Decision-makers consult multiple forecasts to account for uncertainties when forming judgments about future events. While prior works have compared unaggregated and highly-aggregated designs for displaying multiple forecasts (e.g., Multiple Forecast Visualizations versus confidence interval plots), it remains unclear how partial aggregation impacts judgment. To investigate the effect of partial aggregation, we curated three designs that partially aggregate multiple forecasts. Through two large-scale studies (Experiment 1 n = 695 and Experiment 2 n = 389) across 14 judgment-related metrics, we observed that one design (Horizon Sampled MFV) significantly enhanced participants’ ability to predict future trends, thereby reducing their surprise when confronted with the actual outcomes. Grounded in empirical evidence, we provide insights into how to design visualizations for multiple forecasts to communicate uncertainty more effectively. Specifically, since no approach excels in all metrics, we advise choosing different designs based on communication goals and prior knowledge of forecasts.
Data visualization literacy is vital for preparing students for higher education in data-driven fields, yet instructions and teaching materials for high school remain limited. To address this gap, we applied participatory design (PD) and cooperative prototyping (CP) with high school teachers and data visualization experts as equal partners. Over a period of 16 months, we conducted an interview and four iterative workshops to identify teaching challenges and co-create solutions. This process produced three prototype sets, culminating in a 17-step teaching unit on the CoTinker platform for computational thinking. The unit uses a music dataset and a developed visualization tool to engage students in collaboratively creating data visualizations through hands-on learning while introducing visualization concepts, visual storytelling, and data ethics. Our work demonstrates how PD and CP can bridge research and practice, and contributes a replicable framework for integrating data visualization into high school by use of co-designed tools grounded in real-world classrooms.
Generative models have substantially expanded video generation capabilities, yet practical thought-to-video creation remains a multi-stage, multi-modal, and decision-intensive process. However, existing tools either hide intermediate decisions behind repeated reruns or expose operator-level workflows that make exploration traces difficult to manage, compare, and reuse. We present T2VTree, a user-centered visual analytics approach for agent-assisted thought-to-video authoring. T2VTree represents the authoring process as a tree visualization. Each node in the tree binds an editable specification (intent, referenced inputs, workflow choice, prompts, and parameters) with the resulting multimodal outputs, making refinement, branching, and provenance inspection directly operable. To reduce the burden of deciding what to do next, a set of collaborating agents translates step-level intent into an executable plan that remains visible and user-editable before execution. We further implement a visual analytics system that integrates branching authoring with in-place preview and stitching for convergent assembly, enabling end-to-end multi-scene creation without leaving the authoring context. We demonstrate T2VTreeVA through two multi-scene case studies and a comparative user study, showing how the T2VTree visualization and editable agent planning support reliable refinement, localized comparison, and practical reuse in real authoring workflows. T2VTree is available at: https://github.com/tezuka0210/T2VTree.
Applications of Implicit Neural Representations (INRs) have emerged as a promising deep learning approach for compactly representing large volumetric datasets. These models can act as surrogates for volume data, enabling efficient storage and on-demand reconstruction via model predictions. However, conventional deterministic INRs only provide value predictions without insights into the model's prediction uncertainty or the impact of inherent noisiness in the data. This limitation can lead to unreliable data interpretation and visualization due to prediction inaccuracies in the reconstructed volume. Identifying erroneous results extracted from model-predicted data may be infeasible, as raw data may be unavailable due to its large size. To address this challenge, we introduce REV-INR, Regularized Evidential Implicit Neural Representation, which learns to predict data values accurately along with the associated coordinate-level data uncertainty and model uncertainty using only a single forward pass of the trained REV-INR during inference. By comprehensively comparing and contrasting REV-INR with existing well-established deep uncertainty estimation methods, we show that REV-INR achieves the best volume reconstruction quality with robust data (aleatoric) and model (epistemic) uncertainty estimates using the fastest inference time. Consequently, we demonstrate that REV-INR facilitates assessment of the reliability and trustworthiness of the extracted isosurfaces and volume visualization results, enabling analyses to be solely driven by model-predicted data.
