Causal inference on human behaviour.

Area of Science:

• Behavioral Science
• Causal Inference
• Research Methodology

Background:

• Human behavior results from intricate interactions between internal states and external factors.
• Establishing causal links in behavior is challenging due to complexity and confounding variables.

Purpose of the Study:

• To provide a conceptual overview of challenges and opportunities in causal inference for human behavior.
• To propose a robust methodology for improving the accuracy of causal claims in behavioral research.

Main Methods:

• Conceptual analysis of challenges in causal inference (ambiguity, control, heterogeneity, interference, timescales, complex treatments).
• Discussion of how methods addressing one challenge can worsen others.
• Proposal of a triangulation approach comparing diverse data sources.

Main Results:

• Identified key challenges hindering accurate causal inference in human behavior studies.
• Demonstrated that optimizing for one methodological challenge can negatively impact others.
• Proposed triangulation as a method to systematically investigate discrepancies in causal estimates.

Conclusions:

• Clearly specified research questions are crucial for enhancing causal inference from data.
• A triangulation approach, comparing (quasi-)experimental and observational data with theoretical assumptions, enables systematic evaluation of causal estimates.