researchers to answer a specific question, and automated analytical methods cannot be applied to these small datasets. Furthermore, the studies are often focused on specific persons or organizations and not always on the modeling or prediction of the behavior of large populations. The tools we design to visualize complex multivariate dynamic networks are unique and suited to typical research questions shared by a large number of researchers. This line of research is supported by the DataIA “HistorIA” project, and by the “IVAN” European project. We currently collaborate with PayAnalytics, an Icelandic company to visualize data to help companies close their gender pay gaps.

4.3 Medicine

We collaborate with the Health-Data-Hub on the analysis and visualization of CNAM Data “parcours de santé” to help referent doctors and epidemiologists make sense of French health data. In particular, we are working on a subset of the CNAM Data focused on urinary problems, and we have received very positive feedback from doctors who can see what happens to the patients treated in France vs. what they thought happened through the literature. This project is starting but is already getting a lot of traction from our partners in medicine, epidemiology, and economy of health.

We are also collaborating with the “Assistance Publique - Hôpitaux de Paris” AP-HP with a newly funded project called URGE, aimed at improving the emergency services for the parisian hospitals. See the press announcement.

5.1 Impact of research results

Aviz' work on illustrative visualization has the potential to be integrated into future teaching materials for students in schools, visitors in museums, or similar.

Aviz' work on visualization of large documents corpora with Cartolabe is used to present the results of the French “Grand débat”, as well as other citizen expressions.

Aviz' work on the gender pay gap aims at improving decision making for closing the adjusted pay gap.

Open science: Aviz regularly shares full research material on the repository of the Center for Open Science to facilitate scrutiny, reuse, and replication:

• Visualization in Motion: A Research Agenda and Two Evaluations  78 — osf.io/km3s2/, osf.io/9c4bz/, osf.io/t748d/, osf.io/t748d/,gitlab.inria.fr/lyao/visinmotion, GRSI Stamp
• Reflections on Visualization in Motion for Fitness Trackers  52 — hal.inria.fr/hal-03775633
• Situated Visualization in Motion for Swimming  77 — hal.inria.fr/hal-03700406
• Situated Visualization in Motion for Video Game  40 — hal.inria.fr/hal-03694019
• Situated Visualization in Motion in Video Game for Different Types of Data  41 — hal.inria.fr/hal-03700418
• Preferences and Effectiveness of Sleep Data Visualizations for Smartwatches and Fitness Bands  50 – osf.io/yz8ar/
• Context Specific Visualizations on Smartwatches  51 – osf.io/vhn43/
• Studying Early Decision Making with Progressive Bar Charts 5 — osf.io/tn2ph/
• Scalability in Visualization  64 — osf.io/xrvu7/
• Six methods for transforming layered hypergraphs to apply layered graph layout algorithms  45 — osf.io/grvwu/
• Pyramid-based Scatterplots Sampling for Progressive and Streaming Data Visualization  44 — ProgressivePyramidBasedSampling and www.yunhaiwang.net/Vis2021/progressive-sampling/

Graphics Replicability Stamp: In 2023, Aviz members received recogniztion for the replicability of their work from the Graphics Replicability Stamp Initiative for several of their journal publications:

• Designing for Visualization in Motion: Embedding Visualizations in Swimming Videos 24
• Eliciting Multimodal and Collaborative Interactions for Data Exploration on Large Vertical Displays 19
• Design Characterization for Black-and-White Textures in Visualization 13
• Visualizing and Comparing Machine Learning Predictions to Improve Human-AI Teaming on the Example of Cell Lineage 15
• V-Mail: 3D-Enabled Correspondence about Spatial Data on (Almost) All Your Devices 20
• BeauVis: A Validated Scale for Measuring the Aesthetic Pleasure of Visual Representations 12

6.1 Events

• Florent Cabric , post-doctoral fellow at Aviz co-organized the French “Visualization Day” (“Journées Visu”) at Univ. Paris-Saclay in June.
• Petra Isenberg , Florent Cabric , and Mickaël Sereno co-organized (“PreVis”) at Univ. Paris-Saclay in June.

