Artificial Intelligence (AI) holds significant potential for supporting clinical decision-making, particularly in high-pressure environments, such as Cardiothoracic Intensive Care Units (CT-ICU). Care teams in these settings face challenges such as alarm fatigue, rapid staff turnover, time-sensitive decisions, and an overwhelming amount of data. AI-driven Clinical Decision-Support Systems (AI-CDSS) can support care teams in overcoming some of these challenges by providing solutions like detecting and reporting risk scores for adverse events that may lead to increased fatalities or re-admissions, enabling timely intervention. One key challenge with risk scores is missing data, which can create considerable uncertainty in risk score values. AI-CDSSs rarely convey the risk score uncertainty, which is important in the effectiveness and reliability of clinical decision-making. In this paper, we describe the interface design process for HEART, an AI-powered system developed collaboratively with clinical and AI experts over a 16-month iterative design process for a hospital’s CT-ICU. The HEART interactive interface integrates understandable visualizations of risk scores and their uncertainty within both a holistic view of all patients in the unit and detailed patient-specific views. We reflect on the user-centered design process, report findings from an expert walkthrough study, and discuss lessons learned as well as broader implications. This work contributes valuable insights into uncertainty visualization design for AI-derived risk scores in a critical care application. Beyond these specific insights, our work illustrates the kind of comprehensive, human-centered design process necessary for responsible AI adoption in critical environments.

Abstract Evidence continues to accrue around the difficulties people have understanding new and complex visualizations, which in turn provides continued incentive to explore additional methods of supporting visualization viewers. Through leveraging viewers' spontaneous visualization sensemaking activities, we introduce a Serendipitous Explanation Approach (SEA). A significant part of SEA's contribution is making active use of a viewer's spontaneous, exploration interactions to offer in situ explanations. SEA adds semantic and structural explanations, making use of visual transformations, and additional representations of visualization elements themselves to communicate meaning through highlights, reconfigurations, and in‐situ annotations. We designed and studied VisTips as an instantiation of SEA. Our study demonstrates the prevalence of spontaneous interaction and appreciation of this approach. Among the explanation types, visual transformations stood out as especially impactful. Our findings also offer practical insights for future use of the ideas in SEA, opening possibilities for seamlessly integrating explanatory visuals that support viewers' natural visualization sensemaking.

Tens of thousands of people have represented data by creating data-encoding textile pieces like blankets, scarves, and more. A prototypical example is the temperature blanket, which represents the weather through rows or blocks of different colors mapped to temperature ranges. While researchers have used fiber arts mediums to create exploratory projects, data visualization and physicalization research has largely not engaged with examples from this enormous and diverse community. We explore the space of data textiles, or fiber arts that encode information, by surveying creators (i.e., data fiber artists) on their projects and processes. We create a corpus of 159 examples of data textiles and present a schema characterizing the data encoding methods used in these projects. We also gather insights into creators' data workflows as well as their motives and discoveries through making with their data. Creators of data textiles use distinct processes to map their data, building fabric from component structures and substructures while using material properties like color and texture. From diverse data-tracking procedures, creators use and relate to data in varied ways. Working on these pieces also contributes to the creators' personal growth and data understanding. Our findings point to new opportunities for visualization, including opportunities to support fiber artists with tools formatted to their needs and opportunities to incorporate concepts from data textiles into other types of visualization (e.g., using texture, structural layouts, colorways).

Enhancing visualization comprehension has been approached through multiple research streams, with significant contributions from areas such as visualization literacy, sensemaking, insight, and graphical perception. While each of these streams independently enhances specific aspects of viewers understanding, they have tended to remain distinct. In this paper, first, we describe the distinct streams of visualization comprehension research and elucidate the interplay between them. Then, we provide design considerations that emerge from these streams. We also outline research opportunities within and at the intersections of these streams. By exploring these connections, we aim to provide a broader perspective on how different approaches to visualization comprehension can inform one another and enhance user engagement with visual data.

Abstract In systems genetics and other multi‐omics research, exploring high‐dimensional relationships among molecular and physiological variables across individuals poses significant challenges. We present the Gridded Trees interface, a novel interactive visualization tool designed to facilitate the exploration of conditional inference trees, which are hierarchical models of relationships in these complex datasets. Traditional static tools struggle to reveal patterns in tree‐structured data, but the Gridded Trees interface provides interactive, coordinated views, allowing users to navigate between overview and detail, filter data dynamically, and compare molecular‐physiological relationships across subgroups. By combining filtering techniques, strip plots, Sankey diagrams, and small multiples, the Gridded Trees interface enhances exploratory data analysis and supports hypothesis generation. In our systems genetics research use case, this tool has revealed significant associations among microbial populations and addiction‐related behavioral traits in genetically diverse mice. The Gridded Trees interface suggests broad potential for visualizing hierarchical and multipartite data across domains. A preprint of this paper as well as Supplemental Materials are available on OSF at https://osf.io/9emn5/ .