Geospalial tasks often require the coordination of various spatial algorithms and operations, k hich are usually performed through tool calling guided by natural language prompts. Crafting effective prompts is challenging due to the inherent complexity and ambiguity of natural language. In this paper, we present GeoPct, visual analytics system designed to simplify the process of prompt engineering to improve the performance of geospatial tool callings on large language models (LLMs). At its core, GeoPei is a sophisticated tool recommendation method that accepts geospatial tasks as input, decomposes them into atomic tasks, relevant tools, and extracts the most relevant tool descript these atomic tasks. The system is designed to support interact prompt engineering of geospatial tool invocations, enabling users to explore the connections between geospatial tasks and established tools and to evaluate their performance. This enables crafting and refining prompts that coordinate LLMs with human expertise. Ger/Pees satisfaction, practicality, usability, and visual design are validated through two case studies and a user study. These demonstrate that the system significantly eases the burden of rapid enpineerinp and skilllially guides LLMs in geospatial tool calling capabilities. By providing a visual and interactive system for prompt engineering, GeoPet helps users navigate complex geospatial tasks and improves the overall efficiency nd accuracy of tool calling for LLMs.
The visualization of chaotic systems plays a crucial role in revealing their dynamic behavior and inherent complexity. However, traditional methods often lack a quantitative assessment of their fidelity in representing chaos and fail to establish intuitive, metaphorical connections for non-expert audiences. To address these gaps, this study proposes a metaphorical visualization approach for one-dimensional chaotic systems, utilizing typographic layouts with “visual disorder” as a metaphor to convey chaos intuitively. The key contributions are twofold: first, a quantitative evaluation of classical chaos visualization techniques is presented, assessing their fidelity in representing chaos intensity; second, a visualization based on typographic layout is introduced and its fidelity and metaphorical appropriateness are quantitatively evaluated. Experimental results reveal a strong linear correlation between the visual disorder in typographic layouts and the Lyapunov Exponents of the chaotic systems, achieving fidelity comparable to that of traditional phase-space plot. At the same time, testers’ perception of the visual metaphor in layout visualization aligns more closely with the essential characteristics of chaos and significantly outperforms traditional visualization. This advantage is expected to help people, especially non-professionals, understand and learn the complex and abstract concept of chaos.
Data visualization is a critical tool for supporting decision-making across various domains. However, its effectiveness in addressing various decision-making problems is yet to be investigated.This study systematically reviewed over 300 papers from major visualization journals and conferences and focused on 40 empirical studies. Using Franz and Kramer’s multidimensional decision-making framework, we classified and analyzed the characteristics of decision problems supported by data visualization.The findings revealed three key insights: (1) data visualization predominantly supports organizational and community-level decision-making, whereas applications at individual and family levels remain limited; (2) support has expanded from structured evaluation problems to include semi-structured and recognition-primed decision-making; and (3) challenges persist in addressing unstructured problems and small-scale contexts.These results highlight the evolving role of data visualization in supporting complex and interdependent decision scenarios. Future studies should address gaps in unstructured decision-making and investigate applications in small-scale contexts to enhance the adaptability and impact of visualization tools.
Understanding user profiling on social media poses significant challenges due to the intertwined complexities of network structures, user interactions, and the multi-dimensional nature of the data. To reduce visual clutter and offer complementary perspectives for engagingly exploring user profiling within these networks, we propose Social Media Island, an interactive 3D metaphoric visualization system. Our system uses 3D mountain metaphors to visualize user profiling, capturing user influence and activity, while tree metaphors visualize the information forwarding process. To support a flexible scope for users to explore in a more intriguing way, we design various interactions such as cutting the mountain to split out a subset with similarity to some extent for further exploration. By using these 3D visualizations with user interactions, Social Media Island facilitates immersion in the data, the fluid exploration of user influence, topic evolution, and the spread of information. The effectiveness of metaphors is evaluated by user studies, and that of the system is evaluated through two case studies.
