Molecules to Humankind: Challenges in Multiscale Visualization

Panel website

Organizers

• Andreas Bueckle, Indiana University
• Christiane V. R. Hütter, University of Vienna
• Jörg Menche, University of Vienna
• Katy Börner, Indiana University and Canadian Institute for Advanced Research (CIFAR)

Panelists

• Alyssa Goodmann, Harvard University and Smithsonian Institution
• G. Elisabeta (Liz) Marai, University of Illinois Chicago
• Angus G. Forbes, NVIDIA
• Han-Wei Shen, The Ohio State University
• Gaël McGill, Harvard Medical School and Digizyme, Inc.

Life's complexity, from molecular interactions to the human body, spans over 10 orders of magnitude, a fundamental challenge for data visualization. Our panel, “Molecules to Humankind: Challenges in Multiscale Visualization”, brings together five experts across astrophysics, computer science, visualization, graphics, and biology to debate the state of the art and identify productive areas of disagreement in this emerging field. While multiscale reasoning appears across many visualization domains, the IEEE VIS community has not yet had a dedicated venue to synthesize common challenges, design principles, and open research questions around multiscale visualization. The panel centers on foundational design tensions. When is bridging scales necessary, and when are explicit breaks preferable to reduce cognitive load? Should multiscale data be integrated into a single coherent environment or connected through specialized views optimized for different data types and tasks? How should visual encodings, interaction techniques, and transitions evolve across scales while staying coherent, interpretable, and scientifically accurate? We will examine competing design philosophies using examples from the panelists' research, which spans interactive molecular visualization, immersive analytics, scientific visualization, and astrophysics, to identify transferable principles as well as unresolved challenges. Rather than seeking consensus, the discussion will emphasize trade-offs between alternative approaches and highlight open questions for the wider VIS community. To encourage active participation, attendees will contribute to a collaborative multiscale basemap by marking the spatial scales of their own research, which will create a shared visualization of the community's expertise that serves as a catalyst for discussion, reflection, and future collaborations.

Visualization as a Foundational Act of Placemaking: Future Places and Alien Worlds

Organizers

• Helen-Nicole Kostis, NASA Scientific Visualization Studio

Panelists

• Rachel Connolly, MIT
• Ryan Watt, American Museum of Natural History
• Gabriela Bila Advincula, MIT Media Lab
• Karthik Kashinath, NVIDIA
• Lee Bingham, NASA Exploration Systems Simulations
• Kwan-Liu Ma, University of California Davis
• Narges Mahyar, City St George's, University of London

Placemaking is the process by which spaces acquire meaning, identity, and human belonging. This panel examines how visualization participates in that process when places are not yet fully formed through lived experience. While placemaking has traditionally depended on accumulated lived experience through movement, memory, and storytelling that transform coordinates into meaningful places, visualization has typically supported this process by helping communities understand and communicate about already inhabited environments. This relationship is now shifting. Future places on Earth such as coastlines reshaped by sea level rise, cities transformed by climate change, and ecosystems in transition remain grounded in human experience but are entering conditions not yet collectively lived or remembered. In contrast, worlds beyond Earth including lunar habitats, Martian surfaces, and off world settlements have only limited and episodic human presence and no sustained accumulated human experience. Yet decisions about both domains are already underway, requiring publics and decision makers to engage with places that are not yet experientially real. We argue that in these contexts visualization must do much of the work of placemaking itself, constructing scale, orientation, emotional resonance, narrative, and trust directly from data. This opens new opportunities at the intersection of visualization, AI, and HCI, including generative methods, new data modalities, and interaction techniques for storytelling and sensemaking, while raising challenges around agency, narrative, and trust. This panel explores visualization not as representation of place, but as a foundational act through which places are brought into being. Focusing on emerging Earth and space futures, it asks how meaning, orientation, and belonging are constructed when place is assembled rather than inherited through experience. The panel builds on recent IEEE VIS discussions and extends them toward a question not yet addressed in visualization research, visualization as a founding act of placemaking in the absence of lived experience.