Today's AI systems for medical decision support often succeed on benchmark datasets in research papers but fail in real-world deployment. This work focuses on the decision making of sepsis, an acute life-threatening systematic infection that requires an early diagnosis with high uncertainty from the clinician. Our aim is to explore the design requirements for AI systems that can support clinical experts in making better decisions for the early diagnosis of sepsis. The study begins with a formative study investigating why clinical experts abandon an existing AI-powered Sepsis predictive module in their electrical health record (EHR) system. We argue that a human-centered AI system needs to support human experts in the intermediate stages of a medical decision-making process (e.g., generating hypotheses or gathering data), instead of focusing only on the final decision. Therefore, we build SepsisLab based on a state-of-the-art AI algorithm and extend it to predict the future projection of sepsis development, visualize the prediction uncertainty, and propose actionable suggestions (i.e., which additional laboratory tests can be collected) to reduce such uncertainty. Through heuristic evaluation with six clinicians using our prototype system, we demonstrate that SepsisLab enables a promising human-AI collaboration paradigm for the future of AI-assisted sepsis diagnosis and other high-stakes medical decision making.

Increasingly, visualization practitioners are working with, using, and studying private and sensitive data. There can be many stakeholders interested in the resulting analyses-but widespread sharing of the data can cause harm to individuals, companies, and organizations. Practitioners are increasingly turning to differential privacy to enable public data sharing with a guaranteed amount of privacy. Differential privacy algorithms do this by aggregating data statistics with noise, and this now-private data can be released visually with differentially private scatterplots. While the private visual output is affected by the algorithm choice, privacy level, bin number, data distribution, and user task, there is little guidance on how to choose and balance the effect of these parameters. To address this gap, we had experts examine 1,200 differentially private scatterplots created with a variety of parameter choices and tested their ability to see aggregate patterns in the private output (i.e. the visual utility of the chart). We synthesized these results to provide easy-to-use guidance for visualization practitioners releasing private data through scatterplots. Our findings also provide a ground truth for visual utility, which we use to benchmark automated utility metrics from various fields. We demonstrate how multi-scale structural similarity (MS-SSIM), the metric most strongly correlated with our study's utility results, can be used to optimize parameter selection.

While the visualization community is increasingly aware that people often find visualizations difficult to understand, there is less information about what we need to do to create comprehensible visualizations. To help visualization creators and designers improve their visualizations, we need to better understand what kind of support people are looking for in their sensemaking process. Empirical studies are needed to tease apart the details of what makes the process of understanding difficult for visualization viewers. We conducted a qualitative study with 14 participants, observing them as they described how they were trying to make sense of 20 information visualizations. We identified the challenges participants faced throughout their sensemaking process and the strategies they employed to help themselves in overcoming the challenges. Our findings show how details and nuances within visualizations can impact comprehensibility and offer research suggestions to help us move toward more understandable visualizations.

Dashboards are no longer mere static displays of metrics; through functionality such as interaction and storytelling, they have evolved to support analytic and communicative goals like monitoring and reporting. Existing dashboard design guidelines, however, are often unable to account for this expanded scope as they largely focus on best practices for visual design. In contrast, we frame dashboard design as facilitating an analytical conversation: a cooperative, interactive experience where a user may interact with, reason about, or freely query the underlying data. By drawing on established principles of conversational flow and communication, we define the concept of a cooperative dashboard as one that enables a fruitful and productive analytical conversation, and derive a set of 39 dashboard design heuristics to support effective analytical conversations. To assess the utility of this framing, we asked 52 computer science and engineering graduate students to apply our heuristics to critique and design dashboards as part of an ungraded, opt-in homework assignment. Feedback from participants demonstrates that our heuristics surface new reasons dashboards may fail, and encourage a more fluid, supportive, and responsive style of dashboard design. Our approach suggests several compelling directions for future work, including dashboard authoring tools that better anticipate conversational turn-taking, repair, and refinement and extending cooperative principles to other analytical workflows.

Chatbots have garnered interest as conversational interfaces for a variety of tasks. While general design guidelines exist for chatbot interfaces, little work explores analytical chatbots that support conversing with data. We explore Gricean Maxims to help inform the basic design of effective conversational interaction. We also draw inspiration from natural language interfaces for data exploration to support ambiguity and intent handling. We ran Wizard of Oz studies with 30 participants to evaluate user expectations for text and voice chatbot design variants. Results identified preferences for intent interpretation and revealed variations in user expectations based on the interface affordances. We subsequently conducted an exploratory analysis of three analytical chatbot systems (text + chart, voice + chart, voice-only) that implement these preferred design variants. Empirical evidence from a second 30-participant study informs implications specific to data-driven conversation such as interpreting intent, data orientation, and establishing trust through appropriate system responses.

Working with data in table form is usually considered a preparatory and tedious step in the sensemaking pipeline; a way of getting the data ready for more sophisticated visualization and analytical tools. But for many people, spreadsheets - the quintessential table tool - remain a critical part of their information ecosystem, allowing them to interact with their data in ways that are hidden or abstracted in more complex tools. This is particularly true for data workers [61], people who work with data as part of their job but do not identify as professional analysts or data scientists. We report on a qualitative study of how these workers interact with and reason about their data. Our findings show that data tables serve a broader purpose beyond data cleanup at the initial stage of a linear analytic flow: users want to see and "get their hands on" the underlying data throughout the analytics process, reshaping and augmenting it to support sensemaking. They reorganize, mark up, layer on levels of detail, and spawn alternatives within the context of the base data. These direct interactions and human-readable table representations form a rich and cognitively important part of building understanding of what the data mean and what they can do with it. We argue that interactive tables are an important visualization idiom in their own right; that the direct data interaction they afford offers a fertile design space for visual analytics; and that sense making can be enriched by more flexible human-data interaction than is currently supported in visual analytics tools.

