Visualization Analysis and Design
... SCOPE is an ordered axis that captures the scale of the data. The ACTION + TARGET pairs follow the model of Munzner [42]: actions are verbs that define user goals, and targets are nouns specifying some aspect of the data that is of interest to the user. Our taxonomy includes targets that are well-known concepts from the graph theory literature, or can be precisely expressed in terms of topological structure. ...
... Some concrete tasks are the combination of multiple simple tasks chained together [12,42]; we call these complex tasks. One example is "Is the largest clique near the graph's periphery?", ...
... In 2013 Brehmer and Munzner [12] proposed bridging from low-level tasks to higher-level tasks with a multi-level typology of abstract tasks. Munzner [42] breaks down tasks into action-target pairs; we extend this approach in our taxonomy. ...
We investigate tasks that can be accomplished with unlabelled graphs, where nodes do not have persistent or semantically meaningful labels. New techniques to visualize these graphs have been proposed, but more understanding of unlabelled graph tasks is required before they can be adequately evaluated. Some tasks apply to both labelled and unlabelled graphs, but many do not translate between these contexts. We propose a taxonomy of unlabelled graph abstract tasks, organized according to the Scope of the data at play, the Action intended by the user, and the Target data under consideration. We show the descriptive power of this task abstraction by connecting to concrete examples from previous frameworks, and connect these abstractions to real-world problems. To showcase the evaluative power of the taxonomy, we perform a preliminary assessment of 6 visualizations for each task. For each combination of task and visual encoding, we consider the effort required from viewers, the likelihood of task success, and how both factors vary between small-scale and large-scale graphs.
... While these models can be used to evaluate each stage of a static multimedia processing pipeline, they are less suited for real-time and interactive multimedia analytics. For usability evaluation, Shneiderman's principles [71] and Munzner's nested model [57] help assess multimedia analytics interaction design but overlook long-term adoption (e.g., critical in longitudinal multimedia forensics). Long-term case studies [70] help to evaluate multimedia analytics solutions over time. ...
... When users analyze multimedia data, they often have high-level tasks they aim to accomplish. The tasks require a sequence of actions that support analytic reasoning and insight generation, and much of the latter occurs as part of human cognition, which has been conceptualized in the Knowledge Generation Model [66] (see subsection 3.3) and the nested model by Munzner [57]. ...
... Traditionally, user actions have been shaped by technical system design, often structured around classical data science tasks such as filtering, searching, or clustering. As such, the actions have operated on different (data) targets, like in the nested model [57], which guides data visualization along the what?, why?, and how? questions. To fit it to multimedia analytics, it requires refinements, in particular to its data (what?) and target definitions (part of why?) as well as the process (how?), mapping it to the broader exploration-search axis [90] that characterizes multimedia analytics. ...
The rapid advances in Foundation Models and agentic Artificial Intelligence are transforming multimedia analytics by enabling richer, more sophisticated interactions between humans and analytical systems. Existing conceptual models for visual and multimedia analytics, however, do not adequately capture the complexity introduced by these powerful AI paradigms. To bridge this gap, we propose a comprehensive multimedia analytics model specifically designed for the foundation model era. Building upon established frameworks from visual analytics, multimedia analytics, knowledge generation, analytic task definition, mixed-initiative guidance, and human-in-the-loop reinforcement learning, our model emphasizes integrated human-AI teaming based on visual analytics agents from both technical and conceptual perspectives. Central to the model is a seamless, yet explicitly separable, interaction channel between expert users and semi-autonomous analytical processes, ensuring continuous alignment between user intent and AI behavior. The model addresses practical challenges in sensitive domains such as intelligence analysis, investigative journalism, and other fields handling complex, high-stakes data. We illustrate through detailed case studies how our model facilitates deeper understanding and targeted improvement of multimedia analytics solutions. By explicitly capturing how expert users can optimally interact with and guide AI-powered multimedia analytics systems, our conceptual framework sets a clear direction for system design, comparison, and future research.