Graph-based Retrieval-Augmented Generation (RAG) has shown great capability in enhancing Large Language Model (LLM)’s answer with an external knowledge base. Compared to traditional RAG, it introduces a graph as an intermediate representation to capture better structured relational knowledge in the corpus, elevating the precision and comprehensiveness of generation results. However, developers usually face challenges in analyzing the effectiveness of GraphRAG on their dataset due to GraphRAG’s complex information processing pipeline and the overwhelming amount of LLM invocations involved during graph construction and query, which limits GraphRAG interpretability and accessibility. This research proposes a visual analysis framework that helps RAG developers identify critical recalls of GraphRAG and trace these recalls through the GraphRAG pipeline. Based on this framework, we develop XGraphRAG, a prototype system incorporating a set of interactive visualizations to facilitate users’ analysis process, boosting failure cases collection and improvement opportunities identification. Our evaluation demonstrates the effectiveness and usability of our approach. Our work is open-sourced and available at https://github.com/Gk0Wk/XGraphRAG.
In basketball, decision-making is one of the core skills for players. For example, when a player is holding the ball, the success of the team’s offense is primarily determined by her/his decisions (i.e., pass, shoot, or dribble) in response to the dynamics of the game. Understanding players’ decision-making processes in changing game situations can help coaches develop effective strategies, which is critical for the success of a team. However, the decision-making process is influenced by various factors (e.g., player’s playing style, opponents’ defense, and time remaining), making understanding a challenging problem. In this study, we propose HoopScouter, a visual analytics system to help understand ball handlers’ decisions in basketball games. Based on a careful investigation of the analysis requirements, we first introduce a representation learning method that characterizes ball handlers’ decision-making styles. We then design a sketch panel with integrated time information to support exploration of player decisions under similar game scenarios. Facet views and coordinated interactions are also provided to identify the strengths and weaknesses of the ball handler’s decision-making, and to understand when and why ball handlers would make certain decisions. To validate the effectiveness of HoopScouter, we conduct two case studies on real-world basketball games and receive positive feedback from domain experts.
Shape-based metrics measure how faithfully a drawing D of a large graph G shows the structure of the graph by comparing the similarity between G and a proximity graph S computed from D. In this paper, we present new degree-constrained shape-based metrics by introducing new proximity graphs dcGG and dcRNG, which constrain the degree of each vertex v in the proximity graph S to be no greater than the degree of v in G. Extensive experiments demonstrate that our new degree-constrained shape-based metrics QdcRNG and QdcGG can more accurately measure the shape-faithfulness than the state-of-the-art degree-sensitive shape-based metrics, with on average 29.1% better metrics for highly shape-faithful layouts. Moreover, we present extensive comparison experiments of ten popular graph layouts using our new shape-based metrics QdcRNG and QdcGG to recommend shape-faithful layouts for large and complex graphs. Furthermore, we present a new shape-faithful graph drawing algorithm dcShFR, based on the most popular force-directed algorithm FR, to optimize the new shape-based metrics QdcRNG and QdcGG. Extensive experiments demonstrate that our dcShFR algorithm computes 14.5% higher shape-faithful drawings than FR on average.
Microbial time-series data usually need to be obtained by professionals through biological experiments, and the amount of data obtained from such manual experiments is very limited. Deep learning generative models leverage the superior learning capabilities of neural networks to generate high-quality synthetic data, which provides a promising approach to address the above shortcomings. However, the black-box nature of deep learning models makes it difficult for domain experts to trust the generated data, limiting the application of these techniques. Thus, we propose MTvis, an interactive visualization system designed to help experts understand and optimize data generated by MT-GAN, which is a microbial time-series data augmentation model proposed by us. MT-GAN not only introduces the temperature lag effect but also captures the dynamics of microbial time-series data via adversarial and joint learning. MTvis provides two exploration modes which enable users to gain insights into the structure of the model and adjust hyperparameters dynamically. The system also provides real-time observation of distribution differences between generated and real data. Experimental results show that MT-GAN effectively improves the fidelity of the generated data, while MTvis enhances the credibility and practicality of synthetic data for domain experts.