When an organization chooses one course of action over alternatives, this task typically falls on a decision maker with relevant knowledge, experience, and understanding of context. Decision makers rely on data analysis, which is either delegated to analysts, or done on their own. Often the decision maker combines data, likely uncertain or incomplete, with non-formalized knowledge within a multi-objective problem space, weighing the recommendations of analysts within broader contexts and goals. As most past research in visual analytics has focused on understanding the needs and challenges of data analysts, less is known about the tasks and challenges of organizational decision makers, and how visualization support tools might help. Here we characterize the decision maker as a domain expert, review relevant literature in management theories, and report the results of an empirical survey and interviews with people who make organizational decisions. We identify challenges and opportunities for novel visualization tools, including trade-off overviews, scenario-based analysis, interrogation tools, flexible data input and collaboration support. Our findings stress the need to expand visualization design beyond data analysis into tools for information management.

Visualization recommendation (VisRec) systems provide users with suggestions for potentially interesting and useful next steps during exploratory data analysis. These recommendations are typically organized into categories based on their analytical actions, i.e., operations employed to transition from the current exploration state to a recommended visualization. However, despite the emergence of a plethora of VisRec systems in recent work, the utility of the categories employed by these systems in analytical workflows has not been systematically investigated. Our article explores the efficacy of recommendation categories by formalizing a taxonomy of common categories and developing a system, Frontier, that implements these categories. Using Frontier, we evaluate workflow strategies adopted by users and how categories influence those strategies. Participants found recommendations that add attributes to enhance the current visualization and recommendations that filter to sub-populations to be comparatively most useful during data exploration. Our findings pave the way for next-generation VisRec systems that are adaptive and personalized via carefully chosen, effective recommendation categories.

Data science is a rapidly growing discipline and organizations increasingly depend on data science work. Yet the ambiguity around data science, what it is, and who data scientists are can make it difficult for visualization researchers to identify impactful research trajectories. We have conducted a retrospective analysis of data science work and workers as described within the data visualization, human computer interaction, and data science literature. From this analysis we synthesis a comprehensive model that describes data science work and breakdown to data scientists into nine distinct roles. We summarise and reflect on the role that visualization has throughout data science work and the varied needs of data scientists themselves for tooling support. Our findings are intended to arm visualization researchers with a more concrete framing of data science with the hope that it will help them surface innovative opportunities for impacting data science work.Data availability:https://osf.io/z2xpd/?view_only=87fa24be486a473884adb9ffbe8db4ec

Handling ambiguity and underspecification of users' utterances is challenging, particularly for natural language interfaces that help with visual analytical tasks. Constraints in the underlying analytical platform and the users' expectations of high precision and recall require thoughtful inferencing to help generate useful responses. In this paper, we introduce a system to resolve partial utterances based on syntactic and semantic constraints of the underlying analytical expressions. We extend inferencing based on best practices in information visualization to generate useful visualization responses. We employ heuristics to help constrain the solution space of possible inferences, and apply ranking logic to the interpretations based on relevancy. We evaluate the quality of inferred interpretations based on relevancy and analytical usefulness.

Dashboards are one of the most common use cases for data visualization, and their design and contexts of use are considerably different from exploratory visualization tools. In this paper, we look at the broad scope of how dashboards are used in practice through an analysis of dashboard examples and documentation about their use. We systematically review the literature surrounding dashboard use, construct a design space for dashboards, and identify major dashboard types. We characterize dashboards by their design goals, levels of interaction, and the practices around them. Our framework and literature review suggest a number of fruitful research directions to better support dashboard design, implementation, and use.

Multimodal interaction offers many potential benefits for data visualization. It can help people stay in the flow of their visual analysis and presentation, with the strengths of one interaction modality offsetting the weaknesses of others. Furthermore, multimodal interaction offers strong promise for leveraging data visualization on diverse display hardware including mobile, AR/VR, and large displays. However, prior research on visualization and interaction techniques has mostly explored a single input modality such as mouse, touch, pen, or more recently, natural language. The unique challenges and opportunities of synergistic multimodal interaction for data visualization have yet to be investigated. This workshop will bring together researchers with expertise in visualization, interaction design, and natural user interfaces. We aim to build a community of researchers focusing on multimodal interaction for data visualization, explore opportunities and challenges in our research, and establish an agenda for multimodal interaction research specifically for data visualization.

Abstract We introduce an approach for explicitly revealing changes between versions of a visualization workbook to support version comparison tasks. Visualization authors may need to understand version changes for a variety of reasons, analogous to document editing. An author who has been away for a while may need to catch up on the changes made by their co‐author, or a person responsible for formatting compliance may need to check formatting changes that occurred since the last time they reviewed the work. We introduce ChangeCatcher, a prototype tool to help people find and understand changes in a visualization workbook, specifically, a Tableau workbook. Our design is based on interviews we conducted with experts to investigate user needs and practices around version comparison. ChangeCatcher provides an overview of changes across six categories, and employs a multi‐level details‐on‐demand approach to progressively reveal details. Our qualitative study showed that ChangeCatcher's methods for explicitly revealing and categorizing version changes were helpful in version comparison tasks.

