Dispersion vs Disparity: Hiding Variability Can Encourage Stereotyping When Visualizing Social Outcomes

Area of Science:

• Data Visualization
• Social Psychology
• Human-Computer Interaction

Background:

• Research often overlooks how chart designs influence perceptions of individuals represented in data.
• Chart designs may interact with social cognitive biases, potentially perpetuating harmful stereotypes.
• Bar charts, common for social inequality data, may promote deficit thinking by attributing disparities to personal factors rather than external causes.

Purpose of the Study:

• To investigate how specific data visualization design choices influence attribution biases.
• To examine if chart designs can mitigate or exacerbate deficit thinking and stereotyping in social inequality contexts.

Main Methods:

• Conducted four experiments using crowdworkers.
• Participants viewed visualizations (bar charts, dot plots, jitter plots, prediction intervals) of social outcomes.
• Evaluated agreement with personal vs. external explanations for visualized disparities.

Main Results:

• Visualizations masking within-group variability (e.g., bar charts) led to greater agreement with personal explanations.
• Visualizations emphasizing within-group variability (e.g., jitter plots) led to less agreement with personal explanations.
• Design choices significantly influenced attribution biases and potential for stereotyping.

Conclusions:

• Data visualizations of social inequity can be misinterpreted, leading to harmful stereotyping.
• Hiding data variability in visualizations increases stereotyping; emphasizing variability reduces it.
• Conscious design choices in data visualization can mitigate biased interpretations and reduce stereotyping.