We make four contributions to lower the overhead of conducting visualisation user studies and promote the reuse and extension of their materials. (i) A declarative Javascript specification lets experimenters describe how studies are assembled from tested visualisations, datasets, tasks and chosen evaluation strategies. (ii) A VisUnit library translates these into sequences of visual stimuli and delivers them to participants. We move away from monolithic evaluation stimuli typical of previous work and construct studies around three ingredients – visual encodings, datasets, and tasks – that can be developed independently and recombined flexibly. (iii) This paves the way for developing benchmark data+tasks test-suites as independent, reusable resources to support multiple studies. (iv) Structuring user studies as “literate” visualisation notebooks brings together in the open all ingredients necessary for replication and scrutiny: formal design specification; underlying materials; participant-facing views; and narratives justifying design and supporting reuse.

We explore the integration of foundation models, such as large language models (LLMs) and multimodal LLMs (MLLMs), into visual analytics (VA) systems through intuitive natural language interactions. We survey current research directions in this emerging field, examining how foundation models have already been integrated into key visualisation-related processes in VA: visual mapping, the creation of data visualisations; visualisation observation, the process of generating a finding through visualisation; and visualisation manipulation, changing the viewport or highlighting areas of interest within a visualisation. We also highlight new possibilities that foundation models bring to VA, in particular, the opportunities to use MLLMs to interpret visualisations directly, to integrate multimodal interactions, and to provide guidance to users. We finally conclude with a vision of future VA systems as collaborative partners in analysis and address the prominent challenges in realising this vision through foundation models. Our discussions in this paper aim to guide future researchers working on foundation model assisted VA systems and help them navigate common obstacles when developing these systems.

Abstract Eye tracking provides a unique perspective on the inherently visual discourse between visualisation systems and their users, and has recently become sufficiently precise and affordable to be integrated as regular input into workstations and virtual or augmented reality headsets alike. As such, real‐time eye tracking can now contribute significantly towards the development of gaze‐aware visualisations that infer and monitor users' needs to actively support their activities. To facilitate such systems we make three contributions. First, we structure and discuss design considerations for gaze‐aware visualisations along four axes: measurable data; inferable data; opportunities for support; and limiting factors to beware. Second, we distill visualisation research challenges that preclude such systems. Finally, we show via three usage scenarios how to apply these design considerations to imagine how existing systems can benefit from real‐time eye tracking. We combined a structured literature analysis, a consideration of suitable places for eye‐tracking integration in the typical visualisation ecosystem, and design space modelling. Eye tracking has significant potential to improve the interactive visual analysis of data across many visualisation domains. Our paper attempts to provide a comprehensive, general survey and conceptual discussion in this promising field, outlining the state‐of‐the‐art and future research opportunities.

Visual analytics (VA) research provides helpful solutions for interactive visual data analysis when exploring large and complex datasets. Due to recent advances in eye tracking technology, promising opportunities arise to extend these traditional VA approaches. Therefore, we discuss foundations for eye tracking support in VA systems. We first review and discuss the structure and range of typical VA systems. Based on a widely used VA model, we present five comprehensive examples that cover a wide range of usage scenarios. Then, we demonstrate that the VA model can be used to systematically explore how concrete VA systems could be extended with eye tracking, to create supportive and adaptive analytics systems. This allows us to identify general research and application opportunities, and classify them into research themes. In a call for action, we map the road for future research to broaden the use of eye tracking and advance visual analytics.

Visualizing network data is applicable in domains such as biology, engineering, and social sciences. We report the results of a study comparing the effectiveness of the two primary techniques for showing network data: node-link diagrams and adjacency matrices. Specifically, an evaluation with a large number of online participants revealed statistically significant differences between the two visualizations. Our work adds to existing research in several ways. First, we explore a broad spectrum of network tasks, many of which had not been previously evaluated. Second, our study uses two large datasets, typical of many real-life networks not explored by previous studies. Third, we leverage crowdsourcing to evaluate many tasks with many participants. This paper is an expanded journal version of a Graph Drawing (GD'17) conference paper. We evaluated a second dataset, added a qualitative feedback section, and expanded the procedure, results, discussion, and limitations sections.

