AI
On Formal Counterfactual Explanations
Publication: cs.AI updates on arXiv.org
Author: Alexey Ignatiev, Peter J. Stuckey, Joao Marques-Silva
Source published: 18/11/2026, 05:00:00 UTC
Briefing created: 18/11/2026, 05:49:23 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 18/11/2026, 05:03:25 UTC.
Briefings
A 2026 arXiv paper bridges closest-world counterfactual explanations with feature-based abductive explanations by introducing conjunctive counterfactuals (CFXps). The study characterizes closest-world counterexamples as disjunctive normal form over minimally distant instances and establishes that every CFXp induces an AXp, with support precisely determined by that AXp. Experiments on tree ensembles and neural networks confirm CFXp extraction is practically efficient, requiring fewer feature changes than point-based explainers and guaranteeing prediction change beyond a single instance.
Key points
- CFXps are a subclass of counterfactual antecedents restricted to conjunctions of feature literals that guarantee prediction change.
- Every CFXp induces an AXp, and every CFXp together with a witness AXp induces a CFXp whose support is precisely determined by that AXp.
Context
This research addresses formal counterfactual explanations in machine learning, bridging closest-world semantics with feature-based abductive explanations to enable efficient computation and minimal hitting-set duality.
Limitations
Evidence is limited to a single arXiv preprint from November 2026; experimental results are based on a representative set of benchmarks using tree ensembles and neural networks, with no further validation or real-world deployment data provided.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 18/11/2026, 05:49:23 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeiguxsqre44lx4aihffiwia5tccnvx6uds7yfe2otwohf3mtjoy52q