... The goal is to achieve an intuitive and useful interpretation of the data rather than a cognitive or information overload when presenting only results [2,11]. Moreover, visual analytics always considers the task that the user wants to perform by means of the visualization, which ensures that the visualization provides the desired insights [11,13]. Through interactive capabilities, the user can see dynamics within the data, gain hidden insights, and adjust the complexity to their needs. ...
... Through interactive capabilities, the user can see dynamics within the data, gain hidden insights, and adjust the complexity to their needs. As a consequence, they can fulfill tasks more effectively and efficiently [9,13]. To systematically evaluate visualizations, Munzner [13] defines an analysis framework with three questions: 'Why', 'What', and 'How' (Fig. 2). ...
... As a consequence, they can fulfill tasks more effectively and efficiently [9,13]. To systematically evaluate visualizations, Munzner [13] defines an analysis framework with three questions: 'Why', 'What', and 'How' (Fig. 2). 'Why' analyzes the reasons for using the visualization (tool) and identifies the performed task. ...
Conformance checking is a sub-discipline of process mining, which compares process execution data with predefined process models to identify deviations between them. Although recognized as the most important feature of process mining tools, conformance checking is currently not widely applied in practice. One reason for this lack of adoption is the absence of process-mining-specific visualizations, which can effectively communicate conformance checking results to practitioners. Although researchers have identified the need for such visualizations, they have left their development to the tool providers, such that available visualizations are highly different and difficult to compare. This inhibits the opportunities to conduct empirical research on conformance checking visualizations, which would be crucial to understanding user preferences. To address this issue and establish a foundation for future empirical research, this paper provides an overview of the existing breadth of characteristics of conformance checking visualizations in the form of a taxonomy. This taxonomy consists of six dimensions, which highlight in a structured manner what information is displayed in conformance checking visualizations and how this is visualized in different academic and commercial tools. Our research enhances the comprehension of visual analytics in process mining, particularly for conformance checking, and highlights promising avenues for future empirical research.
... From over 119 visualizations published at the time of our comment analysis, 27 maps were identified, out of which we chose 13 maps, detailed in Table S1 of the supplementary material. These 13 maps represent four key map types, according to the classification proposed by Munzner [86]: divergent color maps, sequential color maps, categorical color maps, and proportional symbol maps [86,128]. For the 13 graphs relevant to our research, we retrieved the series' online comments in December 2022 using the Selenium crawler in a custom Python script. ...
... From over 119 visualizations published at the time of our comment analysis, 27 maps were identified, out of which we chose 13 maps, detailed in Table S1 of the supplementary material. These 13 maps represent four key map types, according to the classification proposed by Munzner [86]: divergent color maps, sequential color maps, categorical color maps, and proportional symbol maps [86,128]. For the 13 graphs relevant to our research, we retrieved the series' online comments in December 2022 using the Selenium crawler in a custom Python script. ...
... Color usage has been shown to be significant for visualization perception, as viewers are influenced by brightness, saturation, and hue choices [86]. In both the comment analysis and the interviews, viewers often referenced regions by colors and highlighted their semantic understanding of color hues, which have been shown to be crucial for first visualization impressions [29,48] and for cartography in general [4,13,30,128]. ...
Crisis maps are regarded as crucial tools in crisis communication, as demonstrated during the COVID-19 pandemic and climate change crises. However, there is limited understanding of how public audiences engage with these maps and extract essential information. Our study investigates the sensemaking of young, digitally native viewers as they interact with crisis maps. We integrate frameworks from the learning sciences and human-data interaction to explore sensemaking through two empirical studies: a thematic analysis of online comments from a New York Times series on graph comprehension, and interviews with 18 participants from German-speaking regions. Our analysis categorizes sensemaking activities into established clusters: inspecting, engaging with content, and placing, and introduces responding personally to capture the affective dimension. We identify friction points connected to these clusters, including struggles with color concepts, responses to missing context, lack of personal connection, and distrust, offering insights for improving crisis communication to public audiences.