6.2 Awards

• Tobias Isenberg: 14 Years SciVis Test-of-Time Award at IEEE VIS 2023
• Petra Isenberg: Induction into the IEEE Visualization Academy

7.1 New software

7.2 New platforms

7.3 Open data

8.1 Visualization of Socio-Demographic Data

Participants: Florent Cabric [LISN, Université Paris-Saclay, CNRS, Inria, Saclay, France, correspondent], Margrét Vilborg Bjarnadóttir[Univ. of Maryland], Anne-Flore Cabouat [LISN, Université Paris-Saclay, CNRS, Inria, Saclay, France], Petra Isenberg [LISN, Université Paris-Saclay, CNRS, Inria, Saclay, France].

Visualization of the typical color used to represent gender in scientific publications for two different time periods.

We started investigations into the visualization of sociodemographic data. Sociodemographic data is a common part of datasets related to people, including institutional censuses, health data systems, and human-resources files. This data is sensitive, and its collection, sharing, and analysis require careful consideration. For instance, the European Union, through the General Data Protection Regulation (GDPR), protects the collection and processing of any personal data, including sexual orientation, ethnicity, and religion. Data visualization of sociodemographic data can reinforce stereotypes, marginalize groups, and lead to biased decision-making. It is, therefore, critical that these visualizations are created based on good, equitable design principles. In a first opinion paper 2, we discuss and provide a set of open research questions around the visualization of sociodemographic data. Our work contributes to an ongoing refection on representing data about people and highlights some important future research directions for the Visualization community. A version of this paper and its fgures are available online at osf.io/a2u9c.

8.2 Design Characterization for Black-and-White Textures in Visualization

Participants: Tingying He [Inria, Paris-Saclay, correspondent], Yuanyang Zhong [Inria, Paris-Saclay], Petra Isenberg [Inria, Paris-Saclay], Tobias Isenberg [Inria, Paris-Saclay].

We investigate the use of 2D black-and-white textures for the visualization of categorical data and contribute a summary of texture attributes, and the results of three experiments that elicited design strategies as well as aesthetic and effectiveness measures 3. Black-and-white textures are useful, for instance, as a visual channel for categorical data on low-color displays, in 2D/3D print, to achieve the aesthetic of historic visualizations, or to retain the color hue channel for other visual mappings. We specifically study how to use what we call geometric and iconic textures. Geometric textures use patterns of repeated abstract geometric shapes, while iconic textures use repeated icons that may stand for data categories. We parameterized both types of textures and developed a tool for designers to create textures on simple charts by adjusting texture parameters. 30 visualization experts used our tool and designed 66 textured bar charts, pie charts, and maps. We then had 150 participants rate these designs for aesthetics. Finally, with the top-rated geometric and iconic textures, our perceptual assessment experiment with 150 participants revealed that textured charts perform about equally well as non-textured charts, and that there are some differences depending on the type of chart.

8.3 Visualizing and Comparing Machine Learning Predictions to Improve Human-AI Teaming on the Example of Cell Lineage

Participants: Jiayi Hong [Inria, Paris-Saclay, correspondent], Ross Maciejewski [Arizona State University], Alain Trubuil [InraE Paris-Saclay], Tobias Isenberg [Inria, Paris-Saclay].

We visualize the predictions of multiple machine learning models to help biologists as they interactively make decisions about cell lineage—the development of a (plant) embryo from a single ovum cell. Based on a confocal microscopy dataset, traditionally biologists manually constructed the cell lineage, starting from this observation and reasoning backward in time to establish their inheritance. To speed up this tedious process, we make use of machine learning (ML) models trained on a database of manually established cell lineages to assist the biologist in cell assignment. Most biologists, however, are not familiar with ML, nor is it clear to them which model best predicts the embryo’s development. We thus have developed a visualization system that is designed to support biologists in exploring and comparing ML models, checking the model predictions, detecting possible ML model mistakes, and deciding on the most likely embryo development. To evaluate our proposed system, we deployed our interface with six biologists in an observational study. Our results show that the visual representations of machine learning are easily understandable, and our tool, LineageD+, could potentially increase biologists’ working efficiency and enhance the understanding of embryos.

9.1 Bilateral contracts with industry

Participants: Jean-Daniel Fekete, Katerina Batziakoudi.

CIFRE PhD fellowship of Katerina Batziakoudi with the Company Berger-Levrault (2023–2026).

Participants: Petra Isenberg, Florent Cabric.

Contract with PayAnalytics for joint work on the Inria exploratory action called “Equity Analytics” (2022-2024)

10.1 International research visitors

10.2 National initiatives

• Program:
  ANR PRC (ANR-19-CE33-0012)
• Project acronym:
  EMBER
• Project title:
  Situated Visualizations for Personal Analytics
• Duration:
  2020 – 2023. Total funding: 712 k€
• Coordinator:
  Pierre Dragicevic
• Other partners:
  Inria Bordeaux, Sorbonne Université
• Participants:

  Participants: Petra Isenberg, Lijie Yao.

• Abstract:
  The Ember project studies how situated data visualization systems can help people use their personal data (e.g., fitness and physiological data, energy consumption, banking transactions, online social activity,…) for their own benefit. Although personal data is generated in many areas of daily life, it remains underused by individuals. Rarely is personal data subjected to an in-depth analysis and used to inform daily decisions. This research aims to empower individuals to improve their lives by helping them become advanced consumers of their own data. This research builds on the area of personal visual analytics, which focuses on giving the general public effective and accessible tools to get insights from their own data. Personal visual analytics is a nascent area of research, but has so far focused on scenarios where the data visualization is far removed from the source of the data it refers to. The goal of this project is to address the limitations of traditional platforms of personal data analytics by exploring the potential of situated data visualizations. In a situated data visualization, the data is directly visualized near the physical space, object, or person it refers to. Situated data visualizations have many potential benefits: they can surface information in the physical environment and allow viewers to interpret data in-context; they can be tailored to highlight spatial connections between data and the physical environment, making it easier to make decisions and act on the physical world in response to the insights gained; and they can embed data into physical environments so that it remains visible over time, making it easier to monitor changes, observe patterns over time and collaborate with other people. Website: ember.inria.fr/.

Other projects:

• Health Data Hub “ParcoursVis” post-doctoral funding. Duration: 24 months. Total funding: 220k€. Partners: Inria Saclay, Health Data Hub. See www.aviz.fr/Research/ParcoursVis.

• Participants: Jean-Daniel Fekete, Mickaël Sereno.

  Participants:

• AP-HP / Inria joint project URGE. Duration 48 months. Total funding: 520k€. Partners: Inria, Hôpital Saint Antoine, Sorbonne Université. Coordinators: Xavier Leroy (Inria), Youri Yordanov (AP-HP, Inserm, Sorbonne Université).

• Participants: Jean-Daniel Fekete, Mickaël Sereno.

  Participants:

• Action Exploratoire: EquityAnalytics: Visual Analytics for Pay Equity, Action Exploratoire. Post-Doctoral Funding. Duration 24 months. Total funding: 106k€. Coordinator: Petra Isenberg . Partners: PayAnalytics.

• Participants: Petra Isenberg, Florent Cabric.

