Challenges in Synchronous & Remote Collaboration Around Visualization

This paper presents a framework of 16 challenges faced in synchronous and remote collaborative visualization, derived from the insights of 29 international experts across five key collaborative activities and organized to guide future research in technological choices, social factors, AI assistance, and evaluation.

The Gist

Imagine you and a group of friends are trying to solve a giant, complex jigsaw puzzle together. In the old days, you'd all sit around a big table, pointing at pieces, moving them around, and laughing together. That's co-located collaboration.

Now, imagine that same group is scattered across the globe. You're on a video call, sharing your screen, but you can't touch the puzzle pieces. You can only point at them with a laser pointer on a screen, and sometimes the internet lags, or your friend has a tiny phone while you have a giant VR headset. That's remote, synchronous collaboration around visualization.

This paper, written by a massive team of 29 experts from around the world, is essentially a "Survival Guide" for this new way of working. They've identified 16 major headaches (challenges) we face when trying to work together on data and charts from different locations, especially with new tech like AI and Virtual Reality (VR) entering the mix.

Here is the breakdown of their findings, translated into everyday language:

The Five Main "Games" We Play

The authors say we usually do five specific types of things with data. Think of these as the different "games" we are trying to play together:

1. Detective Work (Exploratory Analysis): Trying to figure out what the data is telling us.
2. Brainstorming (Divergent Ideation): Creating new ideas or designs together.
3. Storytelling (Presentations): Showing a story to an audience using charts.
4. Voting & Deciding (Joint Decision Making): Using data to make a hard choice together.
5. Watching the Dashboard (Real-time Monitoring): Keeping an eye on live data, like a stock market ticker or weather radar.

The Four Big Categories of Headaches

The paper groups the 16 challenges into four buckets. Here is what they mean, using some creative metaphors:

1. The Tech Toolbox (Technological Challenges)

• The Problem: We are trying to use tools built for different purposes to do a new job. It's like trying to fix a car engine using a Swiss Army knife.
• The "Looking Back" Challenge: We have old tools designed for people sitting in the same room. Do they work when everyone is in different time zones?
• The "Looking Outward" Challenge: We need to borrow ideas from video games and VR. If you put on a VR headset, you feel like you're in the same room. But can we do that for complex data without making people dizzy or confused?
• The "Asymmetry" Challenge: Imagine you are wearing a $3,000 VR headset, but your teammate is on a $200 laptop. The game needs to work for both of you, or the person on the laptop feels left out.
• The "Transferability" Challenge: Just because a tool works for doctors looking at X-rays doesn't mean it works for marketers looking at sales charts. We need to figure out what tricks can be copied from one field to another.

2. The Human Element (Social Challenges)

• The Problem: Technology is easy; people are hard.
• The "Scaling" Challenge: It's easy to collaborate with 3 people. What happens when you have 300? It's like trying to have a conversation at a dinner table vs. a stadium concert. The rules change completely.
• The "Dynamic Roles" Challenge: In a meeting, you might be the boss, the note-taker, the skeptic, and the presenter all in one hour. The tools need to let you switch hats instantly without breaking the flow.
• The "Trust & Agency" Challenge: If I can't see what you're doing, do I trust you? If I feel like I have no control over the shared screen, I stop caring. We need to make everyone feel like they own the puzzle, not just the person holding the mouse.
• The "Accessibility" Challenge: If the data is only visual, what about the person who is blind? If it's only audio, what about the person who is deaf? We need to build bridges so everyone can play the game, regardless of their abilities.

3. The Robot Helper (AI Challenges)

• The Problem: AI is joining the party, but it's a bit of a "black box."
• The "Paradigm" Challenge: Right now, we treat AI like a vending machine (you ask, it gives). But in a team, the AI should be a teammate. It should know when to speak up, when to stay quiet, and how to read the room.
• The "Provenance" Challenge: If the AI draws a chart, how do we know why it drew it that way? We need a "receipt" that shows the AI's thought process, or we won't trust its work.
• The "Reliability" Challenge: AI makes mistakes and can be biased. If the AI suggests a bad decision, how do we catch it before we ruin the project? We need to know when to trust the robot and when to ignore it.
• The "Privacy" Challenge: To make the AI helpful, it needs to know a lot about us (our habits, our data). How do we get that help without the AI spying on us or stealing our secrets?