Industry Meets Academia: Data and Visual Analytics in the Agentic Era

Organizers

• Yuan "Charles" Cui, Microsoft
• Gustavo Soares, Microsoft
• Fumeng Yang, University of Maryland College Park

Panelists

• Nicole Sultanum, Tableau Research
• Fritz Lekschas, Ridge AI
• Zahra Ashktorab, Microsoft
• Lane Harrison, Worcester Polytechnic Institute
• Dominik Moritz, Carnegie Mellon University

Agentic AI is reshaping how data and visual analytics products are designed, built, used, and evaluated. Major visual analytics software (e.g., Microsoft Excel, Tableau) and a new wave of AI-native startups are introducing natural-language interfaces and LLM-based autonomous agents, opening a rich design and research space that the VIS community is uniquely positioned to shape. This 90-minute panel brings together three industry practitioners building agentic visual analytics products (at Tableau Research, Microsoft, and the startup Ridge AI) and two professors in academia to discuss: (1) what industry is building in the agentic era of data and visual analytics, (2) what research the VIS community can contribute that industry most needs, and (3) how should universities train VIS students to be impactful in building real-world data and visual analytics solutions.

Friction in Visualization: When Should We Slow People Down?

Panel website

Organizers

• Katerina Batziakoudi, Inria Université Paris-Saclay
• Anne-Flore Cabouat, Inria Université Paris-Saclay
• Lane Harrison, Worcester Polytechnic Institute
• Carolina Nobre, University of Toronto
• Filip Sadlo, Heidelberg University
• Antonia Schlieder, Heidelberg University
• Karen Schloss, University of Wisconsin–Madison
• Michael Wybrow, Monash University

Panelists

• Menna El-Assady, ETH Zurich
• Carolina Nobre, University of Toronto
• Arnaud Prouzeau, Inria
• Stacy Shaw, Worcester Polytechnic Institute
• Florian Windhager, Donau-Universität Krems

Visualization research has traditionally sought to reduce barriers between data and insight through more efficient representations, interactions, and analytical workflows. Yet reducing effort might not always be beneficial. Across areas such as uncertainty communication, sensemaking, learning, critical visualization, and human–AI interaction, researchers are increasingly examining situations in which friction can be leveraged instead: slowing users down, encouraging reflection, or introducing productive effort may improve understanding and decision-making. Although these discussions often employ different terminology, they raise a common question: when is friction beneficial in visualization? This panel brings together researchers from diverse backgrounds to discuss the role of friction in designing, evaluating, and teaching visualization. By connecting currently fragmented conversations, we aim to clarify key concepts, identify productive tensions, and explore whether friction can serve as a useful lens for future visualization research.

From Correlation to Causality: Future Directions in Visual Causal Inference

Panel website

Organizers

• Arran Zeyu Wang, University of North Carolina-Chapel Hill
• David Borland, RENCI and University of North Carolina-Chapel Hill
• David Gotz, University of North Carolina-Chapel Hill

Panelists

• Cindy Xiong Bearfield, Georgia Institute of Technology
• David Gotz, University of North Carolina-Chapel Hill
• Alex Kale, University of Chicago
• Bum Chul Kwon, IBM Research
• Klaus Muller, Stony Brook University

Visualizations excel at revealing correlational patterns, yet human viewers frequently and naturally draw causal conclusions from these representations despite the well-known maxim that correlation does not imply causation. This persistent tension has catalyzed a rapidly growing Visual Causal Inference (VCI) research community, with 35 full papers published in major visualization venues over the past five years—including a VIS best paper award and three VIS honorable mentions. However, the field faces significant fragmentation: current VCI methods, tools, and experiments largely operate in isolation, lacking clear connections across research threads and a unified theoretical foundation. It remains unclear to build a theory of VCI in exploratory data analysis. Critically, the field lacks standardized evaluation benchmarks and metrics for assessing the quality of causal inferences made through visual analytics. This VCI panel brings together leading researchers with diverse expertise in causal models, causal reasoning, cognitive science, generative AI, and agentic systems, and aims to discuss four core objectives: developing actionable frameworks for integrating causal inference into visual analytics, establishing empirically-rigorous foundations grounded in how people infer causality from visualizations, identifying opportunities to transform EDA tool design through VCI principles, and inspiring interdisciplinary research directions. This panel will promote the development of a coherent theory for VCI, one that relates when, where, why, and how causal inferences are made in EDA, enabling more robust reasoning and decision-making through visualization.

