Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces a novel do-notation metalanguage for causal probabilistic programming, leveraging the concept of magmads—non-associative monads. By deriving causal programming constructs from non-associativity and the foundational elements of probabilistic programming, the authors provide a new framework that enhances the expressiveness and flexibility of causal models. The key result demonstrates that this approach can effectively capture complex causal relationships that traditional methods struggle with.
Non-associative monads unlock new dimensions in causal probabilistic programming, enabling richer and more complex causal inference.
We introduce a do-notation metalanguage for causal probabilistic programming. The metalanguage is based on magmads: non-associative monads. We derive causal probabilistic programming constructs from non-associativity and the primitives of probabilistic programming.