... A reliable dataset-characterized by accuracy, validity, uniformity, completeness, and consistency-is essential for predictive accuracy. With technological advancements, visual approaches (e.g., histograms for continuous variables, bar charts for categorical data) now dominate data exploration, as visual patterns are more intuitive than numerical summaries (Munzner, 2014) [15] . Researchers must first develop a clear understanding of variable types (e.g., continuous vs. categorical, input vs. output) before proceeding to data exploration. ...
... A reliable dataset-characterized by accuracy, validity, uniformity, completeness, and consistency-is essential for predictive accuracy. With technological advancements, visual approaches (e.g., histograms for continuous variables, bar charts for categorical data) now dominate data exploration, as visual patterns are more intuitive than numerical summaries (Munzner, 2014) [15] . Researchers must first develop a clear understanding of variable types (e.g., continuous vs. categorical, input vs. output) before proceeding to data exploration. ...
This review paper highlights the evolution of data analysis techniques in service quality research, drawing on service quality studies from the SCOPUS database (2013-2022). Building on Khan & Adil (2013) review paper on data analysis technique, this paper extends the discourse by summarizing technique advancements and emerging trends. The analysis reveals a pronounced shift from traditional methods (e.g., descriptive statistics, regression analysis, t-tests) to advanced multivariate approaches, particularly structural equation modelling. Post-2019 trends highlight a shift toward PLS-SEM for testing model and also for testing complex mediation and moderation effects. Factor analysis (EFA and CFA) remains pivotal for validating service quality constructs. Importance-performance analysis has gained prominence as a valuable tool for evaluating key factors in service quality research. The emergence of data mining reflects a growing trend of interdisciplinary innovation. Despite the dominance of SEM-based techniques, there is persistent reliance on regression and ANOVA. By summarising decade-long trends, this study offers a roadmap for researchers to navigate methodological choices, emphasizing context-specific adaptability. This review advocates for emerging computational tools (softwares) in service quality research while retaining core psychometric principles. Future research should prioritize methodological transparency, reproducibility, and the integration of mixed-methods designs to bridge theoretical and practical gaps.
... Zhang and Koppaka (2007) used a compound network that places cases citing and cases cited by a case under analysis above and below the central node, delving into subnetworks focusing on a specific legal issue. Compound networks are multi-level networks where the original network is augmented with a derived cluster hierarchy (Munzner 2014). The time dimension was represented by the positioning of networks above and below the central node, representing cases decided before and after the case under analysis. ...
... Shneiderman and Aris's (2006) visuals were scatter plots distributed on non-overlapping regions where node placement was based on node attributes. Such regions are indicated by containment marks (Munzner 2014) encompassing network nodes. They used node labels, link labels, directed networks, node attributes (coloring, sizing, and grouping), and link attributes (coloring and thickness) to encode information. ...
... The HCID approach combines Norman's human-centered design (HCD) process [17] with Munzner's what-why-how visualization framework [15], offering a theory-informed method to design LA indicators systematically. The HCID process is composed of four iterative stages that support the design of the appropriate indicator: (1) Define Goal/Question, (2) Ideate, (3) Prototype, and (4) Test. ...
... An ISC aims to allow quick and low-cost design of low-fidelity LA indicators. It follows the Goal-Question-Indicator (GQI) approach [16] to design LA indicators that meet users' goals and applies information visualization (InfoVis) guidelines from Munzner's what-why-how visualization framework [15]. Concretely, it describes a systematic workflow from the why? (i.e., user goal/question) to the how? (i.e., visualization). ...