Three dimensional (3D) scanned point clouds offer an invaluable tool for the digital preservation of cultural heritage, capturing intricate and complex 3D structures. Transparent visualization, i.e., see-through visualization, is essential for effectively understanding these data. However, since 3D scanned point clouds do not include normal vector information, applying shading effects to reveal complex shapes is challenging. In this paper, we present a method for assigning normal vectors to 3D scanned points using a statistical approach based on principal component analysis. To resolve the ambiguity of normal vector orientations, specifically the front or back of local surfaces, we exploit the characteristics of transparent visualization. Each normal vector is oriented toward the viewpoint, enabling consistent shading across all visible points. We then apply dynamic shading by varying the light source position sequentially, which highlights different regions of the visualized object. This approach allows for a shift in focus between surface features and deeper internal structures, enhancing the understanding of the overall structure of complex 3D cultural heritage objects. We demonstrate the effectiveness of dynamic shading through its application to real 3D scanned data of cultural heritage objects with intricate internal structures.
Image -generative models have gained popularity over the last years with their ability to create realistic artwork. Realizing complex artworks with specific creative ideas often requires iterative optimization of specialized prompts, but may still result in inadequate images. The inclusion of reference images and adapting model specific parameters can help in steering the model and fostering the creative intent of the user. But by providing text prompts, initial images, and adapting model parameters, users face a vast design space for creating images. To navigate through this space, we propose a visualization approach that combines an interactive Provenance Graph, parameter visualizations, and high-dimensional embeddings. Our approach helps pursue multiple parallel creation paths, makes workflows traceable and parameter changes transparent, and facilitates the reporting of image editing steps. In addition to prompt formulation, we focus on targeted generation by probing parameters, image compositions, and editing details. We integrate the generative process into existing image editing software, enabling users to compose artwork in collaboration with the model. The presented approach is evaluated in a user experiment (n=9) for generating artwork. The results show that users with different levels of experience can create targeted artwork but use different strategies when working with the Provenance Graph.
In mobile edge computing (MEC), one optimization strategy for mobile applications is to offload heavy computing tasks to cloud and edge servers. Constructing partitioning algorithms involves modeling individual methods through static code profilers, but exploiting dynamic user-driven execution patterns is also crucial. This paper introduces PartFlow, an interactive visualization system that supports comprehensive analysis of mobile application components and aids researchers in developing partitioning and offloading algorithms using real human behavioral data. PartFlow collects application component data remotely through binary instrumentation of mobile applications. Interactive diagrams are designed to evaluate component performance and illustrate transition patterns using the collected data. Additionally, PartFlow integrates a deep learning (DL)-based approach for multi-step forecasting of component states to improve accuracy and user experience in algorithm design. A case study and user feedback demonstrate PartFlow’s effectiveness in assisting researchers and engineers in creating offloading strategies.
Fiber surfaces and fiber lines, being the preimages of a bivariate function to a so-called control polygon in the function’s range, are the adaptation of isosurfaces and isolines to bivariate scalar fields. Previous works have proposed use cases for fiber surface and fiber line extraction using algorithmically generated control polygons with many line segments, but their authors have either not presented results or noted that the computation becomes prohibitive slow. We present an algorithm that speeds up fiber surface extraction by using dual bounding volume hierarchy (BVH) traversal, as well as a variation that combines the benefits of single and dual BVH traversal. We study the influence of the number of line segments in the control polygon on the performance of single and dual BVH traversal algorithms using data sets from various application domains and types of control polygons. We find that dual BVH traversal is several times faster in test cases where single BVH traversal is slow, facilitating the interactive exploration of fiber surfaces in many cases where existing methods are too slow.
