Invited commentary: Agent-based models for causal inference—reweighting data and theory in epidemiology

Area of Science:

• Epidemiology
• Causal Inference
• Computational Modeling

Background:

• The balance between empirical data and theoretical assumptions in causal inference differs across scientific fields.
• Complex research questions often necessitate a greater reliance on theory and modeling.

Purpose of the Study:

• To explore the role of agent-based modeling in epidemiology for causal inference.
• To highlight the parametric g-formula as a transitional method towards more complex modeling approaches in epidemiology.

Main Methods:

• Discussion of the evolving landscape of causal inference in epidemiology.
• Introduction of agent-based modeling as a tool for integrating theory and data.
• Positioning the parametric g-formula as an intermediate step between traditional methods and agent-based modeling.

Main Results:

• Agent-based modeling allows for a greater incorporation of theory alongside empirical data in epidemiological studies.
• The parametric g-formula provides a practical pathway for epidemiologists to adopt modeling-intensive causal inference techniques.

Conclusions:

• Epidemiology is moving towards more complex questions, requiring advanced causal inference methods.
• Agent-based modeling and the parametric g-formula represent key advancements in this transition, enhancing the integration of data and theory.