Emerging research on human-centered learning analytics (HCLA) has demonstrated the importance of involving diverse stakeholders in co-designing learning analytics (LA) systems. However, there is still a demand for effective and efficient methods to co-design LA dashboards and indicators. Indicator Specification Cards (ISCs) have been introduced recently to facilitate the systematic co-design of indicators by different LA stakeholders. In this paper, we strive to enhance the user experience and usefulness of the ISC-based indicator design process. Towards this end, we present the systematic design, implementation, and evaluation details of the ISC Creator, an interactive LA tool that allows low-cost and flexible design of LA indicators. Our findings demonstrate the importance of carefully considered interactivity and recommendations for orienting and supporting non-expert LA stakeholders to design custom LA indicators.
... However, like many complex systems, whether in physics, biology, or economics, deep structural patterns may not be obvious when considering only individual elements or local properties. Data visualization methods provide a powerful tool for revealing such hidden patterns and correlations [1,2]. ...
... Однако, подобно многим сложным системам, будь то в физике, биологии или экономике, глубинные структурные закономерности могут быть не очевидны при рассмотрении только отдельных элементов или локальных свойств. Методы визуализации данных предоставляют мощный инструмент для выявления таких скрытых паттернов и корреляций [1,2]. ...
Pythagorean triples (PTs), integer solutions to the equation a² + b² = c², are a classic object of number theory. Despite their simplicity, they possess a rich internal structure that often remains hidden under a purely algebraic approach. In this work, we present a study of primitive Pythagorean triples (PPTs) using methods of multidimensional data visualization and draw parallels with concepts from quantum mechanics. We generate sequences of PPTs and analyze the distribution of their properties, such as the values of the legs, the last digits of the hypotenuse in different number systems (base 10 and base 16), and the sequential generation index. Visualizations include 3D scatter plots, analysis of network structures based on common elements of the triples, distribution histograms ('density of states'), and projections onto geometric surfaces (sphere and cylinder). The observed patterns-discrete 'layers', 'allowed transitions' between triples , peaks in the 'density of states', and specific geometric distributions on the sphere and cylinder-demonstrate striking analogies with quantized energy levels, selection rules for transitions, density of states, and the spatial distribution of quantum states (e.g., atomic orbitals). Changing the number system visually resembles a change of observation basis in quantum theory. Although these analogies are heuristic in nature , they highlight deep structural patterns in the set of Pythagorean triples, revealed through visualization, and suggest the potential value of an interdisciplinary approach for studying numerical systems.
... To visualize property changes in language instances, we need to consider their visibility. Munzner's book [ 24] recommends automatic highlighting with varied colors, shapes, or positions can emphasize distinctions between properties. The "pop-out" effect in Munzner's book helps users spot differences quickly without focused attention. ...
... Like mentioned in Munzner's book [ 24], we aim to create a solution that addresses the invisibility of property changes and enhances user experience by implementing these highlighting strategies while considering potential conflicts with user-defined materials, overlapping color use, or making components transparent for highlighting. We opted against animations due to their potential to cause change blindness, distracting users from subtle property changes. ...
Cyber-Physical Systems are designed and developed using multi-disciplinary teams that require handovers from one discipline to another. These handovers often involve text documents written in nat-ural language, which can be imprecise, ambiguous, and lead to errors. To address this issue, we created a textual Domain Specific Language with immediate graphical feedback. This language allows mechanical and mechatronic engineers to communicate more effectively during handovers by providing a formalized system description that can be easily visu-alized in real-time. Our approach also incorporates multiple industry standards, which enables bi-directional navigation between languages, making it easier for teams to collaborate across different disciplines and prevent errors from being made. We applied and evaluated our approach in relation to medical robots at Philips IGT.
... Meanwhile, the Internet platform can realize the multidimensional presentation and rapid sharing of information (Liu et al., 2021). According to the theory of information visualization, when information that is difficult to be directly displayed is transformed into forms such as graphics and videos for presentation, it can greatly enhance the individual's reception degree of information and expand the sources of information (Munzner, 2014). Rural residents through using of the internet enrich their knowledge and vision, absorb fresh ideas, so as to improve the happiness of them (Zhang et al., 2024). ...