4. The Report Card (Evaluation Challenges)

• The Problem: How do we know if our new tools actually work?
• The "Scope" Challenge: Most tests are done in a lab with 4 students. But the real world is messy, with 50 people, bad internet, and high stakes. We need to test in the wild, not just in a cage.
• The "Questions" Challenge: We need to ask better questions. Instead of "Did they finish the task fast?", we should ask "Did they feel heard? Did they trust each other?"
• The "Logistics" Challenge: Recording a meeting where everyone is in different countries, using different devices, and talking at once is a logistical nightmare. We need better ways to record and study these moments without being creepy.
• The "Analysis" Challenge: We will collect so much data (video, audio, clicks, eye movements) that it will be overwhelming. We need new ways to make sense of this mountain of information.

The Big Takeaway

The authors aren't just complaining; they are handing out a map for the future.

They are saying: "We have amazing new tech (VR, AI), but our tools for working together on data are still stuck in the past. We need to build systems that feel like we are sitting at the same table, even when we are oceans apart. We need to make sure everyone can play, trust the robot helpers, and actually make good decisions together."

It's a call to action for designers, engineers, and researchers to stop patching together old tools and start building a new, inclusive, and intelligent way for humans to collaborate on the data that shapes our world.

Technical Summary

Here is a detailed technical summary of the paper "Challenges in Synchronous & Remote Collaboration Around Visualization" by Brehmer et al. (CHI '26).

1. Problem Statement

The paper addresses the critical gap between the rapid adoption of remote and synchronous work (accelerated by the pandemic) and the current limitations of collaboration tools in supporting complex data visualization and analytics.

• Current Limitations: Existing teleconference tools (e.g., Zoom, Teams) and productivity suites function as a "patchwork" of disjointed applications. They lack deep interoperability with analytical artifacts (dashboards, notebooks, 3D models), leading to fragmented workflows, cognitive overload, and inefficiencies.
• Loss of Nuance: Remote settings struggle to replicate the subtle interpersonal cues, non-verbal communication, and shared situational awareness found in co-located collaborative analysis.
• Emerging Disruptions: The landscape is further complicated by the rise of Extended Reality (XR) and Artificial Intelligence (AI). While XR offers immersive potential and AI offers automation, their integration into collaborative visualization workflows is nascent, creating new challenges regarding trust, privacy, and interaction paradigms.
• Research Gap: Prior literature on collaborative visualization largely focuses on asynchronous collaboration or co-located synchronous work, predating the current convergence of remote work, social XR, and generative AI.

2. Methodology

The authors employed a consensus-building methodology involving a large, interdisciplinary group of experts to characterize challenges and define a research agenda.

• Participants: 29 experts from 19 institutions across 9 countries (North America, Europe, Asia, Australia). The group spans Human-Computer Interaction (HCI), Visualization, CSCW, XR, and AI sub-communities, including both academic researchers and industry practitioners.
• Process:
  1. NII Shonan Seminar (June 2024): A 4-day seminar with 12 sessions involving breakout groups and plenary discussions.
  2. Working Sessions: Weekly virtual meetings (Dec 2024 – Mar 2025) to refine themes, reconcile top-down (challenge enumeration) and bottom-up (literature review) perspectives, and reach consensus.
  3. Artifacts: The process generated digital/physical whiteboards, annotated bibliographies, and meeting recordings, which were synthesized into the final challenge framework.
• Framework: The challenges are anchored in five core collaborative activities involving visualization:
  1. ☼ Exploratory Data Analysis: Understanding patterns in data.
  2. 👉 Divergent Ideation: Collaborative generation of new ideas/data.
  3. × Presentations: Data-rich storytelling to an audience.
  4. ± Joint Decision Making: Negotiating actions based on data.
  5. g Real-time Data Monitoring: Responding to dynamic, changing data streams.

3. Key Contributions

The primary contribution is a structured set of 16 actionable challenges organized into four thematic pillars. These serve as a roadmap for researchers and practitioners.

Theme 1: Technological Challenges (Sections 5.1–5.4)

Focuses on the viability of interface technologies and techniques.

