Aug 06 2025 The Illusion of Causality in Charts

• Correlation ≠ Causation: Humans often misinterpret correlational data as causal, and charts can amplify this bias depending on design choices
• Visual Encoding Effects: Bar graphs and text summaries tend to evoke stronger causal inferences compared to line or scatter plots, which typically lead to weaker assumptions of causality
• Data Aggregation Bias: Charts that group data into fewer categories—especially two-bar bar charts—encourage stronger causal interpretations than less aggregated visuals
• Chart Type Susceptibility: Scatter plots are most prone to misinterpretation as causal, followed by line charts, while bar charts (especially simple ones) are most likely to mislead without proper context
• Design Recommendations:
• Avoid two-bar bar charts when showing correlations
• Prefer raw-data visuals (e.g. scatter plots with minimal aggregation) to reduce causal illusion
• Consider presentation context: choice of layout, arrows, ordering and animations can implicitly suggest causality
• Human and Contextual Factors: Prior beliefs and cognition biases interact with chart design to shape causal interpretation, meaning visuals alone don't determine perception
• Broader Implications: Especially in business and policy contexts, misleading charts can trigger poor decisions—visual design has ethical consequences

The full post is available here.