By the AGI Safety Analysis Team @ DeepMind
Probability trees are one of the simplest models of causal generative processes. They possess clean semantics and are strictly more general than causal Bayesian networks, as they are able to e.g. represent causal relations that causal Bayesian networks cannot. Yet, they have received little attention from the AI and ML community. Here we present new algorithms for causal reasoning in discrete probability trees that cover the entire causal hierarchy (association, intervention, and counterfactuals), and operate on arbitrary propositional and causal events. Our work expands the domain of causal reasoning to a very general class of discrete stochastic processes.
For details, see our paper Algorithms for Causal Reasoning in Probability Trees.
The accompanying colab is available here:
If you use the code here please cite this paper.
Tim Genewein*, Tom McGrath*, Grégoire Delétang*, Vladimir Mikulik*, Miljan Martic, Shane Legg, Pedro A. Ortega. [arXiv]