Interactive visual data analysis is most productive when users can focus on answering the questions they have about their data, rather than focusing on how to operate the interface to the analysis tool. One viable approach to engaging users in interactive conversations with their data is a natural language interface to visualizations. These interfaces have the potential to be both more expressive and more accessible than other interaction paradigms. We explore how principles from language pragmatics can be applied to the flow of visual analytical conversations, using natural language as an input modality. We evaluate the effectiveness of pragmatics support in our system Evizeon, and present design considerations for conversation interfaces to visual analytics tools.

Visualization researchers and practitioners engaged in generating or evaluating designs are faced with the difficult problem of transforming the questions asked and actions taken by target users from domain-specific language and context into more abstract forms. Existing abstract task classifications aim to provide support for this endeavour by providing a carefully delineated suite of actions. Our experience is that this bottom-up approach is part of the challenge: low-level actions are difficult to interpret without a higher-level context of analysis goals and the analysis process. To bridge this gap, we propose a framework based on analysis reports derived from open-coding 20 design study papers published at IEEE InfoVis 2009-2015, to build on the previous work of abstractions that collectively encompass a broad variety of domains. The framework is organized in two axes illustrated by nine analysis goals. It helps situate the analysis goals by placing each goal under axes of specificity (Explore, Describe, Explain, Confirm) and number of data populations (Single, Multiple). The single-population types are Discover Observation, Describe Observation, Identify Main Cause, and Collect Evidence. The multiple-population types are Compare Entities, Explain Differences, and Evaluate Hypothesis. Each analysis goal is scoped by an input and an output and is characterized by analysis steps reported in the design study papers. We provide examples of how we and others have used the framework in a top-down approach to abstracting domain problems: visualization designers or researchers first identify the analysis goals of each unit of analysis in an analysis stream, and then encode the individual steps using existing task classifications with the context of the goal, the level of specificity, and the number of populations involved in the analysis.

Histopathology laboratories devote considerable time and effort to sorting tissue sample slides. We observed slide sorting in a typical urban hospital to understand the existing workflow and explore how it might be supported by an interactive computer support system. We observed 8.5 hours of slide sorting activity through a video camera mounted above a laboratory workbench. Through detailed video analysis, we characterised the process, examined which activities took the most time, and explored design considerations. We found that a very large proportion (23.5%) of the slide sorting time involved managing paper documents. We suggest that an interactive computer support system could automatically detect which slides are sorted into which folders and digitally list additional slides to include with these sets; this would support the workflow of technicians, while eliminating paper management and manual barcode reading operations, leading to time savings of approximately 30%. Additional recommendations for the design of such a support system include focusing on case management (e.g. how many slides belong to each case, whether a complete case will fit within the current folder, and which slides associated with a case are still missing), supporting recovery from disruptions, and enabling a flexible rather than a highly scripted workflow.

Data analysis involves constantly formulating and testing new hypotheses and questions about data. When dealing with a new dataset, especially one with many dimensions, it can be cumbersome for the analyst to clearly remember which aspects of the data have been investigated (i.e., visually examined for patterns, trends, outliers etc.) and which combinations have not. Yet this information is critical to help the analyst formulate new questions that they have not already answered. We observe that for tabular data, questions are typically comprised of varying combinations of data dimensions (e.g., what are the trends of Sales and Profit for different Regions?). We propose representing analysis history from the angle of dimension coverage (i.e., which data dimensions have been investigated and in which combinations). We use scented widgets [30] to incorporate dimension coverage of the analysts' past work into interaction widgets of a visualization tool. We demonstrate how this approach can assist analysts with the question formation process. Our approach extends the concept of scented widgets to reveal aspects of one's own analysis history, and offers a different perspective on one's past work than typical visualization history tools. Results of our empirical study showed that participants with access to embedded dimension coverage information relied on this information when formulating questions, asked more questions about the data, generated more top-level findings, and showed greater breadth of their analysis without sacrificing depth.

Mobile devices have begun to raise questions around the potential for overuse when in the presence of family or friends. As such, we conducted a diary and interview study to understand how people use mobile devices in the presence of others at home, and how this shapes their behavior and household dynamics. Results show that family members become frustrated when others do non-urgent activities on their phones in the presence of others. Yet people often guess at what others are doing because of the personal nature of mobile devices. In some cases, people developed strategies to provide a greater sense of activity awareness to combat the problem. Mobile phone usage was sometimes perceived as beneficial by providing a mechanism for needed disengagement from family members. These findings suggest several opportunities for redesigning mobile device software to mitigate emergent frustrations, and open up new opportunities for nurturing social interactions among family members.

Feedback tools help people to monitor information about themselves to improve their health, sustainability practices, or personal well-being. Yet reasoning about personal data (e.g., pedometer counts, blood pressure readings, or home electricity consumption) to gain a deep understanding of your current practices and how to change can be challenging with the data alone. We integrate quantitative feedback data within a personal digital calendar; this approach aims to make the feedback data readily accessible and more comprehensible. We report on an eight-week field study of an on-calendar visualization tool. Results showed that a personal calendar can provide rich context for people to reason about their feedback data. The on-calendar visualization enabled people to quickly identify and reason about regular patterns and anomalies. Based on our results, we also derived a model of the behavior feedback process that extends existing technology adoption models. With that, we reflected on potential barriers for the ongoing use of feedback tools.

Natural language interfaces for visualizations have emerged as a promising new way of interacting with data and performing analytics. Many of these systems have fundamental limitations. Most return minimally interactive visualizations in response to queries and often require experts to perform modeling for a set of predicted user queries before the systems are effective. Eviza provides a natural language interface for an interactive query dialog with an existing visualization rather than starting from a blank sheet and asking closed-ended questions that return a single text answer or static visualization. The system employs a probabilistic grammar based approach with predefined rules that are dynamically updated based on the data from the visualization, as opposed to computationally intensive deep learning or knowledge based approaches.

Data surrounds each and every one of us in our daily lives, ranging from exercise logs, to archives of our interactions with others on social media, to online resources pertaining to our hobbies. There is enormous potential for us to use these data to understand ourselves better and make positive changes in our lives. Visualization (Vis) and visual analytics (VA) offer substantial opportunities to help individuals gain insights about themselves, their communities and their interests; however, designing tools to support data analysis in non-professional life brings a unique set of research and design challenges. We investigate the requirements and research directions required to take full advantage of Vis and VA in a personal context. We develop a taxonomy of design dimensions to provide a coherent vocabulary for discussing personal visualization and personal visual analytics. By identifying and exploring clusters in the design space, we discuss challenges and share perspectives on future research. This work brings together research that was previously scattered across disciplines. Our goal is to call research attention to this space and engage researchers to explore the enabling techniques and technology that will support people to better understand data relevant to their personal lives, interests, and needs.

When people work together to analyze a data set, they need to organize their findings, hypotheses, and evidence, share that information with their collaborators, and coordinate activities amongst team members. Sharing externalizations (recorded information such as notes) could increase awareness and assist with team communication and coordination. However, we currently know little about how to provide tool support for this sort of sharing. We explore how linked common work (LCW) can be employed within a `collaborative thinking space', to facilitate synchronous collaborative sensemaking activities in Visual Analytics (VA). Collaborative thinking spaces provide an environment for analysts to record, organize, share and connect externalizations. Our tool, CLIP, extends earlier thinking spaces by integrating LCW features that reveal relationships between collaborators' findings. We conducted a user study comparing CLIP to a baseline version without LCW. Results demonstrated that LCW significantly improved analytic outcomes at a collaborative intelligence task. Groups using CLIP were also able to more effectively coordinate their work, and held more discussion of their findings and hypotheses. LCW enabled them to maintain awareness of each other's activities and findings and link those findings to their own work, preventing disruptive oral awareness notifications.

Conflicting results are reported in the literature on whether dynamic visualizations are more effective than static visualizations for learning and mastering 3-D tasks, and only a few investigations have considered the influence of the spatial abilities of the learners. In a study with 117 participants, we compared the benefit of static vs. dynamic visualization training tools on learners with different spatial abilities performing a typical 3-D task (specifically, creating orthographic projections of a 3-D object). We measured the spatial abilities of the participants using the Mental Rotation Test (MRT) and classified participants into two groups (high and low abilities) to examine how the participants' abilities predicted change in performance after training with static versus dynamic training tools. Our results indicate that: 1) visualization training programs can help learners to improve 3-D task performance, 2) dynamic visualizations provide no advantages over static visualizations that show intermediate steps, 3) training programs are more beneficial for individuals with low spatial abilities than for individuals with high spatial abilities, and 4) training individuals with high spatial abilities using dynamic visualizations provides little benefit.

Maintaining an awareness of collaborators' actions is critical during collaborative work, including during collaborative visualization activities. Particularly when collaborators are located at a distance, it is important to know what everyone is working on in order to avoid duplication of effort, share relevant results in a timely manner and build upon each other's results. Can a person's brushing actions provide an indication of their queries and interests in a data set? Can these actions be revealed to a collaborator without substantially disrupting their own independent work? We designed a study to answer these questions in the context of distributed collaborative visualization of tabular data. Participants in our study worked independently to answer questions about a tabular data set, while simultaneously viewing brushing actions of a fictitious collaborator, shown directly within a shared workspace. We compared three methods of presenting the collaborator's actions: brushing & linking (i.e. highlighting exactly what the collaborator would see), selection (i.e. showing only a selected item), and persistent selection (i.e. showing only selected items but having them persist for some time). Our results demonstrated that persistent selection enabled some awareness of the collaborator's activities while causing minimal interference with independent work. Other techniques were less effective at providing awareness, and brushing & linking caused substantial interference. These findings suggest promise for the idea of exploiting natural brushing actions to provide awareness in collaborative work.

To verify cluster separation in high-dimensional data, analysts often reduce the data with a dimension reduction (DR) technique, and then visualize it with 2D Scatterplots, interactive 3D Scatterplots, or Scatterplot Matrices (SPLOMs). With the goal of providing guidance between these visual encoding choices, we conducted an empirical data study in which two human coders manually inspected a broad set of 816 scatterplots derived from 75 datasets, 4 DR techniques, and the 3 previously mentioned scatterplot techniques. Each coder scored all color-coded classes in each scatterplot in terms of their separability from other classes. We analyze the resulting quantitative data with a heatmap approach, and qualitatively discuss interesting scatterplot examples. Our findings reveal that 2D scatterplots are often 'good enough', that is, neither SPLOM nor interactive 3D adds notably more cluster separability with the chosen DR technique. If 2D is not good enough, the most promising approach is to use an alternative DR technique in 2D. Beyond that, SPLOM occasionally adds additional value, and interactive 3D rarely helps but often hurts in terms of poorer class separation and usability. We summarize these results as a workflow model and implications for design. Our results offer guidance to analysts during the DR exploration process.