• Looking Back: Re-evaluating legacy collaborative visualization techniques (e.g., Sense.us) for modern synchronous/remote contexts.
• Looking Outward: Assessing commercial tools and CSCW techniques (e.g., XR, augmented video) for their ability to support visualization-specific needs.
• Asymmetries: Addressing disparities in devices, display capabilities, and connectivity between collaborators (e.g., HMD vs. desktop) and designing for heterogeneous environments.
• Transferability: Determining which techniques are domain-specific versus those that can be generalized across different data types and activities.

Theme 2: Social Challenges (Sections 6.1–6.4)

Focuses on human dynamics, roles, and inclusivity.

• Scaling: Adapting techniques for varying group sizes (from pairs to thousands) and managing the shift from small-group analysis to large-scale presentations.
• Dynamic Roles: Supporting fluid transitions between roles (presenter, analyst, decision-maker) and managing asymmetries in expertise and authority.
• Agency & Trust: Ensuring all participants feel a sense of ownership and trust in shared data artifacts, particularly when roles or device capabilities differ.
• Accessibility: Creating inclusive designs for users with visual, auditory, or cognitive impairments, ensuring multimodal interaction is accessible to all.

Theme 3: AI Assistance Challenges (Sections 7.1–7.4)

Focuses on integrating AI agents as collaborators, mediators, or assistants.

• Paradigms: Moving beyond command-and-response to proactive collaboration (e.g., AI as a co-analyst or moderator) and defining appropriate interaction models.
• Provenance: Capturing and visualizing the history of interactions (human and AI) to maintain continuity and understand the semantics of shared actions.
• Reliability: Establishing trust in AI outputs, managing expectations of LLM capabilities, and mitigating risks of bias and hallucination in real-time collaboration.
• Privacy: Balancing the data collection required for personalization (e.g., context-aware AI) with strict privacy and security requirements, especially in sensitive domains.

Theme 4: Evaluation Challenges (Sections 8.1–8.4)

Focuses on how to scientifically study these complex systems.

• Scope: Expanding evaluation beyond individual task performance to include group dynamics, long-term ethnographic studies, and ecological validity.
• Questions & Metrics: Developing research questions and metrics that capture both individual and group experiences, moving beyond standard usability tests.
• Logistics: Overcoming the difficulties of instrumenting remote, multi-device, and multi-modal sessions for data collection.
• Analysis: Synthesizing rich, multimodal data (video, logs, speech, gaze) to identify patterns in group dynamics without being overwhelmed by data volume.

4. Results (Synthesized Findings)

The paper does not present experimental data but rather a synthesis of expert consensus resulting in a prioritized research agenda:

• Interoperability is Critical: Current tools fail because they treat visualization as a static image rather than an interactive object. Future systems must allow deep, multi-user interaction with analytics artifacts.
• Asymmetry is Inevitable: Designers must assume collaborators will use different devices (XR vs. 2D screens) and have different levels of authority; systems must support "selective asymmetric visibility" rather than forcing uniformity.
• AI Requires New Paradigms: AI cannot simply be a chatbot; it must be an embodied or multimodal agent that understands the context of the visualization, manages provenance, and operates with transparency regarding its reliability.
• Evaluation Needs Evolution: Traditional controlled experiments are insufficient. The community must adopt mixed-method approaches, including longitudinal studies and field observations, to capture the complexity of remote collaborative sensemaking.

5. Significance

• Community Unification: This work bridges the gap between the Visualization community (focused on data representation) and the CSCW community (focused on group dynamics), creating a shared vocabulary for remote data work.
• Strategic Roadmap: It provides a concrete, organized framework (the 16 challenges) for funding agencies, industrial partners, and researchers to direct resources toward high-impact problems.
• Future-Proofing: By explicitly addressing XR and AI, the paper anticipates the next decade of collaborative technology, moving beyond the limitations of current "Zoom fatigue" tools.
• Societal Impact: The authors emphasize that effective remote data collaboration is essential for global resilience in fields like public health, climate science, and urban planning, where decisions must be made collectively across borders using data.

In summary, the paper serves as a manifesto and research agenda for the next generation of collaborative visualization systems, arguing that technical innovation must be matched by social design, ethical AI integration, and rigorous, context-aware evaluation.