A Visual Analytics Approach for Exploratory Causal Analysis: Exploration, Validation, and Applications.

Area of Science:

• Data Science
• Human-Computer Interaction
• Causal Inference

Background:

• Causal relations are crucial for decision-making across diverse fields like marketing and medicine.
• Existing statistical models for inferring causality lack effective visual interfaces for practitioners.
• Domain experts need better tools to interpret and apply complex causal relationships in their workflows.

Purpose of the Study:

• To develop and evaluate a visualization tool that bridges the gap between causal inference models and practical decision-making.
• To provide domain practitioners with an intuitive interface for exploring, validating, and applying causal relations.
• To address the need for uncertainty-aware visualization of causal graphs derived from high-dimensional data.

Main Methods:

• Conducted interview studies with domain experts to understand decision-making workflows and needs.
• Employed an iterative design process to develop a visualization tool.
• Integrated uncertainty-aware causal graph visualization with interactive controls for "what-if" analyses and action planning.

Main Results:

• Developed a visualization tool enabling exploration, validation, and application of causal relations.
• The tool visualizes large sets of causal relations inferred from high-dimensional data.
• Case studies in marketing and student advising demonstrated effective use for goal-oriented action planning.

Conclusions:

• The developed visualization tool empowers analysts to effectively interpret and utilize causal relations for decision-making.
• The tool facilitates "what-if" scenario analysis and the creation of actionable plans.
• This approach enhances the practical application of causal inference in real-world scenarios across various domains.