Interactions of information transfer along separable causal paths

Area of Science:

• Complex systems analysis
• Information theory
• Causal inference

Background:

• Complex systems exhibit interdependencies between variables, often visualized in time-series graphs.
• Existing methods like transfer entropy and information partitioning capture net information transfer, obscuring pathway-specific causal interactions.
• Discerning causal interactions within specific subgraphs requires methods that can isolate information flow through distinct pathways.

Purpose of the Study:

• To develop a framework for quantifying information partitioning along separable causal paths in complex systems.
• To expand the concept of momentary information transfer to characterize synergistic, unique, and redundant information flow.
• To analyze the impact of separable/nonseparable paths, causality structure, and noise on information partitioning.

Main Methods:

• Building upon momentary information transfer along causal paths.
• Developing a framework for quantifying information partitioning along separable causal paths.
• Utilizing graphical models and synthetic data from coupled logistic equations for analysis.

Main Results:

• The proposed framework quantifies synergistic, unique, and redundant information transfer through separable causal paths.
• Analysis reveals the impact of separable and nonseparable causal paths, graph causality structure, and noise on information partitioning.
• Demonstrated the utility of the approach using synthetic data from coupled logistic models.

Conclusions:

• The developed framework offers a method for autonomous information partitioning along separable causal paths.
• This approach provides a valuable reference for understanding causal subgraphs influencing a target variable.
• Enables a more granular understanding of information flow in complex systems by dissecting causal pathways.