A key compositional technique, leading lines guide the observer’s gaze across visual scenes. This study presents a novel approach to the visual analysis of leading lines in artworks by applying discrete Morse theory to saliency maps refined through persistent homology. The saliency map’s gradient field is abstracted into a structured representation using the maximum graph, enabling the identification of both explicit and implicit leading lines. To enhance the analysis, texture gradient cues are incorporated to embed depth information into the saliency map. Our visual analysis system also features semiautomatic thresholding and tools to depict the direction of visual guidance. An experimental study with participants from an art university validated the system’s feasibility by analyzing various Western paintings. This research contributes to the field of digital and public humanities and computational art analysis by offering an environment in which to explore visual composition.
Image-based person re-identification (Re-ID) aims to identify and track individuals across multiple camera views using query images. While machine learning methods have made progress, their real-world performance remains limited. A key challenge lies in the nature of person retrieval, which requires users to perform fine-grained matching and filtering. This process often involves manually comparing and evaluating a large number of candidate images, resulting in low retrieval efficiency and being time-consuming. To address these issues, we introduce textual information to assist users in performing fine-grained retrieval tasks. Specifically, we utilize vision-language models fine-tuned with domain knowledge to generate hierarchical textual descriptions as retrieval cues. We also provide a visual analysis tool, which adopts multi-view and adjustable visual encodings to aggregate and present image data, supporting interactive browsing and retrieval. Finally, we conduct a case study and visual analysis experiments to evaluate the effectiveness of the textual retrieval cues. The evaluation results reveal the potential of textual information in optimizing person retrieval and offers insights for future work.
Data -visualizations are now commonly used in online press articles which often supports engaging data -driven stories. However, due to its visual nature, this type of content inherently lacks accessibility (e.g. when one wants to consume those visualizations using conversational agents, hearing them in audible formats, or using screen reader). Writing alternative texts is the recommended standard in order to provide text descriptions associated with an image. However, newsrooms rarely produce them for data -visualizations, or when they do, these are overly simplistic. Several intertwined limitations explain that situation like the limited amount of time journalists have to produce- these-expected detailed descriptions or the lack of precise and standardized writing guidelines for describing visualizations. To address this issue, we propose a new approach to help journalists generate descriptions of data-visualizations, based on a set of generated question and answer pairs (hereafter referred to as Q/A). Due to the previously enumerated limitations, our method first generates those Q/As using a generative AI model of Natural Language Processing (NLP). This approach alleviates and homogenizes the writing task workload and allows for a systematic and more exhaustive exploration of the possible Q/As for a given visualization..However, among the critical challenges of using AI -based generative tools in a journalism context is the risk of publishing unreliable or biased information. Therefore, the methodology proposed in this paper gives the journalist user a high level of control over the Al-generated Q/As. To enable and optimize this mandatory validation task, we design an interface where Q/As are grouped in terms of semantic and textual content, and accessibility interest. Visual cues are also displayed to improve the journalist's decision -making. To evaluate this proposed methodology, that we call GenQA, we conducted a comparative design study that gathered journalists from two different Canadian newsrooms and teachers. We observed that GenQA was efficiently used by those users and helped them to produce detailed visualization descriptions that met their expectations in terms of quality and workload. This study also showed that GenQA triggered significant serendipity potential, allowing users to explore and produce Q/As that cover aspects they might not have considered.
The increasing complexity and adoption of machine learning (ML) pipelines has led to a rising demand for effective visualization tools. This paper presents a comprehensive review of existing tools for visualizing data flow in machine learning (ML) pipelines. We highlight the tools’ purposes, integration methods, and visualization techniques. We collected and analyzed 22 open-source tools and concepts, analyzing their features and classifying them based on their primary purpose. Our analysis revealed five main purposes of visualization tools: exploration, explanation, visual development, comparison, monitoring. We provide an analysis of their integration methods, from standalone visual interfaces to code-level libraries, as well as a review of various visualization techniques, including Directed Acyclic Graphs (DAGs), pipeline matrices, and annotated visualizations. Our findings highlight the importance of visualization in enhancing the interpretability and efficiency of ML workflows. Moreover, the paper provides key limitations and challenges in current visualization methods to promote future research directions enhancing the usability and functionality of ML pipeline visualization tools.
