Home > Research > Causality

Causality

Causal inference and machine learning are two important fields of study in data science. Causal inference is the process of identifying and measuring the causes of events, while machine learning is the process of teaching computers to learn from observational data. These two fields are complementary, and together they allow us to better understand and make predictions about complex phenomena. However, the estimation of causal queries has been limited by techniques defined around estimands, formulas that translate causal expressions into observational terms that can be computed from observational data.

In other to tackle these limitations, we have developed Deep Causal Graphs (DCGs), a general estimand-agnostic framework capable of answering any identifiable causal query from one single model. Trained only once per dataset using the same general procedure, it can adapt to many kinds of causal queries through three simple methods.