Abstract We provide two contributions, a taxonomy of visual cluster separation factors in scatterplots, and an in‐depth qualitative evaluation of two recently proposed and validated separation measures. We initially intended to use these measures to provide guidance for the use of dimension reduction (DR) techniques and visual encoding (VE) choices, but found that they failed to produce reliable results. To understand why, we conducted a systematic qualitative data study covering a broad collection of 75 real and synthetic high‐dimensional datasets, four DR techniques, and three scatterplot‐based visual encodings. Two authors visually inspected over 800 plots to determine whether or not the measures created plausible results. We found that they failed in over half the cases overall, and in over two‐thirds of the cases involving real datasets. Using open and axial coding of failure reasons and separability characteristics, we generated a taxonomy of visual cluster separability factors. We iteratively refined its explanatory clarity and power by mapping the studied datasets and success and failure ranges of the measures onto the factor axes. Our taxonomy has four categories, ordered by their ability to influence successors: Scale, Point Distance, Shape, and Position. Each category is split into Within‐Cluster factors such as density, curvature, isotropy, and clumpiness, and Between‐Cluster factors that arise from the variance of these properties, culminating in the overarching factor of class separation. The resulting taxonomy can be used to guide the design and the evaluation of cluster separation measures.

Co-located collaborative tasks allow teams to leverage the skills of each individual member. While numerous guidelines exist to develop visualizations for individuals working on desktops, very little is known about how groups of individuals interpret and comprehend diverse types of visual constructs on larger displays. To study whether group size impacts the collective understanding of relationships in three-dimensional (3D) spatial structures when using different types of presentation, we carried out three experiments. We compared individual performance at structure understanding tasks to performance of groups containing two or four members. We consider two alternate visualization techniques for extracting 3D structure information: a 3D view with animated rotations and a combination of one static 3D plus three static two-dimensional (2D) projection views. In general our studies suggest that as group size increases, so does accuracy but with a cost in efficiency. Our results also suggest that beyond a threshold limit in group size, performance on certain tasks begins to degrade. Regardless of group size, participants performed better when the display was presented in the animation condition instead of the multiple static views, except when large groups needed to relate the visualization to a physical counterpart. We summarize our results in terms of Steiner's model for explaining the effects of group size and task characteristics on group performance.

It remains challenging for information visualization novices to rapidly construct visualizations during exploratory data analysis. We conducted an exploratory laboratory study in which information visualization novices explored fictitious sales data by communicating visualization specifications to a human mediator, who rapidly constructed the visualizations using commercial visualization software. We found that three activities were central to the iterative visualization construction process: data attribute selection, visual template selection, and visual mapping specification. The major barriers faced by the participants were translating questions into data attributes, designing visual mappings, and interpreting the visualizations. Partial specification was common, and the participants used simple heuristics and preferred visualizations they were already familiar with, such as bar, line and pie charts. We derived abstract models from our observations that describe barriers in the data exploration process and uncovered how information visualization novices think about visualization specifications. Our findings support the need for tools that suggest potential visualizations and support iterative refinement, that provide explanations and help with learning, and that are tightly integrated into tool support for the overall visual analytics process.

We introduce eSeeTrack, an eye-tracking visualization prototype that facilitates exploration and comparison of sequential gaze orderings in a static or a dynamic scene. It extends current eye-tracking data visualizations by extracting patterns of sequential gaze orderings, displaying these patterns in a way that does not depend on the number of fixations on a scene, and enabling users to compare patterns from two or more sets of eye-gaze data. Extracting such patterns was very difficult with previous visualization techniques. eSeeTrack combines a timeline and a tree-structured visual representation to embody three aspects of eye-tracking data that users are interested in: duration, frequency and orderings of fixations. We demonstrate the usefulness of eSeeTrack via two case studies on surgical simulation and retail store chain data. We found that eSeeTrack allows ordering of fixations to be rapidly queried, explored and compared. Furthermore, our tool provides an effective and efficient mechanism to determine pattern outliers. This approach can be effective for behavior analysis in a variety of domains that are described at the end of this paper.

PhotoScope visualizes the spatiotemporal coverage of photos in a photo collection. It extends the standard photo browsing paradigm in two main ways: visualizing spatial coverage of photos, and indexing photos by a combination of spatial coverage, time, and content specifications. This approach enables users to browse and search space- and time-indexed photos more effectively. We designed PhotoScope specifically to address challenges in the construction management industry, where large photo collections are amassed to document project progress. These ideas may also apply to any photo collection that is spatially constrained and must be searched using spatial, temporal, and content criteria. We describe the design choices made when developing PhotoScope and the results of user evaluation.

Abstract Project schedules are effectively represented by Gantt charts, but comparing multiple versions of a schedule is difficult. To compare versions with current methods, users must search and navigate through multiple large documents, making it difficult to identify differences. We present two novel visualization techniques to support the comparison of Gantt charts. First, we encode two Gantt charts in one view by overlapping them to show differences. Second, we designed an interactive visual technique, the ‘TbarView’, that allows users to compare multiple schedules within one single view. We evaluated the overlap and TbarView techniques via a user study. The study results showed that our design provided a quick overview of the variances among two or more schedules, and the techniques also improved efficiency by minimizing view switching. Our visual techniques for schedule comparison could be combined with other resource analysis tools to help project teams identify and resolve errors and problems in project schedules.

