Do You "Trust" This Visualization? An Inventory to Measure Trust in Visualizations

Area of Science:

• Data Visualization
• Human-Computer Interaction
• Psychology

Background:

• Trust is crucial in visual data communication and decision-making.
• Existing trust measures in visualization research are inconsistent, hindering cross-study comparisons.
• A unified understanding of 'trust' in visualizations is needed.

Purpose of the Study:

• To operationally define trust in data visualizations through a data-driven approach.
• To develop and validate a reliable and valid inventory for measuring trust in visualizations.
• To provide a standardized tool for future visualization research.

Main Methods:

• Compiled and adapted trust-related statements from existing inventories.
• Collected reader responses to visualizations with varying trustworthiness.
• Utilized exploratory factor analysis to derive an operational definition of trust.
• Developed an eight-item inventory (four core, four optional).
• Assessed reliability (McDonald's omega) and validity (content and criterion) through trust games.

Main Results:

• An operational definition of trust emerged: credible information, comprehensibility, and usability.
• The developed eight-item inventory demonstrated strong reliability and validity.
• The inventory effectively measures trust in visualizations across different contexts.
• Criterion validity was confirmed through trust games with real-world stakes.

Conclusions:

• A standardized inventory for measuring trust in data visualizations has been established.
• This tool enables consistent evaluation of how design, tasks, and domains impact visualization trust.
• Future research can use this inventory to foster appropriate trusting behaviors in human-data interactions.