Eye-tracking data is traditionally analyzed by looking at where on a visual stimulus subjects fixate, or, to facilitate more advanced analyses, by using area-of-interests (AOI) defined onto visual stimuli. Recently, there is increasing interest in methods that capture what users are looking at rather than where they are looking. By instrumenting visualization code that transforms a data model into visual content, gaze coordinates reported by an eye-tracker can be mapped directly to granular data shown on the screen, producing temporal sequences of data objects that subjects viewed in an experiment. Such data collection, which is called gaze to object mapping (GTOM) or data-of-interest analysis (DOI), can be done reliably with limited overhead and can facilitate research workflows not previously possible. Our paper contributes to establishing a foundation of DOI analyses by defining a DOI data model and highlighting its differences to AOI data in structure and scale; by defining and exemplifying a space of DOI enabled tasks; by describing three concrete examples of DOI experimentation in three different domains; and by discussing immediate research challenges in creating a framework of visual support for DOI experimentation and analysis.

Eye-tracking data is currently analyzed in the image space that gaze-coordinates were recorded in, generally with the help of overlays such as heatmaps or scanpaths, or with the help of manually defined areas of interest (AOI). Such analyses, which focus predominantly on where on the screen users are looking, require significant manual input and are not feasible for studies involving many subjects, long sessions, and heavily interactive visual stimuli. Alternatively, we show that it is feasible to collect and analyze eye-tracking information in data space. Specifically, the visual layout of visualizations with open source code that can be instrumented is known at rendering time, and thus can be used to relate gaze-coordinates to visualization and data objects that users view, in real time. We demonstrate the effectiveness of this approach by showing that data collected using this methodology from nine users working with an interactive visualization, was well aligned with the tasks that those users were asked to solve, and similar to annotation data produced by five human coders. Moreover, we introduce an algorithm that, given our instrumented visualization, could translate gaze-coordinates into viewed objects with greater accuracy than simply binning gazes into dynamically defined AOIs. Finally, we discuss the challenges, opportunities, and benefits of analyzing eye-tracking in visualization and data space.

We present the design and evaluation of a method for estimating gaze locations during the analysis of static visualizations using crowdsourcing. Understanding gaze patterns is helpful for evaluating visualizations and user behaviors, but traditional eye-tracking studies require specialized hardware and local users. To avoid these constraints, we developed a method called Fauxvea, which crowdsources visualization tasks on the Web and estimates gaze fixations through cursor interactions without eye-tracking hardware. We ran experiments to evaluate how gaze estimates from our method compare with eye-tracking data. First, we evaluated crowdsourced estimates for three common types of information visualizations and basic visualization tasks using Amazon Mechanical Turk (MTurk). In another, we reproduced findings from a previous eye-tracking study on tree layouts using our method on MTurk. Results from these experiments show that fixation estimates using Fauxvea are qualitatively and quantitatively similar to eye tracking on the same stimulus-task pairs. These findings suggest that crowdsourcing visual analysis tasks with static information visualizations could be a viable alternative to traditional eye-tracking studies for visualization research and design.

Abstract We present GraphUnit, a framework and online service that automates the process of designing, running and analyzing results of controlled user studies of graph visualizations by leveraging crowdsourcing and a set of evaluation modules based on a graph task taxonomy. User studies play an important role in visualization research but conducting them requires expertise and is time consuming. GraphUnit simplifies the evaluation process by allowing visualization designers to easily configure user studies for their web‐based graph visualizations, deploy them online, use Mechanical Turk to attract participants, collect user responses and store them in a database, and analyze incoming results automatically using appropriate statistical tools and graphs. We demonstrate the effectiveness of GraphUnit by replicating two published evaluation studies on network visualization, and showing that these studies could be configured in less than an hour. Finally, we discuss how GraphUnit can facilitate quick evaluations of alternative graph designs and thus encourage the frequent use of user studies to evaluate design decisions in iterative development processes.

Abstract Performing typical network tasks such as node scanning and path tracing can be difficult in large and dense graphs. To alleviate this problem we use eye‐tracking as an interactive input to detect tasks that users intend to perform and then produce unobtrusive visual changes that support these tasks. First, we introduce a novel fovea based filtering that dims out edges with endpoints far removed from a user's view focus. Second, we highlight edges that are being traced at any given moment or have been the focus of recent attention. Third, we track recently viewed nodes and increase the saliency of their neighborhoods. All visual responses are unobtrusive and easily ignored to avoid unintentional distraction and to account for the imprecise and low‐resolution nature of eye‐tracking. We also introduce a novel gaze‐correction approach that relies on knowledge about the network layout to reduce eye‐tracking error. Finally, we present results from a controlled user study showing that our methods led to a statistically significant accuracy improvement in one of two network tasks and that our gaze‐correction algorithm enables more accurate eye‐tracking interaction.