  Participants:

11.1 Promoting scientific activities

• Jean-Daniel Fekete has introduced Graphs and their visualization to high-school students during the Science Ouverte Seminar, Bobigny, Jun. 26, 2023

11.2 Teaching - Supervision - Juries

12.1 Major publications

• 1 inproceedingsL.Leilani Battle, P.Philipp Eichmann, M.Marco Angelini, T.Tiziana Catarci, G.Giuseppe Santucci, Y.Yukun Zheng, C.Carsten Binnig, J.-D.Jean-Daniel Fekete and D.Dominik Moritz. Database Benchmarking for Supporting Real-Time Interactive Querying of Large Data.SIGMOD '20 - International Conference on Management of DataProceedings of the 2020 ACM SIGMOD International Conference on Management of DataPortland, OR, United StatesACMJune 2020, 1571-1587HALDOI
• 2 inproceedingsF.Florent Cabric, M.Margrét Vilborg Bjarnadóttir, A.-F.Anne-Flore Cabouat and P.Petra Isenberg. Open Questions about the Visualization of Sociodemographic Data.IEEE VIS: Visualization & Visual Analytics Conference - Workshop on Visualization for Social Good (VIS4Good)VIS4Good 2023 - IEEE Workshop on Visualization for Social GoodMelbourne, AustraliaOctober 2023HALDOIback to text
• 3 articleT.Tingying He, P.Petra Isenberg, R.Raimund Dachselt and T.Tobias Isenberg. BeauVis: A Validated Scale for Measuring the Aesthetic Pleasure of Visual Representations.IEEE Transactions on Visualization and Computer Graphics291January 2023, 363–373HALDOIback to text
• 4 bookB.Bongshin Lee, R.Raimund Dachselt, P.Petra Isenberg and E. K.Eun Kyoung Choe. Mobile Data Visualization.0Chapman and Hall/CRCDecember 2021, 346HALDOIback to text
• 5 articleA.Ameya Patil, G.Gaëlle Richer, C.Christopher Jermaine, D.Dominik Moritz and J.-D.Jean-Daniel Fekete. Studying Early Decision Making with Progressive Bar Charts.IEEE Transactions on Visualization and Computer Graphics291January 2023, 407-417HALDOIback to textback to text
• 6 articleG.Gaëlle Richer, A.Alexis Pister, M.Moataz Abdelaal, J.-D.Jean-Daniel Fekete, M.Michael Sedlmair and D.Daniel Weiskopf. Scalability in Visualization.IEEE Transactions on Visualization and Computer GraphicsDecember 2022HALDOI
• 7 articleW.Wesley Willett, B. A.Bon Adriel Aseniero, S.Sheelagh Carpendale, P.Pierre Dragicevic, Y.Yvonne Jansen, L.Lora Oehlberg and P.Petra Isenberg. Perception! Immersion! Empowerment! Superpowers as Inspiration for Visualization.IEEE Transactions on Visualization and Computer Graphics281January 2022, 22-32HALDOI
• 8 articleL.Lijie Yao, A.Anastasia Bezerianos, R.Romain Vuillemot and P.Petra Isenberg. Visualization in Motion: A Research Agenda and Two Evaluations.IEEE Transactions on Visualization and Computer Graphics2810October 2022, 3546-3562HALDOIback to text

