DeepMind Introduces Algorithms for Causal Reasoning in Probability Trees

Thanks for sharing.

What are the areas where this approach would provide the most benefit Vs SCMs and Bayesian Networks? What are the trade-offs?

Linking to their tutorial notebook: https://colab.research.google.com/github/deepmind/deepmind_r...

Questions: 1. Has anyone used a package like "bnlearn" in R to guide the structure learning of one of these starting from data? 2. Would be curious about their relationship to CART models, Random Forest models and Causal Decision Trees (http://nugget.unisa.edu.au/jiuyong/CDT-Revision.pdf) 3. This feels like an important limitation Vs SCMs: "A system of transformations akin to the do-calculus is entirely absent and left for future work. Such a system is necessary for addressing questions about identification, i.e. whether a causal effect can be estimated from observation alone"