We present the results of evaluating four techniques for displaying group or cluster information overlaid on node-link diagrams: node coloring, GMap, BubbleSets, and LineSets. The contributions of the paper are three fold. First, we present quantitative results and statistical analyses of data from an online study in which approximately 800 subjects performed 10 types of group and network tasks in the four evaluated visualizations. Specifically, we show that BubbleSets is the best alternative for tasks involving group membership assessment; that visually encoding group information over basic node-link diagrams incurs an accuracy penalty of about 25 percent in solving network tasks; and that GMap's use of prominent group labels improves memorability. We also show that GMap's visual metaphor can be slightly altered to outperform BubbleSets in group membership assessment. Second, we discuss visual characteristics that can explain the observed quantitative differences in the four visualizations and suggest design recommendations. This discussion is supported by a small scale eye-tracking study and previous results from the visualization literature. Third, we present an easily extensible user study methodology.

Methodologies for complex systems diagnosis and verification have long been studied as part of systems engineering research. Most techniques involve decomposing the complex system into smaller connected components, and analyzing those. We describe the implementation and evaluation of a visualization tool which enhances drawings of physical components with information visualization artifacts for analyzing complex systems and their operating capability. Our visualization was designed to help users navigate through complex systems composed of multiple layers of components, identify if the system is ready to complete a task based on availability and performance of its components, and efficiently diagnose system malfunctions. A formal evaluation shows that our visualization tool enables users to diagnose complex systems faster than using conventional workflows.

Subtle changes in analysis system interfaces can be used purposely to alter users' analytic behaviors. In a controlled study subjects completed three analyses at one-week intervals using an analysis support system. Control subjects used one interface in all sessions. Test subjects used modified versions in the last two sessions: a first set of changes aimed at increasing subjects' use of the system and their consideration of alternative hypotheses; a second set of changes aimed at increasing the amount of evidence collected. Results show that in the second session test subjects used the interface 39% more and switched between hypotheses 19% more than in the first session. They then collected 26% more evidence in the third than in the second session. These increases differ significantly (p<0.05) from near constant control rates. We hypothesize that this approach can be used in many real applications to guide analysts unobtrusively towards improved analytic strategies.

We present an ethnographic study of design differences in visual presentations between academic disciplines. Characterizing design conventions between users and data domains is an important step in developing hypotheses, tools, and design guidelines for information visualization. In this paper, disciplines are compared at a coarse scale between four groups of fields: social, natural, and formal sciences; and the humanities. Two commonplace presentation types were analyzed: electronic slideshows and whiteboard "chalk talks". We found design differences in slideshows using two methods - coding and comparing manually-selected features, like charts and diagrams, and an image-based analysis using PCA called eigenslides. In whiteboard talks with controlled topics, we observed design behaviors, including using representations and formalisms from a participant's own discipline, that suggest authors might benefit from novel assistive tools for designing presentations. Based on these findings, we discuss opportunities for visualization ethnography and human-centered authoring tools for visual information.

We introduce two-dimensional neural maps for exploring connectivity in the brain. For this, we create standard streamtube models from diffusion-weighted brain imaging data sets along with neural paths hierarchically projected into the plane. These planar neural maps combine desirable properties of low-dimensional representations, such as visual clarity and ease of tract-of-interest selection, with the anatomical familiarity of 3D brain models and planar sectional views. We distribute this type of visualization both in a traditional stand-alone interactive application and as a novel, lightweight web-accessible system. The web interface integrates precomputed neural-path representations into a geographical digital-maps framework with associated labels, metrics, statistics, and linkouts. Anecdotal and quantitative comparisons of the present method with a recently proposed 2D point representation suggest that our representation is more intuitive and easier to use and learn. Similarly, users are faster and more accurate in selecting bundles using the 2D path representation than the 2D point representation. Finally, expert feedback on the web interface suggests that it can be useful for collaboration as well as quick exploration of data.

