In the recent rather pedantic exchange regarding inference of causality from data that I had with Matt, it became apparent that what seems obvious to me is not so obvious to even someone as otherwise knowledgeable and reasonable as Matt. While I certainly can, and, in fact, do attribute this to Matt suffering from Crimestop, he is certainly not alone. Resistance to reforming the natural sciences with the Algorithmic Information Criterion for causal model selection is so widespread that it isn't even considered worthy of resisting!
So I went looking around for literature on the topic that may be sufficiently pedantic to, in a sense, "speak the lingo" of pedantry. What I came up with was the widely cited "Causal inference using the algorithmic Markov condition <https://arxiv.org/abs/0804.3678>" by Janzing and Schölkopf. That paper argues for a method, based on computable approximation of Kolmogorov Complexity, to select from among different directed Acyclic graphs that model a standard dataset. This is, of course, reminiscent of the, in my opinion, profoundly misleading "MDL principle" as set forth by Rissanen in which the descriptive codes were not Turing complete <https://doi.org/10.1016%2F0005-1098%2878%2990005-5>. However, like Judea Pearl's approach to causality, Janzing and Schölkopf are stuck with the directed Acyclic graph -- which is quite a puzzle since such models are incapable of so much as a *fiction* of Turing complete codes (as in computer instruction sets) that execute on a finite state machine. But at least their pedantry addresses some of Matt's and does so even without the requirement of directed Cyclic graphs to model datasets. My conversation with 3.7 Sonnet explores the nuances: https://claude.ai/share/0ad9d742-fd44-4824-a3eb-f1cb3e4742a2 ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T0f47884dae19d52d-M9eb451aca570de3f5fb4e8eb Delivery options: https://agi.topicbox.com/groups/agi/subscription
