Section Ethics Spotlight: Avoiding Misleading Visualizations
Visualizations can powerfully communicate data insights, but they can also mislead viewers if not created responsibly. Understanding common ways visualizations can mislead helps us create more ethical data presentations.
Common misleading visualization practices include:
- Truncated Axes
Starting bar charts or line charts at values other than zero can exaggerate differences. While non-zero baselines can be appropriate for line charts showing trends, they should be clearly indicated.
- Manipulated Aspect Ratios
Changing the height-to-width ratio of a chart can make trends appear steeper or flatter than they actually are.
- Cherry-Picking Data
Selectively including or excluding data points to support a particular narrative rather than showing the complete picture.
- Problematic Color Scales
Using colors that imply judgment (e.g., red/green) for neutral data, or using color scales that perceptually distort the data.
- 3D Charts for 2D Data
Using 3D effects for 2D data can distort proportions and make accurate comparisons difficult.
- Inappropriate Chart Types
Using chart types that are not suited to the data or question, leading to misinterpretation.
- Missing Context
Failing to provide necessary context, such as sample sizes, time periods, or relevant comparisons.
To create ethical visualizations:
Present data completely and accurately
Choose appropriate scales and chart types
Provide necessary context and uncertainty information
Use color and design elements thoughtfully
Consider diverse audience perspectives
Be transparent about data sources and processing
Example 94. Ethical Considerations in Community Health Visualization.
When visualizing our Community Health data, ethical considerations might include:
Ensuring bar charts comparing health outcomes across neighborhoods start at zero to avoid exaggerating differences
Including confidence intervals or other uncertainty indicators when showing health metrics based on small sample sizes
Using culturally sensitive color schemes when mapping data by neighborhood to avoid reinforcing stereotypes
Providing context about historical factors that might explain environmental disparities rather than just showing current differences
Being careful about implying causation when visualizing correlations between environmental factors and health outcomes
Including complete data rather than cherry-picking neighborhoods or time periods that support a particular narrative
Checkpoint 95. Identifying Misleading Visualizations.
For each visualization description, identify the primary way it might mislead viewers.
Activity 28. Creating Ethical Alternatives.
In this activity, you’ll practice identifying and correcting misleading visualizations.
(a)
Working with a partner, create a deliberately misleading visualization using your dataset (or the Community Health dataset) in CODAP. Use at least one of the misleading techniques discussed.
(b)
Exchange your misleading visualization with another pair of students. For the visualization you receive:
Identify how the visualization might mislead viewers
Create an ethical alternative that presents the same data more accurately
Explain why your alternative is more ethical and effective
(c)
Discuss as a class: What responsibility do data scientists have to create ethical visualizations? How can we balance creating impactful visualizations with ensuring they’re not misleading?
