Probabilities of causation (PoCs) quantify how likely one event is to have caused another. The concept originated with Pearl's binary measures and has since matured into a broader theory, with Shu et al. recently extending it to non-binary settings. A new arXiv paper, 'Unified Sharp Bounds on Probabilities of Causation,' aims to consolidate these developments.
According to the abstract, the paper offers unified sharp bounds on these probabilities. Sharp bounds are the tightest possible limits consistent with the available data, so the result is significant for any application that relies on causal inference. The abstract notes that PoCs have found applications in a range of fields, though the specific domains are not detailed in the preview.
Because the abstract is brief, the exact mathematical techniques and proof strategies are not described. The paper's contribution appears to be theoretical: it brings together earlier binary and non-binary approaches under a single bounding framework.