Digital village construction (DVC) is a new form of development that uses emerging digital technologies and brings change to rural residents. This paper delves into the influence and mechanism of DVC on rural resident happiness utilizing the ordered Probit model, drawing on data from the China Family Panel Studies (CFPS). Our findings indicate a marked enhancement in rural residents well-being as a consequence of DVC, and the results remain robust after replacing OLS regression, changing the measurement of explanatory and explained variables, adding other control variables, and using CMP estimation. Heterogeneity analysis shows that the enhancement effect is significant in male, middle-aged and highly educated groups, while the effect is insignificant for young groups, low and middle education groups. The mechanism tests show that the DVC can improve the rural residents’ happiness by increasing household income and reducing the loneliness. Government might amplify its investment and bolster digital infrastructure, enabling a greater number of rural residents to fully benefit from the development of a digital countryside.
... We examined the canvas photographs to categorize the different types of data visualizations based on their level of customization relative to the provided examples. The first author reviewed all images and classified the collaborative visualization according to modification of their visualization elements, including channels, marks, and data types [45]. We used the interview data to understand data patterns participants identified from these visualizations. ...
... We observed that the design of information visualization depends not only on the information presented but also on the way this information is transmitted. According to Munzner [30], the design must be focused on the user and aligned with the target audience's specific needs. Goal 2: Contextualized Metrics -metrics aligned with the three pillars of sustainability (presented in Section 2.2) support the evaluation of the efficiency and effectiveness of each idea. ...
... Using digital technology to publish typical cases related to human settlement governance as educational and publicity content can greatly increase farmers' opportunities to obtain relevant information online and enhance farmers' awareness of environmental protection and governance. According to information visualization theory, vivid and intuitive educational forms such as videos and pictures can help reduce the difficulty for farmers to understand educational content (Munzner, 2014), thereby improving farmers' cognitive level and willingness to participate. In addition, digital technology provides new ways for farmers to participate in voting, evaluating, and providing feedback on issues in human settlement governance, thereby effectively protecting farmers' rights to know, participate, supervise, and speak. ...
Digital technology has emerged as a significant factor influencing farmers’ participation in human settlements governance, and it shows obvious community heterogeneity. The Technology-Organization-Environment (TOE) theory illustrates that the performance of digital technology is not solely dependent on technical conditions, but also closely linked to the context in which it is implemented. This study incorporates digital technology and community capacity into a unified analytical framework. Utilizing data from a survey of 1,043 farmers, the Ordered Probit model was employed to empirically examine the influence of these two factors on farmers’ participation in human settlements governance. Additionally, the propensity score matching method (PSM), conditional mixed process method (CMP), and threshold effect model were utilized for robustness testing and further analysis. The results indicate that (1) Both digital technology and community capacity have significant positive effects on farmers’ participation, and the two have complementary effects. (2) The higher the community capacity (threshold value 0.682), the more obvious the positive impact of digital technology on farmers’ participation. (3) The impact of digital technology and community capacity on farmers’ participation is heterogeneous based on total income and regional distribution, showing that the two have significant impacts on the participation behavior of middle-income farmers and farmers in the central region respectively. Therefore, this paper proposes that it is crucial to harmonize the efficient utilization of digital technology with the strengthening of community capacity. Additionally, rural areas should adopt a tailored approach to managing their communities based on their unique characteristics.
... A visualização é adequada quando existe a necessidade de aumentar as capacidades humanas em vez de substituir as pessoas por métodos computacionais de tomada de decisão. O espaço de design de possíveis expressões de visualização é enorme, e inclui as considerações de como criar e como interagir com representações visuais (Munzner, 2014). ...
... To ensure that InsightMap effectively facilitates insight exploration, we surveyed existing research on insight mining [10,11,23,29], visualization recommendation [7,19,24,35,36], principles for visualization design [20,25] and exploratory search [2,15,31], and further compile a list of design considerations after multiple iterations of discussions and designs. The major considerations include: C1 Provide an overview of data distribution. ...
