Beyond Correlation: Incorporating Counterfactual Guidance to Better Support Exploratory Visual Analysis

Area of Science:

• Data Visualization
• Human-Computer Interaction
• Causal Inference

Background:

• Effective guidance in exploratory visual analytics is crucial for high-dimensional data.
• Correlation-based guidance can lead to misinterpretations of causal relationships.

Purpose of the Study:

• To propose and evaluate a novel counterfactual guidance method for exploratory visual analytics.
• To enhance causal inference performance and user understanding in complex datasets.

Main Methods:

• Developed a simple and efficient counterfactual guidance technique.
• Integrated the method into an exploratory visual analytics system.
• Conducted a comparative user study using synthetic causal data.

Main Results:

• Counterfactual guidance significantly improved visual causal inference accuracy.
• Users exhibited different exploratory behaviors compared to correlation-based guidance.
• The method demonstrated potential in reducing false positives in causal interpretations.

Conclusions:

• Counterfactual guidance offers a promising approach to enhance exploratory visual analytics.
• Further research is needed to explore its full potential and address implementation challenges.