Abstract Visual fixation on one's tool(s) takes much attention away from one's primary task. Following the belief that the best tools ‘disappear’ and become invisible to the user, we present a study comparing visual fixations (eye gaze within locations on a graphical display) and performance for mouse, pen, and physical slider user interfaces. Participants conducted a controlled, yet representative, color matching task that required user interaction representative of many data exploration tasks such as parameter exploration of medical or fuel cell data. We demonstrate that users may spend up to 95% fewer visual fixations on physical sliders versus standard mouse and pen tools without any loss in performance for a generalized visual performance task.

Spatialization displays use a geographic metaphor to arrange non-spatial data. For example, spatializations are commonly applied to document collections so that document themes appear as geographic features such as hills. Many common spatialization interfaces use a 3-D landscape metaphor to present data. However, it is not clear whether 3-D spatializations afford improved speed and accuracy for user tasks compared to similar 2-D spatializations. We describe a user study comparing users' ability to remember dot displays, 2-D landscapes, and 3-D landscapes for two different data densities (500 vs. 1000 points). Participants' visual memory was statistically more accurate when viewing dot displays and 3-D landscapes compared to 2-D landscapes. Furthermore, accuracy remembering a spatialization was significantly better overall for denser spatializations. These results are of benefit to visualization designers who are contemplating the best ways to present data using spatialization techniques.

PartyVote is a democratic music jukebox designed to give all participants an equal influence on the music played at social gatherings or parties. PartyVote is designed to provide appropriate music in established social groups with minimal user interventions and no pre-existing user profiles. The visualization uses dimensionality reduction to show song similarity and overlays information about how votes affect the music played. Visualizing voting decisions allows users to link music selections with individuals, providing social awareness. Traditional group norms can subsequently be leveraged to maintain fair system use and empower users.

We use a haptically enhanced mixing board with a video projector as an interface to various data visualization tasks. We report results of an expert review with four participants, qualitatively evaluating task), parallel coordinates interface (multi-dimensional combinato-rial search), and ExoVis (3D spatial navigation). Our investigation sought to determine the strengths of this physical input given its ca-pability to facilitate bimanual interaction, constraint maintenance, tight coupling of input and output, and other features. Participants generally had little difficulty with the mappings of parameters to sliders. The graspable sliders apparently reduced the mental exer-tion needed to acquire control, allowing participants to attend more directly to understanding the visualization. Participants often des-ignated specific roles for each hand, but only rarely moved both hands simultaneously.

Spatializations represent non-spatial data using a spatial layout similar to a map. We present an experiment comparing different visual representations of spatialized data, to determine which representations are best for a non-trivial search and point estimation task. Primarily, we compare point-based displays to 2D and 3D information landscapes. We also compare a colour (hue) scale to a grey (lightness) scale. For the task we studied, point-based spatializations were far superior to landscapes, and 2D landscapes were superior to 3D landscapes. Little or no benefit was found for redundantly encoding data using colour or greyscale combined with landscape height. 3D landscapes with no colour scale (height-only) were particularly slow and inaccurate. A colour scale was found to be better than a greyscale for all display types, but a greyscale was helpful compared to height-only. These results suggest that point-based spatializations should be chosen over landscape representations, at least for tasks involving only point data itself rather than derived information about the data space.

Designing collaborative interfaces for tabletops remains difficult because we do not fully understand how groups coordinate their actions when working collaboratively over tables. We present two observational studies of pairs completing independent and shared tasks that investigate collaborative coupling, or the manner in which collaborators are involved and occupied with each other's work. Our results indicate that individuals frequently and fluidly engage and disengage with group activity through several distinct, recognizable states with unique characteristics. We describe these states and explore the consequences of these states for tabletop interface design.

We describe a series of experiments that compare 2D displays, 3D displays, and combined 2D/3D displays (orientation icon, ExoVis, and clip planes) for relative position estimation, orientation, and volume of interest tasks. Our results indicate that 3D displays can be very effective for approximate navigation and relative positioning when appropriate cues, such as shadows, are present. However, 3D displays are not effective for precise navigation and positioning except possibly in specific circumstances, for instance, when good viewing angles or measurement tools are available. For precise tasks in other situations, orientation icon and ExoVis displays were better than strict 2D or 3D displays (displays consisting exclusively of 2D or 3D views). The combined displays had as good or better performance, inspired higher confidence, and allowed natural, integrated navigation. Clip plane displays were not effective for 3D orientation because users could not easily view more than one 2D slice at a time and had to frequently change the visibility of individual slices. Major factors contributing to display preference and usability were task characteristics, orientation cues, occlusion, and spatial proximity of views that were used together.

Displays combining both 2D and 3D views have been shown to support higher performance on certain visualization tasks. However, it is not clear how best to arrange a combination of 2D and 3D views spatially in a display. In this study, we analyzed the eyegaze strategies of participants using two arrangements of 2D and 3D views to estimate the relative position of objects in a 3D scene. Our results show that the 3D view was used significantly more often than individual 2D views in both displays, indicating the importance of the 3D view for successful task completion. However, viewing patterns were significantly different between the two displays: transitions through centrally-placed views were always more frequent, and users avoided saccades between views that were far apart. Although the change in viewing strategy did not result in significant performance differences, error analysis indicates that a 3D overview in the center may reduce the number of serious errors compared to a 3D overview placed off to the side.