Automated data insight mining and visualization have been widely used in various business intelligence applications (e.g., market analysis and product promotion). However, automated insight mining techniques often output the same mining results to different analysts without considering their personal preferences, while interactive insight discovery requires significant manual effort. This paper fills the gap by integrating automated insight mining with interactive data visualization and striking a proper balance between them to facilitate insight discovery and exploration. Specifically, we regard data insights as a special type of data and further present InsightMap, a novel visualization approach that uses the map metaphor to provide a quick overview and in-depth exploration of different data insights, where a metric is proposed to measure the similarity between different insights. The effectiveness and usability of InsightMap are demonstrated through extensive case studies and in-depth user interviews.
... Meanwhile, virtual reality (VR) has proven to be a valid and effec-tive tool for performing spatial tasks, providing a better understanding of navigation mechanisms [3,11]. Furthermore, VR allows transient variations in scale [56] and data visualization from new perspectives, generating new interpretations [31,38]. This enhances visualization characteristics, helping users to identify volumetric samples more effectively than traditional computer visualizations [19]. ...
Navigating topographic charts involves nuanced skills and real-world correspondence, influenced by individual learning styles. Traditionally, 2D and 3D representations bridge map-reality gaps, but sandboxes often introduce manual errors and reliance on personal interpretation. This paper examines VR's potential in military planning, focusing on terrain interpretation, data visualization, and scale transitions. A study with 36 army cadets investigates VR's effectiveness in enhancing spatial perception and real-world task performance. Findings suggest that VR improves position choices and grades and reduces result disparities, particularly benefiting users with lower spatial skills. The research also evaluates the impact of different scales on VR planning, offering insights into potential advantages and challenges. Future studies should explore control issues and completion times in real-world scenarios.
... display size, mobility, and interaction modalities. Therefore, a range of computing environments beyond the traditional desktop environment [59] have been studied. For example, such as display walls [6], [8] and tabletops [21] for the larger display size, smartphones for situated analysis [45], and smartwatches [31] for on-body input capability. ...
We built a spatial hybrid system that combines a personal computer (PC) and virtual reality (VR) for visual sensemaking, addressing limitations in both environments. Although VR offers immense potential for interactive data visualization (e.g., large display space and spatial navigation), it can also present challenges such as imprecise interactions and user fatigue. At the same time, a PC offers precise and familiar interactions but has limited display space and interaction modality. Therefore, we iteratively designed a spatial hybrid system (PC+VR) to complement these two environments by enabling seamless switching between PC and VR environments. To evaluate the system's effectiveness and user experience, we compared it to using a single computing environment (i.e., PC-only and VR-only). Our study results (N=18) showed that spatial PC+VR could combine the benefits of both devices to outperform user preference for VR-only without a negative impact on performance from device switching overhead. Finally, we discussed future design implications.
This article presents alternative concepts for rendering thematic cartograms, using primitive shapes arranged under different concepts of spatial distribution, aiming to meet the criteria explained in the proposal in a more intuitive way. initially defended by Barreto, Kosminsky, and Esperança(2018). Graphic rendering algorithms, developed for the automatic generation of thematic cartograms according to the concepts presented here, and following the methodology of Speckmann and Verbeek (2010)’s work, are also presented.
From a Human-Computer Interaction perspective, data visualizations are visual representations of data that improve users' cognitive capabilities during a task. In particular, UX data visibility can raise a team's engagement with the UX design and better inform product decisions. However, researchers and professionals lack a foundation to build new UX data visualizations. In this context, this paper describes a Systematic Mapping of the Literature that aims to consolidate the state of the art on UX data visualizations. To guide the open coding of the findings, we defined ten questions that span the Visual Information Seeking Mantra and the four levels of Munzner's analysis framework. We identified 28 well-known and seven custom chart formats, with the node-link diagram arising as the most popular. Most of the visualized data comes from software logs, and there is a lack of exploration of UX metrics, acoustic data, and demographic data as data sources. Regarding the Visual Information Seeking Mantra, visualizations had a zoom and filter function and a details on demand function for most chart formats. However, most chart formats lacked overview functions. Our findings provide a broad overview of the literature that can support the creation of new UX data visualizations.
