Brighton Summary - October 2024

Uncover & mitigate bias in data visualisations

Key Takeaways

  1. Three Core Areas of Bias in Data Visualization:

    • Data
    • Cognitive
    • Visual

  2. Data Bias:

    • Survivorship bias often occurs due to incomplete datasets
    • Understanding accuracy and precision helps identify survivorship bias
    • Critical thinking is crucial to identify the extent of survivorship bias

  3. Cognitive Bias:

    • The human brain processes 35,000 decisions daily, leading to cognitive shortcuts
    • Key cognitive biases: a. Framing effect b. Causal illusion c. Anchoring effect

  4. Visual Bias:

    • Can be mitigated with a high level of control
    • Key visual biases: a. Scale manipulation b. Correlation illusion c. Visual distortion d. Color choice

  5. Accessibility in Data Visualization:

    • Critical for mitigating visual bias
    • Key accessibility features: a. Data markers b. Patterns c. Alt text d. Tables and shapes

Action Items

  1. Assess Data Quality:

    • Evaluate datasets for completeness and potential survivorship bias
    • Use accuracy and precision metrics to understand data quality

  2. Implement Critical Thinking:

    • Develop a framework for critically assessing data, especially for sources like Google Analytics

  3. Address Cognitive Biases:

    • Be aware of framing effects when presenting data
    • Avoid implying causation without proper evidence
    • Provide full context to mitigate anchoring effects

  4. Improve Visual Representations:

    • Use appropriate scales in charts to avoid trend exaggeration
    • Be cautious with trend lines and ensure they accurately represent data
    • Choose chart types that minimize visual distortion (e.g., bar charts over pie charts)

  5. Enhance Color Usage:

    • Select colors that don't introduce cognitive biases
    • Ensure color choices are accessible to all users, including those with color vision deficiencies

  6. Implement Accessibility Features:

    • Add symbol markers to charts for visual distinction
    • Use patterns in bar charts to differentiate data points
    • Provide alt text for all data visuals
    • Use tables and shapes to aid in data comprehension

  7. Practice Balanced Reporting:

    • Present both positive and negative aspects of data trends
    • Include relevant context when showcasing improvements or declines

  8. Verify Correlations:

    • Use statistical methods (e.g., p-values) to confirm correlations before presenting them
    • Investigate potential confounding variables in apparent correlations

  9. Standardize Visualization Practices:

    • Develop guidelines for data visualization within your organization
    • Ensure consistent timeframes and scales when comparing data

  10. Continuous Learning:

    • Stay updated on best practices in data visualization
    • Regularly review and update your visualization techniques

Remember: While we cannot eliminate bias entirely, we can work to uncover and mitigate it through careful analysis, thoughtful presentation, and accessible design in our data visualizations.

For more information: