Historical development and key researchers
The concept of using tree-based methods for causal inference has evolved over the past few decades, influenced by advancements in both machine learning and statistical theory.
In that spirit, we’ve come a long way with decision trees in machine learning since the 1980s. They started as simple tools for prediction but have evolved into powerful instruments for understanding cause and effect. In the early days, algorithms such as CART and ID3 made decision trees popular for classification and regression tasks [1, 2, 8]. Over time, they became more versatile, handling complex scenarios such as multi-label learning.
The real game-changer came in the 2000s when researchers started exploring how decision trees could uncover causal relationships. In 2016, Susan Athey and Guido Imbens introduced causal trees, which revolutionized how we partition data based on treatment effects rather than just outcomes [3, 9]. Building on this, Athey...
