Search papers, labs, and topics across Lattice.
This paper introduces a formal framework for understanding causal relationships through the lens of symmetries in data, aiming to extend causal reasoning beyond the limitations of IID data. By establishing a general mathematical language for causal effect identification, the authors provide a robust methodology that aligns with existing theoretical results while addressing complex causal queries that traditional methods struggle to capture. The findings reveal that this new approach not only clarifies well-known causal structures but also enhances the analysis of missing data and the robustness of causal models in diverse scenarios.
Uncovering causal relationships through symmetries could revolutionize how we approach complex machine-learning tasks beyond IID constraints.
In the natural sciences, symmetries and cause-effect relationships are ubiquitous. Yet for complex machine-learning tasks, like world-modeling in reinforcement learning, they appear difficult to harness. We propose a formal description of statistical systems based on symmetries in data leaving causal mechanisms invariant. The result is an abstract, simple and general mathematical language for causal reasoning. This paper provides formal descriptions of models and queries, setting up this language, and the formal infrastructure and strategies for their mathematically rigorous identification from data within this formalism. This approach reproduces and matches standard theoretical results on IID data and transport of experimental and non-experimental data. But its main purpose is to unify and substantially extend the scope of causal reasoning, in going beyond IID data and in approaching complex causal queries not captured by do- or soft-interventions. This new perspective on causally relevant aspects of data-modeling additionally sheds new light on well-known structures like c-components or hedges but also includes aspects of missing data and is inherently well-suited for the description of transfer and robustness properties.