To make a spectral representation of color practicable for volume rendering, a new low-dimensional subspace method is used to act as the carrier of spectral information. With that model, spectral light material interaction can be integrated into existing volume rendering methods at almost no penalty. In addition, slow rendering methods can profit from the new technique of postillumination-generating spectral images in real-time for arbitrary light spectra under a fixed viewpoint. Thus, the capability of spectral rendering to create distinct impressions of a scene under different lighting conditions is established as a method of real-time interaction. Although we use an achromatic opacity in our rendering, we show how spectral rendering permits different data set features to be emphasized or hidden as long as they have not been entirely obscured. The use of postillumination is an order of magnitude faster than changing the transfer function and repeating the projection step. To put the user in control of the spectral visualization, we devise a new widget, a "light-dial," for interactively changing the illumination and include a usability study of this new light space exploration tool. Applied to spectral transfer functions, different lights bring out or hide specific qualities of the data. In conjunction with postillumination, this provides a new means for preparing data for visualization and forms a new degree of freedom for guided exploration of volumetric data sets.

We present a user interface, based on parallel coordinates, that facilitates exploration of volume data. By explicitly representing the visualization parameter space, the interface provides an overview of rendering options and enables users to easily explore different parameters. Rendered images are stored in an integrated history bar that facilitates backtracking to previous visualization options. Initial usability testing showed clear agreement between users and experts of various backgrounds (usability, graphic design, volume visualization, and medical physics) that the proposed user interface is a valuable data exploration tool.

We compare 2D/3D combination displays to displays with 2D and 3D views alone. Combination displays we consider are: orientation icon (i.e., side-by-side), in-place methods (e.g., clip planes), and a new method called ExoVis. We specifically analyze performance differences (i.e., time and accuracy) for 3D orientation and relative position tasks. Empirical results show that 3D displays are effective for approximate navigation and relative positioning whereas 2D/3D combination displays (orientation icon and ExoVis) are useful for precise orientation and position tasks. Combination 2D/3D displays had as good or better performance as 2D displays. Clip planes were not effective for a 3D orientation task, but may be useful when only one slice is needed.

Visualization can provide valuable assistance for data analysis and decision making tasks. However, how people perceive and interact with a visualization tool can strongly influence their understanding of the data as well as the system's usefulness. Human factors therefore contribute significantly to the visualization process and should play an important role in the design and evaluation of visualization tools. Several research initiatives have begun to explore human factors in visualization, particularly in perception-based design. Nonetheless, visualization work involving human factors is in its infancy, and many potentially promising areas have yet to be explored. Therefore, this paper aims to 1) review known methodology for doing human factors research, with specific emphasis on visualization, 2) review current human factors research in visualization to provide a basis for future investigation, and 3) identify promising areas for future research.

Must we continue to define a difference between information and scientific visualization? Scientific visualization evolved first in the late 1980’s while information visualization matured in the mid-1990’s. Scientific visualization is frequently considered to focus on the visual display of spatial data associated with scientific processes such as the bonding of molecules in computational chemistry. Information visualization examines developing visual metaphors for non-inherently spatial data such as the exploration of text-based document databases. This panel examines the effective, productive, and perhaps confusing tension between these subfields of visualization by highlighting the following issues:

2D and 3D views are used together in many visualization domains, such as medical imaging, flow visualization, oceanographic visualization, and computer aided design (CAD). Combining these views into one display can be done by: (1) orientation icon (i.e., separate windows), (2) in-place methods (e.g., clip and cutting planes), and (3) a new method called ExoVis. How 2D and 3D views are displayed affects ease of mental registration (understanding the spatial relationship between views), an important factor influencing user performance. This paper compares the above methods in terms of their ability to support mental registration. Empirical results show that mental registration is significantly easier with in-place displays than with ExoVis, and significantly easier with ExoVis than with orientation icons. Different mental transformation strategies can explain this result. The results suggest that ExoVis may be a better alternative to orientation icons when in-place displays are not appropriate (e.g., when in-place methods hide data or cut the 3D view into several pieces).

Computing devices within current work and play environments are relatively static. As the number of 'networked' devices grows, and as people and their devices become more dynamic, situations will commonly arise where users will wish to use 'that device there' instead of navigating through traditional user interface widgets such as lists. This paper describes a process for identifying devices through a pointing gesture using custom tags and a custom stylus called the gesturePen. Implementation details for this system are provided along with qualitative and quantitative results from a formal user study. As ubiquitous computing environments become more pervasive, people will rapidly switch their focus between many computing devices. The results of our work demonstrate that our gesturePen method can improve the user experience in ubiquitous environments by facilitating significantly faster interactions between computing devices.

We present the problem of visualizing time-varying medical data. Two medical imaging modalities are compared-MRI and dynamic SPECT. For each modality, we examine several derived scalar and vector quantities such as the change in intensity over time, the spatial gradient, and the change of the gradient over time. We compare several methods for presenting the data, including isosurfaces, direct volume rendering, and vector visualization using glyphs. These techniques may provide more information and context than methods currently used in practice; thus it is easier to discover temporal changes and abnormalities in a data set.