Uncertainty in Science is Malleable. Advocating for User-Agency in Defining Uncertainty in Visualizations: a Case Study in Geology

Research Background and Issues

• Problems or Challenges Identified by the Authors:

  1. Uncertainty is an inherent element of scientific research, particularly when studying natural phenomena and data requiring subjective interpretation (e.g., volcanic deposit research in geology). Not only is assessing uncertainty itself challenging, but defining uncertainty also poses a significant issue.
  2. Traditional visualization systems often assume that uncertainty can be objectively measured and predefined, neglecting the evolving nature of uncertainty definitions as domain knowledge develops over time.

• Significance of the Problem:

  1. Uncertainty is critical to the construction and validation of scientific knowledge. In volcanic deposit research, errors or uncertainties in sample attribution can affect the analysis of nearby volcanic eruption histories, thereby compromising the reliability of natural hazard mitigation.
  2. The lack of adequate support for complex and dynamic uncertainty factors limits the efficiency and accuracy of scientific research.

• Research Motivation and Related Work:

  1. Previous work on uncertainty visualization has focused on representing direct (quantifiable) and indirect (subjective or implicit) uncertainties, but little attention has been paid to how users dynamically construct and modify uncertainty definitions.
  2. Traditional design approaches have failed to account for the dynamic nature of introducing new uncertainty factors based on disciplinary context and the needs of domain experts.

Proposed Solution

• Method or Solution Proposed by the Authors: The authors propose a dynamic visualization design approach for uncertainty that emphasizes granting users the ability to construct uncertainty definitions (user agency) and supports the dynamic characteristics of uncertainty through a flexible and adaptive system. The approach involves two steps:

  1. Co-design with geologists (tephra research experts) to explore the primary sources of uncertainty in sample attribution and develop an initial design prototype based on existing visualization systems.
  2. Validate and expand the initial design through technical probes and interactions with a broader group of domain experts, achieving new iterations through a series of interactive validations.

• Innovative Aspects of the Solution:

  1. Highlights the "malleability" of uncertainty, meaning that uncertainty definitions dynamically change with shifts in domain knowledge, research objectives, and data context.
  2. Advocates for visualization systems that not only display uncertainty but also enable users to define and manipulate dimensions of uncertainty.
  3. Treats uncertainty not as a static logical structure but as a resource for driving scientific dialogue and collaboration.

• Implementation Steps and Key Techniques:

  1. First Design Iteration (User Immersive Co-Design):
     • Identify key sources of uncertainty considered by domain experts: geographic proximity, geochemical similarity, and sample quantity.
     • Design multiple visualization prototypes (e.g., heatmaps, transparent overlays, and regression line distance visualizations) and test their applicability.
  2. Second Design Iteration (Redesign for Technical Probes):
     • Integrate additional uncertainty factors such as "data provenance" and "analysis criteria" based on the initial foundation.
     • Introduce user-adjustable uncertainty mappings and filters.
     • Employ more intuitive annotations (e.g., encoding uncertainty through marker size) to reduce complexity and ensure broader user acceptance.

Research Outcomes

• Specific Outcomes:

  1. Developed a new system prototype for customizing and dynamically defining uncertainty, integrating various sources of uncertainty while simplifying visualization complexity.
  2. Identified 61 specific sources of uncertainty and categorized them into three major groups: data-related factors, subjective interpretation factors, and other contextual non-data sources.

• Advantages Over Existing Solutions:

  1. Provides greater user flexibility, allowing users to dynamically adjust or define new sources of uncertainty.
  2. Transforms the system from a traditional knowledge retrieval tool into a bridge for fostering new research questions and promoting scientific collaboration.
  3. Better accommodates the evolving nature of knowledge over time, supporting the reevaluation of historical results.

• Experimental or Evaluation Results:

  1. Through expert interviews and technical probe validation, the system was deemed effective and valuable in visualizing multiple sources of uncertainty.
  2. Most participants appreciated the flexibility of filtering and uncertainty mapping but suggested improvements for some quantitative representation methods.

• Limitations and Future Directions:

  1. Dataset scale and regionalization issues: The current tool primarily targets the southern Andes region, and more experienced users may require additional data support for specific regions.
  2. Conservative representation: To align with users' familiarity with visualization formats, the system has been cautious in adopting innovative annotations and dynamic representations.
  3. Future research should explore how to support cross-disciplinary or cross-domain sharing and collaboration in defining uncertainty, as well as expand support for the dynamic evolution of uncertainty.

Through this study, the authors not only reveal the fluidity and complexity of uncertainty in the field of geology but also provide practical design principles and application prospects for building dynamic visualization systems better suited to scientific practice. This lays the foundation for future exploration in areas such as uncertainty definition, scientific collaboration, and user customization.