Thematic maps, like choropleth maps, symbol maps or cartograms, are commonly used to visualize spatial quantitative data. Many studies have been conducted to compare the different approaches and thus define the best strategies to produce suitable and efficient maps. When analyzing spatial data, it is also often necessary to visualize and compare several variables on the same map. Therefore, the question arises of how to best associate two variables in a single representation without one of them prevailing over the other, while avoiding overloading the map and making it difficult to interpret. In this article, we propose a comparison of five types of bivariate maps based on a user study. Participants performed a set of tasks using different maps produced from multiple datasets. Our analysis is based on three approaches: (1) quantitative analysis of user answer accuracy, (2) quantitative analysis of user answer times, and (3) quantitative and qualitative analysis of user feedback. The results suggest that combining symbol and choropleth maps is the most effective approach among those tested, while combining cartograms with any technique is the worst.
Prescriptive process monitoring methods recommend interventions during the execution of a process to maximize its success rate. Current research in this field focuses on algorithms to learn intervention policies that maximize the expected payoff of the interventions under certain statistical assumptions. In contrast, there has been limited attention on how to aid process stakeholders in understanding the outputs of these algorithms. In this research, we set to develop an interface to provide end users with relevant information to guide the decision on where and when to trigger interventions in a process. We draw upon an analysis of existing solutions and a review of the literature to elicit information items for a user interface for prescriptive process monitoring. Thereon, we develop a user interface concept and evaluate it with experts. The evaluation confirms the informational needs covered by the user interface concept. In addition, the evaluation shows that different end-user groups (operational users, tactical managers, and process analysts) can benefit from the information items included in the interface.
Adversarial machine learning (AML) studies attacks that can fool machine learning algorithms into generating incorrect outcomes as well as the defenses against worst-case attacks to strengthen model robustness. Specifically for image classification, it is challenging to understand adversarial attacks due to their use of subtle perturbations that are not human-interpretable, as well as the variability of attack impacts influenced by diverse methodologies, instance differences, and model architectures. Through a design study with AML learners and teachers, we introduce AdvEx , a multi-level interactive visualization system that comprehensively presents the properties and impacts of evasion attacks on different image classifiers for novice AML learners. We quantitatively and qualitatively assessed AdvEx in a two-part evaluation including user studies and expert interviews. Our results show that AdvEx is not only highly effective as a visualization tool for understanding AML mechanisms, but also provides an engaging and enjoyable learning experience, thus demonstrating its overall benefits for AML learners.
The use of machine learning in decision-making has become increasingly pervasive across various fields, from healthcare to finance, enabling systems to learn from data and improve their performance over time. The transformative impact of these new technologies warrants several considerations that demand the development of modern solutions through responsible artificial intelligence—the incorporation of ethical principles into the creation and deployment of AI systems. Fairness is one such principle, ensuring that machine learning algorithms do not produce biased outcomes or discriminate against any group of the population with respect to sensitive attributes, such as race or gender. In this context, visualization techniques can help identify data imbalances and disparities in model performance across different demographic groups. However, there is a lack of guidance towards clear and effective representations that support entry-level users in fairness analysis, particularly when considering that the approaches to fairness visualization can vary significantly. In this regard, the goal of this work is to present a comprehensive analysis of current tools directed at visualizing and examining group fairness in machine learning, with a focus on both data and binary classification model outcomes. These visualization tools are reviewed and discussed, concluding with the proposition of a focused set of visualization guidelines directed towards improving the comprehensibility of fairness visualizations.
