The Impact of Uncertainty Visualization on Trust in Thematic Maps

Paper Title

The Impact of Uncertainty Visualization on Trust in Thematic Maps

Publication Info

• Topic area: The effect of uncertainty visualization on trust in thematic maps among non-expert audiences.
• Keywords: Uncertainty visualization, thematic maps, trust, geovisualization, MAPTRUST framework, non-expert audiences, attribute uncertainty, cognitive trust, cartography, data visualization.

Background and Problem

• Problem / challenge: Uncertainty in thematic maps is rarely visualized, and its impact on trust remains underexplored. Prior studies offer mixed perspectives on whether visualizing uncertainty fosters or reduces trust.
• Significance: Understanding how uncertainty visualization affects trust is critical for designing credible maps, especially as thematic maps are widely used in public communication.
• Motivation and related work: Previous research has focused on expert users or specific domains like meteorology, often treating trust as an implicit outcome. This study addresses the gap by explicitly measuring trust in thematic maps among non-expert audiences using the MAPTRUST framework.

Solution

• Proposed approach: A between-subjects experiment using the MAPTRUST framework to measure trust in thematic maps with and without uncertainty visualization, varying the levels of uncertainty (low, medium, high).
• Novelty:
  1. Systematic measurement of trust in thematic maps using a validated multidimensional scale.
  2. Examination of how different levels of uncertainty visualization affect trust.
  3. Focus on non-expert audiences and public-facing thematic maps.
  4. Use of a controlled experimental design to isolate the effects of uncertainty visualization.
• Procedure and key techniques:
  • Participants (N = 161) were divided into two groups: one viewing maps without uncertainty and the other viewing maps with uncertainty visualized using symbol fuzziness at three levels (low, medium, high).
  • Six thematic topics were represented across maps: social, environment, health, crime, economic, and housing.
  • Trust was measured using 12 MAPTRUST adjectives rated on a 7-point Likert scale.
  • Statistical analysis employed Cumulative Link Mixed Models (CLMM) to assess the effects of uncertainty visualization and its levels on trust.

Results

• Concrete findings:
  • Visualizing uncertainty reduced trust, with greater reductions observed as uncertainty levels increased.
  • Low levels of uncertainty visualization were trust-neutral, showing no significant difference from maps without uncertainty.
  • Uncertainty visualization primarily affected perceptions of accuracy, while perceptions of dependability and reliability were less impacted.
• Advantage over baselines:
  • Maps with no uncertainty visualization received higher trust ratings than those with moderate or high uncertainty.
  • Low uncertainty visualization did not significantly differ from no uncertainty in trust ratings.
• Experiments / evaluation:
  • Participants rated six maps each, with trust measured using the MAPTRUST framework.
  • Statistical models controlled for map themes and participant-level variability.
  • Results were consistent across six thematic topics and three uncertainty levels.
• Limitations and future work:
  • Results may not generalize to other geographic scales, controversial topics, or alternative uncertainty visualization techniques.
  • Synthetic uncertainty levels were used, which may differ from real-world patterns.
  • Trust was measured through self-reported ratings, which may not fully capture real-world decision-making.
  • Future work should explore numeric uncertainty representations, real-world data, and longitudinal trust formation.

Summary

This study investigates how visualizing uncertainty in thematic maps affects trust among non-expert audiences. Using the MAPTRUST framework, the authors find that uncertainty visualization generally reduces trust, with greater reductions at higher uncertainty levels. Low uncertainty visualization is trust-neutral, while moderate and high levels significantly erode trust, particularly in accuracy-related dimensions. These findings highlight the need for careful design of uncertainty representations to balance transparency and credibility. Future research should explore broader contexts, alternative visualization techniques, and real-world applications to refine these insights.