12.2 Publications of the year

12.3 Cited publications

• 37 inproceedingsS.Sameer Agarwal, H.Henry Milner, A.Ariel Kleiner, A.Ameet Talwalkar, M.Michael Jordan, S.Samuel Madden, B.Barzan Mozafari and I.Ion Stoica. Knowing when You'Re Wrong: Building Fast and Reliable Approximate Query Processing Systems.Proceedings of the 2014 ACM SIGMOD International Conference on Management of DataSIGMOD '14New York, NY, USASnowbird, Utah, USAACM2014, 481--492URL: http://doi.acm.org/10.1145/2588555.2593667DOIback to text
• 38 articleS. K.Sriram Karthik Badam, N.Niklas Elmqvist and J.-D.Jean-Daniel Fekete. Steering the Craft: UI Elements and Visualizations for Supporting Progressive Visual Analytics.Computer Graphics Forum363June 2017, 491--502HALDOIback to text
• 39 articleA.Anastasia Bezerianos, P.Pierre Dragicevic, J.-D.Jean-Daniel Fekete, J.Juhee Bae and B.Ben Watson. GeneaQuilts: A System for Exploring Large Genealogies.IEEE Transactions on Visualization and Computer Graphics166October 2010, 1073-1081HALDOIback to text
• 40 miscF.Federica Bucchieri, L.Lijie Yao and P.Petra Isenberg. Situated Visualization in Motion for Video Games.PosterJune 2022HALDOIback to text
• 41 inproceedingsF.Federica Bucchieri, L.Lijie Yao and P.Petra Isenberg. Visualization in Motion in Video Games for Different Types of Data.Journée Visu 2022Bordeaux, FranceJune 2022HALback to text
• 42 bookS. K.Stuart K. CardJ. D.Jock D. MackinlayB.Ben ShneidermanReadings in information visualization: using vision to think.San Francisco, CA, USAMorgan Kaufmann Publishers Inc.1999back to text
• 43 unpublishedJ.Jian Chen, P.Petra Isenberg, R. S.Robert S Laramee, T.Tobias Isenberg, M.Michael Sedlmair, T.Torsten Mōller and H.-W.Han-Wei Shen. Not As Easy As You Think-Experiences and Lessons Learnt from Creating a Visualization Image Typology.October 2022, working paper or preprintHALDOIback to text
• 44 articleX.Xin Chen, J.Jian Zhang, C.-W.Chi-Wing Fu, J.-D.Jean-Daniel Fekete and Y.Yunhai Wang. Pyramid-based Scatterplots Sampling for Progressive and Streaming Data Visualization.IEEE Transactions on Visualization and Computer Graphics281January 2022, 593-603HALDOIback to text
• 45 articleS.Sara Di Bartolomeo, A.Alexis Pister, P.Paolo Buono, C.Catherine Plaisant, C.Cody Dunne and J.-D.Jean-Daniel Fekete. Six Methods for Transforming Layered Hypergraphs to Apply Layered Graph Layout Algorithms.Computer Graphics Forum413June 2022, 1467-8659HALDOIback to text
• 46 articleJ.-D.Jean-Daniel Fekete, D.Danyel Fisher, A.Arnab Nandi and M.Michael Sedlmair. Progressive Data Analysis and Visualization (Dagstuhl Seminar 18411).Dagstuhl Reports8102018, 1--32URL: https://doi.org/10.4231/DagRep.8.1.1DOIback to text
• 47 unpublishedJ.-D.Jean-Daniel Fekete and R.Romain Primet. Progressive Analytics: A Computation Paradigm for Exploratory Data Analysis.July 2016, https://arxiv.org/abs/1607.05162 - working paper or preprintHALback to text
• 48 bookM.Minos Garofalakis, J.Johannes Gehrke and R.Rajeev Rastogi. Data stream management: processing high-speed data streams.Springer2016back to text
• 49 bookB. K.Bhaskar Kumar Ghosh and P. K.Pranab Kumar Sen. Handbook of sequential analysis.CRC Press1991back to text
• 50 inproceedingsA.Alaul Islam, R.Ranjini Aravind, T.Tanja Blascheck, A.Anastasia Bezerianos and P.Petra Isenberg. Preferences and Effectiveness of Sleep Data Visualizations for Smartwatches and Fitness Bands.CHI 2022 - Conference on Human Factors in Computing SystemsNew Orleans, LA, United StatesApril 2022HALDOIback to text
