Position: Probabilistic Modelling is Sufficient for Causal Inference

Bruno Mlodozeniec, David Krueger, Richard E Turner

International Conference on Machine Learning 2025 · Oral

This talk, presented by Bruno Mlodozeniec and co-authored with David Krueger and Richard E Turner, challenges a foundational debate in machine learning and statistics: whether specialized "causal tools" are strictly necessary for answering causal questions. The central thesis is that standard probabilistic modeling, when applied rigorously and comprehensively, is entirely sufficient for performing causal inference. This contentious claim directly confronts the long-held views of prominent researchers like Judea Pearl, who has famously advocated for a distinct causal-statistical divide and the necessity of operators like the `do-operator`.

AI review

A competent and clearly argued philosophical position paper that challenges Pearl's causal-statistical distinction by reframing causal inference as a special case of probabilistic modeling over expanded variable sets. The central thesis — that the do-operator and SCMs are 'syntactic sugar' over a sufficiently general joint distribution — is not technically wrong, but it is also not new. The work occupies a well-trodden space between Dawid's decision-theoretic approach, Rubin's potential outcomes framework, and the long-standing observation that interventional distributions can be represented…