Graphs, generally used as data structures in computer science applications, have steadily shown a growth in mapping various types of relationships, from maps to computer networks to social networks. As graph layouts and visualizations have been at the forefront of graph drawing research for decades, it consequently led to aesthetic heuristics that not only generated better visualizations and aesthetically appealing graphs but also improved readability and understanding of the graphs. A variety of approaches examines aesthetics of nodes, edges, or graph layout, and related readability metrics. In this paper we focus on the edge crossing problem and propose a solution that incorporates Gestalt principles to improve graph aesthetics and readability. We introduce the concept of breaks in edges at edge crossings. A break is a gap in an edge drawing occurring in the vicinity of an edge crossing. At every edge crossing, one of the incident edges is broken, which will prevent any unintentional Gestalts that occur at edge crossings that reduce the readability of a graph drawing. We present our preliminary results and user studies that show that this technique could play a role in improving graph readability.

We introduce several novel visualization and interaction paradigms for visual analysis of published protein-protein interaction networks, canonical signaling pathway models, and quantitative proteomic data. We evaluate them anecdotally with domain scientists to demonstrate their ability to accelerate the proteomic analysis process. Our results suggest that structuring protein interaction networks around canonical signaling pathway models, exploring pathways globally and locally at the same time, and driving the analysis primarily by the experimental data, all accelerate the understanding of protein pathways. Concrete proteomic discoveries within T-cells, mast cells, and the insulin signaling pathway validate the findings. The aim of the paper is to introduce novel protein network visualization paradigms and anecdotally assess the opportunity of incorporating them into established proteomic applications. We also make available a prototype implementation of our methods, to be used and evaluated by the proteomic community.

We introduce the concept of coloring close and crossing edges in graph drawings with perceptually opposing colors making them individually more distinguishable and reducing edge-crossing effects. We define a "closeness" metric on edges as a combination of distance, angle and crossing. We use the inverse of this metric to compute a color embedding in the L*a*b* color space and assign "close" edges colors that are perceptually far apart. We present the following results: a distance metric on graph edges, a method of coloring graph edges, and anecdotal evidence that this technique can improve the reading of graph edges.

We present a visual exploration paradigm that facilitates navigation through complex fiber tracts by combining traditional 3D model viewing with lower dimensional representations. To this end, we create standard streamtube models along with two two-dimensional representations, an embedding in the plane and a hierarchical clustering tree, for a given set of fiber tracts. We then link these three representations using both interaction and color obtained by embedding fiber tracts into a perceptually uniform color space. We describe an anecdotal evaluation with neuroscientists to assess the usefulness of our method in exploring anatomical and functional structures in the brain. Expert feedback indicates that, while a standalone clinical use of the proposed method would require anatomical landmarks in the lower dimensional representations, the approach would be particularly useful in accelerating tract bundle selection. Results also suggest that combining traditional 3D model viewing with lower dimensional representations can ease navigation through the complex fiber tract models, improving exploration of the connectivity in the brain.

An information hierarchy is a collection of relational information that is arranged in a ranking organization where each entity is subject to a single other entity, except for the top (root) element. The usefulness of a visualization of an information hierarchy depends on its capability of conveying the information quickly and clearly. The interaction with the information hierarchy allows a user to further analyze its underlying structures and relationships, which is essential for the effectiveness of the visualization. In this paper we present a novel method to interactively visualize information hierarchies in real-time. We use the World Wide Web as an application example of our techniques. The result is a novel Web browsing and visualization method with an innovative combination of features: (i) Web data is retrieved and displayed in real-time (i.e. Web data is not pre-recorded), (ii) browsing and visualization are synchronized together in the same interface, (iii) tree-based visualization engine, and (iv) space-efficient display of visualization. Our study shows that users are able to orient themselves better in cyberspace and locate Web pages of interest faster.

Many algorithms have been designed to visualize binary trees efficiently with respect to a quality measure. While each algorithm is suitable for drawing particular categories of binary trees, an effort to compile these algorithms to maximize the quality of drawings has not been realized. Our first step is to create a system that determines the type of a binary tree and then selects an algorithm to draw the tree depending upon the specified quality measures. Currently, our system recognizes six types of binary trees (AVL, Complete, Fibonacci, Random, Unbalanced-tothe- left, Unbalanced-to-the-right) and allows the user to choose from eleven quality measures (Area, Aspect Ratio, Total Edge Length, Maximum Edge Length, Uniform Edge Length, Closest Leaf, Farthest Leaf, Size, Minimum Angle Size, Average Angle Size, Angular Resolution). Experiments show that our adaptive visualization system outperforms any system using a single binary tree drawing algorithm. In addition, our approach allows the user to select multiple quality measures and automatically detects the best available binary tree drawing algorithm.