• 51 miscA.Alaul Islam, T.Tanja Blascheck and P.Petra Isenberg. Context Specific Visualizations on Smartwatches.PosterJune 2022HALDOIback to text
• 52 inproceedingsA.Alaul Islam, L.Lijie Yao, A.Anastasia Bezerianos, T.Tanja Blascheck, T.Tingying He, B.Bongshin Lee, R.Romain Vuillemot and P.Petra Isenberg. Reflections on Visualization in Motion for Fitness Trackers.MobileHCI 2022 Workshop on New Trends in HCI and SportsVancouver, CanadaSeptember 2022HALback to text
• 53 phdthesisY.Yvonne Jansen. Physical and Tangible Information Visualization.Université Paris Sud-Paris XI2014back to text
• 54 articleJ.Jaemin Jo, J.Jinwook Seo and J.-D.Jean-Daniel Fekete. PANENE: A Progressive Algorithm for Indexing and Querying Approximate k-Nearest Neighbors.IEEE Transactions on Visualization and Computer Graphics262February 2020, 1347-1360HALDOIback to text
• 55 articleJ.Jaemin Jo, F.Frédéric Vernier, P.Pierre Dragicevic and J.-D.Jean-Daniel Fekete. A Declarative Rendering Model for Multiclass Density Maps.IEEE Transactions on Visualization and Computer Graphics251January 2019, 470-480HALDOIback to text
• 56 miscE.Eric Jones, T.Travis Oliphant, P.Pearu Peterson and others. SciPy: Open source scientific tools for Python.[Online; accessed <today>]2001, URL: http://www.scipy.org/back to text
• 57 bookD.Daniel Kahneman. Thinking, Fast and Slow.Farrar, Straus and Giroux2011back to text
• 58 articleS.S. Latif and F.F. Beck. VIS Author Profiles: Interactive Descriptions of Publication Records Combining Text and Visualization.IEEE Transactions on Visualization and Computer Graphics251January 2019, 152-161DOIback to text
• 59 articleH.Haichao Miao, E.Elisa De Llano, T.Tobias Isenberg, M. E.M. Eduard Gröller, I.Ivan Barišic and I.Ivan Viola. DimSUM: Dimension and Scale Unifying Maps for Visual Abstraction of DNA Origami Structures.Computer Graphics Forum373June 2018, 403-413HALDOIback to text
• 60 articleM.M Miao, E.Elisa De Llano, J.Johannes Sorger, Y.Yasaman Ahmadi, T.Tadija Kekic, T.Tobias Isenberg, M. E.M. Eduard Gröller, I.Ivan Barišic and I.Ivan Viola. Multiscale Visualization and Scale-Adaptive Modification of DNA Nanostructures.IEEE Transactions on Visualization and Computer Graphics241January 2018, 1014--1024HALDOIback to text
• 61 articleH.Haichao Miao, T.Tobias Klein, D.David Kouřil, P.Peter Mindek, K.Karsten Schatz, M. E.M. Eduard Gröller, B.Barbora Kozl\'iková, T.Tobias Isenberg and I.Ivan Viola. Multiscale Molecular Visualization.Journal of Molecular Biology4316March 2019, 1049--1070HALDOIback to text
• 62 inproceedingsR. B.Robert B. Miller. Response Time in Man-computer Conversational Transactions.Proc. of the Fall Joint Computer Conference, Part ISan Francisco, CaliforniaACM1968, 267--277URL: http://doi.acm.org/10.1145/1476589.1476628DOIback to textback to text
• 63 articleA.Alexis Pister, P.Paolo Buono, J.-D.Jean-Daniel Fekete, C.Catherine Plaisant and P.Paola Valdivia. Integrating Prior Knowledge in Mixed Initiative Social Network Clustering.IEEE Transactions on Visualization and Computer Graphics272February 2021, 1775 - 1785HALDOIback to text
• 64 articleG.Gaëlle Richer, A.Alexis Pister, M.Moataz Abdelaal, J.-D.Jean-Daniel Fekete, M.Michael Sedlmair and D.Daniel Weiskopf. Scalability in Visualization.IEEE Transactions on Visualization and Computer GraphicsDecember 2022HALDOIback to text