Narrative visualization is a method for conveying data or information in an engaging storytelling form using interactive visualizations. This method integrates storytelling and visualization techniques to represent information through organized narratives for effective communication. Narrative visualization has gained increasing interest in many data-driven domains like business, healthcare, governance, and education. However, the literature indicates limited insights and organization regarding advances in narrative visualization, particularly in its design elements and methodologies. This survey paper aims to present a comprehensive overview of the taxonomy, methodologies, challenges, and emerging opportunities in recent narrative visualization research. The paper firstly introduces the fundamental concepts of narrative visualization and its application trends across domains. Emerging trends and taxonomy were derived from the survey to provide an overview of advances in narrative visualization. The derived taxonomy classifies narrative visualization works with regards to the types of design, genre, narrative, and enabler technologies. Next, summarizations of common methodologies adopted by recent works particularly on design requirement identification and design evaluation are presented. Subsequently, major challenges addressed by recent works are classified based on three aspects: application, performance, and technical. Lastly, this survey highlights the emerging opportunities in narrative visualizations for future undertaking or exploration by the visualization community. The findings presented in this paper can offer significant advantages for both visualization practitioners and researchers in generating effective narrative visualization solutions.
Background
Recent technological advances in wearable devices offer new potential for measuring mobility in real-world contexts. Mobilise-D has validated digital mobility outcomes to provide novel outcomes and end points in clinical research of 4 different long-term health conditions (Parkinson disease, multiple sclerosis, chronic obstructive pulmonary disease, and proximal femoral fracture). These outcomes also provide unique information that is important to patients; however, there is limited literature that explores the optimal methods to achieve this, such as the best way to visualize patients’ data.
Objective
This study aimed to identify meaningful outcomes for each condition and how to best visualize them from the perspective of end users.
Methods
Using a Delphi-type protocol with patients as subject matter experts, we gathered iterative feedback on the cocreation of visualizations through 3 rounds of questionnaires. An open-ended questionnaire was used in round 1 to understand what aspects of mobility were most influenced by their health condition. These responses were mapped onto relevant digital mobility outcomes and walking experiences and then prioritized for visualization. Using patient responses, we worked alongside researchers, clinicians, and a patient advisory group to develop visualizations that depicted a week of mobility data. During rounds 2 and 3, participants rated usefulness and ease of understanding on a 5-point Likert scale and provided unstructured feedback in comment boxes for each visualization. Visualizations were refined using the feedback from round 2 before receiving further feedback in round 3.
Results
Participation varied across rounds 1 to 3 (n=48, n=79, and n=78, respectively). Round 1 identified important outcomes and contexts for each health condition, such as walking speed and stride length for people with Parkinson disease or multiple sclerosis and number of steps for people with chronic obstructive pulmonary disease or proximal femoral fracture. The consensus was not reached for any visualization reviewed in round 2 or 3. Feedback was generally positive, and some participants reported that they were able to understand the visualization and interpret what the visualization represented.
Conclusions
Through the feedback provided and existing data visualization principles, we developed recommendations for future visualizations of mobility- and health-related data. Visualizations should be readable by ensuring that large and clear fonts are used and should be friendly for people with vision impairments, such as color blindness. Patients have a strong understanding of their own condition and its variability; hence, adding additional factors into visualizations is recommended to better reflect the nuances of a condition. Ensuring that outcomes and visualizations are meaningful requires close collaboration with patients throughout the development process.
We investigate the perception of visual variables on wall-sized tiled displays within an immersive environment. We designed and conducted two formal user studies focusing on elementary visualization reading tasks in VR. The first study compared three different virtual display arrangements (Flat, Cylinder, and Cockpit). It showed that participants made smaller errors on virtual curved walls (Cylinder and Cockpit) compared to Flat. Following that, we compared the results with those from a previous study conducted in a real-world setting. The comparative analysis showed that virtual curved walls resulted in smaller errors than the real-world flat wall display, but with longer task completion time. The second study evaluated the impact of four 3D user interaction techniques (Selection, Walking, Steering, and Teleportation) on performing the elementary task on the virtual Flat wall display. The results confirmed that interaction techniques further improved task performance. Finally, we discuss the limitations and future work.
ResearchGate has not been able to resolve any references for this publication.