Facts & All the Feels: Data Visualization Through the Affective Lens

Panel website

Organizers

• Roxanne Ziman, University of Bergen
• Ester Scheck, TU Wien
• Kajetan Enge, Independent Researcher
• Jakob Kusnick, University of Bergen

Panelists

• Jonathan Schwabisch, PolicyViz
• Narges Mahyar, City St. George's, University of London
• Laura Koesten
• Crystal Lee, MIT
• Benjamin Bach

Although data visualization has long been associated with objectivity and neutrality, we are reckoning with the impact of design choices on how data is not only understood, but also how it is felt. In particular, visualizations used in high-stakes public advocacy contexts (e.g., climate change, public health, social-civic issues, and science communication) are often designed to emotionally engage as well as to inform. Designers must strike a balance between these goals, as they often also shape attitudes and behavioural intentions. Affective visualization research elucidates the role of emotions, individual characteristics, and sociocultural contexts in how information is processed and acted upon. The rhetorical nature of visualization design and subjective audience experience requires us to examine ethical questions that go beyond merely avoiding the design of misleading visualizations; we must also consider how the data visualization pipeline, from data collection and processing to design and dissemination, carry emotional weight and influence trust and receptivity amongst diverse audiences. If visualizations are meant to engage and inform public audiences, what does responsible engagement actually look like? When is it appropriate or not to design for emotion in visualization, and to what end? What responsibility do visualization designers and media platforms have when visualizations circulate through our complex media ecosystem to influence public discourse? This panel will draw the visualization community into discussion about how we should account for, and can more effectively measure, the impact of emotional techniques applied in visualizations for public communication, with the aim of guiding us toward responsible practices

Beyond the Last Mile: What Should Visualization Own in the Age of Data Systems and AI Agents?

Organizers

• Dylan Wootton, MIT
• Vidya Setlur, University of Michigan and Tableau Research

Panelists

• Vidya Setlur, University of Michigan and Tableau Research
• Leilani Battle, University of Washington
• Evan Peck, University of Colorado Boulder
• Eugene Wu, Columbia University

This panel asks what role visualization should play when analytical work increasingly extends beyond the chart. While visualization research has traditionally excelled at representation, perception, and interaction, many of today's analytical bottlenecks occur upstream: interpreting data semantics, assessing trust and uncertainty, reasoning about provenance, validating transformations, and supervising automated analytical work. As database, AI, and systems communities invest heavily in these problems, the panel examines whether visualization research is under-recognized, under-positioned, or insufficiently engaged in shaping this emerging landscape. Bringing together researchers from visualization, databases, HCI, and commercial analytics, the discussion will debate whether VIS should reframe its contributions around analytical reasoning and knowledge construction, strengthen its connections to adjacent communities, or expand its scope to encompass the broader human-centered infrastructure of data analysis.

Challenges and Opportunities for Evaluating AI-Assisted Generation of Visualizations

Organizers

• Devin Lange, Purdue University
• Astrid van den Brandt, Harvard Medical School
• Huyen N. Nguyen, Harvard Medical School
• Nils Gehlenborg, Harvard Medical School

Panelists

• Alexander Lex, Graz University of Technology
• Anamaria Crisan, University of Waterloo
• Fritz Lekschas, Ridge AI
• Niklas Elmqvisst, Aarhus University

Evaluating which visualization is correct for a given situation is a longstanding question in the visualization research community, and even claiming that there are correct and incorrect visualizations is debated. With the advancing capabilities of large language models, visualizations are now being constructed, on demand, in response to natural language requests. But how do we know whether such systems produce acceptable visualizations? In this panel, we will probe this high-level question and explore its implications for the visualization research community. We will explore different aspects that make this a challenging and relevant question: scale, plurality, technical ecosystems, generalizability, and trust. These dimensions provide structure to the wicked problem of evaluation. Through this discussion we will highlight the immediate practical challenges this question is surfacing and will discuss the epistemological underpinnings of benchmarking and visualization research.