• 65 inproceedingsV.Vanessa Serrano Molinero, B.Benjamin Bach, C.Catherine Plaisant, N.Nicole Dufournaud and J.-D.Jean-Daniel Fekete. Understanding the Use of The Vistorian: Complementing Logs with Context Mini-Questionnaires.Visualization for the Digital HumanitiesPhoenix, United StatesOctober 2017HALback to text
• 66 articleB.Ben Shneiderman. Response Time and Display Rate in Human Performance with Computers.ACM Comput. Surv.163September 1984, 265--285URL: http://doi.acm.org/10.1145/2514.2517DOIback to text
• 67 bookR.Robert Spence. Information Visualization.Boston, MA, USAAddison-Wesley Publishing Company2000back to text
• 68 miscTableau Software.Last accessed Feb 20182018, URL: http://www.tableau.comback to text
• 69 manualR. C.R Core Team. R: A Language and Environment for Statistical Computing.R Foundation for Statistical ComputingVienna, Austria2019, URL: https://www.R-project.orgback to text
• 70 articleC.Cagatay Turkay, N.Nicola Pezzotti, C.Carsten Binnig, H.Hendrik Strobelt, B.Barbara Hammer, D. A.Daniel A. Keim, J.-D.Jean-Daniel Fekete, T.Themis Palpanas, Y.Yunhai Wang and F.Florin Rusu. Progressive Data Science: Potential and Challenges.CoRRabs/1812.080322018, URL: http://arxiv.org/abs/1812.08032back to text
• 71 articleP. R.Paola R Valdivia, P.Paolo Buono, C.Catherine Plaisant, N.Nicole Dufournaud and J.-D.Jean-Daniel Fekete. Analyzing Dynamic Hypergraphs with Parallel Aggregated Ordered Hypergraph Visualization.IEEE Transactions on Visualization and Computer Graphics271January 2021, 1-13HALDOIback to textback to text
• 72 incollectionI.Ivan Viola, M.Min Chen and T.Tobias Isenberg. Visual Abstraction.Foundations of Data VisualizationSpringer International PublishingAugust 2020, 15--37HALDOIback to text
• 73 articleI.Ivan Viola and T.Tobias Isenberg. Pondering the Concept of Abstraction in (Illustrative) Visualization.IEEE Transactions on Visualization and Computer Graphics249September 2018, 2573--2588HALDOIback to text
• 74 inproceedingsX.Xiyao Wang, L.Lonni Besançon, F.Florimond Guéniat, M.Mickael Sereno, M.Mehdi Ammi and T.Tobias Isenberg. A Vision of Bringing Immersive Visualization to Scientific Workflows.Proceedings of the Conference on Human Factors in Computing Systems (CHI) - Workshop on Interaction Design & Prototyping for Immersive AnalyticsGlasgow, United KingdomMay 2019HALback to text
• 75 bookC.Colin Ware. Information Visualization -- Perception for Design.Morgan Kaufmann Series in Interactive TechnologiesSan Francisco, CA, USAMorgan Kaufmann Publishers2000back to textback to text
• 76 articleW.Wesley Willett, Y.Yvonne Jansen and P.Pierre Dragicevic. Embedded Data Representations.IEEE Transactions on Visualization and Computer Graphics2017, 461 - 470HALDOIback to text
• 77 inproceedingsL.Lijie Yao, A.Anastasia Bezerianos, R.Romain Vuillemot and P.Petra Isenberg. Situated Visualization in Motion for Swimming.Journée Visu 2022Bordeaux, FranceJune 2022HALback to text
• 78 articleL.Lijie Yao, A.Anastasia Bezerianos, R.Romain Vuillemot and P.Petra Isenberg. Visualization in Motion: A Research Agenda and Two Evaluations.IEEE Transactions on Visualization and Computer Graphics2810October 2022, 3546-3562HALDOIback to text
• 79 articleM.Matthew van der Zwan, W.Wouter Lueks, H.Henk Bekker and T.Tobias Isenberg. Illustrative Molecular Visualization with Continuous Abstraction.Computer Graphics Forum303May 2011, 683--690URL: https://tobias.isenberg.cc/VideosAndDemos/Zwan2011IMVDOIback to text