Computing Actual Causes for Neural Network Predictions under Structured Causal Inputs
Explaining the predictions of neural networks is a central challenge in trustworthy AI. Existing explanation methods, such as those based on feature attribution or minimal sufficient sets, typically treat input features as independent, which can yield misleading explanations when inputs exhibit structured dependencies. We address this by formalizing explanations as Halpern-Pearl (HP) actual causes, modeling input dependencies using Boolean Structural Causal Models (SCMs). We compute HP causes by applying bound propagation and branch-and-bound techniques, while providing formal guarantees of completeness and minimality. Our experiments show that we substantially outperform brute-force and ILP baselines in scalability, and outperform heuristic search as graph size grows, computing all minimal actual causes on instances with search spaces of up to $2.3\times10^{13}$ candidate (cause, contingency) pairs, on SCMs with up to 28 nodes, within a 180s per-instance budget. In a case study, we further show that ignoring input dependencies inflates the number of reported causes, 14.9% of which are spurious under our SCM.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
Efficient Discovery of Actual Causality with Uncertainty
Identifying the actual cause of events in engineered systems is a fundamental challenge in system analysis. Finding such causes becomes more challenging in the presence of noise and uncertainty in real-world systems. In …
Evaluating the extent to which homeostatic plasticity learns to compute prediction errors in unstructured neuronal networks
The brain is believed to operate in part by making predictions about sensory stimuli and encoding deviations from these predictions in the activity of "prediction error neurons." This principle defines the widely influen…
PredictionDevelopment of a Medical Incident Report Corpus with Intention and Factuality Annotation
Medical incident reports (MIRs) are documents that record what happened in a medical incident. A typical MIR consists of two sections: a structured categorical part and an unstructured text part. Most texts in MIRs descr…
Measuring Causal Effects of Data Statistics on Language Model's `Factual' Predictions
Large amounts of training data are one of the major reasons for the high performance of state-of-the-art NLP models. But what exactly in the training data causes a model to make a certain prediction? We seek to answer th…
From Causes for Database Queries to Repairs and Model-Based Diagnosis and Back
In this work we establish and investigate connections between causes for query answers in databases, database repairs wrt. denial constraints, and consistency-based diagnosis. The first two are relatively new research ar…